Energy storage power station safety early warning system operation method based on battery base layer coupling model

Abstract

The invention discloses an energy storage power station safety early warning system operation method based on a battery base layer coupling model. Comprising the steps that a sensor arranged on a single battery of an energy storage power station collects single battery voltage, single battery temperature and electrolyte leakage detection information, and a battery module detects impedance of a battery through an impedance analyzer to obtain internal resistance change of the module; and judging according to the number of the isolated monomers in combination with a detection numerical value obtained by the electrolyte detection sensor for correspondingly collecting the monomers under a fixed time length, and if the detection numerical value does not have a decreasing trend or the decreasing amplitude is smaller than a fixed value, generating a command packet for starting the spraying device and sending the command packet to a power environment system to execute the command.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an operating method of a safety early warning system for an energy storage power station based on a battery-based coupling model. Background Art

[0002] The present invention relates to safety warning and emergency response technology in the battery management system (BMS) of energy storage power stations, especially the rapid location, risk assessment and response measures for electrolyte leakage in battery cells or modules. This technology is applicable to large-scale energy storage systems, such as grid-level energy storage power stations, to improve the safety, reliability and response speed of the system.

[0003] Traditional battery management systems mainly rely on conventional parameters such as voltage and temperature to monitor battery status, and lack effective early warning mechanisms for sudden failures such as electrolyte leakage. In addition, existing BMSs often fail to accurately locate and respond quickly when facing multiple point failures, leading to the escalation of accidents. In addition, there are the following defects: 1. The existing system has a certain delay in data collection and transmission. Especially in large-scale energy storage power stations, the amount of data in the sensor network is huge, resulting in insufficient real-time performance and affecting the rapid response to sudden failures. 2. It is difficult to accurately locate the specific location of electrolyte leakage with traditional methods, especially in complex systems with multiple modules and multiple monomers, which are prone to misjudgment or omission; 3. Most existing systems rely only on a single parameter (such as voltage, temperature) for fault judgment, and fail to make full use of multi-dimensional data (such as internal resistance change, electrochemical impedance spectroscopy, etc.) for comprehensive analysis, resulting in low fault prediction accuracy; 4. Once a leak is detected, the emergency response mechanism of the existing system is relatively simple, lacking intelligent isolation measures and linkage control, and cannot effectively prevent the accident from further expanding; 5. The existing system fails to fully consider the development trend and diffusion speed of electrolyte leakage, making it difficult to provide scientific emergency rescue guidance, which increases the difficulty of accident handling. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes an operating method of a safety early warning system for an energy storage power station based on a battery-based coupling model, which aims to achieve early warning and precise positioning of electrolyte leakage through multi-dimensional data collection, intelligent analysis and rapid response, and ensure the safe operation of the energy storage power station through intelligent emergency response measures.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions to achieve: A method for operating a safety early warning system for an energy storage power station based on a battery-based coupling model, the method comprising the following steps: a) Sensors deployed in battery cells of energy storage power stations are used to collect battery cell voltage, battery cell temperature and electrolyte leakage detection information; and the battery module detects the impedance of the battery through an impedance analyzer to obtain the change in the module internal resistance; b) Obtain the data packet size of each battery cell voltage, battery cell temperature, electrolyte leakage detection information, and module internal resistance change, calculate the time required to upload the voltage and temperature data of all cells and the internal resistance change data of each module in combination with the throughput specified by the hardware network protocol of the network communication module, and obtain the clock synchronization data packet size through synchronization with the clock deployed on the module, and obtain the communication network delay information of the first data packet according to the time required for data transmission and the size of the clock synchronization packet; c) Use linear regression algorithm to process the changes of battery cell voltage, battery cell temperature and module internal resistance. If the changes of battery cell voltage, temperature and module internal resistance are linear within a certain period of time, the data will be directly input into the leakage risk prediction model for processing to determine whether the battery module has a tendency of electrolyte leakage; d) For modules that detect electrolyte leakage tendency, the battery cells are tested by electrochemical impedance spectroscopy (EIS) to obtain impedance spectrum data, and the obtained impedance spectrum data is compared with the impedance spectrum data of the single battery cells in normal battery status to determine the battery cells corresponding to the impedance spectrum with differences. The impedance spectrum data has a significant frequency spectrum change trend, indicating that multiple batteries have electrolyte leakage tendency; e) Establish a multi-input convolutional network, receive the voltage data, temperature data and electrolyte leakage data of a single battery cell, analyze it in the time domain analysis module, perform wavelet analysis for frequency domain analysis, and build a deep neural network model for leakage judgment by calculating the fast Fourier transform; compare the state change value input by the electrolyte detection sensor with the set threshold, and obtain an electrolyte leakage alarm signal that exceeds the threshold; f) If the electrolyte detection sensor repeatedly issues an alarm signal within a fixed period of time and determines that the electrolyte has a tendency to leak, the slope is obtained by calculating the numerical relationship between the voltage drop amplitude and the temperature rise amplitude within the same sampling time length, and the voltage mutation amplitude and change time point are obtained as the data source for triggering the safety isolation mechanism; g) According to the obtained battery cell position information exceeding the safety threshold, trigger and output the corresponding safety isolation measure signal to the actuator, and the actuator starts the corresponding isolation actuator to cut off the circuit connection of the battery cell according to the received control signal; determine the single battery to be isolated through the safety isolation mechanism, and trigger the corresponding exhaust device to operate; h) According to the number of isolated cells combined with the detection value obtained by the electrolyte detection sensor of the cell under the fixed time length, if the detection value has no decreasing trend or the decreasing amplitude is less than the fixed value, a command packet for starting the spray device is generated and sent to the power environment system to execute the command; i) By judging and analyzing the status of multiple batteries with electrolyte leakage detection, a thermal map of the battery leakage location is obtained. The thermal maps obtained multiple times are input into the same battery leakage accident prediction network model based on the thermal map for calculation. The development trend of the electrolyte leakage tendency and the diffusion speed result data packet are generated according to the prediction results of the previous thermal map combined with the current thermal map. The emergency rescue command package is sent to the fire department for execution according to the speed.

[0006] In a preferred embodiment, step a) further comprises: a.1) Obtain the voltage and temperature of the battery cell and electrolyte leakage detection information, pre-process the collected information, remove abnormal values, and obtain effective detection data; a.2) According to the preset voltage, temperature and leakage threshold, determine whether the working status of each battery cell is normal. If it exceeds the threshold range, it is marked as an abnormal battery cell; a.3) Obtain the internal resistance test result of the battery module and calculate the change rate of the module internal resistance. If the change rate exceeds the preset range, the module performance is determined to be abnormal; a.4) For battery cells marked as abnormal, the support vector machine (SVM) algorithm is used to identify the fault type and obtain the cause of the cell fault; a.5) For battery modules with abnormal performance, the decision tree algorithm is used to determine whether the module needs to be repaired or replaced in combination with the single fault type in the module; a.6) According to the single failure causes and module abnormality judgment results, the association rule mining algorithm is used to analyze the correlation and regularity of battery failures in energy storage power stations; a.7) Based on the fault correlation rules, formulate energy storage power station battery operation and maintenance strategies, including regular inspections, cell replacement and module maintenance, to improve battery reliability.

[0007] In a preferred embodiment, step b) further comprises: b.1) Determine the corresponding data packet size based on the voltage, temperature, electrolyte leakage detection information of the battery cell and the module internal resistance change data; b.2) Obtain the throughput specified by the hardware network protocol of the network communication module, and calculate the transmission time required to upload all monomer data and module internal resistance data in combination with the data packet size; b.3) Obtain the data packet size required for clock synchronization through the clock deployed on the module; b.4) determining whether the communication network delay of the first data packet exceeds a preset threshold value based on the time required for data transmission and the size of the clock synchronization packet; b.5) ​​If the network delay exceeds the threshold, a data compression algorithm is used to compress the uploaded data to reduce the data packet size and shorten the transmission time; b.6) If the network delay does not exceed the threshold, the original data packet is directly uploaded to the server; b.7) After receiving the data packet, the server uses the Kalman filter algorithm to fuse the data according to the timestamp and sequence number of the data packet to obtain accurate battery status information.

[0008] In a preferred embodiment, step c) further comprises: c.1) The acquisition device periodically obtains the voltage value, temperature value of each battery cell in the battery module and the internal resistance value of the entire battery module, and each type of data is recorded as a time series; c.2) Perform data preprocessing on each time series, calculate the change value of each cell voltage series and each cell temperature series within a period of time, and the change value of the module internal resistance series, and determine whether the three types of change value time series data satisfy a linear relationship; c.3) If the voltage change, temperature change and internal resistance change are all linear, samples are continuously intercepted from the processed three types of change value data sequences according to the fixed window length through the sliding window technology to obtain multiple sample sets, each of which is a subsequence containing three change value data; c.4) Use the principal component analysis (PCA) algorithm to process the three types of change value data in each sample set. Only one main component is extracted from each set of data as the key feature representing the characteristics of the set, and finally a feature vector group is obtained; c.5) After obtaining the feature vector group, combined with the operation history of the battery module, search for the historical operation stage similar to the feature vector group, analyze whether there is an electrolyte leakage record of the battery module after the corresponding historical operation stage, and construct a binary classification mapping relationship between the characteristics of each stage and the leakage state label to obtain a data set for judging the battery leakage risk; c.6) Train a long short-term memory network (LSTM) for binary classification based on the constructed data set, and obtain the probability value of whether the corresponding battery module is in an electrolyte leakage state based on the LSTM; c.7) If the probability value of the battery module leaking electrolyte exceeds the set threshold, an alarm is issued; if the probability value is lower than the set threshold, it is considered that the battery module has no leakage risk temporarily.

[0009] In a preferred embodiment, step d) further comprises: d.1) Obtain electrochemical impedance spectroscopy data of all single cells in the battery module to form an initial data set. The initial data set of each single cell includes impedance values ​​at different frequencies. The initial data sets of all single cells constitute the module data set; d.2) Compare each set of impedance spectrum data in the module data set with the reference value in the pre-established single cell battery status normal database. If the absolute value of the difference between the two is within the preset threshold range, the impedance spectrum data is determined to be a normal value. If the absolute value of the difference between the two is greater than the preset threshold, the impedance spectrum data is determined to be an abnormal value, and abnormal impedance spectrum data is obtained; d.3) By calculating the abnormal impedance spectrum data, the change rate of the abnormal impedance spectrum data is obtained to form a change rate data set, and the impedance value at different frequencies in the abnormal impedance spectrum data of each single cell is used as a variable, and the difference between the impedance values ​​of adjacent frequencies is divided by the difference in frequency as the change rate of the frequency segment, and multiple change rates constitute the change rate data of the single cell; d.4) Extract features from the change rate data set to obtain multiple pre-established change trend models, each of which stores a different range of change rate values, and substitute the change rate data of each single cell into different change trend models for fitting; d.5) If the change rate data of a single cell has a fitting degree with a change trend model higher than a preset threshold, it indicates that the single cell has a change trend indicated by the model; if the change rate data of multiple single cells have a fitting degree with a change trend model higher than a preset threshold, it indicates that these single cells have a tendency to leak electrolyte; d.6) Generate an electrolyte leakage risk map of the battery module based on the location information of the single battery with electrolyte leakage tendency in the battery module, and use the number of the single battery as an identifier; d.7) Continuously track and monitor the electrolyte leakage risk map of battery modules with electrolyte leakage tendency, collect changes in the electrolyte leakage risk map at each preset time interval, obtain the electrolyte leakage risk level of different battery modules, and store it in the battery module database.

[0010] In a preferred embodiment, step e) further comprises: e.1) Collect voltage data, temperature data and electrolyte leakage data of a single battery cell to generate an initial data set; e.2) Using an electrolyte detection sensor to detect electrolyte leakage data to obtain numerical changes and generate data A; e.3) Preprocess the initial data set to obtain a preprocessed data set; e.4) Perform amplitude transformation on data A to obtain data B; e.5) The preprocessed data set enters the time domain analysis module, and the fluctuations of the data at multiple time points of the preprocessed data set are calculated to obtain the time domain envelope data C; e.6) Compare data B with the threshold value set in the preset threshold value table to obtain comparison result data; e.7) Perform wavelet transform on the time domain envelope data C to obtain time-frequency domain feature data, which enters the frequency domain analysis module; e.8) The data is Fourier transformed in the frequency domain analysis module to obtain frequency domain envelope data D. The comparison result data is used to determine whether there is an over-threshold state. If there is an over-threshold state, an alarm light signal is issued; e.9), constructing data set E according to frequency domain envelope data D and time domain envelope data C; e.10) Extract feature information from data set E using convolution operation to generate convolution feature map data; e.11) Process the convolution feature map data, use the nonlinear activation layer function to obtain nonlinear feature data, and use a multi-layer deep neural network to perform feature fusion on the nonlinear feature data; e.12) Obtain fusion feature data F; e.13) The fused feature data F enters the fully connected layer for classification, obtains the output layer result data, analyzes the output layer result data, determines whether there is a leakage, and determines the output of the final electrolyte leakage result.

[0011] In a preferred embodiment, step f) further comprises: f.1) After the electrolyte detection sensor detects data during the acquisition period, a predetermined amount of voltage data is acquired to form a first data set; f.2) obtaining a plurality of differential voltages as the second data set according to the voltage values ​​at different time points in the first data set, and counting the time when each voltage data collection occurs; f.3) obtaining multiple differential voltages greater than a preset voltage change threshold in the second data set, and sorting the statistical differential voltages according to their time, obtaining the termination time and occurrence time of multiple target differential voltage collections, where the target differential voltage is all the voltages greater than the preset voltage change threshold; f.4), obtaining the end time and occurrence time interval of all target differential voltages and constructing them into a third data set, obtaining the target time intervals corresponding to all target differential voltages and arranging them in order; f.5) and setting a predetermined ratio, taking the target time intervals with a specific ratio in the target time interval arrangement result as reference time intervals; f.6) obtaining a difference voltage matching each reference time interval according to the reference time interval, obtaining a preceding time point in the first data set corresponding to the difference voltage matching each reference time interval, constructing multiple target data sets from the voltages of the acquisition time occurrence point in all the first data sets before the preceding time point, and performing data standardization processing on each target data set; f.7) Calculate the standard deviation of voltage data in multiple target data sets after standardization, obtain the minimum standard deviation of all target data sets and compare it with the preset standard deviation threshold. If the standard deviation of this target data set is greater than the threshold, it is determined that the electrolyte is in a leakage state, and obtain the temperature data set of the electrolyte in the voltage drop sampling time period and the voltage detection synchronization time period; f.8) Take this target data set and the synchronized temperature data set as input, obtain the two parameter set data through the long short-term memory network model, perform specific function calculation to obtain the change rate, and if the change rate is greater than the change rate threshold, it is determined that the power-off protection measure needs to be triggered.

[0012] In a preferred embodiment, step g) further comprises: g.1) Obtain the real-time voltage data of all battery cells in the battery pack and compare it with the preset safety threshold to determine whether there is a battery cell exceeding the threshold; g.2) If there is a battery cell that exceeds the safety threshold, determine the target cell that needs to be isolated based on the location information of the cell; g.3) Generate an isolation control signal for the target monomer and transmit the signal to the corresponding actuator; g.4) After receiving the isolation control signal, the actuator triggers the internal isolation actuator to cut off the positive and negative circuit connections of the target monomer; g.5) According to the position information of the target monomer, determine the exhaust device corresponding to it, and generate a control instruction to start the exhaust device; g.6) After receiving the start-up command, the exhaust device starts to operate and continuously ventilates and dissipates heat for the isolated target monomer; g.7) Continuously monitor the temperature changes of the isolated cells. When the temperature drops to a safe range, determine again whether the voltage of the cell has returned to normal. If the voltage is normal and the temperature is safe, release the isolation and stop the exhaust device to reconnect the cell to the battery pack.

[0013] In a preferred embodiment, step h) further comprises: h.1) Collect electrolyte detection data through sensors deployed at battery cells, obtain the electrolyte detection data of the battery cell at each moment, and construct a battery cell time series detection data set; h.2) According to the battery cell time series detection data set, collect fixed-length data segments and input them into the pre-established machine learning prediction model for training to obtain the prediction value sequence of the corresponding time range; h.3) The real data of the time series segment and the predicted value sequence within the time range are calculated to obtain the data packet of the variation range of the battery cell within the time range; h.4) Obtain a fixed threshold parameter and compare whether the data in the variation range data packet is less than the fixed threshold; h.5) If the value is lower than the fixed threshold, it is determined that there is a battery cell with hidden danger; if the value is higher, it is determined that there is no hidden danger battery cell; h.6) If the battery cell is identified as having a hidden danger, a heat dissipation parameter solution is calculated based on the number and distribution information of all hidden danger battery cells; h.7) If there is no hidden danger battery cell, return to step h.1 and repeat the work; h.8) Communicate through the power environment system interface according to the heat dissipation parameter scheme; h.9) Generate the corresponding heat dissipation spray command set; h.10) According to the power environment system operation log, monitor the system command status data to obtain whether the command has been delivered; h.11) If the command is not delivered, retry the delivery until the command is delivered.

[0014] In a preferred embodiment, step i) further comprises: i.1) Acquire multiple battery electrolyte leakage detection data, determine whether multiple batteries are in an electrolyte leakage state through pre-established rules, and if the detection result is that electrolyte leakage occurs, obtain leakage location information and occurrence time information, and generate a first thermal map of the battery leakage location at that moment according to the battery structure information and leakage location information; i.2) storing each generated first heat map of the battery leakage location, determining the size of the tensor input to the model by the number of the first heat maps, using at least one first heat map tensor to input into the same heat map-based battery leakage accident prediction network model for calculation, and obtaining the first prediction result corresponding to the group of first heat maps; i.3) Compare the time of generating the second heat map with the time of the first heat map. If the time of the first heat map is later than that of the second heat map, the first heat map replaces the second heat map as the comparison heat map. If the time of the first heat map is earlier than that of the second heat map, combine the first heat map with the second heat map, fuse the second heat map into the first heat map, and input the fused first heat map tensor into the battery leakage accident prediction network model based on the heat map to obtain the second prediction result; i.4) Perform difference calculation using the first prediction result corresponding to the first thermogram and the second prediction result corresponding to the second thermogram to obtain the change value between the two prediction results, determine the leakage level of the electrolyte according to the size of the change value, and obtain the information on the development trend of the electrolyte leakage tendency according to the leakage level and the time of the second thermogram; i.5) Taking the time difference between the first thermogram and the second thermogram as the time interval, dividing the spatial position change between the first prediction result and the second prediction result by the time interval, obtaining the electrolyte diffusion velocity value, and packaging the electrolyte diffusion velocity value and the leakage tendency development trend information together to generate an electrolyte leakage information data packet; i.6) Set a preset threshold and compare it with the electrolyte diffusion speed. If the electrolyte diffusion speed is greater than the preset threshold, an alarm message is generated. If the electrolyte diffusion speed is lower than the preset threshold, no alarm message is generated. According to the alarm message, the preset emergency rescue information is obtained and combined with the alarm information into an alarm data packet; i.7) Determine the integrity of the alarm data packet. If the data packet is missing, return the alarm data packet acquisition instruction. If the alarm data packet is complete, obtain the sending destination according to the pre-set destination list, determine the sending data channel according to the destination, and transmit the alarm data packet to the specified receiving terminal.

[0015] An operation method of a safety early warning system for an energy storage power station based on a battery-based coupling model, the beneficial effects of which include but are not limited to the following: 1. The present invention uses a sensor network deployed on battery cells in energy storage power stations to monitor the voltage, temperature and electrolyte leakage status of each battery cell in real time, and conducts a comprehensive evaluation based on the internal resistance change of the battery module. The linear regression algorithm is used to pre-process multi-dimensional data to quickly identify whether the battery module has a tendency to leak. This method not only improves the detection sensitivity of leakage risks, but also can issue early warnings in the early stages of a fault to avoid further expansion of the accident. 2. For modules suspected of leakage, the present invention uses electrochemical impedance spectroscopy (EIS) technology to conduct in-depth analysis of single cells and accurately identify specific battery cells with leakage tendency. Compared with the traditional single parameter monitoring method, EIS technology can provide more detailed internal battery status information to ensure the accuracy of fault location. In addition, the multi-input convolutional network further improves the accuracy of leakage judgment by analyzing the time domain and frequency domain of single cell voltage, temperature and leakage data, ensuring that the system can accurately identify and locate the problem battery; 3. The present invention optimizes the sensor network architecture and communication protocol, reduces data transmission delays, and ensures that the system can collect and analyze key data in the shortest possible time. Through the clock synchronization mechanism, the consistency and reliability of data transmission are guaranteed, allowing the system to respond quickly to sudden failures. Especially in large-scale energy storage power stations, this efficient real-time monitoring capability is crucial, which can effectively shorten the time difference from fault detection to emergency response and reduce the risk of accidents; 4. Once the electrolyte leakage is confirmed, the present invention immediately triggers the safety isolation mechanism, cuts off the circuit connection of the problematic battery, and prevents the fault from spreading. At the same time, the system will automatically start the exhaust device to reduce the concentration of combustible gas in the battery compartment and reduce the risk of fire. If the leakage is not effectively controlled, the system will further start the sprinkler device to extinguish the fire and reduce the temperature. This multi-level, intelligent emergency response mechanism can effectively prevent the further expansion of the accident and ensure the safe operation of the energy storage power station; 5. The present invention generates a thermal map of multi-point electrolyte leakage detection data to predict the development trend and diffusion speed of the leakage. Based on these data, the system can generate a detailed leakage accident prediction report to provide a scientific basis for emergency rescue. The fire department can formulate a rescue plan in advance based on the prediction results, reasonably allocate resources, and ensure the efficiency and safety of the rescue operation. In addition, the system can dynamically adjust the emergency measures according to the speed and scope of the leakage to ensure the flexibility and pertinence of the rescue work. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is a general flow chart of an operation method of a safety early warning system for an energy storage power station based on a battery-based coupling model. DETAILED DESCRIPTION

[0017] Example 1: Electrolyte leakage warning and positioning based on linear regression and electrochemical impedance spectroscopy; A method for early warning and positioning of electrolyte leakage for battery cells and modules of energy storage power stations, combining linear regression algorithm and electrochemical impedance spectroscopy (EIS) technology, aims to achieve early warning and precise positioning of electrolyte leakage risks. The specific implementation steps are as follows: S1. Deploy sensors on each battery cell in the energy storage power station to collect voltage, temperature and electrolyte leakage detection information in real time. At the same time, use an impedance analyzer to detect the internal resistance change of the battery module; S1.1. Obtain the voltage and temperature of the battery cell and the electrolyte leakage detection information, pre-process the collected information, remove abnormal values, and obtain effective detection data; S1.2. Determine whether the working status of each battery cell is normal based on the preset voltage, temperature and leakage threshold; if it exceeds the threshold range, mark it as an abnormal battery cell; S1.3, obtain the internal resistance test result of the battery module, calculate the module internal resistance change rate, if the change rate exceeds the preset range, the module performance is determined to be abnormal; S1.4. For the battery cells marked as abnormal, the support vector machine (SVM) algorithm is used to identify the fault type and obtain the cause of the cell fault; S1.5. For battery modules with abnormal performance, the decision tree algorithm is used to determine whether the module needs to be repaired or replaced in combination with the single fault type in the module; S1.6. Based on the single-cell failure causes and module abnormality judgment results, an association rule mining algorithm is used to analyze the correlation and regularity of battery failures in energy storage power stations; S1.7. Based on the fault correlation rules, formulate the energy storage power station battery operation and maintenance strategy, including regular inspections, cell replacement and module maintenance, to improve battery reliability; S2. Use a linear regression algorithm to process the changes in battery cell voltage, battery cell temperature and module internal resistance. If the changes in battery cell voltage, temperature and internal resistance are linear within a certain period of time, the data is directly input into the leakage risk prediction model for processing to determine whether the battery module has a tendency to leak electrolyte; S2.1. The acquisition device periodically acquires the voltage value, temperature value of each battery cell in the battery module and the internal resistance value of the entire battery module, and each type of data is recorded as a time series; S2.2. Perform data preprocessing on each time series, calculate the change value of each cell voltage series and each cell temperature series within a period of time, and the change value of the module internal resistance series, and determine whether the three types of change value time series data satisfy a linear relationship; S2.3, if the voltage change, temperature change and internal resistance change are all linear, samples are continuously intercepted from the processed three types of change value data sequences according to the fixed window length through the sliding window technology to obtain multiple sample sets, each of which is a subsequence containing three change value data; S2.4. Use the principal component analysis (PCA) algorithm to process the three types of change value data in each sample set. Only one main component is extracted from each set of data as the key feature representing the characteristics of the set, and finally a feature vector group is obtained; S2.5. After obtaining the feature vector group, combine the operation history of the battery module to search for the historical operation stage similar to the feature vector group, analyze whether there is an electrolyte leakage record of the battery module after the corresponding historical operation stage, and construct a binary classification mapping relationship between the characteristics of each stage and the leakage state label to obtain a data set for judging the battery leakage risk; S2.6. Train a long short-term memory network (LSTM) for binary classification based on the constructed data set, and obtain the probability value of whether the corresponding battery module is in an electrolyte leakage state according to the LSTM; S2.7. If the probability value of the battery module leaking electrolyte exceeds the set threshold, an alarm is issued; if the probability value is lower than the set threshold, it is considered that the battery module has no leakage risk temporarily; S3. For the module that detects the tendency of electrolyte leakage, the battery cell is tested by electrochemical impedance spectroscopy (EIS) to obtain impedance spectrum data, and the obtained impedance spectrum data is compared with the impedance spectrum data of the single battery in normal battery state to determine the battery cell corresponding to the different impedance spectrum. The impedance spectrum data has a significant frequency spectrum change trend, indicating that multiple batteries have a tendency of electrolyte leakage; S3.1. Obtain electrochemical impedance spectroscopy data of all single cells in the battery module to form an initial data set. The initial data set of each single cell includes impedance values ​​at different frequencies. The initial data sets of all single cells constitute a module data set. S3.2, comparing each set of impedance spectrum data in the module data set with the reference value in the pre-established single cell battery status normal database, if the absolute value of the difference between the two is within the preset threshold range, the impedance spectrum data is determined to be a normal value; if the absolute value of the difference between the two is greater than the preset threshold, the impedance spectrum data is determined to be an abnormal value, and abnormal impedance spectrum data is obtained; S3.3. By calculating the abnormal impedance spectrum data, the change rate of the abnormal impedance spectrum data is obtained to form a change rate data set. The impedance values ​​at different frequencies in the abnormal impedance spectrum data of each single cell are used as variables. The difference between the impedance values ​​of adjacent frequencies is divided by the difference in frequency as the change rate of the frequency segment. Multiple change rates constitute the change rate data of the single cell. S3.4, extracting features from the change rate data set to obtain a plurality of pre-established change trend models, each of which stores a different range of change rate values, and substituting the change rate data of each single cell into different change trend models for fitting; S3.5. If there is a single cell whose rate of change data and a change trend model have a fitting degree higher than a preset threshold, it means that the single cell has a change trend indicated by the model; if there are multiple single cells whose rate of change data and a change trend model have a fitting degree higher than a preset threshold, it is obtained that these single cells have a tendency to leak electrolyte; S3.6. Generate an electrolyte leakage risk map of the battery module based on the location information of the single battery with a tendency to electrolyte leakage in the battery module, and use the number of the single battery as an identifier; S3.7. Continuously track and monitor the electrolyte leakage risk map of battery modules with electrolyte leakage tendency, collect changes in the electrolyte leakage risk map at each preset time interval, obtain the electrolyte leakage risk level of different battery modules, and store them in the battery module database; S4. According to the obtained battery cell position information exceeding the safety threshold, trigger and output the corresponding safety isolation measure signal to the actuator. The actuator starts the corresponding isolation actuator to cut off the circuit connection of the battery cell according to the received control signal, determines the single battery to be isolated through the safety isolation mechanism, and triggers the corresponding exhaust device to operate; S4.1. Obtain the real-time voltage data of all battery cells in the battery pack and compare it with the preset safety threshold to determine whether there are battery cells that exceed the threshold. S4.2. If there is a battery cell that exceeds the safety threshold, determine the target cell that needs to be isolated based on the location information of the cell; S4.3, generating an isolation control signal for the target monomer, and transmitting the signal to the corresponding actuator; S4.4, after receiving the isolation control signal, the actuator triggers the internal isolation actuator to cut off the positive and negative circuit connections of the target monomer; S4.5. Determine the exhaust device corresponding to the target unit according to the position information of the target unit, and generate a control instruction for starting the exhaust device; S4.6, after receiving the start command, the exhaust device starts to operate and continuously ventilates and dissipates heat for the isolated target monomer; S4.7. Continuously monitor the temperature change of the isolated cell. When the temperature drops to a safe range, determine again whether the voltage of the cell has returned to normal. If the voltage is normal and the temperature is safe, remove the isolation and stop the exhaust device to reconnect the cell to the battery pack. Through the above method, this embodiment can significantly improve the early warning capability of electrolyte leakage risk, accurately identify and locate problematic battery cells, and effectively prevent and control the occurrence of electrolyte leakage accidents. In addition, the intelligent emergency response mechanism can quickly take measures after the accident occurs to prevent the situation from escalating and ensure the safe operation of the energy storage power station.

[0018] Example 2: Electrolyte leakage prediction and emergency rescue guidance based on multi-input convolutional network and thermal map analysis; A method for predicting electrolyte leakage and providing emergency rescue guidance by using multi-input convolutional network (CNN) and heat map analysis technology; the method aims to generate a heat map of the battery leakage location by comprehensively analyzing the data of multiple electrolyte leakage detection points, and predict the development trend and diffusion speed of the leakage, so as to provide a scientific basis for emergency rescue. The specific implementation steps are as follows: S1. Deploy sensors on each battery cell in the energy storage power station to collect voltage, temperature and electrolyte leakage detection information in real time. At the same time, use an impedance analyzer to detect the internal resistance change of the battery module; S1.1. Obtain the voltage and temperature of the battery cell and the electrolyte leakage detection information, pre-process the collected information, remove abnormal values, and obtain effective detection data; S1.2. Determine whether the working status of each battery cell is normal based on the preset voltage, temperature and leakage threshold; if it exceeds the threshold range, mark it as an abnormal battery cell; S1.3, obtain the internal resistance test result of the battery module, calculate the module internal resistance change rate, if the change rate exceeds the preset range, the module performance is determined to be abnormal; S1.4. For the battery cells marked as abnormal, the support vector machine (SVM) algorithm is used to identify the fault type and obtain the cause of the cell fault; S1.5. For battery modules with abnormal performance, the decision tree algorithm is used to determine whether the module needs to be repaired or replaced in combination with the single fault type in the module; S1.6. Based on the single-cell failure causes and module abnormality judgment results, an association rule mining algorithm is used to analyze the correlation and regularity of battery failures in energy storage power stations; S1.7. Based on the fault correlation rules, formulate the energy storage power station battery operation and maintenance strategy, including regular inspections, cell replacement and module maintenance, to improve battery reliability; S2. Establish a multi-input convolutional network, receive the voltage data, temperature data and electrolyte leakage data of a single battery cell, analyze it in the time domain analysis module, perform wavelet analysis for frequency domain analysis, build a deep neural network model for leakage judgment by calculating the fast Fourier transform, and compare the state change value input by the electrolyte detection sensor with the set threshold to obtain an electrolyte leakage alarm signal that exceeds the threshold; S2.1. Collect voltage value data, temperature value data and electrolyte leakage value data of a single battery cell to generate an initial data set; S2.2, using an electrolyte detection sensor to detect electrolyte leakage data to obtain a numerical change and generate data A; S2.3, preprocessing the initial data set to obtain a preprocessed data set; S2.4, performing amplitude transformation on data A to obtain data B; S2.5, the preprocessed data set enters the time domain analysis module, and the fluctuation of the data at multiple time points of the preprocessed data set is calculated to obtain the time domain envelope data C; S2.6, comparing data B with the threshold value set in the preset threshold value table to obtain comparison result data; S2.7, perform wavelet transform on the time domain envelope data C to obtain time-frequency domain feature data, which enters the frequency domain analysis module; S2.8. The data is Fourier transformed in the frequency domain analysis module to obtain frequency domain envelope data D. The comparison result data is used to determine whether there is an over-threshold state. If there is an over-threshold state, an alarm light signal is issued; S2.9, constructing a data set E according to the frequency domain envelope data D and the time domain envelope data C; S2.10, extracting feature information from the data set E using a convolution operation to generate convolution feature map data; S2.11, processing the convolution feature map data, using the nonlinear activation layer function to obtain nonlinear feature data, and using a multi-layer deep neural network to perform feature fusion on the nonlinear feature data; S2.12, obtaining fusion feature data F; S2.13, the fused feature data F enters the fully connected layer for classification, obtains the output layer result data, analyzes the output layer result data, determines whether there is a leakage, and determines to output the final electrolyte leakage result; S3. By judging and analyzing the status of multiple batteries detected for electrolyte leakage, a thermal map of the battery leakage location is obtained. The thermal maps obtained multiple times are input into the same battery leakage accident prediction network model based on the thermal map for calculation. The development trend and diffusion speed result data packet of the electrolyte leakage tendency are generated according to the prediction results of the previous thermal map combined with the current thermal map. The emergency rescue command package is sent to the fire department for execution according to the speed. S3.1. Acquire electrolyte leakage detection data of multiple batteries, and determine whether the multiple batteries are in an electrolyte leakage state through pre-established rules. If the detection result is that electrolyte leakage occurs, obtain leakage location information and occurrence time information, and generate a first thermal map of the battery leakage location at that moment according to the battery structure information and the leakage location information; S3.2, storing each generated first heat map of the battery leakage location, determining the size of the tensor input to the model according to the number of the first heat maps, using at least one first heat map tensor to input into the same heat map-based battery leakage accident prediction network model for calculation, and obtaining the first prediction result corresponding to the group of first heat maps; S3.3. Compare the time of generating the second heat map with the time of the first heat map. If the time of the first heat map is later than that of the second heat map, the first heat map replaces the second heat map as the comparison heat map. If the time of the first heat map is earlier than that of the second heat map, combine the first heat map with the second heat map, fuse the second heat map into the first heat map, and input the fused first heat map tensor into the battery leakage accident prediction network model based on the heat map to obtain the second prediction result. S3.4. Perform a difference operation using the first prediction result corresponding to the first thermogram and the second prediction result corresponding to the second thermogram to obtain a change value between the two prediction results, determine the leakage level of the electrolyte according to the change value, and obtain information on the development trend of the electrolyte leakage tendency according to the leakage level and the time of the second thermogram; S3.5. Taking the time difference between the first thermogram and the second thermogram as the time interval, dividing the spatial position change between the first prediction result and the second prediction result by the time interval, obtaining the electrolyte diffusion velocity value, and packaging the electrolyte diffusion velocity value and the leakage tendency development trend information together to generate an electrolyte leakage information data packet; S3.6. Set a preset threshold and electrolyte diffusion speed for comparison. If the electrolyte diffusion speed is greater than the preset threshold, an alarm message is generated. If the electrolyte diffusion speed is lower than the preset threshold, no alarm message is generated. According to the alarm message, obtain the preset emergency rescue information and combine it with the alarm message into an alarm data packet. S3.7, determine the integrity of the alarm data packet. If the data packet is missing, return the alarm data packet acquisition instruction. If the alarm data packet is complete, obtain the sending destination according to the pre-set destination list, determine the sending data channel according to the destination, and transmit the alarm data packet to the specified receiving terminal; S4. Based on the generated electrolyte leakage information data packet, the fire department can formulate a rescue plan in advance according to the prediction results, reasonably allocate resources, and ensure the efficiency and safety of the rescue operation. In addition, the system can dynamically adjust the emergency measures according to the speed and scope of the leakage to ensure the flexibility and pertinence of the rescue work; Through the above method, this embodiment can not only accurately predict the development trend and diffusion speed of electrolyte leakage, but also provide a scientific basis for emergency rescue to ensure the effectiveness of rescue operations. In addition, the analysis method based on the heat map can intuitively display the location and severity of battery leakage, help operation and maintenance personnel make decisions quickly, and improve the safety management level of energy storage power stations.

[0019] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be regarded as exemplary and non-restrictive from all points of view, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention; any reference numerals in the claims should not be regarded as limiting the claims involved.

Claims

1. A method for operating a safety early warning system for an energy storage power station based on a battery-based coupling model, characterized in that: The system operation method includes the following steps: a) Sensors deployed in battery cells of energy storage power stations are used to collect battery cell voltage, battery cell temperature and electrolyte leakage detection information; and the battery module detects the impedance of the battery through an impedance analyzer to obtain the change in the module internal resistance; b) Obtain the data packet size of each battery cell voltage, battery cell temperature, electrolyte leakage detection information, and module internal resistance change, calculate the time required to upload the voltage and temperature data of all cells and the internal resistance change data of each module in combination with the throughput specified by the hardware network protocol of the network communication module, and obtain the clock synchronization data packet size through synchronization with the clock deployed on the module, and obtain the communication network delay information of the first data packet according to the time required for data transmission and the size of the clock synchronization packet; c) Use linear regression algorithm to process the changes of battery cell voltage, battery cell temperature and module internal resistance. If the changes of battery cell voltage, temperature and module internal resistance are linear within a certain period of time, the data will be directly input into the leakage risk prediction model for processing to determine whether the battery module has a tendency of electrolyte leakage; d) For modules that detect electrolyte leakage tendency, the battery cells are tested by electrochemical impedance spectroscopy (EIS) to obtain impedance spectrum data, and the obtained impedance spectrum data is compared with the impedance spectrum data of the single battery cells in normal battery status to determine the battery cells corresponding to the impedance spectrum with differences. The impedance spectrum data has a significant frequency spectrum change trend, indicating that multiple batteries have electrolyte leakage tendency; e) Establish a multi-input convolutional network, receive the voltage data, temperature data and electrolyte leakage data of a single battery cell, analyze it in the time domain analysis module, perform wavelet analysis for frequency domain analysis, and build a deep neural network model for leakage judgment by calculating the fast Fourier transform; compare the state change value input by the electrolyte detection sensor with the set threshold, and obtain an electrolyte leakage alarm signal that exceeds the threshold; f) If the electrolyte detection sensor repeatedly issues an alarm signal within a fixed period of time and determines that the electrolyte has a tendency to leak, the slope is obtained by calculating the numerical relationship between the voltage drop amplitude and the temperature rise amplitude within the same sampling time length, and the voltage mutation amplitude and change time point are obtained as the data source for triggering the safety isolation mechanism; g) According to the obtained battery cell position information exceeding the safety threshold, trigger and output the corresponding safety isolation measure signal to the actuator, and the actuator starts the corresponding isolation actuator to cut off the circuit connection of the battery cell according to the received control signal; determine the single battery to be isolated through the safety isolation mechanism, and trigger the corresponding exhaust device to operate; h) According to the number of isolated cells combined with the detection value obtained by the electrolyte detection sensor of the cell under the fixed time length, if the detection value has no decreasing trend or the decreasing amplitude is less than the fixed value, a command packet for starting the spray device is generated and sent to the power environment system to execute the command; i) By judging and analyzing the status of multiple batteries with electrolyte leakage detection, a thermal map of the battery leakage location is obtained. The thermal maps obtained multiple times are input into the same battery leakage accident prediction network model based on the thermal map for calculation. The development trend and diffusion speed result data packet of the electrolyte leakage tendency are generated according to the prediction results of the previous thermal map combined with the current thermal map. The emergency rescue command package is sent to the fire department for execution according to the speed.

2. The method for operating the energy storage power station safety early warning system based on the battery base coupling model according to claim 1 is characterized in that: Step a) further comprises: a.1) Obtain the voltage and temperature of the battery cell and electrolyte leakage detection information, pre-process the collected information, remove abnormal values, and obtain effective detection data; a.2) According to the preset voltage, temperature and leakage threshold, determine whether the working status of each battery cell is normal. If it exceeds the threshold range, it is marked as an abnormal battery cell; a.3) Obtain the internal resistance test result of the battery module and calculate the change rate of the module internal resistance. If the change rate exceeds the preset range, the module performance is determined to be abnormal; a.4) For battery cells marked as abnormal, the support vector machine (SVM) algorithm is used to identify the fault type and obtain the cause of the cell fault; a.5) For battery modules with abnormal performance, the decision tree algorithm is used to determine whether the module needs to be repaired or replaced in combination with the single fault type in the module; a.6) According to the single failure causes and module abnormality judgment results, the association rule mining algorithm is used to analyze the correlation and regularity of battery failures in energy storage power stations; a.7) Based on the fault correlation rules, formulate energy storage power station battery operation and maintenance strategies, including regular inspections, cell replacement and module maintenance, to improve battery reliability.

3. The method for operating the energy storage power station safety early warning system based on the battery-based coupling model according to claim 1 is characterized in that: Step b) further comprises: b.1) Determine the corresponding data packet size based on the voltage, temperature, electrolyte leakage detection information of the battery cell and the module internal resistance change data; b.2) Obtain the throughput specified by the hardware network protocol of the network communication module, and calculate the transmission time required to upload all monomer data and module internal resistance data in combination with the data packet size; b.3) Obtain the data packet size required for clock synchronization through the clock deployed on the module; b.4) determining whether the communication network delay of the first data packet exceeds a preset threshold value based on the time required for data transmission and the size of the clock synchronization packet; b.5) ​​If the network delay exceeds the threshold, a data compression algorithm is used to compress the uploaded data to reduce the data packet size and shorten the transmission time; b.6) If the network delay does not exceed the threshold, the original data packet is directly uploaded to the server; b.7) After receiving the data packet, the server uses the Kalman filter algorithm to fuse the data according to the timestamp and sequence number of the data packet to obtain accurate battery status information.

4. The method for operating the energy storage power station safety early warning system based on the battery-based coupling model according to claim 1 is characterized in that: Step c) further comprises: c.1) The acquisition device periodically obtains the voltage value, temperature value of each battery cell in the battery module and the internal resistance value of the entire battery module, and each type of data is recorded as a time series; c.2) Perform data preprocessing on each time series, calculate the change value of each cell voltage series and each cell temperature series within a period of time, and the change value of the module internal resistance series, and determine whether the three types of change value time series data satisfy a linear relationship; c.3) If the voltage change, temperature change and internal resistance change are all linear, samples are continuously intercepted from the processed three types of change value data sequences according to the fixed window length through the sliding window technology to obtain multiple sample sets, each of which is a subsequence containing three change value data; c.4) Use the principal component analysis (PCA) algorithm to process the three types of change value data in each sample set. Only one main component is extracted from each set of data as the key feature representing the characteristics of the set, and finally a feature vector group is obtained; c.5) After obtaining the feature vector group, combined with the operation history of the battery module, search for the historical operation stage similar to the feature vector group, analyze whether there is an electrolyte leakage record of the battery module after the corresponding historical operation stage, and construct a binary classification mapping relationship between the characteristics of each stage and the leakage state label to obtain a data set for judging the battery leakage risk; c.6) Train a long short-term memory network (LSTM) for binary classification based on the constructed data set, and obtain the probability value of whether the corresponding battery module is in an electrolyte leakage state based on the LSTM; c.7) If the probability value of the battery module leaking electrolyte exceeds the set threshold, an alarm is issued; if the probability value is lower than the set threshold, it is considered that the battery module has no leakage risk temporarily.

5. The method for operating the energy storage power station safety early warning system based on the battery-based coupling model according to claim 1 is characterized in that: Step d) further comprises: d.1) Obtain electrochemical impedance spectroscopy data of all single cells in the battery module to form an initial data set. The initial data set of each single cell includes impedance values ​​at different frequencies. The initial data sets of all single cells constitute the module data set; d.2) Compare each set of impedance spectrum data in the module data set with the reference value in the pre-established single cell battery status normal database. If the absolute value of the difference between the two is within the preset threshold range, the impedance spectrum data is determined to be a normal value. If the absolute value of the difference between the two is greater than the preset threshold, the impedance spectrum data is determined to be an abnormal value, and abnormal impedance spectrum data is obtained; d.3) By calculating the abnormal impedance spectrum data, the change rate of the abnormal impedance spectrum data is obtained to form a change rate data set, and the impedance value at different frequencies in the abnormal impedance spectrum data of each single cell is used as a variable, and the difference between the impedance values ​​of adjacent frequencies is divided by the difference in frequency as the change rate of the frequency segment, and multiple change rates constitute the change rate data of the single cell; d.4) Extract features from the change rate data set to obtain multiple pre-established change trend models, each of which stores a different range of change rate values, and substitute the change rate data of each single cell into different change trend models for fitting; d.5) If the change rate data of a single cell has a fitting degree with a change trend model higher than a preset threshold, it indicates that the single cell has a change trend indicated by the model; if the change rate data of multiple single cells have a fitting degree with a change trend model higher than a preset threshold, it indicates that these single cells have a tendency to leak electrolyte; d.6) Generate an electrolyte leakage risk map of the battery module based on the location information of the single battery with electrolyte leakage tendency in the battery module, and use the number of the single battery as an identifier; d.7) Continuously track and monitor the electrolyte leakage risk map of battery modules with electrolyte leakage tendency, collect changes in the electrolyte leakage risk map at each preset time interval, obtain the electrolyte leakage risk level of different battery modules, and store it in the battery module database.

6. The method for operating the energy storage power station safety early warning system based on the battery-based coupling model according to claim 1 is characterized in that: Step e) further comprises: e.1) Collect voltage data, temperature data and electrolyte leakage data of a single battery cell to generate an initial data set; e.2) Using an electrolyte detection sensor to detect electrolyte leakage data to obtain numerical changes and generate data A; e.3) Preprocess the initial data set to obtain a preprocessed data set; e.4) Perform amplitude transformation on data A to obtain data B; e.5) The preprocessed data set enters the time domain analysis module, and the fluctuations of the data at multiple time points of the preprocessed data set are calculated to obtain the time domain envelope data C; e.6) Compare data B with the threshold value set in the preset threshold value table to obtain comparison result data; e.7) Perform wavelet transform on the time domain envelope data C to obtain time-frequency domain feature data, which enters the frequency domain analysis module; e.8) The data is Fourier transformed in the frequency domain analysis module to obtain frequency domain envelope data D. The comparison result data is used to determine whether there is an over-threshold state. If there is an over-threshold state, an alarm light signal is issued; e.9), constructing data set E according to frequency domain envelope data D and time domain envelope data C; e.10) Extract feature information from data set E using convolution operation to generate convolution feature map data; e.11) Process the convolution feature map data, use the nonlinear activation layer function to obtain nonlinear feature data, and use a multi-layer deep neural network to perform feature fusion on the nonlinear feature data; e.12) Obtain fusion feature data F; e.13) The fused feature data F enters the fully connected layer for classification, obtains the output layer result data, analyzes the output layer result data, determines whether there is a leakage, and determines the output of the final electrolyte leakage result.

7. The method for operating the energy storage power station safety early warning system based on the battery-based coupling model according to claim 1 is characterized in that: Step f) further comprises: f.1) After the electrolyte detection sensor detects data during the acquisition period, a predetermined amount of voltage data is acquired to form a first data set; f.2) obtaining a plurality of differential voltages as the second data set according to the voltage values ​​at different time points in the first data set, and counting the time when each voltage data collection occurs; f.3) obtaining multiple differential voltages greater than a preset voltage change threshold in the second data set, and sorting the statistical differential voltages according to their time, obtaining the termination time and occurrence time of multiple target differential voltage collections, where the target differential voltage is all the voltages greater than the preset voltage change threshold; f.4), obtaining the end time and occurrence time interval of all target differential voltages and constructing them into a third data set, obtaining the target time intervals corresponding to all target differential voltages and arranging them in order of size; f.5) and setting a predetermined ratio, taking the target time intervals with a specific ratio in the target time interval arrangement result as reference time intervals; f.6) obtaining a difference voltage matching each reference time interval according to the reference time interval, obtaining a preceding time point in the first data set corresponding to the difference voltage matching each reference time interval, constructing multiple target data sets from the voltages of the acquisition time occurrence point in all the first data sets before the preceding time point, and performing data standardization processing on each target data set; f.7) Calculate the standard deviation of voltage data in multiple target data sets after standardization, obtain the minimum standard deviation of all target data sets and compare it with the preset standard deviation threshold. If the standard deviation of this target data set is greater than the threshold, it is determined that the electrolyte is in a leakage state, and obtain the temperature data set of the electrolyte in the voltage drop sampling time period and the voltage detection synchronization time period; f.8) Take this target data set and the synchronized temperature data set as input, obtain the two parameter set data through the long short-term memory network model, perform specific function calculation to obtain the change rate, and if the change rate is greater than the change rate threshold, it is determined that the power-off protection measure needs to be triggered.

8. The method for operating the energy storage power station safety early warning system based on the battery base coupling model according to claim 1 is characterized in that: Step g) further comprises: g.1) Obtain the real-time voltage data of all battery cells in the battery pack and compare it with the preset safety threshold to determine whether there is a battery cell exceeding the threshold; g.2) If there is a battery cell that exceeds the safety threshold, determine the target cell that needs to be isolated based on the location information of the cell; g.3) Generate an isolation control signal for the target monomer and transmit the signal to the corresponding actuator; g.4) After receiving the isolation control signal, the actuator triggers the internal isolation actuator to cut off the positive and negative circuit connections of the target monomer; g.5) According to the position information of the target monomer, determine the exhaust device corresponding to it, and generate a control instruction to start the exhaust device; g.6) After receiving the start-up command, the exhaust device starts to operate and continuously ventilates and dissipates heat for the isolated target monomer; g.7) Continuously monitor the temperature changes of the isolated cells. When the temperature drops to a safe range, determine again whether the voltage of the cell has returned to normal. If the voltage is normal and the temperature is safe, release the isolation and stop the exhaust device to reconnect the cell to the battery pack.

9. The method for operating the energy storage power station safety early warning system based on the battery-based coupling model according to claim 1 is characterized in that: Step h) further comprises: h.1) Collect electrolyte detection data through sensors deployed at battery cells, obtain the electrolyte detection data of the battery cell at each moment, and construct a battery cell time series detection data set; h.2) According to the battery cell time series detection data set, collect fixed-length data segments and input them into the pre-established machine learning prediction model for training to obtain the prediction value sequence of the corresponding time range; h.3) The real data of the time series segment and the predicted value sequence within the time range are calculated to obtain the data packet of the variation range of the battery cell within the time range; h.4) Obtain a fixed threshold parameter and compare whether the data in the variation range data packet is less than the fixed threshold; h.5) If the value is lower than the fixed threshold, it is determined that there is a battery cell with hidden danger; if the value is higher, it is determined that there is no hidden danger battery cell; h.6) If the battery cell is identified as having a hidden danger, a heat dissipation parameter solution is calculated based on the number and distribution information of all hidden danger battery cells; h.7) If there is no hidden danger battery cell, return to step h.1 and repeat the work; h.8) Communicate through the power environment system interface according to the heat dissipation parameter scheme; h.9) Generate the corresponding heat dissipation spray command set; h.10) According to the power environment system operation log, monitor the system command status data to obtain whether the command has been delivered; h.11) If the command is not delivered, retry the delivery until the command is delivered.

10. The method for operating the energy storage power station safety early warning system based on the battery base coupling model according to claim 1 is characterized in that: Step i) further comprises: i.1) Acquire multiple battery electrolyte leakage detection data, determine whether multiple batteries are in an electrolyte leakage state through pre-established rules, and if the detection result is that electrolyte leakage occurs, obtain leakage location information and occurrence time information, and generate a first thermal map of the battery leakage location at that moment according to the battery structure information and leakage location information; i.2) storing each generated first heat map of the battery leakage location, determining the size of the tensor input to the model by the number of the first heat maps, using at least one first heat map tensor to input into the same heat map-based battery leakage accident prediction network model for calculation, and obtaining the first prediction result corresponding to the group of first heat maps; i.3) Compare the time of generating the second heat map with the time of the first heat map. If the time of the first heat map is later than that of the second heat map, the first heat map replaces the second heat map as the comparison heat map. If the time of the first heat map is earlier than that of the second heat map, combine the first heat map with the second heat map, fuse the second heat map into the first heat map, and input the fused first heat map tensor into the battery leakage accident prediction network model based on the heat map to obtain the second prediction result; i.4) Perform difference calculation using the first prediction result corresponding to the first thermogram and the second prediction result corresponding to the second thermogram to obtain the change value between the two prediction results, determine the leakage level of the electrolyte according to the size of the change value, and obtain the information on the development trend of the electrolyte leakage tendency according to the leakage level and the time of the second thermogram; i.5) Taking the time difference between the first thermogram and the second thermogram as the time interval, dividing the spatial position change between the first prediction result and the second prediction result by the time interval, obtaining the electrolyte diffusion velocity value, and packaging the electrolyte diffusion velocity value and the leakage tendency development trend information together to generate an electrolyte leakage information data packet; i.6) Set a preset threshold and compare it with the electrolyte diffusion speed. If the electrolyte diffusion speed is greater than the preset threshold, an alarm message is generated. If the electrolyte diffusion speed is lower than the preset threshold, no alarm message is generated. According to the alarm message, the preset emergency rescue information is obtained and combined with the alarm information into an alarm data packet; i.7) Determine the integrity of the alarm data packet. If the data packet is missing, return the alarm data packet acquisition instruction. If the alarm data packet is complete, obtain the sending destination according to the pre-set destination list, determine the sending data channel according to the destination, and transmit the alarm data packet to the specified receiving terminal.

Images