Energy storage device control system

Abstract

A control system for energy storage equipment belongs to the technical field of energy storage equipment control systems and is designed to solve the problem that the charging process of energy storage equipment becomes unstable and unpredictable. The control system of the present invention mainly includes a data acquisition module, a data processing and storage module, a prediction analysis module, a decision support module, a control execution module and an interactive configuration module. The data acquisition module collects information from multiple data sources, the data processing and storage module cleans, processes and stores the collected data, and preserves the processed data for a long time. The prediction analysis module uses data analysis technology and prediction algorithms to conduct in-depth analysis of long-term power grid loads, renewable energy generation trends, etc., predicts energy supply and demand in the future, and adjusts the operation mode of the energy storage equipment to achieve efficient use of energy and stable operation of the system.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage device control systems, and in particular to an energy storage device control system. Background Art

[0002] The application of energy storage equipment in power systems mainly includes peak and valley regulation, frequency regulation, stabilizing power systems, improving power quality, etc. In the field of renewable energy, energy storage equipment is also widely used, mainly to solve the intermittent and instability problems of renewable energy and improve the utilization rate and reliability of renewable energy. The energy storage equipment control system is a system used to manage and control the operation of energy storage equipment. It can monitor the status of energy storage equipment, control the charging and discharging process of energy storage equipment, and work in coordination with other systems to achieve efficient utilization and optimized management of energy.

[0003] Energy storage equipment plays a key role in peak load shifting and valley filling in the power system, and can release electric energy to supplement the grid supply during power shortage periods. However, its charging process relies on renewable energy such as photovoltaic power generation or excess power in the grid, which makes the charging and discharging strategy of energy storage batteries affected by multiple factors, including grid demand fluctuations, intermittent and unstable photovoltaic power generation, etc. Due to these factors, the sustainability of energy storage equipment faces challenges.

[0004] Specifically, photovoltaic power generation is significantly affected by climate and weather conditions, and its output has significant volatility. When the sun is sufficient, photovoltaic power generation can generate a large amount of electricity, providing sufficient charging source for energy storage equipment. However, during periods of cloud cover or at night, the output of photovoltaic power generation will drop significantly or even be zero, which makes the charging process of the energy storage equipment unstable and unpredictable. Summary of the invention

[0005] The purpose of the present invention is to provide an energy storage device control system, which solves the problem in the background art that the charging process of the energy storage device becomes unstable and unpredictable.

[0006] To achieve the above object, the present invention provides the following technical solution: an energy storage device control system, comprising:

[0007] Data acquisition module: used to collect information from multiple data sources, including: energy storage equipment status, historical solar power generation data, weather conditions and grid operation status;

[0008] Data processing and storage module: used for data cleaning, data processing and data storage, identifying and correcting outliers and missing values, ensuring data accuracy and reliability, and performing conversion and normalization operations, as well as long-term storage of processed data for subsequent analysis and prediction;

[0009] Prediction and analysis module: Using data analysis technology and prediction algorithms, the system can conduct in-depth analysis of long-term grid load and renewable energy generation trends. By establishing a prediction model, it can predict energy supply and demand in the future and provide a basis for decision-making.

[0010] Furthermore, the data acquisition module includes a state acquisition module and a communication gateway, wherein the state acquisition module is used to acquire energy storage device information and historical solar power generation.

[0011] Furthermore, the energy storage device related information includes the state of charge (SOC), state of health (SOH), input and output power, voltage, current and temperature, which are used to monitor the health of the device, optimize the charging and discharging strategy and prevent failures.

[0012] Furthermore, the communication gateway is used to obtain recent weather forecast information from the meteorological station and recent power grid supply and demand from the State Grid, and the weather forecast information includes the total amount of medium and low clouds and the corresponding cloudy and sunny data.

[0013] Furthermore, the prediction and analysis module selects two indicators, Euclidean distance and cosine similarity, to represent the similarity. Before prediction, the date with similar environmental factors to the prediction day is found as a data sample of similar days. Finally, the two indicators "Euclidean distance" and "cosine similarity" are combined into one indicator to find similar days, and then the approximate power generation data of the prediction day is predicted. By predicting the power generation data and obtaining the supply and demand relationship of the power grid, it can be determined when to charge and discharge.

[0014] Furthermore, the Euclidean distance calculation formula is:

[0015]

[0016] The above formula describes the description of any i-th day X i and prediction day j X j The overall difference of meteorological factors between them, k is the number of eigenvectors, m is the number of eigenvectors, and the Euclidean distance is a value greater than 0. The smaller the distance, the more similar it is.

[0017] Furthermore, the cosine similarity calculation formula is:

[0018]

[0019] The above formula describes any i-th day X i and prediction day j X j The similarity of the changing trends between them, k is the number of the feature vector, and m is the number of feature vectors;

[0020] The cosine similarity value is in the interval [0,1]. The closer it is to 1, the more similar it is. In order to be consistent with the cosine similarity, the Euclidean distance is converted so that the converted distance is also a value in the interval [0,1], and the closer it is to 1, the more similar it is. The formula after conversion is:

[0021]

[0022] Where max(d ij ) refers to all d ij The maximum value in .

[0023] The Euclidean distance and cosine similarity are used to represent the similarity synthesis.

[0024] S=αD ij +(1-α)Dcos ij ;

[0025] Where α is the empirical weight coefficient, and its value will be different under different weather conditions.

[0026] Furthermore, the prediction analysis module predicts the photovoltaic power generation through the decision tree algorithm of SVM-RFE, uses the fast principal component analysis algorithm and the principal component analysis algorithm based on the kernel function to perform feature fusion on the data to reduce the dimension, and then establishes a decision tree model to predict the photovoltaic power generation. The prediction steps are as follows:

[0027] S1: Obtain the data set from the database and preprocess it, selecting features such as temperature, wind direction, date and time, radiation, capacitance and voltage;

[0028] S2: Use the fastPCA algorithm to reduce the dimension of the ultra-high-dimensional feature vector to optimize the time performance, apply the KPCA algorithm to process nonlinear features to improve the dimension reduction effect, and fuse the results of fastPCA and KPCA to obtain the feature set after comprehensive dimension reduction;

[0029] S3: Apply the SVM-RFE algorithm to select features from the fused multimodal feature set, remove unimportant features based on model accuracy, build a decision tree model using the CART algorithm, select feature variables and divide nodes based on the principle of minimizing squared errors, and set the maximum depth of the decision tree, the minimum number of samples required for internal node re-division, and the maximum number of leaf nodes.

[0030] Furthermore, the SVM-RFE algorithm formula is:

[0031]

[0032]

[0033] Where M is the total number of kernel functions, α in the formula i and β m The coefficients are obtained from the training data through quadratic programming, β m is the weight of the mth kernel function, and also the importance of the group feature using the kernel function, thereby completing the group feature selection; assuming that there is a set of feature sets s(1), s(2), ···, s(T) for rumor detection, containing T samples, assuming that each sample feature set s contains three features: local image features, global image features, and visual repetition rate, which are expressed as x(t)l, x(t)g, x(t)vrr, and the sample label set y(t)∈{Truth,Rumor}. The objective function is a dual objective function with both maximum and minimum terms.

[0034] Furthermore, the control system further comprises:

[0035] Decision support module: used to provide charging and discharging decisions for operators and automatic control systems, including a user interface to present analysis results and recommended decisions;

[0036] Control execution module: Based on the output of the decision module, it is responsible for issuing specific control commands to the energy storage equipment, regulating the charging and discharging process, and interacting with the power grid;

[0037] Interactive Configuration Module: Used to provide an interface for system users, grid operators or independent power producers to monitor system status, adjust settings, and connect or configure new energy storage devices.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention provides a control system for energy storage equipment. The control system mainly includes a data acquisition module, a data processing and storage module, a prediction analysis module, a decision support module, a control execution module and an interactive configuration module. The data acquisition module collects information from multiple data sources. The data processing and storage module cleans, processes and stores the collected data and preserves the processed data for a long time. The prediction analysis module uses data analysis technology and prediction algorithms to conduct in-depth analysis of long-term power grid loads, renewable energy generation trends, etc., predicts energy supply and demand in the future, and adjusts the operation mode of the energy storage equipment to achieve efficient use of energy and stable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the control system module of the present invention;

[0041] Figure 2 It is a schematic diagram of a data acquisition module of the present invention;

[0042] Figure 3 It is a schematic diagram of the energy storage device information module of the present invention;

[0043] Figure 4 The process of the photovoltaic power generation prediction model of the present invention is Figure 1 ;

[0044] Figure 5 The process of the photovoltaic power generation prediction model of the present invention is Figure 2 . DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In order to solve the technical problem that the charging process of energy storage equipment becomes unstable and unpredictable, such as Figure 1-Figure 5 As shown, the following preferred technical solutions are provided:

[0047] An energy storage device control system, comprising:

[0048] Data acquisition module: used to collect information from multiple data sources, including: energy storage equipment status, historical solar power generation data, weather conditions and grid operation status;

[0049] Data processing and storage module: used for data cleaning, data processing and data storage, identifying and correcting outliers and missing values, ensuring data accuracy and reliability, and performing conversion and normalization operations, as well as long-term storage of processed data for subsequent analysis and prediction;

[0050] Prediction and analysis module: Using data analysis technology and prediction algorithms, the system can conduct in-depth analysis of long-term grid load and renewable energy generation trends. By establishing a prediction model, it can predict energy supply and demand in the future and provide a basis for decision-making;

[0051] Decision support module: used to provide charging and discharging decisions for operators and automatic control systems, including a user interface to present analysis results and recommended decisions;

[0052] Control execution module: Based on the output of the decision module, it is responsible for issuing specific control commands to the energy storage equipment, regulating the charging and discharging process, and interacting with the power grid;

[0053] Interactive Configuration Module: Used to provide an interface for system users, grid operators or independent power producers to monitor system status, adjust settings, and connect or configure new energy storage devices.

[0054] Data collection: The data collection module collects information from multiple data sources, including energy storage device status, historical solar power generation data, weather conditions, and grid operation status. These data can be obtained through sensors, monitoring equipment, weather stations, etc.

[0055] The data acquisition module includes a status acquisition module and a communication gateway, wherein the status acquisition module is used to obtain energy storage equipment information and historical solar power generation. The energy storage equipment related information includes charging status, health status, input and output power, voltage, current and temperature, which are used to monitor the health status of the equipment, optimize the charging and discharging strategy and prevent faults. The communication gateway is used to obtain recent weather forecast information from the meteorological station and recent power grid supply and demand from the State Grid. The weather forecast information includes the total amount of medium and low clouds and the corresponding cloudy and sunny data.

[0056] Data processing and storage: The data processing and storage module cleans, processes and stores the collected data. First, the data is cleaned to identify and correct outliers and missing data to ensure the accuracy and reliability of the data. Then, the data is processed, including conversion and normalization operations, to make it suitable for subsequent analysis and prediction. Finally, the processed data is stored for a long time to facilitate subsequent data analysis and decision making.

[0057] Predictive analysis: The predictive analysis module uses data analysis technology and prediction algorithms to conduct in-depth analysis of long-term grid load, renewable energy generation trends, etc. By establishing a prediction model, it predicts the energy supply and demand in the future and provides a basis for decision-making. This can help optimize charging and discharging strategies and adjust the operation mode of energy storage equipment to maximize the use of renewable energy and balance the grid load.

[0058] Similarity analysis: In the prediction analysis, the system measures the similarity between different dates by selecting two indicators, namely Euclidean distance and cosine similarity. According to the environmental factors of the prediction day, similar historical dates are found as the data samples of similar days. By calculating the Euclidean distance and cosine similarity, the similarity synthesis index is obtained to predict the approximate power generation data of the prediction day.

[0059] Decision support: The decision support module provides charging and discharging decision suggestions to operators and automatic control systems based on prediction results and other information, presents analysis results through the user interface, and supports operators in decision making, including optimizing charging and discharging strategies and adjusting energy storage equipment.

[0060] Control execution: Based on the output of the decision module, the control execution module issues specific control commands to the energy storage device, regulates the charging and discharging process, and interacts with the power grid. According to the decision results and real-time power grid conditions, the control execution module adjusts the operation mode of the energy storage device to achieve efficient energy utilization and stable operation of the system.

[0061] Interactive Configuration: The interactive configuration module provides an interface for system users, grid operators or independent power producers to monitor system status, adjust settings, and connect or configure new energy storage devices. Users can use this interface to monitor the operation of energy storage devices, set parameters and optimize adjustments to meet different needs and requirements.

[0062] If we can find the date with similar environmental factors as the predicted date as the data sample of similar days before prediction, the prediction accuracy can be greatly improved. In order to find the date with similar environmental factors and change patterns as the predicted date to form a time series, we select "Euclidean distance" and "cosine similarity" as the two indicators to represent the similarity.

[0063] The Euclidean distance calculation formula is:

[0064]

[0065] The above formula describes the description of any i-th day X i and prediction day j X j The overall difference of meteorological factors between them, k is the number of eigenvectors, m is the number of eigenvectors, and the Euclidean distance is a value greater than 0. The smaller the distance, the more similar it is.

[0066] The cosine similarity calculation formula is:

[0067]

[0068] The above formula describes any i-th day X i and prediction day j X j The similarity of the changing trends between them, k is the number of the feature vector, and m is the number of feature vectors.

[0069] The cosine similarity value is in the interval [0,1]. The closer it is to 1, the more similar it is. In order to be consistent with the cosine similarity, the Euclidean distance is converted so that the converted distance is also a value in the interval [0,1], and the closer it is to 1, the more similar it is. The formula after conversion is:

[0070]

[0071] Where max(d ij ) refers to all d ij The maximum value in .

[0072] The formula for the similarity synthesis of the two indicators of Euclidean distance and cosine similarity is:

[0073] S=αD ij +(1-α)Dcos ij ;

[0074] Among them, α is the empirical weight coefficient, and its value will be different under different weather conditions. If the weather and humidity have an obvious changing trend within a day, α should be close to 0, otherwise it should be close to 1. After the dates with similar environmental factors to the predicted day are used as similar days to form data samples, the historical solar power generation is collected through the data acquisition module, and the power generation on the predicted day can be further predicted.

[0075] Specifically, the data acquisition module collects information from multiple data sources, including the status of energy storage equipment, historical solar power generation data, weather conditions and grid operation status, etc., and obtains this data through sensors, monitoring equipment, meteorological stations, etc.; then, the data processing and storage module cleans, processes and stores the collected data. First, the data is cleaned, outliers and missing data are identified and corrected to ensure the accuracy and reliability of the data. Then, the data is processed, including conversion and normalization operations, so that it is suitable for subsequent analysis and prediction. Finally, the processed data is stored for a long time to facilitate subsequent data analysis and decision-making; next, the predictive analysis module uses data analysis technology and prediction algorithms to conduct in-depth analysis of long-term grid loads, renewable energy generation trends, etc., and by establishing a prediction model, it predicts the energy supply and demand in the future, providing a basis for decision-making, which helps to optimize the charging and discharging strategy and adjust the operation mode of energy storage equipment to maximize the use of renewable energy and balance the grid load; in the predictive analysis, the system selects two indicators, Euclidean distance and cosine similarity, to measure different dates. The similarity between them is determined according to the environmental factors of the forecast day. The historical dates similar to them are found as the data samples of similar days. The similarity synthesis index is obtained by calculating the Euclidean distance and cosine similarity, which is used to predict the approximate power generation data of the forecast day. The decision support module provides charging and discharging decision suggestions for operators and automatic control systems based on the forecast results and other information, presents the analysis results through the user interface, and supports operators in making decisions, including optimizing charging and discharging strategies and adjusting energy storage equipment. The control execution module issues specific control commands to the energy storage equipment based on the output of the decision module, regulates the charging and discharging process, and interacts with the power grid. According to the decision results and the real-time power grid situation, the operation mode of the energy storage equipment is adjusted to achieve efficient use of energy and stable operation of the system. Finally, the interactive configuration module provides an interface for system users, power grid operators or independent power producers to monitor the system status, adjust settings, and connect or configure new energy storage equipment. Users can monitor the operation of energy storage equipment, set parameters and optimize adjustments through this interface to meet different needs and requirements.

[0076] Another technical solution provided by the present invention is to predict the power generation of photovoltaic electric field through the decision tree algorithm of SVM-RFE. The model first uses the fast principal component analysis algorithm and the principal component analysis algorithm based on the kernel function to fuse the data features to reduce the dimension, and then establishes a decision tree model to predict the amount of photovoltaic power generation.

[0077] It includes the following steps:

[0078] S1: Obtain the data set from the database and preprocess it, selecting features such as temperature, wind direction, date and time, radiation, capacitance and voltage;

[0079] S2: Use the fastPCA algorithm to reduce the dimension of the ultra-high-dimensional feature vector to optimize the time performance, apply the KPCA algorithm to process nonlinear features to improve the dimension reduction effect, and fuse the results of fastPCA and KPCA to obtain the feature set after comprehensive dimension reduction;

[0080] S3: Apply the SVM-RFE algorithm to select features from the fused multimodal feature set, remove unimportant features based on model accuracy, build a decision tree model using the CART algorithm, select feature variables and divide nodes based on the principle of minimizing squared errors, and set the maximum depth of the decision tree, the minimum number of samples required for internal node re-division, and the maximum number of leaf nodes.

[0081] PCA is a statistical method widely used in data dimensionality reduction. Its main function is to extract implicit relationships between features. However, the original PCA method is not suitable for nonlinear features, while the improved KPCA has better results for nonlinear data dimensionality reduction. For ultra-high-dimensional feature vectors, especially feature sets with dimensions far higher than the number of samples, the time complexity of the original PCA is relatively high. Therefore, the fastPCA algorithm calculates the covariance matrix of the transposed matrix of the sample to reduce the complexity of the PCA algorithm from O(n3) to O(n2), thereby optimizing the time performance of the PCA algorithm.

[0082] The recursive feature elimination (RFE) method obtains the importance of features based on model accuracy. In each round, the least important features are removed recursively, and the model is built with the remaining features. Repeated iterations are used to obtain the optimal feature set.

[0083] The core learner of multi-kernel learning in the FS-MKL method is SVM. In order to achieve the best dimensionality reduction effect, the SVM-RFE algorithm is used to select features from the multimodal feature set;

[0084] Assume that there is a set of feature sets s(1), s(2), ···, s(T) for rumor detection, containing T samples. Assume that each sample feature set s contains three features: local image features, global image features, and visual repetition rate, which are represented as x(t)l, x(t)g, x(t)vrr. Sample label set y(t)∈{Truth,Rumor}, the MKL algorithm uses the label set to optimize the objective function, which is a dual objective function with both maximum and minimum terms, as shown in the formula

[0085]

[0086]

[0087] Where M is the total number of kernel functions, α in the formula i and βm The coefficients are obtained from the training data through quadratic programming, β m is the weight of the mth kernel function, and is also the importance of the group feature using the kernel function, thereby completing the group feature selection;

[0088] Dimensionality reduction based on principal component analysis is more efficient in processing large data sets, but its contribution to the detection model is not strong. The SVM-PFE algorithm uses model accuracy as a feature screening condition, which is more targeted in feature selection, but the iterative execution learner greatly reduces the efficiency of the algorithm. Combining the two to give full play to their advantages, different dimensionality reduction processes will be performed on all group features, the final dimensionality reduction plan will be determined, and then multi-core learning will be used to select group features of the feature set and output the detection results.

[0089] The FS-MKL method uses the fastPCA and KPCA algorithms to reduce the dimension of the feature set respectively, and directly connects and merges the two feature vectors. The two dimension reduction results are fused and input into the SVM-PFE algorithm for secondary dimension reduction to obtain the final optimal feature set.

[0090] The decision tree model is a model that makes decisions based on a tree structure. It divides the feature space into different areas for prediction and uses the CART algorithm to predict the power generation of photovoltaic power generation.

[0091] Assume that X and Y are input and output variables respectively, and Y is a continuous variable. Given a training data set:

[0092] N={(x 1 ,y 1 ), (x 2 ,y 2 ),…,(x N ,y N )}

[0093] Traverse each characteristic variable and its corresponding characteristic value in turn. Assuming that the current segmentation variable is the jth variable and the corresponding segmentation characteristic value is s, two regions can be divided and defined:

[0094]

[0095] Through the above steps, the input space will be continuously divided into L subspaces α1, α2, …, αL. Each subspace αl contains part of the sample data and the output value βl. At this time, the solution of the model can be expressed as:

[0096]

[0097] Select the optimal segmentation attribute j and attribute value s to divide the input space, expressed as:

[0098]

[0099] Traverse all input feature variables and their eigenvalues, find the current optimal split point (j, s), and then divide the current space into two sub-areas according to the split point. At this time, if the two sub-areas cannot be divided any further, the corresponding optimal output value can be obtained, which is expressed as:

[0100]

[0101] According to the above steps, if the area can still be divided, repeat the above steps for the area until the entire decision tree is completely indivisible. The constructed decision tree uses the minimization of square error as the splitting criterion to select feature variables, the maximum depth is 4, the minimum number of samples required for the internal node (i.e., judgment condition) to be re-divided is 2, and the maximum number of leaf nodes is 10.

[0102] The flowchart of establishing photovoltaic power generation prediction model using decision tree algorithm is as follows: Figure 5 As shown in the figure, firstly, the original samples are sorted and features are extracted, and the training set and test set are divided. Then, the decision tree model is trained.

[0103] By combining fastPCA, KPCA and SVM-RFE algorithms for feature dimensionality reduction and selection, and then using the decision tree model to predict photovoltaic power generation, the prediction accuracy and the generalization ability of the model can be effectively improved.

[0104] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0105] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A control system for energy storage equipment, characterized in that: include: Data acquisition module: used to collect information from multiple data sources, including: energy storage equipment status, historical solar power generation data, weather conditions and grid operation status; Data processing and storage module: used for data cleaning, data processing and data storage, identifying and correcting outliers and missing values, ensuring data accuracy and reliability, and performing conversion and normalization operations, as well as long-term storage of processed data for subsequent analysis and prediction; Prediction analysis module: Using data analysis technology and prediction algorithms, the system can conduct in-depth analysis of long-term grid load and renewable energy generation trends. By establishing a prediction model, it can predict energy supply and demand in the future and provide a basis for decision-making; The prediction analysis module selects two indicators, Euclidean distance and cosine similarity, to represent similarity. Before prediction, the date with similar environmental factors to the prediction day is found as a data sample of similar days. Finally, the two indicators "Euclidean distance" and "cosine similarity" are combined into one indicator to find similar days, and then the approximate power generation data of the prediction day is predicted. By predicting the power generation data and obtaining the supply and demand relationship of the power grid, it can be determined when to charge and discharge. The Euclidean distance calculation formula is: The above formula describes the description of any i-th day X i and prediction day j X j The overall difference of meteorological factors between them, k is the number of eigenvectors, m is the number of eigenvectors, and the Euclidean distance is a value greater than 0. The smaller the value, the more similar it is. The cosine similarity calculation formula is: The above formula describes any i-th day X i and prediction day j X j The similarity of the changing trends between them, k is the number of the feature vector, and m is the number of feature vectors; The cosine similarity value is in the interval [0,1]. The closer it is to 1, the more similar it is. In order to be consistent with the cosine similarity, the Euclidean distance is converted so that the converted distance is also a value in the interval [0,1], and the closer it is to 1, the more similar it is. The formula after conversion is: Where max(d ij ) refers to all d ij The maximum value in ; The Euclidean distance and cosine similarity are used to represent the similarity synthesis. S=αD ij +(1-α)Dcos ij ; Where α is the empirical weight coefficient, and its value will be different under different weather conditions; The prediction analysis module predicts the photovoltaic power generation through the decision tree algorithm of SVM-RFE, uses the fast principal component analysis algorithm and the principal component analysis algorithm based on the kernel function to perform feature fusion on the data to reduce the dimension, and then establishes a decision tree model to predict the photovoltaic power generation. The prediction steps are as follows: S1: Obtain the data set from the database and preprocess it, selecting features such as temperature, wind direction, date and time, radiation, capacitance and voltage; S2: Use the fastPCA algorithm to reduce the dimension of the ultra-high-dimensional feature vector to optimize the time performance, apply the KPCA algorithm to process nonlinear features to improve the dimension reduction effect, and fuse the results of fastPCA and KPCA to obtain the feature set after comprehensive dimension reduction; S3: Apply the SVM-RFE algorithm to select features from the fused multimodal feature set, remove unimportant features based on model accuracy, build a decision tree model using the CART algorithm, select feature variables and divide nodes based on the principle of minimizing squared errors, and set the maximum depth of the decision tree, the minimum number of samples required for internal node re-division, and the maximum number of leaf nodes. The SVM-RFE algorithm formula is: Where M is the total number of kernel functions, α in the formula i and β m The coefficients are obtained from the training data through quadratic programming, β m is the weight of the mth kernel function, and is also the importance of the group feature using the kernel function, thereby completing the group feature selection; assuming that there is a set of feature sets s(1), s(2), ···, s(T) for rumor detection, containing T samples, assuming that each sample feature set s contains three features: local image features, global image features, and visual repetition rate, which are represented as x(t)l, x(t)g, x(t)vrr, and the sample label set y(t)∈{Truth,Rumor}. The objective function is a dual objective function with both maximum and minimum terms; The control system further comprises: Decision support module: used to provide charging and discharging decisions for operators and automatic control systems, including a user interface to present analysis results and recommended decisions; Control execution module: Based on the output of the decision module, it is responsible for issuing specific control commands to the energy storage equipment, regulating the charging and discharging process, and interacting with the power grid; Interactive Configuration Module: Used to provide an interface for system users, grid operators or independent power producers to monitor system status, adjust settings, and connect or configure new energy storage devices.

2. An energy storage device control system as claimed in claim 1, characterized in that: The data acquisition module includes a state acquisition module and a communication gateway, wherein the state acquisition module is used to acquire energy storage device information and historical solar power generation.

3. An energy storage device control system as claimed in claim 2, characterized in that: The energy storage device related information includes the state of charge (SOC), the state of health (SOH), input and output power, voltage, current and temperature, which are used to monitor the health status of the device, optimize the charging and discharging strategy and prevent failures.

4. An energy storage device control system as claimed in claim 3, characterized in that: The communication gateway is used to obtain recent weather forecast information from the meteorological station and recent power grid supply and demand from the State Grid. The weather forecast information includes the total amount of medium and low clouds and the corresponding cloudy and sunny data.

Images