CN120704217A

CN120704217A - Remote intelligent control system for aquaculture pond water quality based on edge computing

Info

Publication number: CN120704217A
Application number: CN202510926283.0A
Authority: CN
Inventors: 龚国围; 李跃; 李军利; 谢梓杰; 吴思彤; 方炫为
Original assignee: Zhuhai City Polytechnic
Current assignee: Zhuhai City Polytechnic
Priority date: 2025-07-07
Filing date: 2025-07-07
Publication date: 2025-09-26

Abstract

The invention belongs to the technical field of intelligent regulation and control, discloses an intelligent remote regulation and control system for water quality of an aquaculture pond based on edge calculation, and aims to solve the problems of delay of data transmission, neglecting coupling effect due to model simplification, lack of cooperativity in equipment regulation and control and the like in traditional water quality regulation and control. The method comprises the steps of acquiring water quality parameter data in real time through a parameter acquisition module, constructing a parameter micro-coupling matrix at an edge node by a micro-coupling analysis module, analyzing a short-term fluctuation relationship and generating dynamic regulation and control weights, and further forming a local regulation and control instruction. And uploading the data to the cloud by the uploading coupling prediction module, constructing a long-term coupling prediction model, and generating a global regulation strategy. Finally, the edge node combines the local instruction and the global strategy to generate a final regulation instruction, and the working states of the oxygenation, bait casting and water circulation equipment are optimized. According to the method, real-time accurate regulation and control are realized through edge calculation, the risk of abnormal water quality is reduced, the efficiency is improved through equipment collaborative optimization, and the resource waste is reduced.

Description

Remote intelligent regulation and control system for water quality of aquaculture pond based on edge calculation

Technical Field

The invention relates to the technical field of intelligent regulation and control, in particular to an intelligent remote regulation and control system for water quality of an aquaculture pond based on edge calculation.

Background

With the rapid development of the aquaculture industry, water quality management of an aquaculture pond becomes a key factor affecting the yield, quality and aquaculture benefit of the aquatic products. There is a complex coupling relationship between water quality parameters such as dissolved oxygen, pH value, ammonia nitrogen, nitrite and the like, and short-term fluctuation and long-term trend of the parameters directly influence the growth health of aquatic organisms and the stability of a culture ecological system. For example, in the case of high temperature weather or excessive feeding, a rapid rise in ammonia nitrogen concentration may lead to a sharp drop in pH, which in turn may trigger a fish stress reaction or even death. Therefore, the real-time monitoring and accurate regulation of water quality parameters become an important requirement of modern aquaculture.

However, the prior art has a particularly disadvantage in terms of complex coupling relation between the treated water quality parameters and real-time performance of remote control. Taking the coupling relation between pH and ammonia nitrogen as an example, in an actual cultivation scene, the rapid rise of ammonia nitrogen concentration often generates a remarkable acid-base disturbance effect on pH, and especially under the condition that water eutrophication is caused by high-temperature weather or excessive feeding, the short-term fluctuation can cause water quality deterioration in a few minutes. However, in the conventional remote system, since the data transmission is limited by the network condition, the ammonia nitrogen data may be delayed for 10 seconds or more to reach the cloud, which results in that the model cannot capture the instant influence of the ammonia nitrogen on the pH in time, and further underestimating the severity of the coupling effect. For example, in a farm where ammonia nitrogen concentration increases rapidly due to excessive feeding, pH may drop to a range where fish is intolerant in a short time, but due to the delay of the data, the system fails to start the oxygenation device or the water circulation device in time, eventually leading to the occurrence of stress reaction and even death of the fish school. In addition, when the traditional system processes data in the cloud, in order to reduce the calculation complexity, a tiny but key coupling effect, such as weak interaction influence of pH and ammonia nitrogen in a specific time period, is often ignored through model simplification, and the simplification is particularly dangerous when the water quality parameter is in a critical state, so that the system can misjudge the water quality state and generate an error regulation instruction. More seriously, the traditional system lacks consideration on equipment cooperativity, for example, when the ammonia nitrogen concentration is increased, only the oxygenation equipment is started, but the water circulation equipment is not regulated at the same time to dilute the pollutant concentration, so that the regulation and control efficiency is reduced, and energy waste and equipment loss can be caused by non-optimal states of equipment operation.

In view of the above, the invention provides an intelligent remote control system for the water quality of an aquaculture pond based on edge calculation to solve the problems.

Disclosure of Invention

In order to overcome the defects in the prior art, the invention provides the technical scheme that the remote intelligent control system for the water quality of the culture pond based on edge calculation comprises the following components:

the parameter acquisition module is used for acquiring pond water quality parameter data of the culture pond in each regulation and control period;

The micro-coupling analysis module is used for constructing a parameter micro-coupling matrix in each regulation period according to the short-term fluctuation relation between any two water quality parameters in the pond water quality parameter data in the edge computing nodes;

The weight determining module is used for determining the dynamic regulation weight of each water quality parameter according to the short-term coupling influence factor and the real-time change trend of the pond water quality parameter data in each regulation period in an edge computing node;

The uploading coupling prediction module is used for uploading the pond water quality parameter data and the parameter micro-coupling matrix to a cloud regulation server, and constructing a long-term coupling prediction model in the cloud regulation server based on the pond water quality parameter data and the parameter micro-coupling matrix of a historical regulation period;

The overall regulation and control module is used for combining the local regulation and control instruction and the overall regulation and control strategy in the edge calculation node to generate a final regulation and control instruction, and regulating and controlling the working states of the oxygenation equipment, the bait casting equipment and the water circulation equipment of the aquaculture pond based on the final regulation and control instruction.

Preferably, the constructing the parameter microcoupling matrix in each regulation period includes:

in each regulation period, carrying out time sequence segmentation on the pond water quality parameter data to obtain a fluctuation difference sequence of any two water quality parameters in each time period;

calculating short-term coupling strength between two corresponding water quality parameters according to the fluctuation amplitude and the fluctuation frequency of each water quality parameter in the fluctuation difference sequence;

acquiring a coupling directivity factor corresponding to the two water quality parameters according to the short-term coupling strength and the value range distribution difference of the two water quality parameters in the time period;

and constructing a parameter micro-coupling matrix in a corresponding regulation and control period by taking the short-term coupling strength as a matrix element value and taking the coupling directivity factor as a matrix element symbol.

Preferably, the determining the dynamic regulation weight of each water quality parameter includes:

calculating short-term coupling influence factors corresponding to the water quality parameters according to matrix row vectors corresponding to each water quality parameter in the parameter micro-coupling matrix;

In each regulation period, carrying out real-time trend analysis on the pond water quality parameter data to obtain the trend change rate of each water quality parameter;

Acquiring the regulation priority of the corresponding water quality parameters according to the product of the short-term coupling influence factor and the trend change rate;

And normalizing the regulation and control priorities of all the water quality parameters to obtain the dynamic regulation and control weights of the corresponding water quality parameters.

Preferably, the generating the local regulation command in the corresponding regulation period includes:

The method comprises the steps of dynamically regulating and controlling the water quality parameters, determining a regulating and controlling target value of each water quality parameter according to the dynamic regulating and controlling weight, calculating the regulating and controlling amplitude of the corresponding water quality parameter based on the difference value between the current value of each water quality parameter in pond water quality parameter data and the regulating and controlling target value, and generating a local regulating and controlling instruction in a corresponding regulating and controlling period according to the regulating and controlling amplitude and a preset equipment regulating and controlling mapping table.

Preferably, the constructing a long-term coupling prediction model includes:

Acquiring pond water quality parameter data and the parameter micro-coupling matrix of a history regulation period in the cloud regulation server, and constructing a history coupling data set;

performing time sequence clustering on the historical coupling data set to obtain long-term coupling modes among water quality parameters under different time scales;

based on the long-term coupling mode, a long-term coupling prediction model based on deep learning is constructed, the long-term coupling prediction model is input into the pond water quality parameter data and the parameter micro-coupling matrix of the current regulation and control period, and the predicted water quality parameter data of the future regulation and control period is output.

Preferably, the generating a global regulation strategy corresponding to a future regulation period includes:

Based on the long-term coupling prediction model, predicted water quality parameter data of a future regulation and control period are obtained;

calculating the regulation demand degree of the corresponding water quality parameters according to the deviation between the predicted value of each water quality parameter in the predicted water quality parameter data and a preset safety threshold value;

and generating a global regulation strategy corresponding to a future regulation period based on the regulation demand degree.

Preferably, the generating the final regulation command includes:

acquiring a regulation target value difference between the local regulation instruction and the global regulation strategy in an edge calculation node;

Judging whether a regulation conflict exists or not according to the regulation target value difference and a preset regulation conflict threshold value;

If the control conflict exists, carrying out weighted fusion on the local control instruction and the global control strategy according to the dynamic control weight to generate a final control instruction;

and if no regulation conflict exists, the local regulation instruction is directly used as a final regulation instruction.

Preferably, said calculating corresponds to a short-term coupling strength between two water quality parameters, comprising:

performing Fourier transformation on the fluctuation difference sequence to obtain fluctuation frequency spectrums corresponding to two water quality parameters;

according to the frequency spectrum overlapping degree of the two water quality parameters in the fluctuation frequency spectrum, calculating frequency coupling factors corresponding to the two water quality parameters;

Calculating amplitude coupling factors corresponding to the two water quality parameters according to the ratio of the fluctuation amplitudes of the two water quality parameters in the fluctuation difference sequence;

And taking the product of the frequency coupling factor and the amplitude coupling factor as the short-term coupling strength between the two corresponding water quality parameters.

Preferably, the performing time series clustering on the historical coupling data set includes:

Carrying out standardized treatment on the pond water quality parameter data in the historical coupling data set to obtain a standardized water quality parameter sequence;

And clustering the standardized water quality parameter sequences by adopting a K-means clustering algorithm according to the similarity to obtain long-term coupling modes under different time scales.

Preferably, the adjusting and controlling the working states of the oxygenation equipment, the bait casting equipment and the water circulation equipment of the aquaculture pond comprises the following steps:

The control signals are sent to the oxygenation equipment, the bait casting equipment and the water circulation equipment through a local controller in an edge computing node;

And monitoring the working states of the oxygenation equipment, the bait casting equipment and the water circulation equipment in real time, and if the working states are abnormal, generating an abnormal alarm signal and uploading the abnormal alarm signal to the cloud regulation server.

The intelligent remote control system for the water quality of the aquaculture pond based on edge calculation has the technical effects and advantages that:

According to the invention, real-time data processing is realized at the edge computing node, and the system can timely capture the tiny coupling effect between parameters such as pH, ammonia nitrogen and the like, so that information loss and regulation hysteresis caused by data transmission delay are avoided. For example, even under the condition of unstable network, the edge node can rapidly analyze the short-term disturbance influence of ammonia nitrogen concentration rise on pH locally and timely generate a regulation command to adjust the equipment state, so that the water quality deterioration seedling head is effectively restrained. And secondly, the coupling relation among the parameters is comprehensively described through a scientific analysis method, so that the problem of neglecting a tiny coupling effect caused by the simplification of a traditional model is avoided, the regulation and control instruction can more accurately reflect the real interaction state among the water quality parameters, for example, the stability measure of timely regulating the pH value when the ammonia nitrogen concentration is rapidly changed, and the ecological risk caused by regulation and control errors is remarkably reduced. In addition, the accurate regulation and control mode not only improves the real-time performance and reliability of water quality management, but also reduces energy waste and operation cost by reducing invalid or wrong equipment operation, enhances the stability of an ecological system of the culture pond, and finally ensures the yield and quality of aquatic products.

Drawings

FIG. 1 is a schematic diagram of an intelligent remote control system for water quality of an aquaculture pond based on edge calculation;

FIG. 2 is a graph showing the comparison of the control effect of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The embodiment of the application provides an intelligent remote control system for the water quality of an aquaculture pond based on edge calculation. The execution main body of the remote intelligent regulation and control system for the water quality of the aquaculture pond based on edge calculation comprises a cloud regulation and control server, a water quality monitoring platform, intelligent oxygenation equipment, bait casting equipment, water circulation equipment and the like which are used for carrying the system, wherein the data processing platform comprises at least one of a water quality parameter acquisition system, a micro-coupling analysis system and a regulation and control optimization system.

Referring to fig. 1, the invention provides an intelligent remote control system for water quality of an aquaculture pond based on edge calculation, comprising:

the parameter acquisition module is used for acquiring pond water quality parameter data of the culture pond in each regulation and control period, wherein the pond water quality parameter data comprise pH value, ammonia nitrogen concentration, dissolved oxygen concentration and water temperature;

The micro-coupling analysis module is used for constructing a parameter micro-coupling matrix in each regulation and control period according to the short-term fluctuation relation between any two water quality parameters in pond water quality parameter data in the edge computing nodes;

The weight determining module is used for determining the dynamic regulation weight of each water quality parameter according to the short-term coupling influence factor and the real-time change trend of the pond water quality parameter data in each regulation period in the edge computing node;

The uploading coupling prediction module is used for uploading the pond water quality parameter data and the parameter micro-coupling matrix to the cloud regulation server, and constructing a long-term coupling prediction model in the cloud regulation server based on the pond water quality parameter data and the parameter micro-coupling matrix of the history regulation period;

the overall regulation and control module is used for generating a final regulation and control instruction by combining a local regulation and control instruction and an overall regulation and control strategy in the edge calculation node, and regulating and controlling the working states of the oxygenation equipment, the bait casting equipment and the water circulation equipment of the aquaculture pond based on the final regulation and control instruction.

According to the method, accurate regulation and control of the water quality of the culture pond are achieved through a layered architecture of edge calculation and cloud cooperation, a parameter micro-coupling matrix is built at the edge end to capture short-term mutual influence relation among water quality parameters, dynamic regulation and control weights are determined based on short-term coupling influence factors and real-time change trends to ensure real-time responsiveness of regulation and control, prospective guidance is provided for cloud analysis historical data construction, and finally an optimal regulation and control instruction is generated through fusion of local regulation and control instructions and global regulation and control strategies by edge calculation nodes, so that accurate maintenance of the water quality parameters and stable optimization of a culture environment are achieved.

In the embodiment of the invention, the parameter acquisition module is used for acquiring pond water quality parameter data of the culture pond in each regulation and control period, and is specifically used for:

The multi-point water quality sensor network is deployed in the culture pond and comprises a pH value sensor, an ammonia nitrogen concentration sensor, a dissolved oxygen sensor and a water temperature sensor to form a multi-dimensional water quality monitoring array, wherein the sensor network adopts a grid layout to ensure the whole coverage of the whole pond water area, and meanwhile, the sensors are vertically distributed in consideration of factors such as water depth, flow direction and the like to capture the parameter changes of different depths of the water body;

Setting data acquisition frequency, and acquiring a complete water quality parameter set usually every 5-15 minutes according to the breed and water quality fluctuation characteristics, wherein the sampling frequency can be dynamically increased to every 1-3 minutes when the water quality is changed drastically (such as after feeding and when weather changes);

the collected original data is transmitted to a data preprocessing unit at the edge of the pond through a low-power consumption wireless sensing network, and the reliability of signal transmission is ensured by adopting an anti-interference technology such as Frequency Hopping Spread Spectrum (FHSS), particularly under a complex cultivation environment;

The data is cleaned and abnormal value detection is carried out in the data preprocessing unit, which comprises the steps of removing data points which obviously exceed a reasonable range, repairing data loss, filtering high-frequency noise and the like, and the data quality is improved by adopting sliding window median filtering and other methods;

Carrying out time window segmentation and organization on the cleaned data according to a regulation and control period (usually 1 hour or less) to form normalized pond water quality parameter data, wherein the normalized pond water quality parameter data comprises four key indexes including complete pH value, ammonia nitrogen concentration, dissolved oxygen concentration and water temperature, and corresponding time stamp and acquisition position information;

And storing the normalized pond water quality parameter data in a local database of the edge computing node, and applying an edge side compression algorithm to reduce data storage and transmission overhead, while preserving the time sequence characteristics of the water quality parameters.

In the embodiment, the comprehensive monitoring of the water quality parameters of the culture pond is realized through the multipoint water quality sensor network, the adaptive sampling frequency is adopted to ensure the data density at the key moment, the data quality is improved and the transmission burden is reduced by the edge side data preprocessing, and finally, the high-quality pond water quality parameter data is formed as the basis for the subsequent analysis and regulation. The process particularly emphasizes the real-time performance and accuracy of data acquisition, and provides reliable data support for subsequent micro-coupling analysis and dynamic regulation.

In the embodiment of the invention, the detailed implementation steps for constructing the parameter micro-coupling matrix in each regulation and control period comprise the following steps:

In each regulation period, carrying out time sequence segmentation on pond water quality parameter data to obtain a fluctuation difference sequence of any two water quality parameters in each time period;

In this embodiment, firstly, the water quality parameter data of each regulation period (usually 1 hour) is segmented according to the time interval or key event (such as feeding, oxygenation starting, etc.), and the typical segmentation is 10-15 minutes/segment, so as to form a continuous time window; in each time period, calculating a fluctuation difference sequence between any two water quality parameters (such as pH value, dissolved oxygen, ammonia nitrogen, water temperature and the like), wherein the method is that after the two parameter sequences are subjected to standardized treatment, the difference change at the same moment is calculated, so that the relative fluctuation relation between the parameters is captured; carrying out fluctuation feature analysis on the obtained fluctuation difference sequence, extracting fluctuation amplitude features (such as standard deviation, peak Gu Chazhi and the like) and fluctuation frequency features (such as fluctuation period, frequency distribution and the like), quantifying short-term coupling strength between any two water quality parameters by applying a correlation analysis method (such as pearson correlation coefficient, mutual information quantity and the like), wherein the larger the coupling strength value is in a [ -1,1] interval, the stronger the coupling is represented by the larger absolute value, analyzing the value domain distribution features of the two water quality parameters including mean value, variance, distribution form and the like, determining the directivity of the coupling by comparing the variation sequence and causality of the two parameters, wherein if the parameter A is changed later, the parameter B is provided with positive coupling direction, the coupling directivity factor is usually represented by +1 (positive coupling) or-1 (negative coupling), and the coupling directivity factor is +1 when the pH value is usually increased due to the increase of dissolved oxygen concentration, taking the calculated coupling strength as the numerical value part of matrix elements, taking the coupling directivity factor as a symbol part of matrix elements, constructing an n multiplied by n parameter microcoupling matrix (n is the quantity of water quality parameters, n=4 in the example), wherein each element M (i, j) in the matrix represents the microcoupling influence intensity and direction of the parameter i on the parameter j.

The complex interaction relation between water quality parameters is quantized through the parameter micro-coupling matrix, the dynamic coupling characteristic between the parameters in a short time (in a regulation period) is particularly captured, and the refined coupling analysis provides a basis for subsequent accurate regulation. The matrix not only reflects the coupling strength, but also contains the coupling direction information, so that the system can understand the causal chain of the water quality parameter change, and the source parameter is processed preferentially during regulation and control.

In the embodiment of the invention, the detailed implementation steps for determining the dynamic regulation weight of each water quality parameter comprise the following steps:

According to the matrix row vector corresponding to each water quality parameter in the parameter micro-coupling matrix, calculating a short-term coupling influence factor corresponding to the water quality parameter;

in each regulation period, carrying out real-time trend analysis on pond water quality parameter data to obtain trend change rate of each water quality parameter;

and carrying out normalization treatment on the regulation and control priorities of all the water quality parameters to obtain the dynamic regulation and control weights of the corresponding water quality parameters.

In the embodiment, firstly, matrix row vectors corresponding to each water quality parameter are extracted from a parameter micro-coupling matrix, and the row vectors represent the influence intensity and direction of the parameter on all other parameters; calculating the norm or weighted sum of each row vector as a short-term coupling influence factor of corresponding water quality parameters, wherein the factor quantifies the influence of the parameters on the whole water quality system in a short period, a large-value method is adopted to determine the influence factor, namely the maximum influence value of the parameters on any other parameters can be used as the influence factor, so that the influence is prevented from being weakened averagely, a sliding window method is adopted to carry out trend analysis on water quality parameter data in the last period (such as the last 30 minutes), a time sequence analysis method such as linear regression, exponential smoothing or ARIMA is adopted to calculate the change slope of each parameter to obtain the trend change rate, the trend change rate not only considers the change direction (ascending or descending), but also considers the change speed (fast or slow), the direction is represented by positive and negative values, the value represents the speed, the maximum influence value of each water quality parameter is multiplied by the trend change rate of the short-term coupling influence factor, the value simultaneously considers the system influence force of the parameters and the current change trend, the high influence force of the parameters are obtained to obtain the regulation priority, the maximum-minimum normalization or the current change trend is considered, the current trend is adopted, the change trend of the water quality parameter is high influence force and the current change parameter is obtained to be high, the priority, the current regulation and the current trend is converted to be the priority is 1, the current and the current dynamic command is converted to be 1, the important current and the current dynamic command is converted to be 1, the current and the current dynamic command is converted to be the important, and the current dynamic command is converted to be 1.

Compared with the prior art, the traditional method may rely on manual intervention or centralized cloud computing, has long response time, particularly under the condition of network delay or manual operation lag, and the traditional fixed weight regulation and control method lacks dynamic regulation capability, which may cause that water quality parameters (such as pH value, ammonia nitrogen concentration, dissolved oxygen concentration and the like) exceed standard, and influence cultivation benefits.

The invention combines the static coupling relation with the dynamic parameter variation trend to generate the dynamic regulation weight reflecting the real-time condition. The method enables the system to dynamically adjust focus on different parameters according to the change characteristics of the current water quality state, ensures that the regulation and control resources are preferentially distributed to the parameters which are the most critical and the most needed to be intervened at present, and improves the accuracy and the efficiency of regulation and control.

In the embodiment of the invention, the detailed implementation steps for generating the local regulation command in the corresponding regulation period comprise:

Determining a regulation target value of each water quality parameter according to the dynamic regulation weight;

calculating the regulation amplitude of the corresponding water quality parameters based on the difference value between the current value and the regulation target value of each water quality parameter in the pond water quality parameter data;

according to the regulation amplitude and a preset equipment regulation mapping table, generating a local regulation instruction in a corresponding regulation period, wherein the local regulation instruction comprises the working power of the oxygenation equipment, the feeding amount of the feeding equipment and the circulation frequency of the water circulation equipment.

In the embodiment, firstly, according to the physiological requirements and growth stage characteristics of cultured varieties (such as grass carp, shrimp and the like), the ideal range of each water quality parameter is preset, such as the dissolved oxygen value of 6.0-8.0mg/L, pH value of 7.2-8.5, the ammonia nitrogen of <0.5mg/L and the water temperature of 25-30 ℃; combining dynamic regulation weights, determining a specific regulation target value of each water quality parameter in a current regulation period, wherein a parameter target value with high weight is closer to a central value of an ideal range, parameters with low weight can be allowed to fluctuate in a wider range, calculating a difference value between the current measured value of each water quality parameter and the regulation target value, wherein the difference value considers both directions (positive and negative) and sizes (absolute values) to form an amplitude to be regulated, multiplying the regulation amplitude by the dynamic regulation weights to obtain a weighted regulation amplitude, reflecting the urgency degree of regulating the parameter under the current condition, inquiring a preset equipment regulation mapping table, wherein the table defines the corresponding relation between the deviation of the water quality parameter and the equipment control parameter, such as that the dissolved oxygen is reduced by 1mg/L, for example, the power of the oxygenation equipment is increased by 20%, calculating specific control parameters of each equipment, such as working power (0-100%), bait casting quantity (kg/time) of the bait casting equipment and circulation frequency (time/hour) of the water circulation equipment, considering a synergistic effect between the regulation of multiple parameters, such as oxygenation and pH value, simultaneously influencing the oxygenation, generating a final balance instruction by optimizing a linear regulation and control instruction of the equipment, such as a linear regulation and control algorithm, including accurate equipment control parameters and execution time schedule, such as "the oxygenation equipment operating power is set to 75% for 30 minutes, the bait casting equipment reduces the bait casting amount by 20%, and the water circulation equipment increases the circulation frequency to 2 times per hour".

The method realizes the fine customization of the regulation and control targets based on the dynamic regulation and control weights, converts abstract water quality parameters into specific equipment control instructions, and establishes a closed-loop feedback mechanism between the water quality state and the regulation and control behaviors. Through the equipment regulation mapping table, the system can convert the water quality requirement into an accurate equipment operation instruction, realize 'regulation on demand', and avoid resource waste and excessive intervention.

In the embodiment of the invention, the detailed implementation steps for constructing the long-term coupling prediction model comprise the following steps:

In a cloud regulation server, acquiring pond water quality parameter data and parameter micro-coupling matrixes of a history regulation period, and constructing a history coupling data set;

based on the long-term coupling mode, a long-term coupling prediction model based on deep learning is constructed, the long-term coupling prediction model is input into a pond water quality parameter data and parameter micro-coupling matrix of the current regulation and control period, and the long-term coupling prediction model is output into predicted water quality parameter data of the future regulation and control period.

In the embodiment, the cloud regulation and control server periodically (e.g. daily) collects pond water quality parameter data and parameter micro-coupling matrixes from a plurality of edge computing nodes, and bandwidth occupation is reduced by adopting an incremental synchronization mode for data transmission; the method comprises the steps of structuring collected data to form a historical coupling data set, wherein the data set comprises original water quality parameter values and a coupling relation matrix among parameters to form a high-dimensional space-time database, applying a time sequence clustering algorithm such as Dynamic Time Warping (DTW) combined hierarchical clustering or density-based clustering (DBSCAN) to the historical coupling data set to identify similar time modes, conducting clustering analysis on multiple time scales, including a intra-day mode (such as early and late rules), a intra-week mode (workday/weekend difference), a month mode (month/month end) and a seasonal mode (spring, summer, autumn and winter change), extracting long-term coupling modes among water quality parameters from clustering results, wherein the modes reflect stable association relations among parameters under different conditions (such as seasons, weather and cultivation stages), constructing a prediction model framework based on deep learning, adopting a neural network structure suitable for time sequence data such as long-short-term memory network (LSTM), gate-controlled cyclic unit (GRU) or trans-former, designing a multi-input multi-output structure, simultaneously considering water quality parameter data and parameter values, capturing a common time sequence of the parameters such as MSE by using a training sequence of the time-domain model, the model parameters are optimized by using evaluation indexes such as average absolute percentage error (MAPE), the trained model can predict the water quality parameter change trend of a plurality of future regulation and control periods (such as 24 hours in the future) based on water quality parameter data and parameter micro-coupling matrixes of the current regulation and control period, the predicted value of each parameter and a confidence interval of each parameter are included, and the model is retrained and updated by using the latest data at a cloud end periodically, so that the prediction accuracy is ensured.

The short-term parameter micro-coupling matrix is combined with long-term historical data to construct a prediction model capable of capturing multi-time scale water quality dynamics. The method not only considers the current water quality state, but also considers the dynamic coupling relation among parameters, so that the prediction result is more accurate and reliable. Through the powerful computing resources of the cloud, the system can analyze a large amount of historical data, identify hidden long-term modes and rules, and provide scientific predictions for future water quality changes, so that prospective regulation and control decisions are supported.

In the embodiment of the invention, the detailed implementation steps of generating the global regulation strategy corresponding to the future regulation period comprise the following steps:

based on a long-term coupling prediction model, acquiring predicted water quality parameter data of a future regulation and control period;

Based on the regulation demand level, a global regulation strategy corresponding to a future regulation period is generated, wherein the global regulation strategy comprises a pre-regulation plan of the oxygenation equipment, a pre-feeding plan of the feeding equipment and a pre-circulation plan of the water circulation equipment.

In the embodiment, the cloud regulation server uses a long-term coupling prediction model, and generates water quality parameter prediction data of a plurality of future regulation periods (such as 6-24 hours in the future) based on the current water quality state and a parameter micro-coupling matrix, wherein the water quality parameter prediction data comprises a time sequence prediction value of pH value, ammonia nitrogen concentration, dissolved oxygen concentration and water temperature; calculating a confidence interval of each predicted value, reflecting the predicted uncertainty degree, providing a basis for risk assessment, comparing predicted water quality parameter data with a safety threshold of a culture variety, wherein the safety threshold generally comprises three layers of a lower warning limit, an optimal range and an upper warning limit, calculating the probability and the severity that each parameter may exceed the safety threshold at a future time point, comprehensively forming a regulation demand degree, wherein the regulation demand degree is a time function reflecting the demand strength of regulation at different future time points, designing an optimal control sequence for a future regulation period by using a Model Predictive Control (MPC) or a rolling time domain optimization method based on the regulation demand degree, designing a pre-scheduling plan of an oxygenation device, comprising the starting time, the running power and the duration of each time point, for example, the 10:00-11:30 am at 50% power, setting a pre-feeding plan of the oxygenation device according to the predicted water quality change and the physiological rhythm of the culture creatures, comprising feeding time points, feeding amount and feeding frequency, such as 4:00 in the water quality cycle time being respectively, 7:00-5:00 in the morning, and water cycle time being respectively, and the predicted water quality cycle time being adjusted according to the predicted water quality cycle time of the predicted time of the water quality of the culture variety, and the water quality of the culture creatures being respectively at the time point of 50%, for example, the water circulation equipment is started at the beginning of the tomorrow at the afternoon of 3:00-5:00, water mixing is carried out at medium intensity, the equipment control plans are integrated into a structured global regulation strategy, execution conditions and emergency plans are added, and the robustness and the adaptability of the strategy are improved.

Based on the prediction data, prospective management of the culture environment is realized, and the passive response is changed into active prevention. The global regulation strategy not only considers the current demand, but also considers the water quality change possibly occurring in the future, so that the regulation is more coherent and systematic. By pre-arranging the equipment operation plan, the system can take preventive measures before the problem occurs, effectively avoid water quality mutation and extreme conditions, and optimize energy use and operation cost.

In the embodiment of the invention, the detailed implementation steps for generating the final regulation instruction comprise:

Acquiring a regulation target value difference between a local regulation instruction and a global regulation strategy in an edge calculation node;

If no control conflict exists, the local control instruction is directly used as a final control instruction.

In the embodiment, an edge computing node receives a global regulation strategy from a cloud and compares the global regulation strategy with a locally generated regulation instruction, calculates differences of equipment control parameters such as oxygenation equipment power difference, bait casting quantity difference and water circulation frequency difference, defines a regulation conflict threshold value such as 'oxygenation equipment power difference > 20%' or 'bait casting quantity difference > 15%' and the like, is used as a standard for judging whether significant conflict exists, judges that regulation conflict exists and needs to be coordinated if the control parameter difference of any equipment exceeds the corresponding conflict threshold value, can directly adopt the local regulation instruction if the control parameter difference of all equipment does not exceed the conflict threshold value, simplifies a processing flow and reduces delay, uses dynamic regulation weights as a weight basis for fusing the local instruction and the global strategy when the regulation conflict exists, is more prone to adopting the local regulation instruction to ensure real-time response, is more prone to adopting the global regulation strategy to keep long-term stability for water quality parameters with low weight, is designed to be a weighted fusion algorithm, is preferably arranged to be higher than the weighting, priority or priority is higher than the local regulation instruction, and is not suitable for generating a stable environment-friendly and stable control instruction, and is not suitable for the final control instruction.

By judging the conflict degree and using dynamic weights to carry out intelligent fusion, the system can maintain the rapid response to the current water quality fluctuation and simultaneously give consideration to the strategic goal of long-term water quality stability. The edge-cloud cooperative regulation and control mode not only plays the low-delay advantage of edge calculation, but also utilizes the strong analysis capability of cloud calculation to form complementary and enhanced effects.

In the embodiment of the invention, the detailed implementation steps for calculating the short-term coupling strength between two corresponding water quality parameters comprise:

According to the frequency spectrum overlapping degree of two water quality parameters in the fluctuation frequency spectrum, calculating frequency coupling factors corresponding to the two water quality parameters;

the product of the frequency coupling factor and the amplitude coupling factor is used as the short-term coupling strength between the two corresponding water quality parameters.

In the embodiment, a Fast Fourier Transform (FFT) algorithm is first applied to the fluctuation difference sequence, and a time domain signal is converted into a frequency domain representation, so as to obtain a fluctuation frequency spectrum corresponding to two water quality parameters; the frequency spectrum shows each frequency component and the amplitude thereof contained in the water quality parameter fluctuation, and reflects the periodic characteristics of the parameter variation; the method comprises the steps of comparing and analyzing frequency spectrums of two water quality parameters, calculating the area or energy proportion of a spectrum overlapping area, enabling the spectrum overlapping degree to be higher, enabling the coupling relation to be tighter, enabling the coupling relation to be higher, enabling the coherence coefficients of the two parameters to be calculated on different frequency points to form a frequency correlation curve, enabling the frequency correlation to be integrated into a single frequency coupling factor in a weighted average or main frequency band focusing mode, enabling the factor to be in a 0,1 interval normally, enabling 1 to represent complete frequency coupling, enabling 0 to represent no frequency coupling, analyzing amplitude characteristics of the two water quality parameters in a fluctuation difference value sequence, enabling the ratio or relative change rate of fluctuation amplitudes of the two parameters to be calculated, enabling a regression analysis method to be used for establishing a functional relation, such as a linear relation, a power law relation and the like, enabling the amplitude coupling degree of the two parameters to be quantized according to the goodness of regression fit and the coefficient size, enabling the amplitude coupling factor to reflect the influence intensity of one parameter on the other parameter, enabling the amplitude coupling factor to be normalized to be in a 0, enabling the frequency coupling factor to be in a 0, enabling the amplitude coupling factor to be in a 0 to be in a range normally, enabling the amplitude coupling factor to be multiplied by the amplitude coupling factor to be in a positive and negative influence on the coupling strength of the fluctuation coefficient to be obtained, determining the influence of the fluctuation amplitude coupling relation, and the influence of the fluctuation amplitude coefficient is considered positive and negative influence on the coupling coefficient is determined on the influence on the fluctuation amplitude and the fluctuation amplitude is considered on the time-fluctuation coefficient, and finally, forming a complete short-term coupling strength index.

The signal processing technology is introduced into the water quality parameter relation analysis, and the coupling characteristic among parameters is comprehensively captured through the double analysis of a frequency domain and a time domain. The frequency coupling factor reflects the rhythmic synchronicity of the parameter variation, and the amplitude coupling factor quantifies the degree of correlation of the variation strength, and the short-term coupling strength formed by multiplying the frequency coupling factor and the amplitude coupling factor provides an accurate measure of the parameter relation. The method can distinguish accidental synchronous change and real causal coupling, and provides scientific basis for constructing parameter microcoupling matrix.

In the embodiment of the invention, the detailed implementation steps of the time sequence clustering of the history coupling data set comprise:

Carrying out standardized treatment on pond water quality parameter data in the historical coupling data set to obtain a standardized water quality parameter sequence;

Calculating the similarity between any two standardized water quality parameter sequences based on a dynamic time warping algorithm;

In the embodiment, firstly, the pond water quality parameter data in a historical coupling data set stored in a cloud end is subjected to standardization processing, including z-score standardization (mean reduction and standard deviation) or min-max standardization (linear mapping to a [0,1] interval); the standardization process eliminates the dimensional differences and numerical range differences between different parameters, so that each parameter has the same weight in the subsequent analysis; the method comprises the steps of carrying out time window segmentation on a standardized water quality parameter sequence to form a large number of time segments with equal length as a basic unit of cluster analysis, selecting proper time window length such as12 hours, 24 hours or 7 days and the like to capture parameter change modes under different time scales, calculating the similarity between any two standardized water quality parameter sequences by using a Dynamic Time Warping (DTW) algorithm, wherein the DTW algorithm can process nonlinear expansion and phase difference in the time sequences, identify sequence modes similar in form but offset on a time axis, constructing a similarity matrix, recording the DTW distance between all time segments, clustering the standardized water quality parameter sequence by using a K-means clustering algorithm based on the similarity matrix, determining the optimal cluster number by adopting evaluation indexes such as profile coefficients (Silhouette Coefficient) or Davies-Bouldin indexes, carrying out cluster analysis on the time windows with different lengths respectively to obtain cluster results under the multiple time scales, analyzing the characteristics of each cluster center including the parameter change mode, the correlation between parameters and the correlation with external factors (such as weather, casting) and the correlation between the parameters, extracting the characteristic of each cluster center such as high-phase coupling pH value and the high-coupling time period such as the dissolution phase coupling between the high-phase value of each cluster time and the high-phase coupling value of the time-length cluster value, and establishing trigger conditions and application ranges of different coupling modes, thereby providing a knowledge base for a long-term coupling prediction model.

A stable mode in the long-term change of the water quality parameters is found through time sequence clustering, and the limitation that the traditional method is difficult to process high-dimensional time sequence data is overcome. The dynamic time warping algorithm enables the system to identify sequence segments that are not perfectly aligned on the time axis but that are similar in pattern, while multi-time scale cluster analysis reveals different periodicity from small temporal level to quaternary level. These long-term coupling modes provide a priori knowledge structure for the prediction model, significantly improving the accuracy and interpretability of the prediction.

In the embodiment of the invention, the detailed implementation steps for regulating and controlling the working states of the oxygenation equipment, the bait casting equipment and the water circulation equipment of the culture pond comprise the following steps:

The control signals are sent to the oxygenation equipment, the bait casting equipment and the water circulation equipment through a local controller in the edge computing node;

and monitoring working states of the oxygenation equipment, the bait casting equipment and the water circulation equipment in real time, and if the working states are abnormal, generating an abnormal alarm signal and uploading the abnormal alarm signal to the cloud regulation server.

In this embodiment, the edge computing node generates specific control signals for each device according to the final regulation command, including PWM power control signals of the oxygenation device, timing feeding and dosage control signals of the bait feeding device, start-stop and speed control signals of the water circulation device, and the like; the control signals accord with communication protocols and control interface specifications of all devices, such as Modbus, CAN bus or 4-20mA analog signals and the like, a local controller in an edge computing node safely transmits the control signals to all execution devices through a wired (such as RS485 bus and Ethernet) or wireless (such as ZigBee, loRa, NB-IoT) communication network, error detection and correction (EDC/ECC) mechanisms are adopted to ensure the reliability of signal transmission, which is particularly important in complex culture environments, the oxygenation devices adjust working power and working time according to the received control signals, such as a variable-frequency waterwheel aerator adjusts rotating speed and impeller depth according to PWM signals to achieve a specified oxygenation effect, the feeding device sets feeding time point, feeding amount and feeding rate according to the control signals, such as an intelligent feeding machine CAN adjust feed particle size, feeding amount and feeding area according to the signals, so as to realize accurate feeding, the water circulation device adjusts the running state and circulation strength according to the control signals, such as a water circulation system CAN control the mixing degree of deep layer and surface layer water, monitors the water layering condition, all devices internally provided with sensors in real-time work states including motor temperature, current, preset rotating speed, regular vibration state of all edge computing devices or other key state information of the devices are triggered by the edge computing system, normal state information of the devices are periodically triggered by the devices, the method comprises the steps of detecting whether working parameters of any equipment exceed a normal range or equipment response is abnormal (such as command execution overtime), immediately generating an abnormal alarm signal, wherein the abnormal alarm signal comprises information such as equipment ID, abnormal type, abnormal value, time stamp, severity and the like, firstly executing local processing by an edge computing node, such as automatic adjustment or degradation operation on slight abnormality, uploading the encrypted abnormal alarm signal to a cloud regulation server to trigger remote monitoring alarm and expert decision support, and collecting equipment state information of a plurality of ponds by the cloud server to perform statistical analysis and fault prediction to provide data support for equipment maintenance and updating.

Not only can accurately control various water quality regulation and control equipment, but also can monitor the working state of the equipment in real time, and timely discover and treat abnormal conditions. Through the rapid response of the edge side and the global coordination of the cloud, a high-reliability equipment control system is formed, the water quality regulation and control measures can be accurately executed, meanwhile, the equipment fault risk is reduced, and the overall stability and reliability of the system are improved.

In the embodiment of the invention, the fluctuation frequency spectrum corresponding to two water quality parameters is obtained by carrying out Fourier transform on the fluctuation difference value sequence, and the method comprises the following steps:

The method comprises the steps of carrying out time window segmentation on a fluctuation difference sequence, reducing frequency spectrum leakage effect by adopting a Hanning window, carrying out a fast Fourier transform algorithm on the fluctuation difference sequence in each time window to obtain local frequency spectrums, carrying out weighted average on all the local frequency spectrums to obtain global frequency spectrums, carrying out smoothing treatment and peak detection on the global frequency spectrums, identifying main frequency components and energy distribution thereof to form a characteristic frequency set, comparing and analyzing the characteristic frequency sets of two water quality parameters, and identifying common frequency components and unique frequency components to form a complete fluctuation frequency spectrum.

In this embodiment, firstly, the fluctuation difference sequence is divided into a plurality of overlapped time windows, the window length is usually 256 or 512 sample points, and the overlapping rate of adjacent windows is 50%, so as to consider both the frequency resolution and the time locality; the method comprises the steps of applying a hanning window function to each time window, reducing spectrum leakage caused by a truncation effect, improving accuracy of spectrum analysis, using a Fast Fourier Transform (FFT) algorithm to transform a sequence subjected to windowing, calculating a complex spectrum, calculating amplitude spectrum and phase spectrum of the spectrum, particularly focusing on energy distribution in the amplitude spectrum, carrying out quality assessment on local frequency spectrum of each window, distributing weight coefficients according to indexes such as signal to noise ratio, data integrity and the like, synthesizing a global frequency spectrum by using a weighted average method, reducing random noise influence, enhancing stable frequency components, carrying out smoothing processing on the global frequency spectrum by using a median filtering method or a wavelet denoising method and the like, improving separation degree of signals and noise, identifying significant peaks in the spectrum by using a peak detection algorithm, wherein the peaks correspond to main periodic components of fluctuation, sorting according to peak energy, screening out frequency points with energy proportion exceeding a threshold (such as 80% of total energy), forming a characteristic frequency set, comparing the characteristic frequency sets of two water quality parameters, calculating coincidence degree or proximity of the frequency points, marking the common frequency components and the frequency components, carrying out smoothing processing by using a weighted average method, carrying out a median filtering method or wavelet denoising method and the like, and carrying out characteristic frequency information, and the characteristic frequency proportion and phase proportion, and phase proportion of fluctuation frequency proportion, and frequency proportion of fluctuation, and structural fluctuation characteristic frequency information and frequency proportion, and frequency characteristic feature information.

The time-frequency analysis method is adopted to deeply mine the frequency domain characteristics of the water quality parameter fluctuation, and the limitation of the traditional FFT on non-stationary signal analysis is overcome through windowing processing and weighted synthesis. The extraction and comparison analysis of the characteristic frequency set enables the system to accurately identify the frequency resonance relation among parameters, and provides scientific basis of frequency domain viewing angle for subsequent coupling strength calculation.

In the embodiment of the invention, the time sequence clustering of the history coupling data set comprises the following steps:

Segmenting the historical coupling data set according to different time scales to form a multi-scale time segment library, wherein the multi-scale time segment library comprises hour-level time segments, day-level time segments, week-level time segments and month-level time segments;

Extracting time domain features, frequency domain features and statistical features from each time segment to construct a multidimensional feature vector, performing dimension reduction processing on the feature vector by applying principal component analysis, and retaining main variation information;

constructing a time sequence distance matrix, calculating the similarity between any two time segments by adopting a dynamic time warping algorithm, performing dimension reduction and grouping on the distance matrix by using a spectral clustering algorithm, and determining the optimal clustering quantity;

and extracting a representative mode from each cluster center to form a coupling mode library under different time scales for subsequent model training.

In the embodiment, firstly, a historical coupling data set is segmented according to a plurality of time scales, an hour-level segment (such as a 3-hour segment) captures short-term fluctuation characteristics, a day-level segment (24-hour segment) captures a day change period, a week-level segment (7-day segment) captures a medium-term change mode, and a month-level segment (30-day segment) captures a long-term trend; extracting multiple features from each time segment, including time domain features (such as trend, seasonal, peak position, etc.), frequency domain features (such as main frequency component, spectral energy distribution, etc.) and statistical features (such as mean, variance, skewness, kurtosis, etc.), taking into consideration cross features between water quality parameters such as correlation coefficients, mutual information amounts, grange causal relationships, etc., capturing interaction modes between parameters, processing high-dimensional feature vectors using Principal Component Analysis (PCA) or t-SNE, etc., reducing computation complexity and retaining primary variation information, selecting a principal component retaining more than 85% of the total variance, typically 20-30% of the original feature dimension, constructing a time sequence distance matrix, calculating similarity between any two time segments using a Dynamic Time Warping (DTW) algorithm, achieving nonlinear time alignment by finding an optimal path, being able to handle sequences of different lengths and time distortions, using an optimization algorithm such as a DTW constraint band, fastDTW or a distance matrix as input, applying a spectral clustering algorithm to perform analysis first, computing a spectral clustering algorithm, and then clustering the spectral coefficients in a matrix using a Laplace matrix in a low-dimensional space of the matrix, and then clustering the spectral coefficients in the Laplace matrix, the method comprises the steps of determining optimal clustering quantity by using a gap statistic or Bayesian information criterion and the like, selecting the most representative time segment (such as an example with the smallest distance from the clustering center) for each clustering center as the mode representation of the class, analyzing the characteristics and the forming conditions of each representative mode, such as a daily variation mode under the condition of high temperature and high illumination in summer, a parameter coupling mode within 4 hours after feeding, and the like, organizing all representative modes into a hierarchical coupling mode library, classifying according to time scale and applicable conditions, and providing a mode identification and matching basis for a long-term coupling prediction model.

The multi-scale and multi-characteristic time sequence mode discovery in the process overcomes the limitation of the traditional clustering method in processing time sequence data, particularly by a dynamic time warping algorithm. The introduction of spectral clustering improves the separation capability of complex nonlinear relations, so that a system can identify complex modes hidden in water quality data. The hierarchical coupling mode library not only supports the training of the prediction model, but also provides an interpretable knowledge structure, so that the system decision is more transparent and reliable.

Referring to fig. 2, for comparison of regulation and control effects, the invention realizes accurate monitoring, microcoupling analysis, dynamic regulation and control and intelligent execution of water quality parameters through the architecture of edge calculation and cloud cooperation, and effectively improves the accuracy, instantaneity and foresight of water quality management of cultivation.

The above is only a preferred embodiment of the present invention, and the present invention is not limited thereto, but it is possible for those skilled in the art to modify the technical solutions described in the above embodiments or to substitute some of the technical features thereof with the detailed description of the present invention with reference to the above embodiments. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

The threshold selection for the present specification is set by those skilled in the art according to the actual circumstances.

Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the spirit and scope of the invention as defined by the appended claims and their equivalents.

Claims

1. Remote intelligent regulation and control system of breed aquatics pond quality of water based on edge calculation, its characterized in that includes:

2. The edge-calculation-based remote intelligent regulation and control system for water quality of an aquaculture pond according to claim 1, wherein said constructing a parameter microcoupling matrix in each regulation and control period comprises:

3. The edge-computing-based remote intelligent regulation system for water quality of an aquaculture pond according to claim 1, wherein said determining the dynamic regulation weights for each water quality parameter comprises:

4. The edge-computing-based remote intelligent regulation and control system for water quality of an aquaculture pond according to claim 1, wherein the generating of the local regulation and control instruction in the corresponding regulation and control period comprises:

5. The edge-calculation-based remote intelligent regulation and control system for water quality of an aquaculture pond according to claim 1, wherein the constructing a long-term coupling prediction model comprises:

6. The edge-computing-based remote intelligent regulation system for water quality in an aquaculture pond according to claim 1, wherein the generating a global regulation strategy corresponding to a future regulation period comprises:

7. The edge-computing-based remote intelligent regulation and control system for water quality of an aquaculture pond according to claim 1, wherein the generating of the final regulation and control instruction comprises:

8. The edge-calculation-based remote intelligent regulation and control system for water quality of an aquaculture pond according to claim 2, wherein the calculation corresponds to a short-term coupling strength between two water quality parameters, comprising:

9. The edge-computing-based remote intelligent regulation and control system for water quality of an aquaculture pond according to claim 5, wherein said time-series clustering of the historical coupling dataset comprises:

10. The edge-calculation-based remote intelligent regulation and control system for water quality of an aquaculture pond according to claim 1, wherein the regulation and control of the working states of oxygenation equipment, bait casting equipment and water circulation equipment of the aquaculture pond comprises: