CN118393265B

CN118393265B - Testing method and system for photovoltaic storage inverter

Info

Publication number: CN118393265B
Application number: CN202410843966.5A
Authority: CN
Inventors: 肖志勇; 熊俊峰; 梁远文
Original assignee: Shenzhen Jia Chuang Dt Science Co ltd
Current assignee: Shenzhen Jia Chuang Dt Science Co ltd
Priority date: 2024-06-27
Filing date: 2024-06-27
Publication date: 2024-09-03
Anticipated expiration: 2044-06-27
Also published as: CN118393265A

Abstract

The application relates to the technical field of testing of optical storage inverters, and discloses a testing method and a testing system of an optical storage inverter. The method comprises the following steps: setting a test parameter set of the light storage inverter; transmitting the test parameter set to a preset three-port test power supply, and controlling a plurality of output circuits in the three-port test power supply to respond to the test parameters to generate initial output circuit parameters; performing operation test on the optical storage inverter, collecting output performance and functional data, and calculating to obtain a plurality of target monitoring indexes; acquiring a plurality of standard monitoring indexes of the light storage inverter, and calculating monitoring index difference data between the plurality of standard monitoring indexes and a plurality of target monitoring indexes; adjusting circuit parameters of the output circuit of the three-port test power supply according to the monitoring index difference data to obtain target output circuit parameters of the three-port test power supply, the application improves the testing accuracy of the light storage inverter and improves the performance and the function of the light storage inverter.

Description

Test method and system of light storage inverter

Technical Field

The application relates to the technical field of testing of optical storage inverters, in particular to a testing method and a testing system of an optical storage inverter.

Background

The light storage inverter is a device for storing and converting electric energy, and is widely applied to the field of renewable energy sources and smart grid systems. In order to ensure proper operation and performance of the light storage inverter, stringent testing and verification is required. At present, a plurality of independent power supplies are required for testing the light storage inverter, which not only increases testing cost and complexity, but also easily causes testing errors, and the energy is absorbed or fed back from the alternating current power grid and the alternating current power grid causes relatively complex power supply topology and lower energy conversion efficiency.

Disclosure of Invention

The application provides a test method and a test system of an optical storage inverter, which are used for improving the test accuracy of the optical storage inverter and improving the performance and the function of the optical storage inverter.

In a first aspect, the present application provides a method for testing an optical storage inverter, the method comprising:

setting a test parameter set of the light storage inverter, wherein the test parameter set comprises output voltage, output frequency and output power;

transmitting the test parameter set to a preset three-port test power supply, and controlling a plurality of output circuits in the three-port test power supply to respond to the test parameters to generate initial output circuit parameters of the three-port test power supply;

Performing operation test on the light storage inverter, collecting output performance and functional data of the light storage inverter, and performing monitoring index calculation on the output performance and the functional data to obtain a plurality of target monitoring indexes;

Acquiring a plurality of standard monitoring indexes of the light storage inverter, and calculating monitoring index difference data between the plurality of standard monitoring indexes and the plurality of target monitoring indexes;

and adjusting circuit parameters of the output circuit of the three-port test power supply according to the monitoring index difference data to obtain target output circuit parameters of the three-port test power supply.

In a second aspect, the present application provides a test system for an optical storage inverter, the test system comprising:

the setting module is used for setting a test parameter set of the light storage inverter, wherein the test parameter set comprises output voltage, output frequency and output power;

The transmission module is used for transmitting the test parameter set to a preset three-port test power supply, controlling a plurality of output circuits in the three-port test power supply to respond to the test parameters and generating initial output circuit parameters of the three-port test power supply;

the acquisition module is used for performing operation test on the optical storage inverter, acquiring output performance and functional data of the optical storage inverter, and performing monitoring index calculation on the output performance and the functional data to obtain a plurality of target monitoring indexes;

The calculation module is used for acquiring a plurality of standard monitoring indexes of the light storage inverter and calculating monitoring index difference data between the standard monitoring indexes and the target monitoring indexes;

And the adjusting module is used for adjusting the circuit parameters of the output circuit of the three-port test power supply according to the monitoring index difference data to obtain the target output circuit parameters of the three-port test power supply.

In the technical scheme provided by the application, (1) the test cost and time are saved, and the requirement on an independent power supply is reduced by integrating a plurality of output circuits; (2) The accuracy and the repeatability of the test are improved, and the test error is avoided through unified control and parameter transmission; (3) The flexibility and the adaptability of the test are improved, and the test requirements of different light storage inverters can be met by adjusting the working state and the parameters of the output circuit; (4) The mode that the input ends of the three output circuits share the direct current bus reduces the load degree of the power supply topology and increases the energy conversion efficiency.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic diagram of an embodiment of a method for testing an inverter of an optical storage device;

Fig. 2 is a schematic diagram of a test system for an optical storage inverter according to an embodiment of the application.

Detailed Description

The embodiment of the application provides a test method and a test system of an optical storage inverter. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

For ease of understanding, a specific flow of an embodiment of the present application is described below with reference to fig. 1, and an embodiment of a method for testing an optical storage inverter in an embodiment of the present application includes:

Step S101, setting a test parameter set of an optical storage inverter, wherein the test parameter set comprises output voltage, output frequency and output power;

it is to be understood that the execution body of the present application may be a test system of an optical storage inverter, and may also be a terminal or a server, which is not limited herein. The embodiment of the application is described by taking a server as an execution main body as an example.

Specifically, the operation requirements and performance requirements of the light storage inverter under different working environments and working states are defined. The output performance of the light storage inverter as a power electronic device is directly related to the stability and efficiency of the whole photovoltaic energy storage system. By analyzing the technical specification and the application scene of the optical storage inverter, key parameters needing to be focused in the test process are determined. The output voltage is considered when setting the test parameter set. The output voltage is an important index for measuring whether the light storage inverter can stably output electric energy. And testing the output voltage under different load conditions, including various working conditions such as no-load, half-load and full-load. Meanwhile, the stability of the output voltage of the inverter under different input voltage conditions is considered. When setting the test parameters, the nominal output voltage of the light storage inverter and the possible fluctuation range should be covered. The output frequency is also a critical test parameter. The light storage inverter needs to convert direct current into alternating current, and the stability of output frequency directly influences the quality of electric energy and the normal operation of the system. Typically, the output frequency of the light storage inverter should be maintained at a fixed value, such as 50Hz or 60Hz. In actual operation, the output frequency may be affected by load variations, input voltage fluctuations, and the like. When the test parameters are set, the output frequencies under different loads and input conditions are tested, so that the output frequencies are ensured to fluctuate within a specified range. The output power of the light storage inverter determines the maximum electrical energy it can provide to the load. To ensure the comprehensiveness and accuracy of the test, the output power of the light storage inverter under different input powers and load conditions is tested. For example, the output power of the inverter under these conditions was tested by adjusting the power output to the photovoltaic cells, simulating different solar radiation intensities. At the same time, the performance of the inverter under overload and underload conditions is taken into account to ensure that it is still operating stably under extreme conditions. In order to achieve the above object, a set of automated test systems is used to transmit a set of set test parameters to a preset three-port test power supply. And responding to the test parameters by controlling a plurality of output circuits in the three-port test power supply, and generating initial output circuit parameters of the three-port test power supply.

Step S102, transmitting the test parameter set to a preset three-port test power supply, and controlling a plurality of output circuits in the three-port test power supply to respond to the test parameters so as to generate initial output circuit parameters of the three-port test power supply;

Specifically, the test parameter set is transmitted to a preset three-port test power supply, which includes an input circuit and three output circuits, each of which is capable of independently providing power to the optical storage inverter. The input circuit is an ACDC module and is used for providing direct current bus voltage for each output circuit module, and meanwhile, the energy compensation or feedback function is realized, so that the stability and the reliability of the test power supply are ensured. Knowing the functions and the configurations of three output circuits, wherein the first output circuit is a PV simulation module which is designed as a unidirectional DCDC converter and is used for simulating the output characteristics of a photovoltaic cell, and the simulation of the photovoltaic power generation working condition of the input end of the photovoltaic inverter is realized by adjusting the output voltage and the output current of the DCDC converter; the second output circuit is a battery simulation module, is designed as a bidirectional DCDC converter, and can simulate the charging and discharging processes of the battery and can also perform bidirectional flow of energy in the testing process so as to ensure the comprehensiveness and accuracy of the test; the third output circuit is a power grid simulation and nonlinear load simulation module, the module is designed into a bidirectional DCAC converter, the power supply condition of the power grid and various nonlinear loads can be simulated, and the performance test of the light storage inverter under the complex load condition is realized by adjusting the output parameters of the DCAC converter. And transmitting the test parameter set to three output circuits controlled by the three-port test power supply to perform test parameter response. The specific operation comprises the steps of adjusting output parameters of the PV simulation module, the battery simulation module and the power grid simulation and nonlinear load simulation module through a control system, so that the output parameters respond to various indexes in a test parameter set. For example, simulation of photovoltaic power generation is achieved by adjusting the output voltage and current of the unidirectional DCDC converter; simulating the working state of the battery by controlling the charge and discharge process of the bidirectional DCDC converter; and regulating the output voltage, frequency and waveform through the bidirectional DCAC converter, and simulating the conditions of power supply of a power grid and nonlinear loads. The three output circuits are correspondingly adjusted according to the test parameter sets respectively to generate initial output circuit parameters of the three-port test power supply, wherein the initial output circuit parameters comprise output voltage, frequency and waveform.

Step S103, performing operation test on the optical storage inverter, collecting output performance and functional data of the optical storage inverter, and performing monitoring index calculation on the output performance and the functional data to obtain a plurality of target monitoring indexes;

specifically, when the operation test is performed on the light storage inverter, the operation state of the light storage inverter under the actual working condition is simulated, and the comprehensive test is performed under different load and input conditions. Output performance data and functional data of the light storage inverter, typically including output voltage, current, power, frequency, and other key performance indicators, are collected in real time during the test. And processing the output performance and the functional data. To ensure accuracy and consistency of data processing, time alignment and curve fitting are performed on the acquired data. The purpose of the timing alignment is to ensure that all data points are compared at the same time reference, while curve fitting is used to eliminate noise and outliers in the data, resulting in a smoother and representative target curve. And respectively carrying out mean value and standard deviation calculation on the plurality of target curves. The mean reflects the central tendency of the data, while the standard deviation describes the degree of dispersion of the data. And (3) calculating the mean value and standard deviation of each target curve to know the performance stability and fluctuation condition of the light storage inverter under different operating conditions. This helps identify the normal operating range of the light storage inverter and also can discover potential performance anomalies. And extracting characteristics of a plurality of target curves according to the mean value and the standard deviation, extracting a representative initial characteristic value set from a large amount of original data, wherein the characteristic values can better reflect the performance condition of the light storage inverter. By feature extraction, the complexity of the data is simplified while retaining information important to performance assessment. And performing principal component dimension reduction on the initial characteristic value set of each target curve to obtain a target characteristic value set of each target curve. The principal component dimension reduction is a data dimension reduction technique that reduces the dimension of data by projecting the data in a high-dimensional feature space into a lower-dimensional space while retaining as much important information as possible in the original data. And inputting the target characteristic value set of each target curve into a preset index analysis model, and performing monitoring index calculation. The index analysis model is based on a statistical method, a machine learning algorithm or other data analysis technology model, and a plurality of target monitoring indexes are calculated according to the input characteristic value set. The monitoring indexes can comprehensively reflect the output performance and the functional state of the light storage inverter.

And inputting the target characteristic value set of each target curve into a preset index analysis model to perform monitoring index calculation, so as to obtain a plurality of target monitoring indexes. The index analysis model comprises an embedded layer and a plurality of Bi-directional long-short-term memory networks (Bi-LSTM), wherein each Bi-LSTM network is connected with one full-connection layer. In the embedded layer, feature normalization processing and vector code conversion are carried out on the target feature value set of each target curve. The purpose of feature normalization is to eliminate the dimension difference between different feature values, so that all the feature values are in the same dimension range, and the training effect and the calculation efficiency of the model are improved. Vector code conversion is carried out on the eigenvalue set after normalization processing, and the eigenvalue set is converted into an input eigenvector which can be processed by a model. A plurality of input feature encoding vectors are respectively input into a plurality of two-way long-short-term memory networks. The two-way long-short-term memory network is a neural network structure capable of capturing the front-back dependency relationship in time series data, and the time sequence characteristics in the data are more comprehensively captured by simultaneously processing input data through the two forward and reverse long-short-term memory network layers. And in each two-way long-short-term memory network, performing time sequence feature analysis on the input feature coding vector to obtain a time sequence feature fusion vector of each two-way long-term memory network. And inputting the obtained time sequence feature fusion vector into the full connection layer. The full connection layer realizes the combination and extraction of the characteristics by connecting all input nodes to all output nodes. In the full connection layer, the ReLU activation function is adopted to carry out nonlinear transformation on the input time sequence feature fusion vector. The ReLU function is a common activation function, can effectively solve the problem of gradient disappearance, and improves the training speed and performance of the model. Through the processing of the full connection layer and the ReLU function, a plurality of target monitoring metrics are calculated, including output power, efficiency, voltage and current stability. Specifically, the output power monitoring index may reflect the power output of the light storage inverter under different operating conditions; the efficiency monitoring index is used for evaluating the energy conversion efficiency of the light storage inverter and ensuring the high-efficiency operation of the light storage inverter in practical application; the voltage and current stability monitoring index can detect fluctuation conditions of output voltage and current of the light storage inverter, and can ensure that the light storage inverter can still stably operate when load changes or input voltage fluctuates.

A plurality of input feature encoding vectors are respectively input into two-way long-short-term memory networks, each two-way long-term memory network is composed of a forward long-term memory network and a backward long-term memory network. The forward long-short term memory network comprises 256 LSTM units connected in one way, the characteristic analysis is carried out from the initial point to the final point of the time sequence, and the backward long-short term memory network also comprises 256 LSTM units connected in one way, and the analysis direction is from the final point to the initial point of the time sequence. In each two-way long-short-term memory network, the forward long-term memory network performs forward time sequence feature analysis on the input feature code vector, and captures the forward dependency of the input data on the time sequence to obtain a forward time sequence feature vector. The forward timing feature vector reflects the law of variation of the input features over time. And meanwhile, the backward long-short-term memory network performs backward time sequence feature analysis on the same input feature code vector, captures backward dependency relationship on time sequence, and obtains a backward time sequence feature vector. Vector splicing is carried out on the two feature vectors, and the forward time sequence feature vector and the backward time sequence feature vector are connected together according to a certain rule to form a time sequence feature fusion vector.

Step S104, a plurality of standard monitoring indexes of the light storage inverter are obtained, and monitoring index difference data between the plurality of standard monitoring indexes and the plurality of target monitoring indexes are calculated;

Specifically, a plurality of standard monitoring indicators of the light storage inverter are obtained. Standard monitoring metrics are a series of performance metrics determined based on light storage inverter design specifications and industry standards, which typically include output power, conversion efficiency, voltage stability, current stability, total harmonic distortion, and the like. And performing operation test on the optical storage inverter to obtain actual output performance and functional data. In the running test process, performance data of the light storage inverter under different working conditions are recorded through test operation and data acquisition means. These data include real-time output voltage, current, power, efficiency, and other key parameters. And comparing and analyzing the obtained standard monitoring index with a target monitoring index obtained by an actual test. And calculating monitoring index difference data between the standard monitoring index and the target monitoring index. Including direct numerical comparison, difference calculation, statistical analysis, etc. And calculating the difference value between each standard monitoring index and the corresponding target monitoring index to obtain difference data. The discrepancy data reflects the deviation of the light storage inverter from design criteria during actual operation, thereby helping to identify potential problems and improvement points. In addition to difference calculations, more complex statistical analysis methods such as mean comparison, analysis of variance, regression analysis, etc. may be employed. For example, by analysis of variance, the influence of different working conditions on the performance of the photovoltaic inverter is evaluated, and the maximum performance deviation under which conditions is found out; and establishing a mathematical model between the standard monitoring index and the target monitoring index through regression analysis, and predicting the performance change condition under different conditions.

And step 105, adjusting circuit parameters of the output circuit of the three-port test power supply according to the monitoring index difference data to obtain target output circuit parameters of the three-port test power supply.

Specifically, the monitoring index difference data is input into a preset circuit parameter compensation model, and the model is composed of an input layer, a plurality of convolution layers, a pooling layer and a parameter optimization layer. And inputting the monitoring index difference data into an input layer for feature extraction. The raw difference data is converted into feature vectors that can be processed by the model. And inputting the index difference characteristics output by the input layer into a plurality of convolution layers to carry out convolution operation. Higher-level convolution features are extracted from the input features by a convolution operation. Each convolution layer contains a plurality of convolution kernels, each of which is responsible for extracting different aspects of the input features, such as local patterns and spatial relationships. By superposition of a plurality of convolution layers, more complex and abstract features are extracted layer by layer, so that the essence of monitoring index difference is better described. The convolved feature vector is input to the pooling layer for downsampling and feature compression. The role of the pooling layer is to reduce the dimensionality of the feature vectors while preserving the most important feature information. Through downsampling operation, complexity of feature vectors is reduced, calculated amount is reduced, and calculation efficiency and stability of the model are improved. The pooling layer generally adopts a mode of maximum pooling or average pooling to process the convolution eigenvectors to obtain pooled eigenvectors. The pooled feature vectors are input to a parameter optimization layer for circuit parameter adjustment and parameter combination. The parameter optimization layer is an output layer of the whole model, and a final circuit parameter adjustment scheme is obtained by processing and optimizing the pooling feature vector. The parameter optimization layer generally adopts a fully-connected network structure, and comprehensively processes the input feature vectors through a nonlinear activation function and an optimization algorithm to output final target circuit parameters.

And defining a fitness function of a particle swarm optimization algorithm in the parameter optimization layer. The fitness function is used for evaluating the quality of each output circuit parameter, so as to guide the optimization process. The fitness function is typically based on performance metrics of the test system, such as output voltage, frequency, and stability and accuracy of the waveform. And carrying out parameter range prediction on the output voltage, the frequency and the waveform according to the pooled feature vector to obtain a parameter range set. And (5) preliminarily determining a reasonable range of each circuit parameter through analysis of the pooled feature vectors. The parameter range set is used as input of a particle swarm optimization algorithm to guide the particle swarm optimization algorithm to search the optimal parameters in a reasonable range. And initializing output circuit parameters according to the ranges by a particle swarm optimization algorithm to generate a corresponding particle swarm. The population of particles is comprised of a plurality of first output circuit parameters, each parameter representing one possible circuit configuration. In the particle swarm optimization algorithm, each particle represents a candidate solution, and the optimal solution is gradually approached by simulating the flight and searching behaviors of the particle in the search space. And respectively calculating the fitness value of each first output circuit parameter through a fitness function. The fitness value reflects the performance quality of each circuit parameter configuration, and the particle swarm is divided into a plurality of first output circuit parameters according to the fitness value, so that a plurality of particle swarms are obtained. And (3) grouping particles with better performance through comparison of fitness values, so that resources are concentrated for optimization in a subsequent optimization process. And generating a plurality of corresponding second output circuit parameters according to the plurality of sub-particle groups, and carrying out optimization solution on the parameters. The particle swarm optimization algorithm enables the particle swarm to gradually trend to the global optimal solution in the search space through iteratively updating the speed and the position of the particles. In each iteration process, the particles adjust positions according to own experience and experience of neighbor particles, so that output circuit parameter configuration is continuously improved. In the optimization solving process, the fitness value of each particle is continuously evaluated and compared through the fitness function, so that the updating direction of the particle is guided. Finally, the target output circuit parameters of the three-port test power supply are obtained through multiple iterations and optimization. The target output circuit parameters are obtained by global searching and optimizing through a particle swarm optimization algorithm on the basis of comprehensively considering a plurality of performance indexes such as output voltage, frequency, waveform and the like.

In the embodiment of the application, (1) test cost and time are saved, and the requirement for an independent power supply is reduced by integrating a plurality of output circuits; (2) The accuracy and the repeatability of the test are improved, and the test error is avoided through unified control and parameter transmission; (3) The flexibility and the adaptability of the test are improved, and the test requirements of different light storage inverters can be met by adjusting the working state and the parameters of the output circuit; (4) The mode that the input ends of the three output circuits share the direct current bus reduces the load degree of the power supply topology and increases the energy conversion efficiency.

In a specific embodiment, the process of executing step S102 may specifically include the following steps:

(1) Transmitting the test parameter set to a preset three-port test power supply, wherein the three-port test power supply comprises: an input circuit and three output circuits, each output circuit can independently provide power for the optical storage inverter, and the input circuit is ACDC module for each output circuit module provides direct current bus voltage and energy compensation or repayment, three output circuits include: the first output circuit is a PV analog module, and the PV analog module is a unidirectional DCDC converter; the second output circuit is a battery simulation module, the battery simulation module is a bidirectional DCDC converter, the third output circuit is a power grid simulation and nonlinear load simulation module, and the power grid simulation and nonlinear load simulation module is a bidirectional DCAC converter;

(2) Controlling a plurality of output circuits in the three-port test power supply to respond to the test parameters, and generating initial output circuit parameters of the three-port test power supply, wherein the initial output circuit parameters comprise: output voltage, frequency and waveform.

In particular, the configuration and the function of the three-port test power supply are ensured to meet the test requirements. The three-port test power supply includes an input circuit and three output circuits, each of which is capable of independently providing power to the optical storage inverter. The input circuit is an ACDC module, and the main function of the input circuit is to provide direct current bus voltage for each output circuit module, and meanwhile, energy compensation or feedback is realized, so that the circuit can maintain stable voltage supply in the test process. The first output circuit is a PV simulation module designed as a unidirectional DCDC converter for simulating the output characteristics of a photovoltaic cell. The PV simulation module simulates the working conditions of the photovoltaic cell under different illumination intensities by adjusting the output voltage and current of the DCDC converter. The second output circuit is a battery analog module designed as a bi-directional DCDC converter. The module not only can simulate the charge and discharge process of the battery, but also can perform bidirectional flow of energy in the test process, and simulate the working characteristics of the battery in different states. The third output circuit is a power grid simulation and nonlinear load simulation module, the module is designed into a bidirectional DCAC converter, the power supply condition of the power grid and various nonlinear loads can be simulated, and the performance test of the light storage inverter under the complex load condition is realized by adjusting the output parameters of the DCAC converter. And transmitting the preset test parameter set to a three-port test power supply. The test parameter set comprises key parameters such as output voltage, frequency and power, and the parameters are determined according to the technical specifications and application requirements of the light storage inverter. These parameters are transmitted to a three-port test power supply via a control system or interface to ensure that each output circuit can operate in accordance with the set parameters. And controlling a plurality of output circuits in the three-port test power supply to respond to the test parameters by the control system. And the PV simulation module adjusts the output voltage and current of the DCDC converter according to the transmitted parameters and simulates the output characteristics of the photovoltaic cells under different illumination conditions. For example, the DCDC converter may output higher voltages and currents when simulating high light intensities, whereas the output voltages and currents are lower when simulating low light intensities. The PV simulation module can simulate different working states of the photovoltaic cell in actual use by adjusting output voltage and current. Meanwhile, the battery simulation module adjusts the charge and discharge process of the bidirectional DCDC converter according to the transmitted parameters. For example, in modeling battery state of charge, the DCDC converter may input energy to a battery modeling module that models the charging process of the battery; when the discharge state of the battery is simulated, the DCDC converter outputs energy to simulate the discharge process of the battery. By the mode, the battery simulation module can truly reflect the working characteristics of the battery in different charge and discharge states, and necessary support is provided for performance test of the light storage inverter. And the power grid simulation and nonlinear load simulation module regulates the output voltage, frequency and waveform of the bidirectional DCAC converter according to the transmitted parameters. The module can simulate power supply conditions of a power grid, for example, when the power grid is simulated to normally supply power, the DCAC converter outputs stable voltage and frequency; when the analog power grid fails or fluctuates, the output voltage and frequency of the DCAC converter change correspondingly. In addition, the module can simulate nonlinear load conditions, and test the performance of the light storage inverter under complex load conditions by outputting currents with different waveform characteristics. After the adjustment is completed, the three-port test power supply generates initial output circuit parameters. These parameters include output voltage, frequency and waveform, and by monitoring and recording these parameters in real time, the performance of the light storage inverter under different operating conditions is assessed. For example, the output voltage and current of the PV simulation module can reflect the working conditions of the photovoltaic cells under different illumination conditions, the charge-discharge characteristics of the battery simulation module can evaluate the energy storage capacity and efficiency of the battery, and the output characteristics of the grid simulation and nonlinear load simulation module can detect the stability and reliability of the light storage inverter under complex grid and load conditions.

In a specific embodiment, the process of executing step S103 may specifically include the following steps:

(1) Performing operation test on the optical storage inverter and collecting output performance and functional data of the optical storage inverter;

(2) Performing time sequence alignment and curve fitting on the output performance and the functional data to obtain a plurality of target curves;

(3) Respectively carrying out mean value and standard deviation calculation on a plurality of target curves to obtain the mean value and standard deviation of each target curve;

(4) Extracting features of a plurality of target curves according to the mean value and the standard deviation to obtain an initial feature value set of each target curve;

(5) Performing main component dimension reduction on the initial characteristic value set of each target curve to obtain a target characteristic value set of each target curve;

(6) And respectively inputting the target characteristic value set of each target curve into a preset index analysis model to perform monitoring index calculation, so as to obtain a plurality of target monitoring indexes.

Specifically, the optical storage inverter is subjected to operation test and output performance and function data of the optical storage inverter are collected. In the actual test process, the optical storage inverter is connected to a simulated power grid and a load system, and the working state of the optical storage inverter in the actual application is simulated by setting different working conditions. The output performance and the functional data of the light storage inverter are monitored and recorded in real time through the data acquisition system, wherein the output performance and the functional data comprise output voltage, current, power, frequency, conversion efficiency and the like. And performing time sequence alignment and curve fitting on the output performance and functional data. The purpose of the timing alignment is to ensure that all data points are compared at the same time reference, avoiding data errors due to time differences. And aligning the data points acquired at different time points through a time sequence alignment technology so as to analyze the data points on the same time axis. Noise and outliers in the data are eliminated by curve fitting techniques, resulting in a smoother and representative target curve. For example, polynomial fitting, spline fitting or other curve fitting methods can be adopted to perform fitting processing on the original data, so as to obtain a smooth curve which meets the practical situation. And carrying out statistical analysis on the target curves, and respectively calculating the mean value and standard deviation of each target curve. The average value reflects the central trend of the target curve, namely the average output performance of the light storage inverter under different working conditions; the standard deviation describes the degree of dispersion of the data, i.e. the fluctuation of the output performance. And (3) calculating the mean value and the standard deviation to know the performance stability and consistency of the light storage inverter under different working conditions. And extracting the characteristics of the plurality of target curves according to the calculated mean value and standard deviation. The purpose of feature extraction is to extract a representative initial set of feature values from a large amount of raw data that better reflect the performance of the light storage inverter. The feature extraction may employ various methods such as time domain feature extraction, frequency domain feature extraction, or time-frequency domain feature extraction. And extracting the characteristics of the target curves to obtain an initial characteristic value set of each target curve. And performing principal component dimension reduction on the initial characteristic value set. The principal component dimension reduction is a data dimension reduction technique that reduces the dimension of data by projecting the data in a high-dimensional feature space into a lower-dimensional space while retaining as much important information as possible in the original data. Through dimension reduction processing, the complexity of data is simplified, the calculation efficiency is improved, and the influence of data noise is reduced. And inputting the target characteristic value set of each target curve into a preset index analysis model, and performing monitoring index calculation. The index analysis model may be a model based on a statistical method, a machine learning algorithm, or other data analysis techniques, and calculates a plurality of target monitoring indexes from the input feature value set. For example, predicting the output performance of the light storage inverter under different operating conditions by a regression analysis model; identifying the working state of the light storage inverter through a classification model; and analyzing the performance characteristic distribution of the light storage inverter through a clustering model. And through analysis of the models, a plurality of target monitoring indexes are obtained, and the monitoring indexes can comprehensively reflect the output performance and the functional state of the light storage inverter.

In a specific embodiment, the executing step inputs the target feature value set of each target curve into a preset index analysis model to perform monitoring index calculation, and the process of obtaining a plurality of target monitoring indexes may specifically include the following steps:

(1) The index analysis model includes: the embedded layer and the two-way long-short-period memory networks are connected with a full-connection layer;

(2) In the embedded layer, respectively carrying out feature normalization and vector code conversion on a target feature value set of each target curve to generate a plurality of input feature code vectors;

(3) Respectively inputting a plurality of input feature coding vectors into a plurality of two-way long-short-term memory networks, and performing time sequence feature analysis on the input feature coding vectors through each two-way long-term memory network to obtain time sequence feature fusion vectors of each two-way long-term memory network;

(4) And inputting the time sequence feature fusion vector into a full-connection layer, and performing monitoring index calculation through a ReLU function in the full-connection layer to obtain a plurality of target monitoring indexes, wherein the plurality of target monitoring indexes comprise output power, efficiency, voltage and current stability.

Specifically, in the embedded layer, feature normalization and vector code conversion are respectively performed on the target feature value set of each target curve. The purpose of feature normalization is to convert the feature values of different dimensions into the same scale range, and eliminate the influence of dimension differences on model training. Common normalization methods include min-max normalization and normalization. The vector code conversion is to convert the normalized eigenvalue set into a vector form which can be processed by a model to generate a plurality of input eigenvectors. A plurality of input feature encoding vectors are respectively input into a plurality of two-way long-short-term memory networks (Bi-LSTM). The Bi-LSTM network is a neural network structure capable of capturing the front-back dependency relationship in time series data, and input data is processed simultaneously through the forward LSTM layer and the backward LSTM layer, so that the time sequence characteristics in the data are more comprehensively captured. And in each Bi-LSTM network, carrying out time sequence feature analysis on the input feature coding vector to obtain a time sequence feature fusion vector of each Bi-LSTM network. And inputting the obtained time sequence feature fusion vector into a full connection layer. The fully connected layer realizes further combination and extraction of features by connecting all input nodes to all output nodes. In the full connection layer, the ReLU activation function is adopted to carry out nonlinear transformation on the input time sequence feature fusion vector. The ReLU function is a common activation function, can effectively solve the problem of gradient disappearance, and improves the training speed and performance of the model. Through the processing of the full connection layer and the ReLU function, a plurality of target monitoring metrics are calculated, including output power, efficiency, voltage and current stability. The input feature encoding vectors are input into a plurality of Bi-LSTM networks, respectively. In each Bi-LSTM network, the forward LSTM layer performs a feature analysis from the initial point to the end point of the time series, and the reverse LSTM layer performs a feature analysis from the end point to the initial point of the time series. The bidirectional processing method can capture all time sequence information of the input data, so that the comprehensiveness and accuracy of feature extraction are improved. And obtaining a time sequence feature fusion vector through Bi-LSTM network processing. The timing feature fusion vector is input into the fully connected layer. In the fully connected layer, nonlinear transformation is performed on the input timing feature fusion vector through a ReLU activation function. The primary function of the ReLU function is to introduce nonlinear characteristics, so that the model can be better fitted with complex input-output relationships. And obtaining final monitoring index output through the processing of the full connection layer, wherein the monitoring indexes comprise output power, conversion efficiency, voltage and current stability of the light storage inverter under different load conditions. In the process of model construction and training, a supervised learning method is adopted to train and optimize the model. Through a large amount of training data and labels, the model is subjected to iterative training, so that the internal parameters are continuously adjusted, and the accuracy and the stability of prediction are improved. In the training process, the technologies of cross verification, early stop and the like are adopted to prevent the model from being fitted excessively, and the optimal model parameters are selected.

In a specific embodiment, the executing step inputs a plurality of input feature encoding vectors into a plurality of bidirectional long-short-term memory networks respectively, and the process of obtaining the time sequence feature fusion vector of each bidirectional long-term memory network by performing time sequence feature analysis on the input feature encoding vectors through each bidirectional long-term memory network may specifically include the following steps:

(1) Respectively inputting a plurality of input feature coding vectors into a plurality of two-way long-short-term memory networks, wherein the two-way long-term memory networks comprise a forward long-term memory network and a backward long-term memory network, the forward long-term memory network comprises 256 LSTM units which are connected in one way, and the backward long-term memory network comprises 256 LSTM units which are connected in one way;

(2) Respectively carrying out forward time sequence feature analysis on the input feature code vectors through a forward long-short-term memory network to obtain forward time sequence feature vectors;

(3) Respectively carrying out backward time sequence feature analysis on the input feature code vectors through a backward long-short-term memory network to obtain backward time sequence feature vectors;

(4) And respectively carrying out vector splicing on the forward time sequence feature vector and the backward time sequence feature vector to obtain the time sequence feature fusion vector of each bidirectional long-short-term memory network.

Specifically, a plurality of input feature encoding vectors are respectively input into a plurality of two-way long-short-term memory networks. Each Bi-LSTM network comprises a forward long-short-term memory network and a backward long-term memory network, wherein the forward long-term memory network comprises 256 LSTM units which are connected in one way, and the backward long-term memory network comprises 256 LSTM units which are connected in one way. The forward long-short-term memory network performs feature analysis from the initial point to the final point of the time sequence, captures forward time sequence features in the input data, and then the backward long-term memory network performs feature analysis from the final point to the initial point of the time sequence, and captures backward time sequence features in the input data. The forward LSTM network processes the input features of each time point starting from the first time point of the time series and passes the processing result to the next time point. Through the gradual processing and transmission, the forward LSTM network can capture the forward change mode of the input characteristic on the whole time sequence, and finally the forward time sequence characteristic vector is obtained. The backward LSTM network processes the input features of each time point starting from the last time point of the time series and passes the processing result to the previous time point. Through the gradual processing and transmission, the backward LSTM network can capture the backward change mode of the input characteristic on the whole time sequence, and finally the backward time sequence characteristic vector is obtained. Vector splicing is carried out on the two feature vectors, and forward and backward time sequence features are combined together to form a comprehensive time sequence feature fusion vector.

In a specific embodiment, the process of executing step S105 may specifically include the following steps:

(1) Inputting the monitoring index difference data into a preset circuit parameter compensation model, wherein the circuit parameter compensation model comprises: the system comprises an input layer, a plurality of convolution layers, a pooling layer and a parameter optimization layer, wherein the input of each layer of the input layer, the plurality of convolution layers and the pooling layer is from the output of all the previous layers;

(2) Inputting the monitoring index difference data into an input layer for feature extraction, and outputting index difference features;

(3) Inputting the index difference characteristics into a plurality of convolution layers to carry out convolution operation to obtain convolution characteristic vectors;

(4) Inputting the convolution feature vector into a pooling layer for downsampling and feature compression to obtain a pooling feature vector;

(5) And inputting the pooled feature vector into a parameter optimization layer for circuit parameter adjustment and parameter combination to obtain the target output circuit parameters of the three-port test power supply.

Specifically, the monitoring index difference data is input into an input layer of the model for feature extraction, and the original data is converted into feature vectors suitable for further processing. And inputting the index difference feature vectors into a plurality of convolution layers to carry out convolution operation, and extracting the local mode and the spatial relation of the input features. Each convolution layer comprises a plurality of convolution kernels, and the convolution kernels perform convolution operation on input feature vectors in a sliding window mode to extract features of different scales. In each convolution layer, the convolution operation is typically followed by a nonlinear activation, such as a ReLU activation function, to enhance the nonlinear capabilities of the feature representation. And by superposing a plurality of convolution layers, gradually extracting a high-level mode in the input characteristics to obtain a more abstract and comprehensive convolution characteristic vector. The convolved feature vector is input to the pooling layer for downsampling and feature compression. The role of the pooling layer is to reduce the dimensionality of the feature vectors by a downsampling operation while preserving the most important feature information. Common pooling operations include maximum pooling and average pooling. By pooling operation, the complexity of the feature vector can be reduced, the calculated amount is reduced, and the calculation efficiency and stability of the model are improved. The pooled feature vectors are input to a parameter optimization layer for circuit parameter adjustment and parameter combination. And generating circuit parameters capable of optimizing the output performance of the three-port test power supply according to the input pooling feature vector. The parameter optimization layer generally adopts a fully connected layer structure, and further combination and extraction of the characteristics are realized by connecting all input nodes to all output nodes. In the parameter optimization layer, an appropriate optimization algorithm, such as particle swarm optimization or genetic algorithm, is adopted to perform optimization processing on the input feature vector, so as to generate a final circuit parameter adjustment scheme.

In a specific embodiment, the step of performing the step of inputting the pooled feature vector to the parameter optimization layer to perform circuit parameter adjustment and parameter combination, and the process of obtaining the target output circuit parameter of the three-port test power supply may specifically include the following steps:

(1) Defining a fitness function of a particle swarm optimization algorithm in a parameter optimization layer;

(2) According to the pooling feature vector, carrying out parameter range prediction on the output voltage, the frequency and the waveform to obtain a parameter range set;

(3) Initializing output circuit parameters according to a parameter range set by a particle swarm optimization algorithm to generate a corresponding particle swarm, wherein the particle swarm comprises a plurality of first output circuit parameters;

(4) Calculating the fitness value of each first output circuit parameter through a fitness function respectively, and dividing the particle swarm of the first output circuit parameters according to the fitness value to obtain a plurality of sub-particle swarms;

(5) And generating a plurality of corresponding second output circuit parameters according to the plurality of sub-particle groups, and carrying out optimization solution on the plurality of second output circuit parameters to obtain target output circuit parameters of the three-port test power supply.

Specifically, an fitness function of a particle swarm optimization algorithm in the parameter optimization layer is defined. The fitness function is typically defined according to an optimization objective, for example in the optimization of a three-port test power supply, the fitness function may reflect the stability and accuracy of the output voltage, frequency and waveform. Assuming that the fitness function is defined as the mean square error between the target output parameter and the expected output parameter, it calculates the difference between the output circuit parameter for each particle and the actual monitoring index and minimizes the difference. And carrying out parameter range prediction on the output voltage, the frequency and the waveform according to the pooled feature vector to obtain a parameter range set. By analyzing the pooled feature vectors, a reasonable range of each circuit parameter is preliminarily determined. for example, assuming that the pooled feature vector contains output data of the light storage inverter under different operating conditions, a range set of each circuit parameter, such as a voltage between 200V and 240V, a frequency between 49Hz and 51Hz, etc., is derived by statistical analysis and machine learning model prediction. And initializing output circuit parameters according to the parameter range set by a particle swarm optimization algorithm to generate a corresponding particle swarm. The population of particles is made up of a plurality of particles, each particle representing one possible circuit configuration. In the particle swarm optimization algorithm, the particles approach the optimal solution step by moving in the search space. During the initialization process, each particle randomly generates its initial position and velocity within the parameter range set. For example, for voltage parameters, the particles may randomly generate different initial values of 200V, 210V, 220V, etc., and for frequency parameters, the particles may generate initial values of 49.5Hz, 50Hz, 50.5Hz, etc. By the guidance of the parameter range set, particle populations covering different combinations of circuit parameters are generated. And respectively calculating the fitness value of each first output circuit parameter through a fitness function. The fitness value reflects the performance advantage of each circuit parameter configuration. Assuming that the circuit parameter of a certain particle is configured to be 220V of output voltage and 50Hz of frequency, calculating the fitness value of the configuration through a fitness function, and the result possibly shows that the configuration is stable in output in practical application and has higher fitness value; While the other particle is configured to output 210V at a frequency of 49.5Hz, the calculation result may indicate that the configuration is unstable in output in practical application, and the fitness value is low. And carrying out particle swarm division on the plurality of first output circuit parameters according to the fitness value to obtain a plurality of sub-particle swarms. The purpose of the particle swarm division is to group the particles with better performance through the comparison of fitness values, so that the resources are concentrated for optimization in the subsequent optimization process. For example, particles with higher fitness values form one sub-particle group, and particles with lower fitness values form another sub-particle group, so that the calculation resources can be more effectively utilized, and the optimization efficiency can be improved. and generating a plurality of corresponding second output circuit parameters according to the plurality of sub-particle groups, and carrying out optimization solution on the parameters. In each iteration process, the particles adjust positions according to own experience and experience of neighbor particles, so that output circuit parameter configuration is continuously improved. By simulating the flight and search behavior of particles in the search space, the globally optimal solution is gradually approached. For example, in one iteration, a certain particle adjusts the voltage parameter from 220V to 225V and the frequency parameter from 50Hz to 50.2Hz according to the historic optimal positions of the particle and the neighbor particle, the fitness value is found to be improved through fitness function evaluation, and the globally optimal output circuit parameter combination is finally found through multiple iterations and adjustment. Finally, the target output circuit parameters of the three-port test power supply are obtained through the process. The target output circuit parameters are obtained by global searching and optimizing through a particle swarm optimization algorithm on the basis of comprehensively considering a plurality of performance indexes such as output voltage, frequency, waveform and the like. For example, over a number of iterations and optimizations, the final target output circuit parameter may be output voltage 230V, frequency 50Hz, waveform distortion is minimal, and such a combination of parameters can ensure stability and reliability of the light storage inverter under different operating conditions.

The method for testing the light storage inverter in the embodiment of the present application is described above, and the system for testing the light storage inverter in the embodiment of the present application is described below, referring to fig. 2, where an embodiment of the system for testing the light storage inverter in the embodiment of the present application includes:

A setting module 201, configured to set a test parameter set of the light storage inverter, where the test parameter set includes an output voltage, an output frequency, and an output power;

The transmission module 202 is configured to transmit the test parameter set to a preset three-port test power supply, and control a plurality of output circuits in the three-port test power supply to perform test parameter response, so as to generate initial output circuit parameters of the three-port test power supply;

The acquisition module 203 is configured to perform operation test on the optical storage inverter, acquire output performance and functional data of the optical storage inverter, and perform monitoring index calculation on the output performance and the functional data to obtain a plurality of target monitoring indexes;

a calculating module 204, configured to obtain a plurality of standard monitoring indexes of the optical storage inverter, and calculate monitoring index difference data between the plurality of standard monitoring indexes and the plurality of target monitoring indexes;

And the adjusting module 205 is configured to adjust circuit parameters of the output circuit of the three-port test power supply according to the monitored index difference data, so as to obtain target output circuit parameters of the three-port test power supply.

Through the cooperative cooperation of the components, (1) test cost and time are saved, and the requirement on an independent power supply is reduced through integrating a plurality of output circuits; (2) The accuracy and the repeatability of the test are improved, and the test error is avoided through unified control and parameter transmission; (3) The flexibility and the adaptability of the test are improved, and the test requirements of different light storage inverters can be met by adjusting the working state and the parameters of the output circuit; (4) The mode that the input ends of the three output circuits share the direct current bus reduces the load degree of the power supply topology and increases the energy conversion efficiency.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random acceS memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. The test method of the light storage inverter is characterized by comprising the following steps of:

Transmitting the test parameter set to a preset three-port test power supply, and controlling a plurality of output circuits in the three-port test power supply to respond to the test parameters to generate initial output circuit parameters of the three-port test power supply; the method specifically comprises the following steps: transmitting the test parameter set to a preset three-port test power supply, wherein the three-port test power supply comprises: an input circuit and three output circuits, each output circuit can independently provide power for the optical storage inverter, the input circuit is an ACDC module for providing dc bus voltage and energy compensation or feedback for each output circuit module, the three output circuits include: the first output circuit is a PV analog module, and the PV analog module is a unidirectional DCDC converter; the second output circuit is a battery simulation module, the battery simulation module is a bidirectional DCDC converter, the third output circuit is a power grid simulation and nonlinear load simulation module, and the power grid simulation and nonlinear load simulation module is a bidirectional DCAC converter; controlling a plurality of output circuits in the three-port test power supply to respond to test parameters, and generating initial output circuit parameters of the three-port test power supply, wherein the initial output circuit parameters comprise: outputting voltage, frequency and waveform;

2. The method of claim 1, wherein the performing the operation test on the optical storage inverter and collecting output performance and function data of the optical storage inverter, and performing monitoring index calculation on the output performance and function data to obtain a plurality of target monitoring indexes, comprises:

Performing operation test on the light storage inverter and collecting output performance and function data of the light storage inverter;

performing time sequence alignment and curve fitting on the output performance and the functional data to obtain a plurality of target curves;

respectively carrying out mean value and standard deviation calculation on the plurality of target curves to obtain the mean value and standard deviation of each target curve;

Extracting features of the plurality of target curves according to the mean value and the standard deviation to obtain an initial feature value set of each target curve;

performing main component dimension reduction on the initial characteristic value set of each target curve to obtain a target characteristic value set of each target curve;

And respectively inputting the target characteristic value set of each target curve into a preset index analysis model to perform monitoring index calculation, so as to obtain a plurality of target monitoring indexes.

3. The method for testing an inverter of claim 2, wherein the step of inputting the target feature value set of each target curve into a preset index analysis model to perform the calculation of the monitoring index to obtain a plurality of target monitoring indexes includes:

The index analysis model includes: the embedded layer and the two-way long-short-period memory networks are connected with a full-connection layer;

in the embedded layer, respectively carrying out feature normalization and vector code conversion on a target feature value set of each target curve to generate a plurality of input feature code vectors;

Respectively inputting the plurality of input feature coding vectors into the plurality of two-way long-short-term memory networks, and performing time sequence feature analysis on the input feature coding vectors through each two-way long-term memory network to obtain time sequence feature fusion vectors of each two-way long-term memory network;

And inputting the time sequence feature fusion vector into the full-connection layer, and performing monitoring index calculation through a ReLU function in the full-connection layer to obtain a plurality of target monitoring indexes, wherein the plurality of target monitoring indexes comprise output power, efficiency, voltage and current stability.

4. The method for testing an optical storage inverter according to claim 3, wherein the inputting the plurality of input feature code vectors into the plurality of bidirectional long-short-term memory networks, respectively, performing a time sequence feature analysis on the input feature code vectors through each bidirectional long-short-term memory network to obtain a time sequence feature fusion vector of each bidirectional long-term memory network, comprises:

Respectively inputting the plurality of input feature encoding vectors into the plurality of two-way long-short-term memory networks, wherein the two-way long-short-term memory networks comprise a forward long-short-term memory network and a backward long-short-term memory network, the forward long-short-term memory network comprises 256 LSTM units which are connected in one way, and the backward long-short-term memory network comprises 256 LSTM units which are connected in one way;

respectively carrying out forward time sequence feature analysis on the input feature code vectors through the forward long-short-term memory network to obtain forward time sequence feature vectors;

Respectively carrying out backward time sequence feature analysis on the input feature code vectors through the backward long-short-term memory network to obtain backward time sequence feature vectors;

and respectively carrying out vector splicing on the forward time sequence feature vector and the backward time sequence feature vector to obtain a time sequence feature fusion vector of each bidirectional long-short-term memory network.

5. The method of claim 1, wherein the performing circuit parameter adjustment on the output circuit of the three-port test power supply according to the monitor index difference data to obtain the target output circuit parameter of the three-port test power supply comprises:

Inputting the monitoring index difference data into a preset circuit parameter compensation model, wherein the circuit parameter compensation model comprises the following components: the system comprises an input layer, a plurality of convolution layers, a pooling layer and a parameter optimization layer, wherein the input of each layer of the input layer, the plurality of convolution layers and the pooling layer is from the output of all the previous layers;

Inputting the monitoring index difference data into the input layer for feature extraction, and outputting index difference features;

inputting the index difference characteristics into the plurality of convolution layers to carry out convolution operation to obtain convolution characteristic vectors;

inputting the convolution feature vector into the pooling layer for downsampling and feature compression to obtain a pooling feature vector;

and inputting the pooling feature vector into the parameter optimization layer for circuit parameter adjustment and parameter combination to obtain the target output circuit parameter of the three-port test power supply.

6. The method of claim 5, wherein inputting the pooled feature vector to the parameter optimization layer for circuit parameter adjustment and parameter combination to obtain the target output circuit parameters of the three-port test power supply comprises:

defining an adaptability function of a particle swarm optimization algorithm in the parameter optimization layer;

According to the pooling feature vector, carrying out parameter range prediction on the output voltage, the frequency and the waveform to obtain a parameter range set;

Initializing output circuit parameters according to the parameter range set by the particle swarm optimization algorithm to generate a corresponding particle swarm, wherein the particle swarm comprises a plurality of first output circuit parameters;

Calculating the fitness value of each first output circuit parameter through the fitness function respectively, and dividing the particle swarm of the plurality of first output circuit parameters according to the fitness value to obtain a plurality of sub-particle swarms;

Generating a plurality of corresponding second output circuit parameters according to the plurality of sub-particle groups, and carrying out optimization solution on the plurality of second output circuit parameters to obtain target output circuit parameters of the three-port test power supply.

7. A test system for an optical storage inverter, for performing the test method of an optical storage inverter as claimed in any one of claims 1-6, the test system comprising: