CN119092852B

CN119092852B - Ternary polymer lithium battery management method of shared charging equipment with high cycle times

Info

Publication number: CN119092852B
Application number: CN202411580936.6A
Authority: CN
Inventors: 潘良春; 宋正江
Original assignee: Shenzhen Besiter Power Bank Co ltd
Current assignee: Shenzhen Besiter Power Bank Co ltd
Priority date: 2024-11-07
Filing date: 2024-11-07
Publication date: 2025-01-10
Anticipated expiration: 2044-11-07
Also published as: CN119092852A

Abstract

The invention relates to a ternary polymer lithium battery management method of shared charging equipment with high cycle times. The method comprises the steps of conducting graph theory modeling on a battery management system in shared charging equipment to construct a topological graph model, conducting depth-first search, identifying a charging and discharging path, calculating a complex impedance value, calculating dynamic voltage spectrum and impedance characteristics to obtain time domain current distribution, conducting multi-factor coupling loss calculation on each ternary polymer lithium battery unit according to the time domain current distribution to obtain dynamic loss distribution data, conducting battery state real-time assessment on the basis of the dynamic loss distribution data, reconstructing a nonlinear charging and discharging characteristic curve, and conducting multi-time scale collaborative scheduling on the basis of the nonlinear charging and discharging characteristic curve to obtain a battery charging and discharging control strategy. The implementation of the invention can accurately model the nonlinear characteristics of the battery, evaluate the state of the battery in real time, and perform intelligent scheduling based on the dynamic prediction result so as to maximize the service life of the battery and the overall efficiency of the system.

Description

Ternary polymer lithium battery management method of shared charging equipment with high cycle times

Technical Field

The invention relates to the technical field of lithium battery management, in particular to a ternary polymer lithium battery management method of shared charging equipment with high cycle times.

Background

Shared charging devices are rapidly spreading as an emerging mobile energy solution. However, as the frequency of use increases, battery management presents a significant challenge. Traditional battery management systems often have difficulty in coping with complex working conditions under high-frequency use occasions, resulting in reduced battery performance, shortened service life and even potential safety hazards. In addition, the existing battery management method is mostly based on a simplified model, and cannot accurately describe the nonlinear characteristics of the battery under the condition of high cycle times, so that the battery capacity estimation deviation and the low charge and discharge efficiency are caused.

Another significant problem is that existing battery management systems lack the ability to adapt to multi-time scale load changes. The usage pattern of the shared charging device has significant randomness and volatility, and the traditional fixed strategy is difficult to balance the contradiction between the instant demand and the long-term performance maintenance, which often results in waste of battery resources or degradation of service quality. Meanwhile, the inconsistency among the battery units is aggravated along with the increase of the use times, and the existing balancing strategy is difficult to effectively handle the dynamic change, so that the performance degradation of the battery pack is further aggravated.

Disclosure of Invention

The main object of the present invention is to provide a method for managing a ternary polymer lithium battery of a shared charging device with high cycle times, which is capable of accurately modeling the non-linear characteristics of the battery, and (3) evaluating the battery state in real time, and performing intelligent scheduling based on the dynamic prediction result so as to maximize the service life of the battery and the overall efficiency of the system.

In order to achieve the above object, the present invention provides a method for managing a ternary polymer lithium battery of a shared charging device with a high cycle number, comprising the steps of:

Performing graph theory modeling on a battery management system in the shared charging equipment to construct a topological graph model comprising vertexes representing the ternary polymer lithium battery unit and the management circuit and directed edges representing the current flow path;

performing depth-first search based on the topological graph model, identifying all charge and discharge paths meeting circuit rules, and calculating a complex impedance value of each charge and discharge path;

Calculating dynamic voltage spectrum and impedance characteristic based on the charge-discharge path and the complex impedance value to obtain time domain current distribution of each ternary polymer lithium battery unit;

according to the time domain current distribution, multi-factor coupling loss calculation is carried out on each ternary polymer lithium battery unit, and dynamic loss distribution data are obtained;

performing real-time battery state evaluation based on the dynamic loss distribution data, and reconstructing a nonlinear charge-discharge characteristic curve;

And carrying out multi-time scale cooperative scheduling based on the nonlinear charge-discharge characteristic curve to obtain a battery charge-discharge control strategy.

The invention also provides a solid-state battery management device of the shared charging equipment with high cycle times, which comprises:

The modeling module is used for carrying out graph theory modeling on the battery management system in the shared charging equipment and constructing a topological graph model comprising vertexes representing the ternary polymer lithium battery unit and the management circuit and directed edges representing the current flow path;

The identification module is used for carrying out depth-first search based on the topological graph model, identifying all charge and discharge paths meeting the circuit rule, and calculating the complex impedance value of each charge and discharge path;

The calculation module is used for calculating dynamic voltage spectrum and impedance characteristics based on the charge-discharge path and the complex impedance value to obtain time domain current distribution of each ternary polymer lithium battery unit;

the processing module is used for carrying out multi-factor coupling loss calculation on each ternary polymer lithium battery unit according to the time domain current distribution to obtain dynamic loss distribution data;

The evaluation module is used for carrying out real-time evaluation on the battery state based on the dynamic loss distribution data and reconstructing a nonlinear charge-discharge characteristic curve;

And the scheduling module is used for carrying out multi-time scale cooperative scheduling based on the nonlinear charge-discharge characteristic curve to obtain a battery charge-discharge control strategy.

The invention also provides a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of any of the methods described above when the computer program is executed.

The invention also provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the method of any of the preceding claims.

In summary, the technical scheme provided by the invention can comprehensively identify the effective charge and discharge paths by abstracting the battery management system into the topological graph model and combining the depth-first search algorithm. The coupling effect of the joule heat loss, the electrochemical reaction loss and the cyclic attenuation loss is considered, and the dynamic loss distribution of the battery can be more accurately estimated by simultaneous solving through a multi-physical field model. Is helpful to optimize the charge and discharge strategy and prolong the service life of the battery. Based on the self-adaptive OCV-SOC mapping model and on-line parameter identification, a nonlinear characteristic equation of the battery is dynamically updated. The method can adapt to the dynamic change of the battery performance, improves the accuracy of SOC estimation, combines the daily load prediction, daily dynamic adjustment and real-time fine adjustment, and realizes the omnibearing scheduling from long-term planning to instant response. The multi-layer scheduling strategy can effectively balance the service efficiency of the battery and prolong the service life of the battery, and improves the overall performance of the system. Deep learning models such as LSTM, one-dimensional convolutional neural network and GRU are adopted to respectively process long-term trend, periodic mode and short-term fluctuation, and the prediction results are integrated through a multi-layer perceptron fusion network. The composite prediction method can more accurately capture the load change characteristics, and continuously update the battery state evaluation model and the nonlinear characteristic curve by continuously recording the data of the charge and discharge process, thereby realizing the self-adaptive optimization of the management strategy. The invention can ensure that the battery management system always maintains the optimal performance and adapts to the change of the whole life cycle of the battery.

Drawings

FIG. 1 is a schematic diagram of a method for managing a ternary polymer lithium battery of a shared charging device with high cycle times according to an embodiment of the present invention;

fig. 2 is a block diagram showing a structure of a solid-state battery management device of a shared charging apparatus with a high cycle number according to an embodiment of the present invention;

Fig. 3 is a block diagram schematically illustrating a structure of a computer device according to an embodiment of the present invention.

The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

Referring to fig. 1, the present embodiment provides a method for managing a high cycle number ternary polymer lithium battery sharing charging equipment, including the following steps:

S1, carrying out graph theory modeling on a battery management system in shared charging equipment, and constructing a topological graph model comprising vertexes representing a ternary polymer lithium battery unit and a management circuit and directed edges representing a current flow path;

the ternary polymer lithium battery unit, the charging circuit, the discharging circuit and the protection circuit of the battery management system in the shared charging equipment are identified, and the type and parameter information of each component, such as the capacity, the internal resistance and the charge state of the battery and the electrical characteristics of each circuit, are obtained. Each of the lithium-ion-ternary battery cells and circuit components is vertex modeled based on the type and parameter information of the respective components. In this process, each cell and circuit assembly is abstracted as a vertex, all of which together make up a vertex set. While constructing the vertex set, the current flow paths between the components in the battery management system are analyzed to determine the interconnection relationship between the components. By analyzing the path of the current flowing from the battery cell through the various circuit components, the direction of current flow between the components is determined, and based thereon, a set of directed edges is established, which represent the channels through which the current flows. And carrying out attribute assignment on each vertex in the vertex set. In the assignment of vertex attributes, each vertex is assigned a particular attribute of its corresponding component. For example, the attribute data for a cell vertex should include its rated capacity, internal resistance, and current state of charge, while for a circuit component vertex, the attributes include the type of component and its electrical parameters. And meanwhile, carrying out attribute assignment on each edge in the directed edge set. The attribute data of the edges mainly comprise the current flowing direction, the maximum allowable current value and the current actual current value, so that the actual flowing condition of the current in the system can be reflected. Based on the vertex set, the directed edge set containing the vertex attribute data and the edge attribute data, a topological graph model of the whole battery management system is finally constructed, and the connection relation between each battery unit and the circuit component in the system and the flow path of current in the whole system are completely described.

S2, performing depth-first search based on a topological graph model, identifying all charge and discharge paths meeting circuit rules, and calculating a complex impedance value of each charge and discharge path;

Specifically, a vertex set in the topological graph model is encoded to obtain a unique identifier of each vertex. At the same time, an access status flag is set for each vertex to track whether the vertex was accessed during the search. Depth-first searching is achieved by constructing a recursive function that will be responsible for progressively exploring all possible paths along the directed edges in the graph, starting from any vertex. Based on the traversal process of the recursive function, all vertexes in the topological graph model are accessed one by one, and each vertex access sequence is recorded. Each time a new vertex is accessed, verification is performed according to circuit rules to ensure that the path corresponding to the access sequence meets the physical and circuit constraints of current flow. Only verified paths will be considered valid paths, which will be collected into a valid set of charge and discharge paths. Along with the continuous execution of the recursive function, all charge and discharge paths meeting the conditions are identified and recorded to form a charge and discharge path set. Based on the effective charge-discharge path set, the resistance, inductance and capacitance of each charge-discharge path are extracted. And analyzing the characteristics of the components on each path to obtain a distribution parameter model related to the path, and reflecting the influence of each component on the path on current transmission. And carrying out mathematical transformation on the distribution parameter model to obtain a complex impedance expression of each path in a frequency domain. The expression integrates the interaction of the resistor, the inductor and the capacitor to form an equivalent model in a frequency domain, and can accurately reflect the impedance characteristics of the path under different frequencies. And calculating the equivalent impedance of the charge-discharge path in the working frequency range based on the frequency domain complex impedance expression of each path. The frequency response characteristics of each component on the path are considered in the calculation process, and the complex impedance value of the path under the actual working condition is obtained through the analysis of the impedance in the frequency domain. The complex impedance value of each path accurately describes the energy loss and other relevant characteristics of the path during current flow.

S3, calculating dynamic voltage spectrum and impedance characteristic based on the charge-discharge path and the complex impedance value to obtain time domain current distribution of each ternary polymer lithium battery unit;

It should be noted that, the topology structure of the charge-discharge path is analyzed, and a time-differential equation set capable of describing the dynamic behavior of the whole battery management system is obtained by analyzing the interaction between the battery unit and each circuit element on the path. The time-differential equation set can reflect the time-varying relation of current and voltage in the system. And converting the time domain differential equation set into an s-domain transfer function to obtain a frequency domain mathematical model of the battery management system. Based on the frequency domain mathematical model and the complex impedance value, a system matrix, an input matrix, an output matrix and a feedforward matrix of the battery management system are constructed. These matrices can characterize the input-output relationship of the system and the flow of current between the components. The matrices are discretized to convert them from continuous time models to discrete time system models, which are suitable for simulation and control in a digitized environment. And carrying out parameter identification on the discrete time system model. By analyzing the model, the characteristic parameters of the system are identified, and the dynamic characteristics of the battery management system are accurately reflected. Based on system characteristic parameters, a broadband excitation signal is applied to the system, which can cover multiple frequency ranges, so as to completely observe the response of the system at each frequency. After the excitation signal is applied, the output of the system is analyzed by utilizing the fast Fourier transform, so that a dynamic voltage spectrum can be calculated, and the dynamic voltage spectrum reflects the voltage response of the battery management system under different frequencies. Based on the dynamic voltage spectrum and the input current information, the impedance characteristics of the battery management system are calculated. The impedance characteristics reflect the electrical characteristics of the battery system at different frequencies, especially the energy loss and current distribution during charge and discharge. In order to convert the result of frequency domain calculation back to the time domain, the impedance characteristics are subjected to inverse Fourier transform, the impedance information of the frequency domain is restored to the time domain, and the time domain current distribution of each ternary polymer lithium battery unit is finally obtained by combining the topology structure information of the circuit.

S4, according to the time domain current distribution, multi-factor coupling loss calculation is carried out on each ternary polymer lithium battery unit, and dynamic loss distribution data are obtained;

Specifically, based on time domain current distribution, square integral operation is performed on the current of each battery unit, and the joule heat loss of the battery unit in the operation process is estimated preliminarily. Because the internal resistance of the battery is dynamically changed, based on the battery internal resistance parameter updated in real time, the temperature correction is carried out on the joule heat loss value calculated initially, so that more accurate joule heat loss data is obtained. Based on the charge and discharge states of the battery cells, charge transfer impedance generated during charge transfer is calculated. This impedance reflects the energy loss due to the charge transfer process in the electrochemical reaction of the cell. And obtaining basic data of electrochemical reaction loss by calculating charge transfer impedance. And carrying out nonlinear compensation on the electrochemical reaction loss basic data to obtain corrected electrochemical reaction loss data. Meanwhile, a battery capacity attenuation model is constructed based on the battery cycle times and the deep discharge state, and the capacity attenuation degree of the battery in each charge and discharge process is estimated according to the battery cycle times and the deep discharge state, so that corresponding cycle attenuation loss data is calculated. And carrying out simultaneous solving on the joule heating loss data, the electrochemical reaction loss data and the cyclic attenuation loss data based on the multi-physical field coupling model. The multi-physical field coupling model can comprehensively consider the comprehensive influence of a plurality of factors such as electricity, heat and chemistry on the battery loss, and the factors are related to each other in the simultaneous solving process to obtain integral coupling loss data, so that the actual loss condition of the battery in the running process is comprehensively reflected. And carrying out time sequence analysis on the coupling loss data, revealing the law of dynamic loss changing along with time through a time sequence analysis method, and obtaining the time-varying characteristic of dynamic loss distribution. Based on the dynamic loss distribution time-varying characteristic, weighting and fusing the joule heating loss data, the electrochemical reaction loss data and the cyclic attenuation loss data, and reasonably integrating the three loss types of data according to the respective weights of the three loss types under different time periods and working conditions to finally obtain complete dynamic loss distribution data.

And S5, carrying out real-time evaluation on the battery state based on the dynamic loss distribution data, reconstructing a nonlinear charge-discharge characteristic curve, and carrying out multi-time scale cooperative scheduling based on the nonlinear charge-discharge characteristic curve to obtain a battery charge-discharge control strategy.

The method comprises the steps of carrying out piecewise linearization on time current distribution, and discretizing original continuous current distribution into a plurality of current sampling sequences. Based on the discretized current sampling sequence, calculating the state of charge (SOC) variation of the battery to obtain an initial SOC estimated value. And compensating and correcting the initial SOC estimation value according to the dynamic loss distribution data to obtain a more accurate corrected SOC value. In order to improve the accuracy of SOC estimation, an adaptive OCV-SOC (open circuit voltage-state of charge) mapping model is constructed based on historical charge-discharge data. The model can dynamically adapt to the change of the battery performance through accumulation and analysis of historical data. In order to ensure the real-time accuracy of the model, the adaptive OCV-SOC mapping model is subjected to online parameter identification, the parameters of the model are continuously updated by using a forgetting factor method, the updated OCV-SOC mapping relation is obtained, and the model can be self-adjusted in the long-term running process of the battery so as to adapt to the change of the battery state. Based on the corrected SOC value and the updated OCV-SOC mapping relation, calculating the instantaneous open-circuit voltage of the battery to obtain a data pair of the SOC-OCV, and reflecting the open-circuit voltage of the battery under different SOC states. And (3) performing curve fitting on the SOC-OCV data pair to obtain a nonlinear characteristic equation of the charge-discharge process. The equation can accurately describe the characteristics of the battery in different charge and discharge states, and can provide a key mathematical model for predicting parameters such as capacitance, internal resistance and the like of the battery in different SOC. Based on the nonlinear characteristic equation, the capacity and the internal resistance of the battery under different SOC states are calculated. The nonlinear charge-discharge characteristic curve reflects the working characteristics of the battery in different states, and can effectively guide the charge-discharge process of the battery in actual working. Through the curve, accurate guidance is provided for a management system of the battery, the maximum working efficiency of the battery under different charge states is ensured, the loss can be reduced, and the service life of the battery is prolonged. Based on the nonlinear charge-discharge characteristic curve, the coordinated scheduling of multiple time scales is realized. Multi-time scale scheduling means that both short-term real-time control and long-term energy scheduling can be optimally adjusted according to the non-linear characteristics of the battery in different time ranges, for example. Through the collaborative scheduling of multiple time scales, the charge and discharge strategies of the battery can be ensured to meet the short-term high-efficiency energy utilization requirement, and long-term battery health management can be optimized, so that a complete battery charge and discharge control strategy is finally obtained.

Wherein, based on the nonlinear charge-discharge characteristic curve, the charge-discharge efficiency and capacity limitation of the battery under different Charge States (SOCs) are determined. By analyzing the operating characteristics of the battery in different SOC ranges, the optimal operating conditions of the battery in each state and the upper performance limit of the battery during charge and discharge are defined. And according to the charge and discharge efficiency and capacity limitation, carrying out daily load prediction by combining historical electricity consumption data to obtain a daily load prediction result. And (3) based on a day-ahead load prediction result, a day-ahead charge-discharge preliminary scheduling plan is formulated, the charge-discharge behavior of the battery is ensured to be matched with the future load demand, and the energy storage and release process is optimized. And performing rolling optimization on the preliminary scheduling plan of charging and discharging in the future. Through rolling optimization, the prediction data is dynamically adjusted along with the time, so that the plan is more fit with the actual load demand, and a daily dynamic adjustment strategy is obtained. Along with the real-time change of the load of the battery in the actual running process, the daily dynamic adjustment strategy also needs to be finely adjusted according to the actual real-time load information. By constantly adjusting the strategy, the management system can generate real-time control instructions for performing specific battery charging and discharging operations. The control instructions are based on the latest load data and the optimized charge and discharge strategies, so that the battery can be ensured to operate efficiently under different load conditions. Meanwhile, various data in the charge and discharge process are recorded in the process of executing the charge and discharge operation. Based on the charge-discharge process data, the battery state estimation model needs to be updated periodically. With the updating of the evaluation model, the original nonlinear charge-discharge characteristic curve is optimized and adjusted, so that the new curve can better reflect the current running state and performance change of the battery. And finally, carrying out multi-time-scale cooperative scheduling adjustment based on the updated nonlinear charge-discharge characteristic curve. In combination with the long-term and short-term load demands, it is ensured that the operation of the battery in different time ranges not only meets the instant energy demands, but also ensures the long-period service life of the battery, and a final battery charge-discharge control strategy is generated.

Further, the historical electricity utilization data is decomposed in time series and is decomposed into three main components, namely a trend item, a season item and a random item. By analyzing the different components, long-term trend, periodic variation and short-term fluctuation are respectively captured. The trend term is input into a long-term trend prediction model for prediction. The long-term trend prediction model adopts an LSTM (long-term memory) network structure, and the network structure has strong memory capacity and can well capture long-term dependency in a time sequence. The LSTM network model consists of an input layer, two hidden layers and an output layer. Each hidden layer contains 128 neurons and uses the tanh activation function to introduce non-linear characteristics, enhancing the predictive power of the model. Through the LSTM model, the load trend of 24 hours in the future can be predicted, and a long-term trend prediction result is obtained. And simultaneously, inputting the seasonal items into the periodic pattern recognition model for processing. Because seasonal variation has obvious periodic characteristics, the model adopts a one-dimensional Convolutional Neural Network (CNN) structure. The network comprises 3 convolution layers and 2 full connection layers, the convolution layers are mapped in a nonlinear way through a ReLU activation function, and batch normalization is used for accelerating the training process and improving the stability of the model. The convolution network is used for identifying a periodic mode in the time sequence, and a periodic load prediction result is obtained by periodically predicting the load of 24 hours in the future. The random term part is input into a short-term fluctuation prediction model for processing. The short-term fluctuation model adopts a GRU (gate control circulation unit) network structure, and the GRU is a simplified RNN structure and can effectively capture short-term change characteristics in a time sequence. The model contains a bi-directional GRU layer containing 64 hidden units for capturing the context of random fluctuations. The model also includes a full link layer for outputting final short-term fluctuation prediction results. And predicting the short-term fluctuation condition of the load in the future 24 hours through a GRU model. The long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result are integrated into a multi-layer perceptron (MLP) fusion network. The fusion network consists of 3 hidden layers, including 64, 32 and 16 neurons, respectively, using LeakyReLU activation functions to handle complex nonlinear relationships of load data, and the last layer using linear activation functions to ensure continuity of the predicted results. And the information from different sources is fused, so that a more accurate and comprehensive daily load prediction result is finally obtained.

In this embodiment, the long-term trend prediction result, the periodic load prediction result, and the short-term fluctuation prediction result are feature-fused. And performing feature stitching on the prediction results of different time scales to form a fusion input vector. The fused input vector is input to a first hidden layer of the multi-layer perceptron. The first hidden layer contains 64 neurons, the output of each neuron being processed by LeakyReLU activation functions. LeakyReLU the activation function can avoid the "dead neurons" problem existing in the conventional ReLU activation function, thereby ensuring better performance of the model when processing complex nonlinear data. And LeakyReLU, obtaining the output of the first hidden layer. In order to improve the stability of the model and accelerate the training process, batch normalization operation is performed on the output of the first hidden layer. The batch normalization eliminates the change of internal covariates by carrying out standardization processing on output data, and improves the convergence rate of the model. The output after batch normalization is added with the initial fusion input vector through residual connection to obtain residual output of the first layer, a residual connection mechanism can relieve the gradient disappearance problem possibly occurring in deep network training, some original information of the input layer is reserved, and the performance of the model is improved. The residual output of the first layer is input to the second hidden layer. The second hidden layer contains 32 neurons, the output of each neuron also being non-linearly processed by LeakyReLU activation functions. To prevent model overfitting, the drop rate is set to 0.3 by doing a dropout operation on its output after the second hidden layer. This means that 30% of neurons are randomly discarded in each training, forcing the network to learn different feature combinations in different training cycles, enhancing the generalization ability of the model. The output after the dropout operation, i.e. the regularized second hidden layer output, is used as input for the third hidden layer. The regularized output of the second hidden layer is input to the third hidden layer. The third hidden layer contains 16 neurons whose outputs are processed using LeakyReLU activation functions. Although the third hidden layer has a smaller number of neurons than the first two layers, it is able to further extract advanced features of the data. The output of the third hidden layer is input to the output layer. The output layer contains 24 neurons representing load predictions within the next 24 hours. To ensure continuity and accuracy of the predicted output, the output layer uses a linear activation function. The activation function is suitable for regression tasks, and does not carry out nonlinear transformation on output values, so that the real range of predicted values is reserved. Through the process, the fusion network combines the information of long-term trend, periodic mode and short-term fluctuation together, and finally a daily load prediction result is obtained.

In one example, graphically modeling a battery management system in a shared charging device, constructing a topology graph model including vertices representing ternary polymer lithium battery cells and management circuitry and directed edges representing current flow paths, includes:

identifying a ternary polymer lithium battery unit, a charging circuit, a discharging circuit and a protection circuit of a battery management system in shared charging equipment to obtain types and parameter information of all components;

Performing vertex modeling on each ternary polymer lithium battery unit and each circuit component based on the type and parameter information of each component to obtain a vertex set;

analyzing the flow paths of the current among all the components according to the circuit topology structure of the battery management system to obtain a directed edge set;

performing attribute assignment on each vertex in the vertex set to obtain vertex attribute data comprising rated capacity, internal resistance, current state of charge, circuit component types and parameters of the ternary polymer lithium battery unit, and performing attribute assignment on each side in the directed side set to obtain side attribute data comprising current flowing direction, maximum allowable current and current value;

and constructing a topological graph model of the battery management system based on the vertex set, the directed edge set and the edge attribute data.

In this example, the identification of the ternary polymer lithium battery cell, charging circuit, discharging circuit, and protection circuit of the battery management system in the shared charging device yields the type and parameter information of each component. For example, the main parameters of a lithium-polymer battery cell include its rated capacity, internal resistance, and current state of charge (SOC), while for charging and discharging circuits, it is necessary to identify their circuit structure, maximum allowable current, and voltage limits. The parameters of the protection circuit generally relate to information such as maximum overvoltage, undervoltage protection values, and overcurrent protection thresholds. Each of the lithium-ion-ternary battery cells and circuit components is vertex modeled based on the type and parameter information of the respective components. Each of the lithium-ion-ternary battery cells and circuit assemblies may be considered an apex in the figure. For example, assuming a battery cell with a rated capacity of 5000mAh, an internal resistance of 50mΩ, and a current state of charge of 80%, the battery cell may be represented as a vertex, and these parameters may be taken as attributes of the vertex. For circuit components, assuming that the maximum allowable current of a certain charging circuit is 10A and the operating voltage range is 3.7V to 4.2V, these information are also assigned to the corresponding vertices. After the vertex modeling is completed, the current flow path in the battery management system is analyzed. The flow path of current between components may be represented as a set of directed edges in the graph. The directed edge represents the path of current flowing from one vertex (component) to another. For example, when current flows from the charging circuit to the battery cell, this may be represented as a directed edge from the charging circuit apex to the battery cell apex. By analyzing the circuit topology, all current paths can be represented by way of directed edges, forming a complete set of directed edges. And carrying out attribute assignment on each vertex in the vertex set. These attributes include the rated capacity, internal resistance, and current state of charge of the battery cell, as well as the type of circuit components and related parameter information. For example, for a ternary polymer lithium battery cell, the SOC may be expressed as:

;

Wherein the method comprises the steps of Representing the current remaining capacity of the battery,Indicating the total capacity of the battery. Assuming that the remaining capacity of the current battery is 4000mAh and the total capacity is 5000mAh, the current SOC may be calculated to be 80%. Attribute assignment is also performed for each edge in the directed edge set. These attributes may include the direction of flow of the current, the maximum allowable current, and the present current value. Assuming that a side represents a path from the charging circuit to the battery cell, with a maximum allowable current of 10A and a current currently actually flowing of 8A, the attribute data of the directed side includes these current parameters. After all the vertex and edge attributes are assigned, a topology graph model of the battery management system is constructed. The topology graph model is composed of a vertex set and a directed edge set, and can describe the interconnection relation of each component in the battery management system and the flow path of current in the system.

In one example, performing a depth-first search based on a topology model, identifying all charge-discharge paths that satisfy circuit rules, and calculating a complex impedance value for each charge-discharge path, comprising:

Encoding the vertex set in the topological graph model to obtain a unique identifier and an access state mark of each vertex, and constructing a recursion function of depth-first search;

Traversing the topological graph model based on a recursion function to obtain all possible vertex access sequences, and carrying out circuit rule verification on each vertex access sequence to obtain an effective charge-discharge path set;

Based on the effective charge-discharge path set, extracting the resistance, inductance and capacitance of each charge-discharge path to obtain a distribution parameter model of the charge-discharge path, and carrying out mathematical transformation on the distribution parameter model of the charge-discharge path to obtain a frequency domain complex impedance expression of each charge-discharge path;

and calculating equivalent impedance of the charge-discharge paths in the working frequency range based on the frequency domain complex impedance expression to obtain a complex impedance value of each charge-discharge path.

In this example, each vertex in the topology model is uniquely encoded and the state flags are initialized. The unique identifier for each vertex may be a unique ID generated based on its location or function, such as using its component name, number, or sequence of locations. For the status flag, whether or not the vertex has been accessed is distinguished, a boolean value representation may typically be used, such as "0" for non-access and "1" for accessed. In this way, each vertex in the graph is given a separate identifier and ensures that access can be tracked during traversal. Based on the above encoding, a recursive function for traversing the topology graph model is constructed. Depth first search is a graph traversal algorithm that can start from any one starting vertex, go deep into each vertex in the graph along a path until it cannot continue, then trace back and continue searching for other paths. The core of the recursive function is to recursively call itself to traverse all possible paths. At each recursive call, the vertex visited is marked as visited to avoid repeated visits. In the directed edge set in the figure, each edge represents the direction of current flow, so the depth-first search recursion process follows the direction of these edges to traverse, starting from the starting vertex, each adjacent vertex in turn, until the complete circuit topology is traversed. During the recursive traversal, the depth-first search algorithm generates all possible vertex-access sequences. Each access sequence represents a potential charge-discharge path. For each vertex access sequence, circuit rule verification is performed. The circuit rules may include current direction, voltage limit of the node, maximum allowed current on the path, etc. For example, some paths may exceed the allowed value of voltage or current at some nodes, at which point the path will be deemed invalid. While paths conforming to the circuit rules will be preserved as effective charge and discharge paths. Based on the effective charge-discharge path set, the resistance, inductance and capacitance of each charge-discharge path are extracted. For each path's resistance, inductance and capacitance, the corresponding circuit element values are used for superposition. After the distribution parameter model of each charge-discharge path is obtained, mathematical transformation is carried out on the parameters, and the parameters are converted into a frequency domain complex impedance expression. Complex impedance is a critical parameter of a circuit under ac conditions, typically a function of frequency. And obtaining an equivalent impedance expression of the charge-discharge path in the frequency domain through mathematical transformation. For a charge-discharge path including resistor, inductor and capacitor, its complex impedanceCan be expressed as:

;

Wherein, Representing the complex impedance of the circuit,Is the resistance of the resistor, which is the resistance of the resistor,Is an inductance of the device, which is a capacitor,Is a capacitor which is used to form a capacitor,Is the angular frequency of the wave form,Is an imaginary unit. The formula shows that the impedance is not dependent on resistance, but also on inductance and capacitance, and that these effects behave differently at different frequencies. By this mathematical transformation, the behavior of the charge-discharge path in the frequency domain can be clearly described and the impedance characteristics of the path at different frequencies revealed. And calculating equivalent impedance of the charge-discharge path in the working frequency range based on the frequency domain complex impedance expression. The equivalent impedance is the total impedance of a plurality of elements when the elements work cooperatively under different frequencies, and the performance of the circuit under the actual working condition can be estimated by calculating the equivalent impedance. For example, in a charging path, as the charging frequency increases, the effect of the inductance increases, while the effect of the capacitance decreases, so that the equivalent impedance of the entire path also changes. By accurately calculating the equivalent impedance, a basis can be provided for subsequent circuit optimization and scheduling. For example, suppose there is a simple system of a ternary polymer lithium battery cell, a charging circuit, a discharging circuit, wherein the internal resistance of the battery cell is 100mThe charging circuit has an inductance of 10 muH and the discharging circuit has a capacitance of 100 muF. In this case, after traversing the depth-first search, an effective path is identified from the charging circuit to the battery cell to the discharging circuit. The resistance, inductance and capacitance of the path are extracted according to parameters in the circuit. Complex impedance at different frequencies is calculated using a complex impedance formula. For example, when the angular frequencyThe complex impedance of the path is:

;

the calculation results are that:

。

in one example, calculating the dynamic voltage spectrum and impedance characteristics based on the charge-discharge path and the complex impedance values results in a time-domain current distribution for each of the lithium-ion-ternary battery cells, comprising:

Analyzing the topological structure of the charge-discharge path to obtain a time domain differential equation set describing the dynamic behavior of the battery management system, and converting the time domain differential equation set into an s-domain transfer function to obtain a frequency domain mathematical model;

Constructing a system matrix, an input matrix, an output matrix and a feedforward matrix of the battery management system based on the frequency domain mathematical model and the complex impedance value, and performing discretization processing to obtain a discrete time system model;

Performing parameter identification on the discrete time system model to obtain system characteristic parameters, applying a broadband excitation signal to a battery management system based on the system characteristic parameters, and obtaining a dynamic voltage spectrum through fast Fourier transform calculation;

Based on the dynamic voltage spectrum and the input current information, the impedance characteristic of the battery management system is obtained through calculation, the impedance characteristic is subjected to inverse Fourier transform, and the time domain current distribution of each ternary polymer lithium battery unit is obtained through combining the circuit topology information.

In this example, the electrical connection relationships between the components in the battery management system are analyzed to extract the time domain behavior of the system. The charge-discharge path of the battery management system may be regarded as an electrical network comprising battery cells, a charging circuit, a discharging circuit, and a protection circuit. In this network, the dynamic behavior of each component, such as resistance, inductance, and capacitance, can be described by a time-domain differential equation. For each charge-discharge path, a system of time-domain differential equations is established using kirchhoff's voltage law and kirchhoff's current law according to the relationship of voltage, current and circuit elements. For example, assume that one charge-discharge path includes a resistorInductanceAnd a capacitorIn the time domain, the current is described by the following equationSum voltage ofRelationship between:

;

Wherein, Is the voltage of the battery management system on that path,Is the current in the path that is to be measured,Is the resistance value on the path and,Is an inductance of the device, which is a capacitor,Is a capacitor. The time-domain differential equation can fully describe the dynamic changes of voltage and current over time during the charge and discharge of the battery. To simplify the analysis and convert to the frequency domain for processing, the time-domain differential equation is converted to a transfer function in the frequency domain using a laplace transform. In the laplace transform, derivative and integral operations are converted into multiplication and division operations in the frequency domain, respectively. For the above equation, after applying the laplace transform, a frequency domain transfer function can be obtained:

;

Wherein, AndThe laplace transform of voltage and current respectively,Is a complex frequency variable representing the complex frequency domain characteristics of the system. The transfer function describes the relationship of voltage and current in the frequency domain, also known as the frequency domain mathematical model of the system. Based on the frequency domain mathematical model, a system matrix, an input matrix, an output matrix and a feedforward matrix of the battery management system are constructed. For a linear time-invariant system, the state space equation can be expressed as:

;

Wherein, Is the time derivative of the system state variable,Is a state vector of the state of the object,Is the input vector which is to be used for the input,Is the output vector of the vector,Is a matrix of the system and,Is an input matrix of the data set,Is the output matrix of the device,Is a feed forward matrix. System matrixDescribes the internal dynamics of a battery management system, input matrixRepresenting the effect of the input signal on the system, an output matrixFeedforward matrixReflecting the interrelationship of the output and direct input of the system, respectively. In order to adapt the model to a digital control system, discretization processing is performed on the model, and a continuous time system is converted into a discrete time system, so that the system can be simulated and controlled through numerical calculation. The discretized system can be represented by a differential equation:

;

Wherein, 、、AndA system matrix, an input matrix, an output matrix and a feed forward matrix of the discrete time system, respectively. By switching, the continuous dynamic behavior of the battery management system can be described and simulated in discrete time steps. And carrying out parameter identification on the discrete time system model, and estimating the sign parameters of the system based on the actual data. The purpose of parameter identification is to determine the system matrixInput matrixAnd the like. By applying a known input signal and measuring a corresponding output signal. To ensure accuracy of parameter identification, a wideband excitation signal is applied to the system, which signal can cover multiple frequency ranges, thereby ensuring that the dynamic behavior of the system at all relevant frequencies can be identified. After the wideband excitation signal is applied, the dynamic voltage spectrum of the system is calculated by fast fourier transform. The fourier transform may convert the time domain signal to a frequency domain signal, revealing the response characteristics of the system at different frequencies. And obtaining a voltage response spectrum of the battery management system in a wide frequency range through fast Fourier transformation. Based on the dynamic voltage spectrum and the input current information, impedance characteristics of the battery management system are calculated. The impedance reflects the impeding effect of the system on current at different frequencies. In the frequency domain, impedanceIs defined as the ratio of voltage to current, i.e.:

;

By the formula, complex impedance values of the battery system are calculated at different frequencies, and the behavior of the system at different working frequencies is analyzed. And converting the impedance characteristic back to the time domain through inverse Fourier transform, reducing impedance information in the frequency domain to a time domain signal, and combining a circuit topological structure to obtain the time domain current distribution of each ternary polymer lithium battery unit. For example, assume that a charge-discharge path includes a resistor An inductorAnd a capacitor. The dynamic behavior of the circuit is described by creating a time-domain differential equation. The time domain equation is converted to a frequency domain transfer function using the laplace transform:

;

Based on the transfer function, a state space equation is constructed and discretization processing is carried out, so that a discrete time system model is obtained. A broadband excitation signal is applied and a voltage spectrum is calculated by fast fourier transformation, and then impedance characteristics at different frequencies are calculated using an impedance formula. And restoring the frequency domain information to the time domain through inverse Fourier transform, and calculating the time domain current distribution of each battery unit by combining the circuit topology information.

In one example, the multi-factor coupling loss calculation is performed on each of the three-polymer lithium battery cells according to the time domain current distribution, to obtain dynamic loss distribution data, including:

Based on time domain current distribution, carrying out current square integral operation on each ternary polymer lithium battery unit to obtain a Joule heat loss initial value, and carrying out temperature correction on the Joule heat loss initial value according to the battery internal resistance parameter updated in real time to obtain Joule heat loss data;

Calculating charge transfer impedance based on the charge and discharge states of the ternary polymer lithium battery unit to obtain electrochemical reaction loss basic data, and performing nonlinear compensation on the electrochemical reaction loss basic data to obtain electrochemical reaction loss data;

based on the battery cycle times and the deep discharge state, constructing a battery capacity attenuation model to obtain cycle attenuation loss data in a single charge and discharge process;

Based on a multi-physical field coupling model, carrying out simultaneous solving on the joule heating loss data, the electrochemical reaction loss data and the cyclic attenuation loss data to obtain coupling loss data;

And carrying out time sequence analysis on the coupling loss data to obtain dynamic loss distribution time-varying characteristics, and carrying out weighted fusion on the focusing ear heat loss data, the electrochemical reaction loss data and the cyclic attenuation loss data based on the dynamic loss distribution time-varying characteristics to obtain dynamic loss distribution data.

In this example, a current square integration operation is performed based on the time-domain current distribution, resulting in an initial value of joule heating loss. Joule heat lossThe calculation can be performed by the following formula:

;

Wherein, Is the current distribution of the battery cells in the time domain,Is the battery is in timeInternal resistance at time. The initial heat loss of the battery due to the internal resistance in the charging and discharging process is obtained through square integration of the current. And carrying out temperature correction according to the battery internal resistance parameter updated in real time. The internal resistance of the battery decreases as the temperature increases, and therefore, it is necessary to adjust the initial value of the joule heat loss by the temperature correction coefficient. Temperature correction typically relies on empirical formulas or experimental data to reflect the non-linear relationship between internal resistance and temperature. The corrected joule heat loss data is more fit to the actual running state. In addition to joule heat loss, the loss of the cell also includes loss of electrochemical reaction. Based on the charge and discharge states of the ternary polymer lithium battery cell, charge transfer impedance is calculated, and basic data of electrochemical reaction loss is obtained from the charge transfer impedance. The charge transfer resistance refers to the resistance of charge transfer between the electrode and electrolyte interface, and is typically obtained by impedance spectroscopy analysis. Loss of electrochemical reactionCan pass through charge transfer impedanceAnd current flowThe relationship of (2) is expressed as:

;

The formula represents the loss due to charge transfer impedance at a given current distribution. And carrying out nonlinear compensation on the electrochemical reaction loss basic data. By nonlinear compensation of the charge transfer impedance, the electrochemical loss characteristics of the battery under different charge and discharge conditions are reflected more accurately. As the capacity gradually decays with repeated charging and discharging of the battery, a capacity decay model needs to be constructed to predict the performance degradation of the battery. The battery capacity fade is generally closely related to the number of cycles and the deep discharge state of the battery. Each deep discharge can cause some damage to the active material of the battery, thereby reducing its usable capacity. By tracking the number of cycles N and the deep discharge parameter DOD of the battery, a capacity fade model is established:

;

Wherein, Is the remaining capacity of the battery and,Is the initial capacity of the device and,AndIs the decay factor associated with the number of cycles and the deep discharge. This model can be used to estimate the capacity fade loss of the battery during each charge and discharge and to gradually update the state of health of the battery over time. And simultaneously solving the joule heat loss data, the electrochemical reaction loss data and the cyclic attenuation loss data based on a multi-physical field coupling model, and comprehensively analyzing the loss condition of the battery in actual operation. By simultaneous solving, the mutual influence among different losses is considered, and more accurate coupling loss data is obtained. For example, joule heat loss can cause the cell to increase in temperature, which in turn can affect the electrochemical reaction rate of the cell, so there is an interaction between these two losses. By solving these coupling effects simultaneously, a more complete description of the battery loss is obtained. And carrying out time sequence analysis on the coupling loss data to reveal the dynamic change characteristic of the loss along with time. Time series analysis can identify long-term trends in loss, seasonal fluctuations, and random short-term fluctuations. By time series analysis, the loss rule of the battery under different use scenes is identified, for example, the loss of the joule heat of the battery is larger in a high-temperature environment, and the loss of the electrochemical reaction is dominant at a low temperature. And carrying out weighted fusion on the joule heating loss, the electrochemical reaction loss and the cyclic attenuation loss based on the time-varying characteristic of the dynamic loss distribution to obtain dynamic loss distribution data. Different weights are given to different types of losses, and reasonable fusion is carried out according to the contribution of the losses to the total loss. Assuming that joule heat loss is the main part at one moment and electrochemical reaction loss is more prominent at the other moment, the actual loss condition of the battery can be reflected more accurately by dynamically adjusting the weight of each loss.

In one example, the method for performing real-time battery state evaluation based on dynamic loss distribution data, reconstructing a nonlinear charge-discharge characteristic curve, and performing multi-time scale cooperative scheduling based on the nonlinear charge-discharge characteristic curve to obtain a battery charge-discharge control strategy includes:

Performing piecewise linearization on the time-domain current distribution to obtain a discretized current sampling sequence, and calculating the state-of-charge variation of the battery based on the discretized current sampling sequence to obtain an initial SOC estimated value;

According to the dynamic loss distribution data, compensating and correcting the initial SOC estimated value to obtain a corrected SOC value;

Constructing a self-adaptive OCV-SOC mapping model based on historical charge and discharge data, carrying out on-line parameter identification on the self-adaptive OCV-SOC mapping model, and updating model parameters by adopting a forgetting factor method to obtain an updated OCV-SOC mapping relation;

calculating the instantaneous open-circuit voltage of the battery based on the corrected SOC value and the updated OCV-SOC mapping relation to obtain an SOC-OCV data pair;

performing curve fitting on the SOC-OCV data pair to obtain a nonlinear characteristic equation of a charging and discharging process, and calculating the battery capacity and internal resistance under different SOCs based on the nonlinear characteristic equation to obtain a nonlinear charging and discharging characteristic curve;

In this example, a piecewise linearization process is performed on the time domain current distribution. In actual operation of the battery, the current is changed continuously along with time, and the original continuous current signal is discretized into a series of sampling points through piecewise linearization. Each sampling point corresponds to a specific current value and a specific time period, and a discretized current sampling sequence is formed. Based on the discretized current sampling sequence, the amount of change in SOC can be calculated by coulomb counting. Coulomb counting estimates the change in state of charge of a battery based on the current flowing through the battery. The specific formula is as follows:

;

Wherein, The amount of change in the SOC is indicated,Is the total capacity of the battery and,Is the firstThe current in the course of the time period,Is the duration of the time period,Is the total number of time periods. An initial SOC estimate of the battery is obtained by accumulating each time period of the discretized current sampling sequence. And compensating and correcting the initial SOC estimation value according to the dynamic loss distribution data. The actual SOC of the battery is affected by loss factors such as joule heat loss, electrochemical reaction loss, and capacity fade loss. And estimating the loss condition of the battery under the current running condition through the dynamic loss distribution data, so as to correct the initial SOC. An adaptive OCV-SOC (open circuit voltage-state of charge) mapping model is constructed based on historical charge-discharge data. The OCV-SOC relationship is an important characteristic of a battery that describes the Open Circuit Voltage (OCV) of the battery at different SOCs. By analyzing the historical charge and discharge data of the battery, an initial OCV-SOC mapping model is constructed and updated online along with time. In order to ensure that the model can adapt to the change of the battery performance, the OCV-SOC model is subjected to online parameter identification. The on-line parameter identification can reflect the current state of the battery by updating the model parameters in real time. The process adopts a forgetting factor method, which is an algorithm for giving decreasing weight to historical data, is used for gradually forgetting early historical data, and more emphasis is given to the latest battery state data. The update formula of the forgetting factor method is as follows:

;

Wherein, Is the firstThe model parameters of the time of day,Is based on the new parameters of the current data,Is a forgetting factor (typically between 0 and 1).The closer to 1, the greater the influence of the history data; The closer to 0, the greater the effect of the latest data. Through the updating mechanism, the OCV-SOC mapping model can be ensured to be adjusted in real time along with the change of the battery state, and the updated OCV-SOC mapping relation is obtained. And after obtaining the corrected SOC value and the updated OCV-SOC mapping relation, calculating the instantaneous open-circuit voltage of the battery. The instantaneous open-circuit voltage is based on the voltage value in the current SOC state, and the corrected SOC value is substituted into the OCV-SOC mapping model to obtain the corresponding open-circuit voltage value, so that a series of SOC-OCV data pairs are generated, and the voltage characteristics of the battery in different SOC states are reflected. And performing curve fitting on the SOC-OCV data pair to obtain a nonlinear characteristic equation of the charge-discharge process. The relationship between SOC and OCV is not typically linear and requires that an optimal equation be found to describe the relationship by a nonlinear fitting technique. The fitted nonlinear characteristic equation can accurately describe the dynamic behavior of the battery in the charge and discharge processes. Common fitting models include polynomial fitting or exponential fitting, which can capture the voltage change law of the battery in different SOC intervals. Based on the nonlinear characteristic equation, the capacity and the internal resistance of the battery under different SOCs are calculated. The capacity and internal resistance of the battery generally change along with the change of the SOC, and the battery parameters under different SOC conditions are estimated through a nonlinear charge-discharge characteristic equation. The change in internal resistance has an important effect on the battery performance because it determines the voltage drop and heat generation of the battery at high currents. And calculating the optimal working parameters of the battery under each state of charge through an accurate SOC-OCV model. And carrying out multi-time scale cooperative scheduling based on the nonlinear charge-discharge characteristic curve. Multi-time scale scheduling refers to optimizing charge and discharge strategies in different time ranges according to the current state of the battery and future requirements. For example, on a short time scale, the system may adjust the charging current in real time to maximize battery efficiency, while on a long time scale, the system may program the charge-discharge cycle of the battery to extend the life of the battery. By the scheduling mode, the battery can be ensured to be in the optimal working state in different time periods, so that the stability of energy supply is ensured, and the service life of the battery is optimized.

In one example, performing multi-time scale collaborative scheduling based on a nonlinear charge-discharge characteristic curve to obtain a battery charge-discharge control strategy includes:

based on the nonlinear charge-discharge characteristic curve, determining charge-discharge efficiency and capacity limitation under different charge states;

according to the charge and discharge efficiency and capacity limitation, historical electricity consumption data are combined to conduct daily load prediction, and a daily load prediction result is obtained;

based on a day-ahead load prediction result, a day-ahead charge-discharge preliminary scheduling plan is formulated, and rolling optimization is carried out on the day-ahead charge-discharge preliminary scheduling plan, so that a day-ahead dynamic adjustment strategy is obtained;

According to the real-time load change, the daily dynamic adjustment strategy is finely adjusted to generate a real-time control instruction, the battery charging and discharging operation is executed based on the real-time control instruction, and the charging and discharging process data is recorded;

updating a battery state evaluation model based on the charge and discharge process data, and optimizing a nonlinear charge and discharge characteristic curve to obtain an updated nonlinear charge and discharge characteristic curve;

And carrying out multi-time scale cooperative scheduling adjustment according to the updated nonlinear charge-discharge characteristic curve to obtain a battery charge-discharge control strategy.

In this example, charge-discharge efficiency and capacity limits of the battery at different states of charge (SOCs) are determined based on the non-linear charge-discharge characteristic curves. The charge and discharge efficiency of a battery is its ability to effectively store or release energy under different SOC conditions, while the capacity limit reflects the maximum energy that the battery can provide in each SOC interval. The nonlinear charge-discharge characteristic curve is generated through historical data and real-time evaluation, and can describe the performance of the battery under different SOCs, including factors such as voltage, internal resistance and efficiency. By analyzing these curves, the optimal operating conditions and limiting capacity of the battery over a particular SOC range are obtained. For example, it is assumed that the charging efficiency of a certain battery is 95% when the SOC is 90%, and the charging efficiency is reduced to 80% when the SOC is 20%. The non-linear change is due to the more difficult charging and reduced efficiency of the battery at high SOC. And through the data, the charge and discharge efficiency and capacity limitation under different SOCs are reasonably determined. And according to the charge and discharge efficiency and capacity limitation, the historical electricity consumption data is combined to predict the load before the day. And predicting the power demand in the future 24 hours according to the past power utilization mode and trend. By time series analysis, future load demands are predicted by models using these historical data, assuming that the electricity usage patterns over time exhibit more regular trends and fluctuations. Based on the predicted load demand, the battery management system makes a charge-discharge plan in advance, ensuring that the battery can release enough energy at high load and charge to restore SOC at low load. The preliminary dispatch plan is mainly formulated according to the current state of the battery and the future demand, but rolling optimization is needed due to the possible uncertainty of the prediction result. The rolling optimization is a dynamic adjustment strategy, and the scheduling plan is corrected in real time according to actual load change in the execution process, so that the scheduling is more accurate. For example, if the actual load does not match the forecast, the system automatically adjusts the charge-discharge rate to ensure that the load demand is met. When the system enters an actual operation stage, the battery management system carries out fine adjustment on the daily dynamic adjustment strategy according to the real-time load change. This means that the system will generate real-time control commands based on current power demand and battery status. These control instructions include information about the rate of charge or discharge, time, and power output, and are intended to ensure that the battery is operating within an optimal efficiency range. For example, if the system detects a sudden increase in current load, the discharge rate may be increased to meet the demand, and conversely the discharge may be decreased or charging may be initiated. The execution of the real-time control command is realized through a specific Battery Management System (BMS), and the BMS can adjust parameters such as current, voltage and the like to ensure the safety and efficiency of the system operation. In the process of executing the charge and discharge operation, process data of charge and discharge, including current, voltage, temperature and change condition of SOC, are recorded, and through these information, the performance of the battery under the current condition is estimated. These data can also be used to optimize the management strategy of the battery. For example, the system may find that the charge efficiency is lower than expected within a certain SOC range, and this information may be used to adjust future charge-discharge strategies. Based on the charge-discharge process data, the battery state evaluation model is updated. The state of the battery can change with time and use condition, so that the state of the battery and the capacity degradation condition can be accurately tracked through continuous state evaluation. As the battery is used, its capacity gradually decreases and its internal resistance gradually increases, and the state of health (SOH) of the battery is estimated from the charge and discharge data. Meanwhile, the data are used for optimizing a nonlinear charge-discharge characteristic curve, so that the curve can accurately reflect the actual working state of the battery. For example, if the system finds from the data that the discharge efficiency of the battery drops significantly below 10% SOC, the system will incorporate this change into the new non-linear charge-discharge characteristic. Over time and data accumulation, the accuracy of the curve will continue to increase, thereby better guiding future battery management. And carrying out multi-time scale cooperative scheduling adjustment through the optimized nonlinear charge-discharge characteristic curve. The multi-time scale scheduling performs collaborative optimization of charge and discharge in both short-term and long-term time ranges according to the current state and future requirements of the battery. For example, on a short time scale, the charge and discharge rate is adjusted in real time to cope with instantaneous load fluctuation, while on a long time scale, the whole charge and discharge period of the battery is planned to avoid frequent deep charge and discharge and prolong the service life of the battery.

In one example, according to the charge-discharge efficiency and capacity limitation, in combination with historical electricity consumption data, a daily load prediction is performed to obtain a daily load prediction result, including:

Carrying out time sequence decomposition on the historical electricity utilization data to obtain trend items, season items and random items;

Inputting a trend item into a long-term trend prediction model, wherein the long-term trend prediction model adopts an LSTM network structure and comprises an input layer, two hidden layers and an output layer, each hidden layer comprises 128 neurons, a tanh activation function is used, and a load trend of 24 hours in the future is predicted through the long-term trend prediction model, so that a long-term trend prediction result is obtained;

Inputting a seasonal item into a periodic pattern recognition model, wherein the periodic pattern recognition model adopts a one-dimensional convolutional neural network structure and comprises 3 convolutional layers and 2 full-connection layers, each convolutional layer uses a ReLU activation function and batch normalization, and the periodic pattern recognition model is applied to a time period of 24 hours in the future to obtain a periodic load prediction result;

The random item is input into a short-term fluctuation prediction model, the short-term fluctuation prediction model adopts a gating circulation unit network structure and comprises a bidirectional GRU layer and a full-connection layer, the GRU layer comprises 64 hidden units, and load fluctuation in the future for 24 hours is predicted through the short-term fluctuation prediction model, so that a short-term fluctuation prediction result is obtained;

And inputting the long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result into a multi-layer perceptron fusion network, wherein the multi-layer perceptron fusion network comprises 3 hidden layers, each layer comprises 64, 32 and 16 neurons respectively, a LeakyReLU activation function is used, and the last layer uses a linear activation function to obtain the daily load prediction result.

In this example, the historical electricity usage data is decomposed to yield trend terms, season terms, and random terms. The long-term trend, the periodic fluctuation and the short-term random fluctuation in the data can be effectively separated through time sequence analysis, and the time sequence is realizedDecomposition into trend termsSeason termRandom termsThe formula is as follows:

;

Wherein, Reflects the long-term electricity consumption trend, such as the trend that the electricity consumption rises or falls along with the change of time; Represents a periodic fluctuation of the electricity consumption, possibly related to the daily, weekly or yearly electricity consumption law, whereas Is a short term random fluctuation that cannot be explained by trend and seasonal patterns. Once the decomposition of the time series is completed, each part is modeled and predicted accordingly. The trend term reflects long-term electricity utilization changes and is suitable for prediction by using a long-term trend prediction model. The long-term trend prediction model can adopt an LSTM (long-term memory) network structure, and the LSTM network can memorize history information in a longer period of time when processing time series data, thereby being suitable for predicting long-term trend. The LSTM network comprises an input layer, two hidden layers and an output layer. Each hidden layer contains 128 neurons and uses the tanh activation function for nonlinear conversion. the tanh activation function is capable of compressing data to the range of [ -1,1] and effectively dealing with nonlinear features in the time series. By inputting the trend term into the LSTM network, the model can predict the long-term electricity utilization trend of 24 hours in the future, and a long-term trend prediction result is obtained. The seasonal items reflect periodic patterns in the electricity usage data. Since seasonal fluctuations typically have a fixed periodic character, such as daily early-late power utilization peaks or weekend power utilization changes, it is suitable to capture such periodic patterns using one-dimensional convolutional neural networks. The periodic pattern recognition model may employ a structure comprising 3 convolutional layers and 2 fully-connected layers. Each convolution layer uses a ReLU activation function and a batch normalization operation. The ReLU activation function can introduce nonlinearities after the convolution operation, while batch normalization can speed up training of the network and improve the stability of the predictions. By identifying the periodic mode of the season items, the model can accurately predict periodic electricity utilization wave within 24 hours in the future, and a periodic load prediction result is obtained. For random terms, it is suitable to use a gated loop unit (GRU) network for prediction of short term fluctuations, since it reflects short term random fluctuations, which cannot be explained by long term trends or periodic patterns. The GRU is a simplified recurrent neural network structure capable of capturing short-term dependencies in a time series with high efficiency. The short-term volatility prediction model may employ a bi-directional GRU layer comprising 64 hidden units and be provided with a fully connected layer for outputting the prediction results. The bi-directional GRU is able to take into account both past and future information of the time series and is therefore more accurate in capturing short term fluctuations. And (3) inputting the random term into a short-term fluctuation prediction model to predict the short-term fluctuation condition in the future 24 hours, so as to obtain a short-term fluctuation prediction result. And inputting the long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result into a multi-layer perceptron fusion network. The fusion network contains 3 hidden layers, 64, 32 and 16 neurons, respectively, each using LeakyReLU activation functions. LeakyReLU is an activation function capable of effectively avoiding neuronal death, and is suitable for solving the task of complex data distribution. The last layer uses a linear activation function to ensure that the output is a continuous load prediction result. And the three prediction results of the long-term trend, the periodic mode and the short-term fluctuation are used as input and are input into the multi-layer perceptron network. The fusion network can generate a more accurate comprehensive load prediction by learning complex relationships between these inputs. The final output layer contains 24 neurons corresponding to load predictions for each of the 24 hours in the future. Assume that the output of the converged network isThen:

;

Wherein, Is the final predicted load value that will be used,Is the result of the long-term trend prediction,Is a result of the prediction of the periodic load,Is the result of the short-term fluctuation prediction,、、Is a fused network learned weight. These weights are dynamically adjusted according to the conditions of different time periods, thereby ensuring the accuracy of the final predicted value.

In one example, the long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result are input into a multi-layer perceptron fusion network, wherein the multi-layer perceptron fusion network comprises 3 hidden layers, each layer comprises 64, 32 and 16 neurons respectively, a LeakyReLU activation function is used, and a linear activation function is used for the last layer to obtain a daily load prediction result, and the method comprises the following steps:

characteristic stitching is carried out on the long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result to obtain a fusion input vector;

Inputting the fusion input vector into a first hidden layer, wherein the first hidden layer comprises 64 neurons, and the output of each neuron is processed by LeakyReLU activation functions to obtain the output of the first hidden layer;

Carrying out batch normalization operation on the first hidden layer output, and adding the batch normalized output and the fusion input vector through residual connection to obtain a first layer residual output;

inputting the residual output of the first layer into a second hidden layer, wherein the second hidden layer comprises 32 neurons, and the output of each neuron is subjected to LeakyReLU activation function processing to obtain the output of the second hidden layer;

carrying out dropout operation on the second hidden layer output, setting the discarding rate to be 0.3, and obtaining the regularized second hidden layer output;

inputting the regularized second hidden layer output into a third hidden layer, wherein the third hidden layer comprises 16 neurons, and the output of each neuron is subjected to LeakyReLU activation function processing to obtain a third hidden layer output;

And outputting the third hidden layer to an input/output layer, wherein the output layer comprises 24 neurons, and the load predicted value corresponds to the load predicted value of 24 hours in the future, and a linear activation function is used to obtain a daily load predicted result.

In this example, the long-term trend prediction result, the periodic load prediction result, and the short-term fluctuation prediction result are feature-stitched, and a plurality of feature vectors are combined into one fusion vector. The long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result represent load change trend, periodic fluctuation and random fluctuation information within 24 hours in the future, respectively. Each prediction can be seen as a vector of length 24, representing the predicted value at each time instant within 24 hours. By feature stitching, the three vectors are combined into one fused input vector, 72 (24×3) in length. The fused input vector is input to the first hidden layer. The first hidden layer contains 64 neurons, each of which performs a linear transformation on the input data, and then applies a nonlinear activation function. The adopted activation function is LeakyReLU (LEAKY RECTIFIED LINEAR Unit), and the activation output with small amplitude can be still kept when the input is negative, so that the phenomenon of 'neuron death' is avoided. LeakyReLU is given by:

;

Wherein, Is an input to which the user is exposed,Is a small constant (typically 0.01) for controlling the magnitude of activation of the negative part. Through nonlinear activation, the first hidden layer can effectively extract complex features in input data to obtain 64 outputs. And carrying out batch normalization operation on the output of the first hidden layer, so that the data has more stable distribution when passing through the neural network, the problem of gradient disappearance or explosion is avoided, the training of the network is accelerated, and the stability of the model is improved. Batch normalization adjusts the data to a standard normal distribution, i.e., 0 for the mean and 1 for the variance, ensuring that the output of each layer is within a similar range of values. After the batch normalization is completed, the output after the batch normalization is added with the initial fusion input vector through residual connection, and residual output of the first layer is obtained. Residual connection is a technology used in deep neural networks and can solve the problem of gradient disappearance caused by the increase of the network depth. Residual connection ensures that the network does not rely too much on the learning results of the intermediate layer by skipping one or more hidden layers, adding the input directly to the output. The formula is:

Residual output ;

Wherein, Is the output of the hidden layer and,Is an input. Through residual connection, the network retains the original input information, so that the features cannot be lost in the deep network, and the performance of the model is improved. The residual output of the first layer is input to the second hidden layer. The second hidden layer contains 32 neurons, with the output of each neuron being processed non-linearly using LeakyReLU activation functions. The number of neurons of the second hidden layer is reduced compared to the first layer in order to progressively compress the feature space and extract higher level feature representations. Through LeakyReLU activation, the second hidden layer can process the input fusion features, and extract key modes and information. To prevent model overfitting, a dropout operation is applied after the output of the second hidden layer. dropout is a regularization technique that enhances the generalization ability of the model by randomly discarding a portion of neurons. In the training process, dropout forces the network to be independent of certain specific neurons by randomly setting the output of part of neurons to 0, so that the performance of the model on different data is improved. The drop rate of dropout is set to 0.3, meaning that 30% of the neuron outputs will be randomly dropped in each training. This operation can effectively prevent the model from overfitting on the training data and promote its performance on unseen data. And outputting and inputting the second hidden layer output subjected to dropout regularization to a third hidden layer. The third hidden layer contains 16 neurons, which are simpler in structure, for compressing features and extracting the most representative patterns. The output of each neuron is also processed by LeakyReLU activation functions to ensure that the model maintains nonlinear characteristics. Through the stepwise processing of the three hidden layers, the input data is converted into a higher level representation of the features. The output of the third hidden layer is input to the output layer. The output layer contains 24 neurons, each corresponding to a load prediction value for the next 24 hours. Since load prediction is a continuous regression task, the output layer uses a linear activation function. The linear activation function does not perform nonlinear conversion on the output value and can directly output the original predicted value. In this way, the output layer is able to generate accurate load prediction results.

Referring to fig. 2, the present embodiment provides a solid-state battery management apparatus of a shared charging device of a high cycle number, including:

a modeling module 1 for graph theory modeling of a battery management system in a shared charging device, and constructing a topological graph model comprising vertices representing ternary polymer lithium battery cells and management circuits and directed edges representing current flow paths;

The identification module 2 is used for carrying out depth-first search based on the topological graph model, identifying all charge and discharge paths meeting the circuit rule, and calculating the complex impedance value of each charge and discharge path;

The calculating module 3 is used for calculating a dynamic voltage spectrum and impedance characteristics based on the charge-discharge path and the complex impedance value to obtain time domain current distribution of each ternary polymer lithium battery unit;

The processing module 4 is used for carrying out multi-factor coupling loss calculation on each ternary polymer lithium battery unit according to the time domain current distribution to obtain dynamic loss distribution data;

and the evaluation module 5 is used for carrying out real-time evaluation on the battery state based on the dynamic loss distribution data, reconstructing a nonlinear charge-discharge characteristic curve, and carrying out multi-time scale cooperative scheduling based on the nonlinear charge-discharge characteristic curve to obtain a battery charge-discharge control strategy.

In this embodiment, for specific implementation of each unit in the above embodiment of the apparatus, please refer to the description in the above embodiment of the method, and no further description is given here.

Referring to fig. 3, in an embodiment of the present invention, there is further provided a computer device, which may be a server, and an internal structure thereof may be as shown in fig. 3. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected by a system bus. Wherein the computer is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used for communicating with an external terminal through a network connection. Which computer program, when being executed by a processor, carries out the above-mentioned method.

It will be appreciated by those skilled in the art that the architecture shown in fig. 3 is merely a block diagram of a portion of the architecture in connection with the present inventive arrangements and is not intended to limit the computer devices to which the present inventive arrangements are applicable.

An embodiment of the present invention also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the above method. It is understood that the computer readable storage medium in this embodiment may be a volatile readable storage medium or a nonvolatile readable storage medium.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium provided by the present invention and used in embodiments may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, apparatus, article, or method that comprises the element.

The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein based on the teachings of the present invention and the accompanying drawings, or any application, directly or indirectly, in other related technical fields.

Claims

1. A ternary polymer lithium battery management method for a shared charging device with a high cycle number, characterized in that it comprises the following steps:

Conduct graph theory modeling of the battery management system in the shared charging device, and construct a topological graph model containing vertices representing ternary polymer lithium battery cells and management circuits and directed edges representing current flow paths;

Performing a depth-first search based on the topology graph model to identify all charge and discharge paths that meet the circuit rules, and calculating the composite impedance value of each charge and discharge path;

Calculating a dynamic voltage spectrum and an impedance characteristic based on the charge and discharge path and the composite impedance value to obtain a time domain current distribution of each ternary polymer lithium battery cell;

According to the time-domain current distribution, a multi-factor coupling loss calculation is performed on each ternary polymer lithium battery unit to obtain dynamic loss distribution data;

Based on the dynamic loss distribution data, the battery state is evaluated in real time, and a nonlinear charge and discharge characteristic curve is reconstructed. Based on the nonlinear charge and discharge characteristic curve, multi-time scale collaborative scheduling is performed to obtain a battery charge and discharge control strategy.

2. The method for managing a ternary polymer lithium battery of a shared charging device with a high number of cycles according to claim 1, characterized in that the battery management system in the shared charging device is modeled by graph theory, and a topological graph model including vertices representing ternary polymer lithium battery cells and management circuits and directed edges representing current flow paths is constructed, comprising:

Identify the ternary polymer lithium battery unit, charging circuit, discharging circuit and protection circuit of the battery management system in the shared charging device to obtain the type and parameter information of each component;

Based on the type and parameter information of each component, vertex modeling is performed on each ternary polymer lithium battery unit and circuit component to obtain a vertex set;

According to the circuit topology of the battery management system, the flow path of the current between the components is analyzed to obtain a directed edge set;

Assigning attributes to each vertex in the vertex set to obtain vertex attribute data including the rated capacity, internal resistance and current state of charge of the ternary polymer lithium battery unit and the type and parameters of the circuit component, and assigning attributes to each edge in the directed edge set to obtain edge attribute data including the current flow direction, the maximum allowable current and the current current value;

Based on the vertex set, the directed edge set and the edge attribute data, a topological graph model of the battery management system is constructed.

3. The ternary polymer lithium battery management method for a shared charging device with a high number of cycles according to claim 2 is characterized in that the depth-first search based on the topological graph model is performed to identify all charge and discharge paths that meet the circuit rules, and the composite impedance value of each charge and discharge path is calculated, including:

Encoding the vertex set in the topological graph model, obtaining a unique identifier and access status flag of each vertex, and constructing a recursive function of depth-first search;

Traversing the topology graph model based on the recursive function to obtain all possible vertex access sequences, and performing circuit rule verification on each vertex access sequence to obtain a set of effective charging and discharging paths;

Based on the effective charge and discharge path set, the resistance, inductance and capacitance of each charge and discharge path are extracted to obtain a distributed parameter model of the charge and discharge path, and the distributed parameter model of the charge and discharge path is mathematically transformed to obtain a frequency domain composite impedance expression of each charge and discharge path;

Based on the frequency domain composite impedance expression, the equivalent impedance of the charge and discharge path within the operating frequency range is calculated to obtain the composite impedance value of each charge and discharge path.

4. The ternary polymer lithium battery management method for a shared charging device with a high number of cycles according to claim 1, characterized in that the dynamic voltage spectrum and impedance characteristics are calculated based on the charge and discharge path and the composite impedance value to obtain the time domain current distribution of each ternary polymer lithium battery cell, including:

Analyzing the topological structure of the charge and discharge path to obtain a time-domain differential equation group describing the dynamic behavior of the battery management system, and converting the time-domain differential equation group into an s-domain transfer function to obtain a frequency-domain mathematical model;

Based on the frequency domain mathematical model and the composite impedance value, a system matrix, an input matrix, an output matrix and a feedforward matrix of the battery management system are constructed, and discretized to obtain a discrete time system model;

Performing parameter identification on the discrete-time system model to obtain system characteristic parameters, applying a broadband excitation signal to the battery management system based on the system characteristic parameters, and obtaining a dynamic voltage spectrum through fast Fourier transform calculation;

Based on the dynamic voltage spectrum and input current information, the impedance characteristics of the battery management system are calculated, and the impedance characteristics are inverse Fourier transformed, and combined with the circuit topology information, the time domain current distribution of each ternary polymer lithium battery unit is obtained.

5. The ternary polymer lithium battery management method for a shared charging device with a high number of cycles according to claim 1, characterized in that the multi-factor coupling loss calculation is performed on each ternary polymer lithium battery unit according to the time domain current distribution to obtain dynamic loss distribution data, including:

Based on the time domain current distribution, the current square integral operation is performed on each ternary polymer lithium battery unit to obtain the initial value of Joule heat loss, and the temperature correction is performed on the initial value of Joule heat loss according to the real-time updated battery internal resistance parameter to obtain Joule heat loss data;

Calculating the charge transfer impedance based on the charge and discharge state of the ternary polymer lithium battery unit to obtain electrochemical reaction loss basic data, and performing nonlinear compensation on the electrochemical reaction loss basic data to obtain electrochemical reaction loss data;

Based on the battery cycle number and deep discharge state, a battery capacity attenuation model is constructed to obtain the cycle attenuation loss data during a single charge and discharge process;

Based on a multi-physics field coupling model, the Joule heat loss data, the electrochemical reaction loss data, and the cycle attenuation loss data are solved simultaneously to obtain coupling loss data;

The coupling loss data is subjected to a time series analysis to obtain a time-varying characteristic of a dynamic loss distribution, and based on the time-varying characteristic of the dynamic loss distribution, the Joule heat loss data, the electrochemical reaction loss data and the cycle attenuation loss data are weightedly fused to obtain dynamic loss distribution data.

6. The ternary polymer lithium battery management method for a shared charging device with a high number of cycles according to claim 1 is characterized in that the battery state is evaluated in real time based on the dynamic loss distribution data, and a nonlinear charge and discharge characteristic curve is reconstructed, and multi-time scale collaborative scheduling is performed based on the nonlinear charge and discharge characteristic curve to obtain a battery charge and discharge control strategy, including:

Performing piecewise linearization processing on the time-domain current distribution to obtain a discrete current sampling sequence, and calculating a battery state of charge change based on the discrete current sampling sequence to obtain an initial SOC estimation value;

According to the dynamic loss distribution data, the initial SOC estimation value is compensated and corrected to obtain a corrected SOC value;

An adaptive OCV-SOC mapping model is constructed based on historical charge and discharge data, and online parameter identification is performed on the adaptive OCV-SOC mapping model, and the model parameters are updated using a forgetting factor method to obtain an updated OCV-SOC mapping relationship;

Based on the corrected SOC value and the updated OCV-SOC mapping relationship, calculating the instantaneous open circuit voltage of the battery to obtain a SOC-OCV data pair;

Performing curve fitting on the SOC-OCV data pair to obtain a nonlinear characteristic equation of the charge and discharge process, and calculating the battery capacity and internal resistance at different SOCs based on the nonlinear characteristic equation to obtain a nonlinear charge and discharge characteristic curve;

Multi-time scale coordinated scheduling is performed based on the nonlinear charge and discharge characteristic curve to obtain a battery charge and discharge control strategy.

7. The ternary polymer lithium battery management method for a shared charging device with a high number of cycles according to claim 6, characterized in that the multi-time scale coordinated scheduling based on the nonlinear charge and discharge characteristic curve is performed to obtain a battery charge and discharge control strategy, including:

Based on the nonlinear charge-discharge characteristic curve, determining the charge-discharge efficiency and capacity limit at different states of charge;

According to the charging and discharging efficiency and the capacity limit, combined with historical electricity consumption data, a day-ahead load forecast is performed to obtain a day-ahead load forecast result;

Based on the day-ahead load forecast result, a day-ahead charge-discharge preliminary scheduling plan is formulated, and the day-ahead charge-discharge preliminary scheduling plan is optimized on a rolling basis to obtain a daily dynamic adjustment strategy;

According to the real-time load change, the intraday dynamic adjustment strategy is fine-tuned to generate real-time control instructions, and based on the real-time control instructions, the battery charging and discharging operations are executed, and the charging and discharging process data are recorded;

Based on the charging and discharging process data, a battery state evaluation model is updated, and the nonlinear charging and discharging characteristic curve is optimized to obtain an updated nonlinear charging and discharging characteristic curve;

Multi-time scale coordinated scheduling adjustment is performed according to the updated nonlinear charging and discharging characteristic curve to obtain the battery charging and discharging control strategy.

8. The method for managing a ternary polymer lithium battery of a shared charging device with a high number of cycles according to claim 7, characterized in that the day-ahead load forecast is performed based on the charge and discharge efficiency and the capacity limit in combination with historical power consumption data to obtain the day-ahead load forecast result, including:

Perform time series decomposition on historical electricity consumption data to obtain trend items, seasonal items, and random items;

The trend item is input into a long-term trend prediction model, which adopts an LSTM network structure, including an input layer, two hidden layers and an output layer, each hidden layer includes 128 neurons, uses a tanh activation function, and predicts the load trend in the next 24 hours through the long-term trend prediction model to obtain a long-term trend prediction result;

The seasonal term is input into a periodic pattern recognition model, which adopts a one-dimensional convolutional neural network structure, including 3 convolutional layers and 2 fully connected layers, each convolutional layer uses a ReLU activation function and batch normalization, and the periodic pattern recognition model is applied to the next 24 hours to obtain a periodic load forecast result;

The random item is input into a short-term fluctuation prediction model, which adopts a gated recurrent unit network structure, including a bidirectional GRU layer and a fully connected layer, and the GRU layer includes 64 hidden units. The load fluctuation in the next 24 hours is predicted by the short-term fluctuation prediction model to obtain a short-term fluctuation prediction result;

The long-term trend forecast result, the periodic load forecast result and the short-term fluctuation forecast result are input into a multi-layer perceptron fusion network, wherein the multi-layer perceptron fusion network comprises three hidden layers, each layer comprises 64, 32 and 16 neurons respectively, and uses a LeakyReLU activation function, and the last layer uses a linear activation function to obtain the day-ahead load forecast result.

9. The ternary polymer lithium battery management method for a shared charging device with a high number of cycles according to claim 8, characterized in that the long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result are input into a multi-layer perceptron fusion network, the multi-layer perceptron fusion network comprises 3 hidden layers, each layer comprises 64, 32 and 16 neurons respectively, uses a LeakyReLU activation function, and the last layer uses a linear activation function to obtain a day-ahead load prediction result, including:

Performing feature splicing on the long-term trend prediction result, the periodic load prediction result and the short-term fluctuation prediction result to obtain a fusion input vector;

Input the fused input vector into a first hidden layer, where the first hidden layer includes 64 neurons, and the output of each neuron is processed by a LeakyReLU activation function to obtain the output of the first hidden layer;

Performing a batch normalization operation on the first hidden layer output, and adding the batch normalized output to the fused input vector through a residual connection to obtain a first layer residual output;

The residual output of the first layer is input into the second hidden layer, where the second hidden layer includes 32 neurons, and the output of each neuron is processed by the LeakyReLU activation function to obtain the output of the second hidden layer;

Perform a dropout operation on the second hidden layer output, with the dropout rate set to 0.3, to obtain the regularized second hidden layer output;

Input the regularized second hidden layer output into the third hidden layer, wherein the third hidden layer comprises 16 neurons, and the output of each neuron is processed by the LeakyReLU activation function to obtain the third hidden layer output;

The output of the third hidden layer is input into the output layer, and the output layer includes 24 neurons corresponding to the load forecast value of the next 24 hours. A linear activation function is used to obtain the day-ahead load forecast result.

10. A solid-state battery management device for a shared charging device with a high number of cycles, characterized in that it is used to implement the steps of the method described in any one of claims 1 to 9, and the device comprises:

A modeling module, used for graph-theoretic modeling of the battery management system in the shared charging device, and constructing a topological graph model including vertices representing ternary polymer lithium battery cells and management circuits and directed edges representing current flow paths;

An identification module, used to perform a depth-first search based on the topological graph model, identify all charge and discharge paths that meet the circuit rules, and calculate the composite impedance value of each charge and discharge path;

A calculation module, used to calculate the dynamic voltage spectrum and impedance characteristics based on the charge and discharge path and the composite impedance value, and obtain the time domain current distribution of each ternary polymer lithium battery unit;

A processing module, used to perform multi-factor coupling loss calculation on each ternary polymer lithium battery unit according to the time domain current distribution to obtain dynamic loss distribution data;

An evaluation module is used to perform real-time battery status evaluation based on the dynamic loss distribution data, reconstruct a nonlinear charge and discharge characteristic curve, and perform multi-time scale collaborative scheduling based on the nonlinear charge and discharge characteristic curve to obtain a battery charge and discharge control strategy.