Early warning method for health degree and residual life of battery of medium-and-large-sized electrochemical energy storage power station

Abstract

An early warning method for battery health degree and residual life of a medium-and-large-sized electrochemical energy storage power station belongs to the field of battery health assessment, and comprises the following steps: firstly, establishing a battery health state grading system, determining threshold ranges of key parameters such as voltage, internal resistance and charge and discharge capacity, training historical data through various algorithms such as a support vector machine to construct a discrimination model, and calculating the battery health degree and residual life of the medium-and-large-sized electrochemical energy storage power station; and the health level is accurately evaluated. The health detection system monitors data in real time, an evaluation model is constructed through preprocessing and correlation analysis, and the health level is judged and a report is generated according to the evaluation model. And for a sub-health battery unit, determining an isolation priority according to fuzzy comprehensive evaluation, unloading and predicting a deterioration trend to formulate a maintenance plan. And during discharging, the battery is selected to discharge and power is adjusted according to the health level. According to the method, historical data are clustered and analyzed by using a machine learning algorithm to construct an evaluation model to predict a new data level, and an energy storage system operation strategy is dynamically adjusted based on the health level, so that economy and reliability are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery health assessment, and in particular relates to a method for early warning of battery health and remaining life in medium-to-large electrochemical energy storage power stations. Background Art

[0002] Medium and large-scale electrochemical energy storage power stations are usually composed of a large number of battery cells, and their operating status directly affects the performance, safety and service life of the energy storage system. During the long-term operation of the energy storage power station, the battery cells will inevitably experience performance degradation, such as capacity decay, increased internal resistance and voltage fluctuations. These problems will not only reduce the charging and discharging efficiency of the energy storage power station, but may also cause safety hazards, such as serious accidents such as thermal runaway.

[0003] At present, traditional battery management systems have certain limitations in terms of battery health monitoring and early warning. Most systems can only monitor and determine thresholds for simple parameters such as voltage and current, and lack the ability to comprehensively and accurately evaluate the health status of batteries. It is difficult to effectively identify and warn in advance of the sub-healthy state of the battery, and it is impossible to take targeted measures in time to prevent further deterioration of the battery. In addition, in terms of the operation strategy of the energy storage system, it is often impossible to dynamically optimize and adjust according to the real-time health status of the battery cells, resulting in uneven use of healthy battery cells and sub-healthy and unhealthy battery cells, which accelerates the performance degradation of the overall battery pack and shortens the service life of the battery cells. At the same time, it also affects the economy and reliability of the energy storage power station in meeting the grid dispatching needs.

[0004] Based on the above analysis, the main technical problem solved by the present invention is: how to classify the health status of batteries in energy storage power stations and make corresponding decisions in different levels. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for early warning of battery health and remaining life in medium and large electrochemical energy storage power stations, establish an accurate battery health status grading system for energy storage power stations, and accurately divide battery cells into three levels: healthy, sub-healthy and unhealthy by comprehensively considering key parameters such as voltage, internal resistance, and charge and discharge capacity. Corresponding operating parameter standards and threshold ranges are set to improve the accuracy of battery health status assessment.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method for early warning of battery health and remaining life of medium and large electrochemical energy storage power stations, comprising the following steps: S1. According to the pre-established energy storage power station battery health status classification system, the battery cells are divided into three levels: healthy, sub-healthy and unhealthy. For each level, corresponding operating parameter standards are set. The operating parameter standards include the threshold ranges of voltage, internal resistance and charge and discharge capacity; S2. Use the health detection system to monitor the operating status information of each battery unit in real time, obtain the real-time data of voltage, internal resistance, and charge and discharge capacity, and judge the health level of each battery unit through internal comparison and analysis; S3. For the sub-healthy battery cells detected, the control system automatically generates a higher priority unloading instruction, and isolates the sub-healthy battery cells from the energy storage system by cutting off the output of the battery cells, thereby preventing the sub-healthy battery cells from further deteriorating; S4. While unloading sub-healthy battery cells, dynamically adjust the working status of other healthy battery cells, and determine the unloading order of different battery cells according to the real-time health rating results to ensure the overall output performance of the energy storage system; S5. When the energy storage system needs to perform a discharge operation, the battery cells with a higher health level are preferentially selected to participate in the discharge. For the battery cells assessed as sub-healthy and unhealthy, the control system gradually reduces their discharge power until the discharge is completely stopped; S6. Analyze the historical operation data of the battery cell using a machine learning algorithm, identify the characteristic patterns of different health levels through a clustering algorithm, and establish a health status assessment model for predicting and classifying the health level of the newly collected battery cell data; S7. Dynamically adjust the operation strategy of the energy storage system based on the health level of the battery cells to maximize the service life of the battery cells while meeting the grid dispatch requirements, thereby improving the overall economy and reliability of the energy storage power station.

[0007] Preferably, the sub-steps of step S1 are: S1.1. According to the pre-established energy storage power station battery health status classification system, obtain the real-time operating parameters of the battery unit, including key parameters such as voltage, internal resistance and charge and discharge capacity; S1.2, comparing the acquired real-time operating parameters of the battery unit with the preset health level threshold range to determine the health status of each battery unit; S1.3. If the real-time operating parameters of the battery cell are all within the threshold range of the health level, the health state of the battery cell is determined as the health level; S1.4. If one or more of the real-time operating parameters of the battery cell exceed the threshold range of the health level, but are all within the threshold range of the sub-health level, the health state of the battery cell is determined to be a sub-health level; S1.5. If one or more of the real-time operating parameters of the battery cell exceed the threshold range of the sub-health level, the health state of the battery cell is determined to be an unhealthy level; S1.6. Based on the determined health status of the battery cells, the corresponding operating parameter standards are adopted to control and optimize the charging and discharging process of the battery cells to ensure the safe and stable operation of the energy storage power station; S1.7. The historical operation data of the battery cell is trained by the support vector machine algorithm to obtain a discriminant model for health status classification, which is used to predict and classify the health status of new battery cell operation data, thereby improving the accuracy and efficiency of battery health status assessment; The training process of the support vector machine algorithm on the historical operation data of the battery unit is as follows: Assume training data set ,in The voltage of the battery cell 、Internal resistance and charge and discharge capacity The characteristic vector of ; They correspond to the three category labels of unhealthy, sub-healthy, and healthy respectively; Its objective function is: ,in are the weight vector components corresponding to voltage, internal resistance, and charge and discharge capacity characteristics, respectively. is the hyperplane intercept, You can set it as an experience value first. is the slack variable; The constraints are: .

[0008] Preferably, the sub-steps of step S2 are: S2.1, obtaining real-time operating status data of each group of battery cells, the operating status data including key parameters such as voltage, internal resistance and charge and discharge capacity; S2.2, transmitting the acquired real-time data to the health detection system and storing it in the corresponding database; S2.3, preprocessing the stored real-time data, including data cleaning and normalization operations to improve data quality; S2.4. Based on the pre-processed real-time data, a correlation analysis algorithm is used to calculate the correlation coefficients between the key parameters and determine the degree of correlation between the parameters; S2.5. Construct a battery health assessment model based on the correlation coefficients of each key parameter, and comprehensively consider the impact of each key parameter on the battery health status; S2.6. Using the constructed battery health assessment model, the health status of each battery cell group is assessed to obtain the health level of each battery cell; S2.7. Generate a battery health report based on the evaluated battery cell health level and send the report to relevant managers so that they can understand the battery health status in time and formulate operation and maintenance strategies in advance.

[0009] Preferably, the sub-steps of step S3 are: S3.1, according to a preset battery health status assessment model, real-time monitoring of each battery cell in the energy storage system is performed to obtain key parameters reflecting the health status of each battery cell; S3.2, if the health status parameter of a battery cell is monitored to be beyond a predetermined threshold range, the battery cell is determined to be in a sub-healthy state and is marked as a sub-healthy battery cell; S3.3, according to the severity of the health status of the sub-healthy battery unit, use the fuzzy comprehensive evaluation method to determine its isolation priority and generate the corresponding isolation instruction; The analysis method of fuzzy comprehensive evaluation method is: Setting factor set ,in The voltage deviation , is the internal resistance change rate , is the charge and discharge capacity decay rate ; Evaluation Set , Indicates low priority isolation, Indicates medium priority isolation, Indicates high priority isolation; Determine the fuzzy relationship matrix : Indicates the voltage deviation Isolate low priority The membership degree of hour, , which is constructed ; Determine the weight set , assuming that the voltage deviation degree is determined by the hierarchical analysis method , internal resistance change rate , charge and discharge capacity decay rate ; Fuzzy comprehensive evaluation results , ( ),according to The maximum membership principle in determines the isolation priority of sub-healthy battery cells; S3.4, sending an isolation instruction to the control unit of the energy storage system, and the control unit cuts off the output port of the sub-healthy battery unit according to the instruction, thereby achieving electrical isolation between the sub-healthy battery unit and the energy storage system; S3.5. Through the battery management system, the isolated sub-healthy battery cells are removed from the logical structure of the energy storage system, and the working status of the remaining healthy battery cells is adjusted to ensure the overall performance of the energy storage system; S3.6. For the isolated sub-healthy battery cells, a degradation prediction algorithm is used to estimate the deterioration trend of their health status in combination with their historical monitoring data, and health status warning information is generated; S3.7. Based on the health status warning information and the design life of the battery unit, determine the maintenance or replacement time of the sub-healthy battery unit and formulate a battery maintenance and replacement plan to avoid its continued deterioration and affect the system.

[0010] Preferably, the sub-steps of step S4 are: S4.1. Obtain real-time health data of each battery cell in the energy storage system, analyze the data through a machine learning algorithm, and obtain a health rating result for each battery cell; S4.2. Based on the health rating results, a decision tree algorithm is used to determine the health status of each battery cell, and battery cells whose health status is lower than a preset threshold are marked as sub-healthy; S4.3. For the battery cells marked as sub-healthy, the battery cells are unloaded from the working circuit of the energy storage system by switching the switch of the control circuit, and the charging and discharging operations thereof are stopped; S4.4. For battery cells in good health, a clustering algorithm is used to classify the battery cells into different health levels according to their health rating results; S4.5. During the operation of the energy storage system, the working status of battery cells with different health levels is dynamically adjusted, and battery cells with higher health levels are preferentially used for charging and discharging operations; S4.6, when it is necessary to unload the sub-healthy battery cells, the optimal unloading sequence is determined by comprehensively considering the health level and charge state of the battery cells through the reinforcement learning algorithm; S4.7. According to the optimized unloading sequence, the unloading process of the battery cells is controlled, and the working status of healthy battery cells is dynamically adjusted to ensure the overall output performance of the energy storage system is stable.

[0011] Preferably, the sub-steps of step S5 are: S5.1. Collect initial battery charge data and battery capacity data of all battery units in the energy storage system; S5.2. Use the measurement algorithm to obtain the voltage data, internal resistance data and operating temperature data of all battery cells, and synchronously transmit all the obtained data to the health model; S5.3, determining battery health level data; According to the health level data and battery capacity data of each battery, the battery state of charge data; By calculating the formula SOH=(Q actual / Q rated )100%, where SOH represents the battery health state, Q actual Represents the maximum capacity of the battery currently available, Q rated Represents the rated capacity of the battery; Use the obtained Q actual Get the Q of each battery rated Data, using a recurrent neural network algorithm to map the pre-established health status relationship mapping library, and input the obtained relevant data into the expert system to obtain the initial battery cell health level data; S5.4. Obtain a sorted list of health levels of all battery cells, and obtain a battery cell sequence set with a maximum health level value by comparing the health levels; S5.5. Determine the discharge order of all battery cells in the battery cell sequence set with the maximum health level value by numerical sorting; S5.6. Discharging with a first depth of discharge through a battery cell sequence set with the largest health level according to the discharge sequence list; S5.7. Set a preset threshold V for each battery cell in the energy storage system min ; The operating monitoring module detects the actual voltage value of the battery cell during the discharge process. If the voltage value of any battery cell is lower than the set threshold value V min , it is determined that the voltage of the battery cell is abnormal; Collect ambient temperature sensor data and operating temperature sensor data, and use Kalman filter fusion algorithm to obtain the fusion value of ambient temperature and operating temperature; If the fusion value is greater than the temperature preset value X, the temperature management module is activated to adjust the operating temperature of the energy storage system. If the fusion value is less than the preset value Y, the real-time information is stored in the database. S5.8, cyclically determine the temperature status; For a battery cell whose voltage is lower than the threshold V min When this state occurs, the preset rule base is updated according to the charge and discharge rate, cycle number, and battery internal resistance value information; S5.9, using a logistic regression algorithm to obtain updated health level data for each battery cell, and storing the updated results in the operation log; The weight levels of the health level update data of different battery cells are obtained through the number of cycles and the charge and discharge rate. When the battery cell with the lowest health level is discharged, a discharge power reduction command will be obtained according to the weight data corresponding to the battery cell. S5.10. For the battery cells that are determined to be unhealthy, a discharging stop operation instruction is obtained.

[0012] Preferably, in step S5.3, the process of the recurrent neural network algorithm is as follows: Hidden state update formula: , , yes The weight matrix of the input to the hidden layer, yes The weight matrix from hidden layer to hidden layer, yes dimensional hidden layer bias vector; Output formula: , yes The weight matrix from the hidden layer to the output layer, the output corresponds to the three categories of health, sub-health, and unhealthy. yes dimensional output layer bias vector; The logistic regression algorithm process is as follows: set up is the input feature vector, where is the voltage change , is the change in internal resistance , is the charge and discharge rate , are model parameters; The prediction function of the logistic regression model is: ; The likelihood function is: ,in is the sample size, For sample The true category label of is solved by the gradient descent optimization algorithm .

[0013] Preferably, the sub-steps of step S6 are: S6.1. Obtain historical operating data of the battery unit and preprocess the data, including operations such as data cleaning, feature extraction, and data standardization, to obtain a data set suitable for machine learning algorithm analysis; S6.2. Perform cluster analysis on the preprocessed historical operation data using the K-means clustering algorithm, divide the data into different categories according to the health status of the battery cells, and obtain the characteristic pattern of each category; S6.3. Construct a support vector machine classification model based on the characteristic patterns of different health levels obtained by cluster analysis, use the characteristic patterns as the input of the model, and the health level as the output of the model to train the health status assessment model; S6.4, obtaining newly collected battery unit operation data, performing the same preprocessing operation on the data as the historical data, and obtaining new data that conforms to the model input format; S6.5, input the pre-processed new data into the trained health status assessment model, and the model predicts the health level of the battery cell according to the characteristic pattern of the input data; S6.6, if the predicted health level is lower than the preset threshold, the battery cell is judged to be in an unhealthy state and needs maintenance or replacement; S6.7. If the predicted health level is higher than the threshold, the battery cell is judged to be in a healthy state and can continue to be used; S6.8. Based on the health status prediction results of the battery cells, classify and manage the battery cells and formulate corresponding maintenance strategies to ensure the reliable operation of the battery system; at the same time, add the newly collected data to the historical data set, regularly update and optimize the health status assessment model to improve the prediction accuracy of the model.

[0014] Preferably, in step S6.2, the K-means clustering algorithm process is as follows: Set data set ,in , we need to aggregate the data into Class, corresponding to the three levels of health, sub-health and unhealthy; Random Initialization Cluster Centers , They represent three levels: healthy, sub-healthy and unhealthy. ,in , , They are the estimated values ​​of voltage, internal resistance, and charge / discharge capacity of the cluster center corresponding to the initial setting of health level; Repeat the following steps until convergence: For each data point , calculate it to each cluster center Distance , Euclidean distance ,Will Assigned to the class of the closest cluster center ; For each class , update the cluster center ,in Is a class The number of data points in .

[0015] Preferably, the sub-steps of step S7 are: S7.1, obtaining health grade data of each battery cell in the energy storage power station, inputting the health grade data into a pre-built battery health grade assessment model, and obtaining a health grade score of each battery cell; S7.2. According to the health level score of each battery cell, the battery cells are grouped using a clustering algorithm, and battery cells with similar health level scores are grouped into the same group; S7.3. For each group of battery cells, a scheduling priority parameter is determined. The battery cell group with a higher health level score has a larger scheduling priority parameter. S7.4. Obtain grid dispatch demand data, combine dispatch priority parameters of each battery unit group, and solve the optimal energy storage system operation strategy through optimization algorithm. While meeting grid dispatch demand, prioritize dispatch of battery unit groups with high health level; S7.5. During the actual operation of the energy storage system, the health status data of each battery cell is obtained in real time. When it is determined that the health level of a battery cell has significantly decreased, the scheduling priority parameters of the group to which the battery cell belongs are dynamically adjusted to reduce the frequency of use of the battery cell group; S7.6. Regularly evaluate the overall operation of the energy storage system, count the health level change trends of each group of battery cells, determine the effect of the optimized operation strategy on extending the service life of the battery cells, and dynamically update the scheduling priority parameters accordingly; S7.7. Continue to optimize the energy storage system operation strategy. On the premise of ensuring that the grid dispatch requirements are met, balance the usage intensity of each battery unit group to the maximum extent possible, slow down the decay rate of the battery unit health level, and thus improve the overall economy and reliability of the energy storage power station.

[0016] The present invention can achieve the following beneficial effects: 1. The present invention establishes an accurate battery health status grading system for energy storage power stations. By comprehensively considering key parameters such as voltage, internal resistance, and charge and discharge capacity, the battery cells are accurately divided into three levels: healthy, sub-healthy, and unhealthy. The corresponding operating parameter standards and threshold ranges are set to improve the accuracy of battery health status assessment.

[0017] 2. The present invention develops efficient health detection systems and algorithm models, such as support vector machine algorithms, fuzzy comprehensive evaluation methods, clustering algorithms, recurrent neural network algorithms, and logistic regression algorithms, etc. These algorithms are used to conduct in-depth analysis of the historical operating data and real-time monitoring data of battery cells, to achieve real-time and accurate prediction and classification of battery health status, and to promptly detect sub-healthy and unhealthy battery cells.

[0018] 3. The present invention designs a reasonable isolation and processing mechanism for sub-healthy battery cells, automatically generates higher-priority unloading instructions according to the severity of their health status, isolates them from the energy storage system, and uses a degradation prediction algorithm to estimate the deterioration trend of their health status, formulate an effective maintenance or replacement plan, avoid their negative impact on the overall performance of the energy storage system, and prevent the occurrence of safety accidents.

[0019] 3. During the operation of the energy storage system, especially during the discharge operation, the present invention dynamically adjusts the working state according to the health level of the battery cell, gives priority to battery cells with higher health levels to participate in the discharge, reasonably controls the discharge power of sub-healthy and unhealthy battery cells, and determines the optimal battery cell unloading sequence through the reinforcement learning algorithm to ensure the overall output performance stability of the energy storage system.

[0020] 4. The present invention dynamically adjusts the operation strategy of the energy storage system based on the health level of the battery unit, combines it with the grid dispatching requirements, solves the optimal operation strategy through the optimization algorithm, balances the usage intensity of each group of battery units, maximizes the service life of the battery units, and improves the overall economy and reliability of the energy storage power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0022] The preferred solution is Figure 1 As shown, a method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station includes the following steps: S1. According to the pre-established energy storage power station battery health status classification system, the battery cells are divided into three levels: healthy, sub-healthy and unhealthy. For each level, the corresponding operating parameter standards are set, including the threshold range of key parameters such as voltage, internal resistance and charge and discharge capacity; specifically: S1.1. Obtain the real-time operating parameters of the battery unit, including key parameters such as voltage, internal resistance, and charge and discharge capacity, based on the pre-established energy storage power station battery health status classification system; S1.2, comparing the acquired real-time operating parameters of the battery unit with the preset health level threshold range to determine the health status of each battery unit; S1.3. If the real-time operating parameters of the battery cell are all within the threshold range of the health level, the health state of the battery cell is determined as the health level; S1.4. If one or more of the real-time operating parameters of the battery cell exceed the threshold range of the health level, but are all within the threshold range of the sub-health level, the health state of the battery cell is determined to be a sub-health level; S1.5. If one or more of the real-time operating parameters of the battery cell exceed the threshold range of the sub-health level, the health state of the battery cell is determined to be an unhealthy level; S1.6. Based on the determined health status of the battery cells, the corresponding operating parameter standards are adopted to control and optimize the charging and discharging process of the battery cells to ensure the safe and stable operation of the energy storage power station; S1.7. The historical operating data of the battery cell is trained through the support vector machine algorithm to obtain a discriminant model for health status classification, which is used to predict and classify the health status of new battery cell operating data, thereby improving the accuracy and efficiency of battery health status assessment.

[0023] The training process of the support vector machine algorithm on the historical operation data of the battery unit is as follows: Assume training data set ,in The voltage of the battery cell 、Internal resistance and charge and discharge capacity The characteristic vectors of key parameters such as ; They correspond to the three category labels of unhealthy, sub-healthy and healthy respectively.

[0024] Its objective function is: ,in are the weight vector components corresponding to voltage, internal resistance, and charge and discharge capacity characteristics, respectively. is the hyperplane intercept, It can be set as an empirical value first, such as (It can be adjusted according to the actual classification effect), is a slack variable. The constraints are: , For example, for health ( ),but S2. Use the health detection system to monitor the operating status information of each battery unit in real time, obtain real-time data of key parameters such as voltage, internal resistance, charge and discharge capacity, and judge the health level of each battery unit through internal comparison and analysis; specifically: S2.1. Obtain the real-time operating status data of each battery unit, including key parameters such as voltage, internal resistance, and charge and discharge capacity; S2.2, transmitting the acquired real-time data to the health detection system and storing it in the corresponding database; S2.3, pre-process the stored real-time data, including data cleaning and normalization, to improve data quality; S2.4. Based on the pre-processed real-time data, a correlation analysis algorithm is used to calculate the correlation coefficients between the key parameters and determine the degree of correlation between the parameters; S2.5. Construct a battery health assessment model based on the correlation coefficients of key parameters, and comprehensively consider the impact of each parameter on the battery health status; S2.6. Using the constructed battery health assessment model, the health status of each battery cell group is assessed to obtain the health level of each battery cell; S2.7. Generate a battery health report based on the assessed battery cell health level and send the report to relevant managers so that they can keep track of the battery health status and formulate operation and maintenance strategies in advance. S3. For sub-healthy battery cells detected, the control system automatically generates a higher priority unloading instruction, cuts off the output of the battery cell, isolates it from the energy storage system, and prevents the sub-healthy battery cell from further deteriorating; specifically: S3.1. According to the preset battery health status assessment model, each battery unit in the energy storage system is monitored in real time to obtain key parameters reflecting its health status; S3.2, if the health status parameter of a battery cell is monitored to be beyond a predetermined threshold range, the battery cell is determined to be in a sub-healthy state and is marked as a sub-healthy battery cell; S3.3, according to the severity of the health status of the sub-healthy battery unit, use the fuzzy comprehensive evaluation method to determine its isolation priority and generate the corresponding isolation instruction; The analysis method of fuzzy comprehensive evaluation method is: Setting factor set ,in The voltage deviation (Ratio of deviation relative to the healthy level voltage threshold range), is the internal resistance change rate (Proportional change compared to the initial internal resistance), is the charge and discharge capacity decay rate . Evaluation Set , Indicates low priority isolation, Indicates medium priority isolation, Indicates high priority quarantine. Determine the fuzzy relationship matrix :For example Indicates the voltage deviation Isolate low priority The degree of membership can be obtained based on expert experience or statistical analysis of a large amount of historical data. hour, etc., thus construct . Determine the weight set , assuming that the voltage deviation degree is determined by the hierarchical analysis method , internal resistance change rate , charge and discharge capacity decay rate . Fuzzy comprehensive evaluation results , ( ),according to The maximum membership principle in determines the isolation priority of sub-healthy battery cells.

[0025] S3.4, sending an isolation instruction to the control unit of the energy storage system, and the control unit cuts off the output port of the sub-healthy battery unit according to the instruction to achieve electrical isolation between the sub-healthy battery unit and the energy storage system; S3.5. Through the battery management system, the isolated sub-healthy battery cells are removed from the logical structure of the energy storage system, and the working status of the remaining healthy battery cells is adjusted to ensure the overall performance of the energy storage system; S3.6. For the isolated sub-healthy battery cells, a degradation prediction algorithm is used to estimate the deterioration trend of their health status in combination with their historical monitoring data, and health status warning information is generated; S3.7. According to the health status warning information and the design life of the battery unit, determine the maintenance or replacement time of the sub-healthy battery unit, form a battery maintenance and replacement plan, and avoid its continued deterioration affecting the system. S4. While unloading the sub-healthy battery unit, dynamically adjust the working status of other healthy battery units, and determine the unloading order of different battery units according to the real-time health rating results to ensure the overall output performance of the energy storage system; specifically: S4.1. Obtain real-time health data of each battery cell in the energy storage system, analyze the data through a machine learning algorithm, and obtain a health rating result for each battery cell; S4.2. Based on the health rating results, a decision tree algorithm is used to determine the health status of each battery cell, and battery cells whose health status is lower than a preset threshold are marked as sub-healthy; S4.3. For the battery cells marked as sub-healthy, the battery cells are unloaded from the working circuit of the energy storage system by switching the switch of the control circuit, and the charging and discharging operations thereof are stopped; S4.4. For battery cells in good health, a clustering algorithm is used to classify the battery cells into different health levels according to their health rating results; S4.5. During the operation of the energy storage system, the working status of battery cells with different health levels is dynamically adjusted, and battery cells with higher health levels are preferentially used for charging and discharging operations; S4.6, when it is necessary to unload the sub-healthy battery cells, the optimal unloading sequence is determined by comprehensively considering factors such as the health level and charge state of the battery cells through the reinforcement learning algorithm; S4.7. According to the optimized unloading sequence, control the unloading process of the battery cells, and dynamically adjust the working status of healthy battery cells to ensure the stability of the overall output performance of the energy storage system. S5. When the energy storage system needs to discharge, give priority to battery cells with higher health levels to participate in the discharge. For battery cells assessed as sub-healthy and unhealthy, the control system gradually reduces their discharge power until the discharge is completely stopped; specifically: S5.1. Collect initial battery charge data and battery capacity data of all battery units in the energy storage system; S5.2. Use the measurement algorithm to obtain the voltage data, internal resistance data and operating temperature data of all battery cells, and synchronously transmit all the obtained data to the health model; S5.3, determining battery health level data; According to the health level data and battery capacity data of each battery, the battery state of charge data; By calculating the formula SOH=(Q actual / Q rated )100%, where SOH represents the battery health state, Q actual Represents the maximum capacity of the battery currently available, Q rated Represents the rated capacity of the battery; Use the obtained Q actual Get the Q of each battery rated Data, using a recurrent neural network algorithm to map the pre-established health status relationship mapping library, and input the obtained relevant data into the expert system to obtain the initial battery cell health level data; The process of the recurrent neural network algorithm is as follows: Hidden state update formula: ,here (respectively time battery cell voltage, internal resistance, charge and discharge capacity), yes The weight matrix of the input to the hidden layer ( is the number of neurons in the hidden layer), yes The weight matrix from hidden layer to hidden layer, yes dimensional hidden layer bias vector.

[0026] Output formula: , yes The weight matrix from the hidden layer to the output layer (the output corresponds to the three categories of health, sub-health, and unhealthy), yes dimensional output layer bias vector.

[0027] S5.4. Obtain a sorted list of health levels of all battery cells, and obtain a battery cell sequence set with a maximum health level value by comparing the health levels; S5.5. Determine the discharge order of all battery cells in the battery cell sequence set with the maximum health level value by numerical sorting; S5.6. Discharging with a first depth of discharge through a battery cell sequence set with the largest health level according to the discharge sequence list; S5.7. Set a preset threshold V for each battery cell in the energy storage system min ; The operating monitoring module detects the actual voltage value of the battery cell during the discharge process. If the voltage value of any battery cell is lower than the set threshold value V min , it is determined that the voltage of the battery cell is abnormal; Collect ambient temperature sensor data and operating temperature sensor data, and use Kalman filter fusion algorithm to obtain the fusion value of ambient temperature and operating temperature; If the fusion value is greater than the temperature preset value X, the temperature management module is activated to adjust the operating temperature of the energy storage system. If the fusion value is less than the preset value Y, the real-time information is stored in the database. S5.8, cyclically determine the temperature status; For a battery cell whose voltage is lower than the threshold V min When this state occurs, the preset rule base is updated according to the charge and discharge rate, cycle number, and battery internal resistance value information; S5.9, using a logistic regression algorithm to obtain updated health level data for each battery cell, and storing the updated results in the operation log; The weight levels of the health level update data of different battery cells are obtained through the number of cycles and the charge and discharge rate. When the battery cell with the lowest health level is discharged, a discharge power reduction command will be obtained according to the weight data corresponding to the battery cell. The logistic regression algorithm process is as follows: set up is the input feature vector, where is the voltage change , is the change in internal resistance , is the charge and discharge rate , is the model parameter.

[0028] The prediction function of the logistic regression model is: . The likelihood function is: ,in is the sample size, For sample The true category label ( or ,For example It means unhealthy. indicates health or sub-health), solved by optimization algorithms such as gradient descent .

[0029] In practical applications, in-depth research and a large amount of experimental data are needed to accurately determine these parameters to ensure the accuracy and effectiveness of the algorithm in battery health assessment and early warning.

[0030] S5.10. Obtain a stop discharge operation instruction for the battery cell that is determined to be unhealthy; S6. Use a machine learning algorithm to analyze the historical operation data of the battery cell, identify the characteristic patterns of different health levels through a clustering algorithm, and establish a health status assessment model for predicting and classifying the health level of the newly collected battery cell data; specifically: S6.1. Obtain historical operating data of the battery unit and preprocess the data, including data cleaning, feature extraction, and data standardization, to obtain a data set suitable for machine learning algorithm analysis; S6.2. Perform cluster analysis on the preprocessed historical operation data using the K-means clustering algorithm, divide the data into different categories according to the health status of the battery cells, and obtain the characteristic pattern of each category; The K-means clustering algorithm process is as follows: Set data set ,in (voltage, internal resistance, and charge and discharge capacity of the battery cell, respectively). Category (corresponding to the three levels of healthy, sub-healthy and unhealthy).

[0031] Random Initialization Cluster Centers ,For example ,in , , They are the estimated values ​​of voltage, internal resistance, and charge / discharge capacity of the initial corresponding health level cluster center (which can be obtained based on the average value of a large number of healthy battery cell data). Similarly, (corresponding to sub-health) and (Corresponding to unhealthy). Repeat the following steps until convergence: For each data point , calculate it to each cluster center Distance , such as the Euclidean distance ,Will Assigned to the class of the closest cluster center - For each class , update the cluster center ,in Is a class The number of data points in .

[0032] S6.3. Construct a support vector machine classification model based on the characteristic patterns of different health levels obtained by cluster analysis, use the characteristic patterns as the input of the model, and the health level as the output of the model to train the health status assessment model; S6.4, obtaining newly collected battery unit operation data, performing the same preprocessing operation on the data as the historical data, and obtaining new data that conforms to the model input format; S6.5, input the pre-processed new data into the trained health status assessment model, and the model predicts the health level of the battery cell according to the characteristic pattern of the input data; S6.6, if the predicted health level is lower than the preset threshold, the battery cell is judged to be in an unhealthy state and needs maintenance or replacement; S6.7. If the predicted health level is higher than the threshold, the battery cell is judged to be in a healthy state and can continue to be used; S6.8. According to the health status prediction results of the battery cells, classify and manage the battery cells and formulate corresponding maintenance strategies to ensure the reliable operation of the battery system; At the same time, the newly collected data is added to the historical data set, and the health status assessment model is regularly updated and optimized to improve the prediction accuracy of the model. S7. Dynamically adjust the operation strategy of the energy storage system based on the health level of the battery unit to maximize the service life of the battery unit while meeting the grid dispatch requirements, and improve the overall economy and reliability of the energy storage power station. Specifically: S7.1. Obtain the health grade data of each battery unit in the energy storage power station, input the health grade data into a pre-built battery health grade assessment model, and obtain the health grade score of each battery unit; S7.2. According to the health level score of each battery cell, the battery cells are grouped using a clustering algorithm, and battery cells with similar health level scores are grouped into the same group; S7.3. For each group of battery cells, a scheduling priority parameter is determined. The battery cell group with a higher health level score has a larger scheduling priority parameter. S7.4. Obtain grid dispatch demand data, combine dispatch priority parameters of each battery unit group, and solve the optimal energy storage system operation strategy through optimization algorithm. While meeting grid dispatch demand, prioritize dispatch of battery unit groups with high health level; S7.5. During the actual operation of the energy storage system, the health status data of each battery cell is obtained in real time. When it is determined that the health level of a battery cell has significantly decreased, the scheduling priority parameters of the group to which the battery cell belongs are dynamically adjusted to reduce the frequency of use of the battery cell group; S7.6. Regularly evaluate the overall operation of the energy storage system, count the health level change trends of each group of battery cells, determine the effect of the optimized operation strategy on extending the service life of the battery cells, and dynamically update the scheduling priority parameters accordingly; S7.7. Continue to optimize the energy storage system operation strategy. On the premise of ensuring that the grid dispatch requirements are met, balance the usage intensity of each battery unit group to the maximum extent possible, slow down the decay rate of the battery unit health level, and thus improve the overall economy and reliability of the energy storage power station.

[0033] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for early warning of battery health and remaining life in medium and large electrochemical energy storage power stations, characterized in that The following steps are involved: S1. According to the pre-established energy storage power station battery health status classification system, the battery cells are divided into three levels: healthy, sub-healthy and unhealthy. For each level, corresponding operating parameter standards are set. The operating parameter standards include the threshold ranges of voltage, internal resistance and charge and discharge capacity; S2. Use the health detection system to monitor the operating status information of each battery unit in real time, obtain the real-time data of voltage, internal resistance, and charge and discharge capacity, and judge the health level of each battery unit through internal comparison and analysis; S3. For the sub-healthy battery cells detected, the control system automatically generates a higher priority unloading instruction, and isolates the sub-healthy battery cells from the energy storage system by cutting off the output of the battery cells, thereby preventing the sub-healthy battery cells from further deteriorating; S4. While unloading sub-healthy battery cells, dynamically adjust the working status of other healthy battery cells, and determine the unloading order of different battery cells according to the real-time health rating results to ensure the overall output performance of the energy storage system; S5. When the energy storage system needs to perform a discharge operation, the battery cells with a higher health level are preferentially selected to participate in the discharge. For the battery cells assessed as sub-healthy and unhealthy, the control system gradually reduces their discharge power until the discharge is completely stopped; S6. Analyze the historical operation data of the battery cell using a machine learning algorithm, identify the characteristic patterns of different health levels through a clustering algorithm, and establish a health status assessment model for predicting and classifying the health level of the newly collected battery cell data; S7. Dynamically adjust the operation strategy of the energy storage system based on the health level of the battery cells to maximize the service life of the battery cells while meeting the grid dispatch requirements, thereby improving the overall economy and reliability of the energy storage power station.

2. According to claim 1, a method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station is characterized by: The sub-steps of step S1 are: S1.

1. According to the pre-established energy storage power station battery health status classification system, obtain the real-time operating parameters of the battery unit, including key parameters such as voltage, internal resistance and charge and discharge capacity; S1.2, comparing the acquired real-time operating parameters of the battery unit with the preset health level threshold range to determine the health status of each battery unit; S1.

3. If the real-time operating parameters of the battery cell are all within the threshold range of the health level, the health state of the battery cell is determined as the health level; S1.

4. If one or more of the real-time operating parameters of the battery cell exceed the threshold range of the health level, but are all within the threshold range of the sub-health level, the health state of the battery cell is determined to be a sub-health level; S1.

5. If one or more of the real-time operating parameters of the battery cell exceed the threshold range of the sub-health level, the health state of the battery cell is determined to be an unhealthy level; S1.

6. Based on the determined health status of the battery cells, the corresponding operating parameter standards are adopted to control and optimize the charging and discharging process of the battery cells to ensure the safe and stable operation of the energy storage power station; S1.

7. The historical operation data of the battery cell is trained by the support vector machine algorithm to obtain a discriminant model for health status classification, which is used to predict and classify the health status of new battery cell operation data, thereby improving the accuracy and efficiency of battery health status assessment; The training process of the support vector machine algorithm on the historical operation data of the battery unit is as follows: Assume training data set ,in The voltage of the battery cell 、Internal resistance and charge and discharge capacity The characteristic vector of ; They correspond to the three category labels of unhealthy, sub-healthy, and healthy respectively; Its objective function is: ,in are the weight vector components corresponding to voltage, internal resistance, and charge and discharge capacity characteristics, respectively. is the hyperplane intercept, You can set it as an experience value first. is the slack variable; The constraints are: .

3. According to claim 1, a method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station is characterized by: The sub-steps of step S2 are: S2.1, obtaining real-time operating status data of each group of battery cells, the operating status data including key parameters such as voltage, internal resistance and charge and discharge capacity; S2.2, transmitting the acquired real-time data to the health detection system and storing it in the corresponding database; S2.3, preprocessing the stored real-time data, including data cleaning and normalization operations to improve data quality; S2.

4. Based on the pre-processed real-time data, a correlation analysis algorithm is used to calculate the correlation coefficients between the key parameters and determine the degree of correlation between the parameters; S2.

5. Construct a battery health assessment model based on the correlation coefficients of each key parameter, and comprehensively consider the impact of each key parameter on the battery health status; S2.

6. Using the constructed battery health assessment model, the health status of each battery cell group is assessed to obtain the health level of each battery cell; S2.

7. Generate a battery health report based on the evaluated battery cell health level and send the report to relevant managers so that they can understand the battery health status in time and formulate operation and maintenance strategies in advance.

4. A method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station according to claim 1, characterized in that: The sub-steps of step S3 are: S3.1, according to a preset battery health status assessment model, each battery cell in the energy storage system is monitored in real time to obtain key parameters reflecting the health status of each battery cell; S3.2, if the health status parameter of a battery cell is monitored to be beyond a predetermined threshold range, the battery cell is determined to be in a sub-healthy state and is marked as a sub-healthy battery cell; S3.3, according to the severity of the health status of the sub-healthy battery unit, use the fuzzy comprehensive evaluation method to determine its isolation priority and generate the corresponding isolation instruction; The analysis method of fuzzy comprehensive evaluation method is: Setting factor set ,in The voltage deviation , is the internal resistance change rate , is the charge and discharge capacity decay rate ; Evaluation Set , Indicates low priority isolation, Indicates medium priority isolation, Indicates high priority isolation; Determine the fuzzy relationship matrix : Indicates the voltage deviation Isolate low priority The membership degree of hour, , which is constructed ; Determine the weight set , assuming that the voltage deviation degree is determined by the hierarchical analysis method , internal resistance change rate , charge and discharge capacity decay rate ; Fuzzy comprehensive evaluation results , ( ),according to The maximum membership principle in determines the isolation priority of sub-healthy battery cells; S3.4, sending an isolation instruction to the control unit of the energy storage system, and the control unit cuts off the output port of the sub-healthy battery unit according to the instruction, thereby achieving electrical isolation between the sub-healthy battery unit and the energy storage system; S3.

5. Through the battery management system, the isolated sub-healthy battery cells are removed from the logical structure of the energy storage system, and the working status of the remaining healthy battery cells is adjusted to ensure the overall performance of the energy storage system; S3.

6. For the isolated sub-healthy battery cells, a degradation prediction algorithm is used to estimate the deterioration trend of their health status in combination with their historical monitoring data, and health status warning information is generated; S3.

7. Based on the health status warning information and the design life of the battery unit, determine the maintenance or replacement time of the sub-healthy battery unit and formulate a battery maintenance and replacement plan to avoid its continued deterioration and affect the system.

5. The method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station according to claim 1, characterized in that: The sub-steps of step S4 are: S4.

1. Obtain real-time health data of each battery cell in the energy storage system, analyze the data through a machine learning algorithm, and obtain a health rating result for each battery cell; S4.

2. Based on the health rating results, a decision tree algorithm is used to determine the health status of each battery cell, and battery cells whose health status is lower than a preset threshold are marked as sub-healthy; S4.

3. For the battery cells marked as sub-healthy, the battery cells are unloaded from the working circuit of the energy storage system by switching the switch of the control circuit, and the charging and discharging operations thereof are stopped; S4.

4. For battery cells in good health, a clustering algorithm is used to classify the battery cells into different health levels according to their health rating results; S4.

5. During the operation of the energy storage system, the working status of battery cells with different health levels is dynamically adjusted, and battery cells with higher health levels are preferentially used for charging and discharging operations; S4.6, when it is necessary to unload the sub-healthy battery cells, the optimal unloading sequence is determined by comprehensively considering the health level and charge state of the battery cells through the reinforcement learning algorithm; S4.

7. According to the optimized unloading sequence, the unloading process of the battery cells is controlled, and the working status of healthy battery cells is dynamically adjusted to ensure the overall output performance of the energy storage system is stable.

6. A method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station according to claim 1, characterized in that: The sub-steps of step S5 are: S5.

1. Collect initial battery charge data and battery capacity data of all battery units in the energy storage system; S5.

2. Use the measurement algorithm to obtain the voltage data, internal resistance data and operating temperature data of all battery cells, and synchronously transmit all the obtained data to the health model; S5.3, determining battery health level data; According to the health level data and battery capacity data of each battery, the battery state of charge data; By calculating the formula SOH=(Q actual / Q rated )100%, where SOH represents the battery health state, Q actual Represents the maximum capacity of the battery currently available, Q rated Represents the rated capacity of the battery; Use the obtained Q actual Get the Q of each battery rated Data, using a recurrent neural network algorithm to map the pre-established health status relationship mapping library, and input the obtained relevant data into the expert system to obtain the initial battery cell health level data; S5.

4. Obtain a sorted list of health levels of all battery cells, and obtain a battery cell sequence set with a maximum health level value by comparing the health levels; S5.

5. Determine the discharge order of all battery cells in the battery cell sequence set with the maximum health level value by numerical sorting; S5.

6. Discharging with a first depth of discharge through a battery cell sequence set with the largest health level according to the discharge sequence list; S5.

7. Set a preset threshold V for each battery cell in the energy storage system min ; The operating monitoring module detects the actual voltage value of the battery cell during the discharge process. If the voltage value of any battery cell is lower than the set threshold value V min , it is determined that the voltage of the battery cell is abnormal; Collect ambient temperature sensor data and operating temperature sensor data, and use Kalman filter fusion algorithm to obtain the fusion value of ambient temperature and operating temperature; If the fusion value is greater than the temperature preset value X, the temperature management module is activated to adjust the operating temperature of the energy storage system. If the fusion value is less than the preset value Y, the real-time information is stored in the database. S5.8, cyclically determine the temperature status; For a battery cell whose voltage is lower than the threshold V min When this state occurs, the preset rule base is updated according to the charge and discharge rate, cycle number, and battery internal resistance value information; S5.9, using a logistic regression algorithm to obtain updated health level data for each battery cell, and storing the updated results in the operation log; The weight levels of the health level update data of different battery cells are obtained through the number of cycles and the charge and discharge rate. When the battery cell with the lowest health level is discharged, a discharge power reduction command will be obtained according to the weight data corresponding to the battery cell. S5.

10. For the battery cells that are determined to be unhealthy, a discharging stop operation instruction is obtained.

7. A method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station according to claim 6, characterized in that: In step S5.3, the process of the recurrent neural network algorithm is as follows: Hidden state update formula: , , yes The weight matrix of the input to the hidden layer, yes The weight matrix from hidden layer to hidden layer, yes dimensional hidden layer bias vector; Output formula: , yes The weight matrix from the hidden layer to the output layer, the output corresponds to the three categories of health, sub-health, and unhealthy. yes dimensional output layer bias vector; The logistic regression algorithm process is as follows: set up is the input feature vector, where is the voltage change , is the change in internal resistance , is the charge and discharge rate , are model parameters; The prediction function of the logistic regression model is: ; The likelihood function is: ,in is the sample size, For sample The true category label of is solved by the gradient descent optimization algorithm .

8. The method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station according to claim 1, characterized in that: The sub-steps of step S6 are: S6.

1. Obtain historical operating data of the battery unit and preprocess the data, including operations such as data cleaning, feature extraction, and data standardization, to obtain a data set suitable for machine learning algorithm analysis; S6.

2. Perform cluster analysis on the preprocessed historical operation data using the K-means clustering algorithm, divide the data into different categories according to the health status of the battery cells, and obtain the characteristic pattern of each category; S6.

3. Construct a support vector machine classification model based on the characteristic patterns of different health levels obtained by cluster analysis, use the characteristic patterns as the input of the model, and the health level as the output of the model to train the health status assessment model; S6.4, obtaining newly collected battery unit operation data, performing the same preprocessing operation on the data as the historical data, and obtaining new data that conforms to the model input format; S6.5, input the pre-processed new data into the trained health status assessment model, and the model predicts the health level of the battery cell according to the characteristic pattern of the input data; S6.6, if the predicted health level is lower than the preset threshold, the battery cell is judged to be in an unhealthy state and needs maintenance or replacement; S6.

7. If the predicted health level is higher than the threshold, the battery cell is judged to be in a healthy state and can continue to be used; S6.

8. Based on the health status prediction results of the battery cells, classify and manage the battery cells and formulate corresponding maintenance strategies to ensure the reliable operation of the battery system; at the same time, add the newly collected data to the historical data set, regularly update and optimize the health status assessment model to improve the prediction accuracy of the model.

9. A method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station according to claim 8, characterized in that: In step S6.2, the K-means clustering algorithm process is as follows: Set data set ,in , we need to aggregate the data into Class, corresponding to the three levels of health, sub-health and unhealthy; Random Initialization Cluster Centers , They represent three levels: healthy, sub-healthy and unhealthy. ,in , , They are the estimated values ​​of voltage, internal resistance, and charge / discharge capacity of the cluster center corresponding to the initial setting of health level; Repeat the following steps until convergence: For each data point , calculate it to each cluster center Distance , Euclidean distance ,Will Assigned to the class of the closest cluster center ; For each class , update the cluster center ,in Is a class The number of data points in .

10. The method for early warning of battery health and remaining life of a medium-to-large electrochemical energy storage power station according to claim 1, characterized in that: The sub-steps of step S7 are: S7.1, obtaining health grade data of each battery unit in the energy storage power station, inputting the health grade data into a pre-built battery health grade assessment model, and obtaining a health grade score of each battery unit; S7.

2. According to the health level score of each battery cell, the battery cells are grouped using a clustering algorithm, and battery cells with similar health level scores are grouped into the same group; S7.

3. For each group of battery cells, a scheduling priority parameter is determined. The battery cell group with a higher health level score has a larger scheduling priority parameter. S7.

4. Obtain grid dispatch demand data, combine dispatch priority parameters of each battery unit group, and solve the optimal energy storage system operation strategy through optimization algorithm. While meeting grid dispatch demand, prioritize dispatch of battery unit groups with high health level; S7.

5. During the actual operation of the energy storage system, the health status data of each battery cell is obtained in real time. When it is determined that the health level of a battery cell has significantly decreased, the scheduling priority parameters of the group to which the battery cell belongs are dynamically adjusted to reduce the frequency of use of the battery cell group; S7.

6. Regularly evaluate the overall operation of the energy storage system, count the health level change trends of each group of battery cells, determine the effect of the optimized operation strategy on extending the service life of the battery cells, and dynamically update the scheduling priority parameters accordingly; S7.

7. Continue to optimize the energy storage system operation strategy. On the premise of ensuring that the grid dispatch requirements are met, balance the usage intensity of each battery unit group to the maximum extent possible, slow down the decay rate of the battery unit health level, and thus improve the overall economy and reliability of the energy storage power station.

Images