Energy storage battery safety early warning method and system based on AI large model, medium and equipment

Abstract

The invention discloses an energy storage battery safety early warning method and system based on an AI large model, a medium and equipment. The method comprises the following steps: acquiring real-time battery characteristic data and historical battery characteristic data of each single battery in a target energy storage system; an AI large model of a self-attention mechanism based on a Transform is constructed; acquiring historical battery characteristic data of each single battery in the reference energy storage system; training the AI large model by using the historical battery characteristic data of each single battery in the reference energy storage system to obtain a trained AI large model; and performing safety early warning on each single battery in the target energy storage system according to the real-time battery characteristic data and the historical battery characteristic data of each single battery in the target energy storage system by using a pre-trained AI large model. According to the method, the accuracy and reliability of identifying the battery thermal runaway risk in advance are greatly improved, the battery thermal runaway early warning precision is improved, the false alarm probability is reduced, and stable operation of each battery cell of a power station is effectively protected.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage battery safety monitoring, and in particular to an energy storage battery safety early warning method, system, medium and equipment based on an AI large model. Background Art

[0002] Energy storage is an important technology and basic equipment supporting the new power system. It is of great significance to promote green energy transformation, respond to extreme events, ensure energy security, promote high-quality energy development, and achieve carbon peak and carbon neutrality goals. With the continuous growth of installed capacity of energy storage systems, fire accidents in energy storage power stations occur frequently and the consequences are fatal. The main cause of the fire is the thermal runaway of the energy storage battery. How to identify the risk of thermal runaway of the battery early and issue early warning to provide effective time for power station maintenance is a current industry problem.

[0003] The current mainstream approach is passive firefighting, that is, when a battery thermal runaway occurs, the corresponding fire-fighting system will not start the fire-fighting work until it spreads to the energy storage compartment and detects the fire. Of course, this process is already very delayed; and there are some studies on energy storage thermal runaway early warning methods, but there is a high probability of false alarms and the inability to accurately identify thermal runaway risk points, let alone accurately identify the location of a battery cell with thermal runaway in advance. Similarly, it is impossible to fundamentally solve the problem of the reliability and effectiveness of active early warning. Summary of the invention

[0004] To solve the technical problems existing in the background technology, the present invention proposes a storage battery safety warning method, system, medium and equipment based on an AI big model.

[0005] The present invention proposes a safety early warning method for energy storage batteries based on an AI large model, comprising: Acquire real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system; wherein the real-time battery characteristic data includes the voltage, temperature and deformation force of each single battery at the current moment; wherein the historical battery characteristic data includes the voltage, temperature and deformation force of each single battery at each moment in the historical time period; Build a large AI model based on Transformer's self-attention mechanism; Obtain historical battery characteristic data of each single battery in the reference energy storage system; The AI ​​big model is trained using the historical battery characteristic data of each single battery in the reference energy storage system to obtain a trained AI big model; A pre-trained AI large model is used to provide safety warnings for each single battery in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system.

[0006] Preferably, the AI ​​big model is trained using the historical battery characteristic data of each single battery in the reference energy storage system to obtain a trained AI big model, specifically including: Preprocessing the historical battery characteristic data of each single battery in the reference energy storage system to obtain preprocessing results: wherein the preprocessing results include the voltage change rate of each single battery in each time period of the charging, discharging and resting period, and the voltage variance and voltage standard deviation between the single batteries at each moment, the temperature change rate of each single battery in each time period of the charging, discharging and resting period, the temperature variance and temperature standard deviation between the single batteries at each moment, the temperature difference change data of the temperature of adjacent single batteries in the charging and discharging period at each moment, and the deformation force change rate of each single battery in each time period of the charging, discharging and resting period, and the deformation force standard deviation between the single batteries at each moment; Setting safety warning control conditions and initial safety warning thresholds; wherein the initial safety warning thresholds include an initial temperature upper limit threshold, an initial temperature difference threshold, and an initial deformation force threshold; Obtaining warning results according to safety warning control conditions, initial safety warning thresholds and preprocessing results; The AI ​​big model is trained according to the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning results to obtain a trained AI big model.

[0007] Preferably, the safety warning control conditions include minute-level advance safety warning control conditions and hour-level advance safety warning control conditions; The specific minute-level advance safety warning control conditions are as follows: If the temperature change rate of a certain i-cell battery exceeds the initial temperature upper limit threshold, and the temperature difference between the i-cell battery and the adjacent cell battery is higher than the initial temperature difference upper limit threshold, and the temperatures of the remaining cells except the i-cell battery are normal; at the same time, the deformation force of the i-cell battery at this time t is greater than the deformation force at the time before this time t, and is greater than the initial deformation force threshold, and the deformation force standard deviation of the i-cell battery at this time t is higher than the deformation force standard deviation at the time before this time t, and the deformation force standard deviation of the i-cell battery at this time t is the largest, then it is judged that the i-cell battery has a thermal runaway risk, and the warning result is abnormal; The hourly early warning conditions are as follows: If the voltage change rate of a certain i-cell battery changes more and more with time and exceeds the change rate threshold, and the difference between the voltage standard deviation of the i-cell battery at this moment t and the voltage standard deviation of the previous moment is large, and this difference continues to increase in the time series after this moment t, and the voltage variance and voltage change rate of the i-cell battery in the charge and discharge period are larger than the voltage variance and voltage change rate of the corresponding historical charge and discharge time; At the same time, the deformation force of the i-cell battery at this moment t is greater than the deformation force at the moment before this moment t, and the deformation force continues to increase in the time series after this moment t, and the deformation force change rate of the i-cell battery during the charging and discharging process is greater than the initial deformation force threshold, and the deformation force standard deviation of the i-cell battery at this moment t is greater than the deformation force standard deviation at the moment before this moment t, and the difference between the deformation force standard deviation of the i-cell battery at this moment t and the deformation force standard deviation at the previous moment is the largest, and the difference continues to increase with time t in the time series at and before this moment t; At the same time, the temperature change rate of the i-cell battery is increasing, while the temperature difference between the i-cell battery and the adjacent cell battery does not exceed the upper temperature difference threshold, and the difference between the temperature variance of the i-cell battery and the temperature variance at the previous moment is increasing with time t, and the temperature variance and temperature change rate of the i-cell battery in the charge and discharge period are larger than the temperature variance and temperature change rate of the corresponding historical charge and discharge time; It is judged that the i single battery has the risk of overcharge, over-discharge and runaway, and the warning result is abnormal.

[0008] Preferably, after the AI ​​big model is trained according to the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning result to obtain the trained AI big model, it also includes: The pre-trained AI big model is fine-tuned using known abnormal battery feature data and known normal battery feature data to obtain a fine-tuned AI big model.

[0009] Preferably, after fine-tuning the pre-trained AI big model using known abnormal battery feature data and known normal battery feature data to obtain the fine-tuned AI big model, the method further includes: The trained AI big model is trained with secondary battery feature quantization to obtain a secondary trained AI big model.

[0010] Preferably, after the trained AI big model is trained for secondary battery feature quantization, the method further includes: The AI ​​model that has been trained twice is trained three times with battery feature pruning to obtain an AI model that has been trained three times.

[0011] The present invention also proposes a storage battery safety early warning system based on an AI large model, comprising: An acquisition module, used to acquire real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system; wherein the real-time battery characteristic data includes the voltage, temperature and deformation force of each single battery at the current moment; wherein the historical battery characteristic data includes the voltage, temperature and deformation force of each single battery at each moment in the historical time period; Model building module, used to build a large AI model based on Transformer's self-attention mechanism; A model training module is used to obtain historical battery characteristic data of each single cell in the reference energy storage system, and use the historical battery characteristic data of each single cell in the reference energy storage system to train the AI ​​big model to obtain a trained AI big model; The early warning module is used to use a pre-trained AI large model to provide safety early warnings for each single battery in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system.

[0012] Preferably, the model training module includes: A preprocessing submodule is used to preprocess the historical battery characteristic data of each single cell in the reference energy storage system to obtain a preprocessing result: wherein the preprocessing result includes the voltage change rate of each single cell in each time period of the charging, discharging and resting period, and the voltage variance and voltage standard deviation between the single cells at each moment, the temperature change rate of each single cell in each time period of the charging, discharging and resting period, the temperature variance and temperature standard deviation between the single cells at each moment, the temperature difference change data of the temperature of adjacent single cells in the charging and discharging period at each moment, and the deformation force change rate of each single cell in each time period of the charging, discharging and resting period and the deformation force standard deviation between the single cells at each moment; The configuration submodule is used to set the safety warning control conditions and the initial safety warning threshold; wherein the initial safety warning threshold includes the initial temperature upper limit threshold, the initial temperature difference threshold, and the initial deformation force threshold; The prediction submodule is used to obtain the warning result according to the safety warning control conditions, the initial safety warning threshold and the preprocessing result; The training submodule is used to train the AI ​​big model based on the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning result to obtain a trained AI big model.

[0013] The present invention also proposes a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the energy storage battery safety warning method based on the AI ​​large model as described in any one of the above.

[0014] The present invention also proposes an electronic device, comprising: a processor and a memory, the memory being used to store one or more programs; when the one or more programs are executed by the processor, the energy storage battery safety warning method based on the AI ​​large model as described in any one of the above is implemented.

[0015] In the present invention, the proposed energy storage battery safety warning method, system, medium and equipment based on the AI ​​big model can use the pre-trained AI big model to perform safety warning on each single battery in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system, and can effectively locate the single battery with thermal runaway or thermal runaway risk in advance, greatly improving the accuracy and reliability of early identification of battery thermal runaway risks, improving the accuracy of battery thermal runaway warning, reducing the probability of false alarms, effectively protecting the stable operation of each battery cell in the power station, significantly reducing the risk of fire in the power station, and improving the safety of power station operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of a safety early warning method for energy storage batteries based on an AI large model in one embodiment of the present invention. DETAILED DESCRIPTION

[0017] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0018] Reference Figure 1 The present invention proposes a safety early warning method for energy storage batteries based on an AI large model, comprising: Acquire real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system; wherein the real-time battery characteristic data includes the voltage, temperature and deformation force of each single battery at the current moment; wherein the historical battery characteristic data includes the voltage, temperature and deformation force of each single battery at each moment in the historical time period; A pre-trained AI large model is used to provide safety warnings for each single battery in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system.

[0019] The present invention can use the pre-trained AI large model to conduct safety warnings on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system, and can locate in advance the single battery with thermal runaway risk or thermal runaway, which greatly improves the accuracy and reliability of early identification of battery thermal runaway risks, improves the accuracy of battery thermal runaway warning, reduces the probability of false alarms, effectively protects the stable operation of each battery cell in the power station, significantly reduces the risk of fire in the power station, and improves the safety of power station operation.

[0020] In order to locate the single battery with thermal runaway risk or thermal runaway in advance, improve the accuracy and reliability of identifying the risk of thermal runaway of the battery in advance, thereby improving the accuracy of battery thermal runaway warning and reducing the probability of false alarms, in this embodiment, before using the pre-trained AI big model to issue a safety warning to each single battery in the target energy storage system according to the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system, it also includes: Build a large AI model based on Transformer's self-attention mechanism; Obtain historical battery characteristic data of each single battery in the reference energy storage system; The AI ​​big model is trained using the historical battery characteristic data of each single battery in the reference energy storage system to obtain a trained AI big model.

[0021] It is important to know that the reference energy storage system has the same model or specification as the target energy storage system. Of course, there are multiple reference energy storage systems to ensure that there are abnormal warning results, thereby ensuring the training effect of the model.

[0022] In order to obtain a trained AI big model, in this embodiment, the historical battery characteristic data of each single battery in the reference energy storage system is used to train the AI ​​big model to obtain a trained AI big model, specifically including: The historical battery characteristic data of each single battery in the reference energy storage system is preprocessed to obtain the preprocessing results: Set safety warning control conditions and initial safety warning thresholds; Obtaining warning results according to safety warning control conditions, initial safety warning thresholds and preprocessing results; The AI ​​big model is trained according to the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning results to obtain a trained AI big model.

[0023] Among them, the initial safety warning thresholds include the initial temperature upper limit threshold, the initial temperature difference threshold, and the initial deformation force threshold.

[0024] Among them, the preprocessing results include the voltage change rate of each single cell in each time period of the charging, discharging and resting period, as well as the voltage variance and voltage standard deviation between the single cells at each moment, the temperature change rate of each single cell in each time period of the charging, discharging and resting period, as well as the temperature variance and temperature standard deviation between the single cells at each moment, the temperature difference change data of the temperature of adjacent single cells in the charging and discharging period at each moment, the deformation force change rate of each single cell in each time period of the charging, discharging and resting period, as well as the deformation force standard deviation between the single cells at each moment.

[0025] The historical battery characteristic data of each single battery in the reference energy storage system is preprocessed to obtain the preprocessing results, which specifically include: According to the historical battery characteristic data, a historical characteristic set of each single battery in the charging process, a historical characteristic set in the discharging process and a historical characteristic set in the static process are obtained respectively; According to the historical feature set of each single cell during the charging process, discharging process and static process, the voltage change rate, temperature change rate and deformation force change rate of each single cell in each time period of the charging, discharging and static period are obtained respectively, the voltage variance and voltage standard deviation, temperature variance, temperature standard deviation and deformation force standard deviation between the single cells at each moment, and the temperature difference of adjacent single cells in the charging and discharging period at each moment are obtained.

[0026] Among them, the historical feature set D of each single battery during the charging process i = [V, T, K, t], where i is the index of the single cell, V is the voltage, T is the temperature, K is the deformation force, and t is the time, and t is between 1 and t his t his is the length of historical data and t his More than 7 days.

[0027] Among them, the historical feature set D of each single battery during the discharge process i = [V, T, K, t], where i is the index of the single cell, and t is between 1 and t his t his is the length of historical data and t his More than 7 days; Among them, the historical feature set D of each single battery during the static process is i = [V, T, K, t], where i is the index of the single cell, and t is between 1 and t his t his is the length of historical data, and t his Greater than N days (N is a positive integer greater than 7).

[0028] Specifically, the time precision of the collection of battery characteristic data is at least within 1 second to improve accuracy.

[0029] Therefore, the voltage change rate of each single battery in each time period during the charging, discharging and resting period is Both are expressed as , and the voltage standard deviation between single cells at each moment for ; The temperature change rate of each single battery in each time period during charging, discharging and resting Both are expressed as , and the temperature standard deviation between the single cells at each moment for , and the temperature difference change data of the temperature of adjacent single cells during the charging and discharging period at each moment for ; The deformation force change rate of each single battery in each time period during charging, discharging and static period Both are expressed as , and the standard deviation of deformation force between single cells at each moment for .

[0030] In this embodiment, the safety warning control condition includes a minute-level advance safety warning control condition, and the minute-level advance safety warning control condition is specifically: If the temperature change rate of a certain i single cell exceeds the initial temperature upper limit threshold, and the temperature difference between the i single cell and the adjacent single cell is higher than the initial temperature difference upper limit threshold, and the temperatures of the remaining single cells except the i single cell are normal; at the same time, the deformation force of the i single cell at this time t is greater than the deformation force at the moment before this time t, and is greater than the initial deformation force threshold, and the standard deviation of the deformation force of the i single cell at this time t is higher than the standard deviation of the deformation force at the moment before this time t, and the standard deviation of the deformation force of the i single cell at this time t is the largest, then it is judged that the i single cell has a risk of thermal runaway, and the warning result is abnormal.

[0031] In a further embodiment, the safety warning control condition further includes an hourly advance safety warning control condition, and the hourly advance warning condition is specifically: If the voltage change rate of a single cell i is The change is getting bigger over time and exceeds the change rate threshold , and the difference between the voltage standard deviation of the i-cell battery at this moment t and the voltage standard deviation at the previous moment is large, and this difference continues to increase in the time series after this moment t, and the voltage variance and voltage change rate of the i-cell battery in the charge and discharge period are larger than the voltage variance and voltage change rate of the corresponding historical charge and discharge time; At the same time, the deformation force of the i-cell battery at this time t is greater than the deformation force at the time before this time t, and the deformation force continues to increase in the time series after this time t, and the deformation force change rate of the i-cell battery during the charging and discharging process is greater than the initial deformation force threshold , and the standard deviation of the deformation force of the i-cell battery at this time t is greater than the standard deviation of deformation force at the moment before this moment t, and the difference between the standard deviation of deformation force of the i-cell at this moment t and the standard deviation of deformation force at the previous moment is the largest, and the difference increases with time t in the time series at this moment t and before; At the same time, the temperature change rate of the i-cell battery continues to increase, and the temperature difference between the i-cell battery and the adjacent cell battery The temperature difference upper threshold is not exceeded , and the difference between the temperature variance of the i-cell battery and the temperature variance at the previous moment increases with time t, and the temperature variance and temperature change rate of the i-cell battery in the charge and discharge period increase compared with the temperature variance and temperature change rate of the corresponding historical charge and discharge time; It is judged that the i single battery has the risk of overcharge, over-discharge and runaway, and the warning result is abnormal.

[0032] Therefore, in the process of using the pre-trained AI big model to conduct safety warning for each single cell in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single cell in the target energy storage system, when the warning result is abnormal and triggers the minute-level advance safety warning control condition, the abnormal warning result is immediately output, and the power circuit corresponding to the i single cell is controlled to be cut off to wait for maintenance; When the warning result is abnormal and triggers the hourly advance safety warning control conditions, the abnormal warning result will be immediately sent to the operation and maintenance personnel so that they can handle it in advance to avoid thermal runaway.

[0033] It can be seen that this embodiment can provide an early warning of at least 1 hour in advance when a single cell battery is at risk of overcharge, over-discharge and thermal runaway, and can provide minute-level warning and cut-off control when there is a risk of thermal runaway, which greatly improves the accuracy of battery thermal runaway warning, reduces the probability of false alarms, reduces the risk of power station fire, and improves the safety of power station operation.

[0034] In this embodiment, after the AI ​​big model is trained according to the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning result to obtain the trained AI big model, it also includes: The trained AI big model is fine-tuned using known abnormal battery feature data and known normal battery feature data to obtain a fine-tuned and trained AI big model.

[0035] During the fine-tuning training process, professional knowledge including known abnormal battery feature data and known normal battery feature data is input into the pre-trained AI big model to train the AI ​​big model in a question-and-answer format until the AI ​​big model gives a 100% accurate answer to the input professional knowledge. The fine-tuning (Fine_tune) training is then terminated to obtain the full AI big model that has been fine-tuned and trained, greatly improving the accuracy of model warnings.

[0036] In a further embodiment, after fine-tuning the pre-trained AI big model using known abnormal battery feature data and known normal battery feature data to obtain the fine-tuned AI big model, it also includes: The fine-tuned AI big model is subjected to secondary battery feature quantization training to obtain a secondary trained AI big model.

[0037] Among them, the process of secondary battery feature quantization training is to input M days of battery feature data into the fine-tuned and trained AI large model, where M is much smaller than N and M approaches 1, to ensure that the training accuracy is not reduced under the condition of minimum input parameters.

[0038] In a further embodiment, after the trained AI big model is trained for secondary battery feature quantization, it also includes: The AI ​​model that has been trained twice is trained three times with battery feature pruning to obtain an AI model that has been trained three times.

[0039] This embodiment can use the AI ​​big model trained three times to provide safety warnings for the real-time battery characteristic data and historical battery characteristic data of the target energy storage system without reducing the prediction accuracy, reducing the complexity and hardware resource usage of the AI ​​big model, so that the AI ​​big model reaches an embedded model that can be run on at least the cortex_A35 architecture hardware platform, facilitating the operation of the early warning system in the system station.

[0040] The present invention also proposes a storage battery safety early warning system based on an AI large model, which is characterized by comprising: An acquisition module, used to acquire real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system; Model building module, used to build a large AI model based on Transformer's self-attention mechanism; A model training module is used to obtain historical battery characteristic data of each single cell in the reference energy storage system, and use the historical battery characteristic data of each single cell in the reference energy storage system to train the AI ​​big model to obtain a trained AI big model; wherein the historical battery characteristic data includes the voltage, temperature and deformation force of each single cell at each moment in the historical time period; The early warning module is used to use a pre-trained AI large model to provide safety early warnings for each single battery in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system.

[0041] In this embodiment, the model training module includes: A preprocessing submodule is used to preprocess the historical battery characteristic data of each single cell in the reference energy storage system to obtain a preprocessing result: wherein the preprocessing result includes the voltage change rate of each single cell in each time period of the charging, discharging and resting period, and the voltage variance and voltage standard deviation between the single cells at each moment, the temperature change rate of each single cell in each time period of the charging, discharging and resting period, the temperature variance and temperature standard deviation between the single cells at each moment, the temperature difference change data of the temperature of adjacent single cells in the charging and discharging period at each moment, and the deformation force change rate of each single cell in each time period of the charging, discharging and resting period, and the deformation force standard deviation between the single cells at each moment; The configuration submodule is used to set the safety warning control conditions and the initial safety warning threshold; wherein the initial safety warning threshold includes the initial temperature upper limit threshold, the initial temperature difference threshold and the initial deformation force threshold; The prediction submodule is used to obtain the warning result according to the safety warning control conditions, the initial safety warning threshold and the preprocessing result; The training submodule is used to train the AI ​​big model based on the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning result to obtain a trained AI big model.

[0042] Among them, the safety warning control conditions include minute-level advance safety warning control conditions and hour-level advance safety warning control conditions; The specific minute-level advance safety warning control conditions are as follows: If the temperature change rate of a certain i-cell battery exceeds the initial temperature upper limit threshold, and the temperature difference between the i-cell battery and the adjacent cell battery is higher than the initial temperature difference upper limit threshold, and the temperatures of the remaining cells except the i-cell battery are normal; at the same time, the deformation force of the i-cell battery at this time t is greater than the deformation force at the time before this time t, and is greater than the initial deformation force threshold, and the deformation force standard deviation of the i-cell battery at this time t is higher than the deformation force standard deviation at the time before this time t, and the deformation force standard deviation of the i-cell battery at this time t is the largest, then it is judged that the i-cell battery has a thermal runaway risk, and the warning result is abnormal; The hourly early warning conditions are as follows: If the voltage change rate of a single cell i is The change is getting bigger over time and exceeds the change rate threshold , and the voltage of the i-cell battery at this time t is different from the standard deviation is larger, and the deviation continues to increase in the time series after time t, and the voltage variance and voltage change rate of the i-cell battery in the charging and discharging period are larger than the voltage variance and voltage change rate of the corresponding historical charging and discharging time; At the same time, the deformation force of the i-cell battery at this time t is greater than the deformation force at the time before this time t, and the deformation force of the time series after this time t continues to increase, and the deformation force change rate of the i-cell battery during the charging and discharging process is greater than the initial deformation force threshold , and the deformation force standard deviation of the i-cell battery at this moment t is greater than the deformation force standard deviation at the moment before this moment t, and the difference between the deformation force standard deviation of the i-cell battery at this moment t and the deformation force standard deviation at the previous moment is the largest, and the difference increases with time t in the time series at this moment t and before; At the same time, the temperature change rate of the i-cell battery continues to increase, and the temperature difference between the i-cell battery and the adjacent cell battery The temperature difference upper threshold is not exceeded , and the difference between the temperature variance of the i-cell battery and the temperature variance at the previous moment increases with time t, and the temperature variance and temperature change rate of the i-cell battery in the charge and discharge period increase compared with the temperature variance and temperature change rate of the corresponding historical charge and discharge time; It is judged that the i single battery has the risk of overcharge, over-discharge and runaway, and the warning result is abnormal.

[0043] Therefore, in the process of using the pre-trained AI big model to conduct safety warning for each single cell in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single cell in the target energy storage system, when the warning result is abnormal and triggers the minute-level advance safety warning control condition, the abnormal warning result is immediately output, and the power circuit corresponding to the i single cell is controlled to be cut off to wait for maintenance; When the warning result is abnormal and triggers the hourly advance safety warning control conditions, the abnormal warning result will be immediately sent to the operation and maintenance personnel so that they can handle it in advance to avoid thermal runaway.

[0044] It can be seen that this embodiment can provide an early warning of at least 1 hour in advance when a single cell battery is at risk of overcharge, over-discharge and thermal runaway, and can provide minute-level warning and cut-off control when there is a risk of thermal runaway, which greatly improves the accuracy of battery thermal runaway warning, reduces the probability of false alarms, reduces the risk of power station fire, and improves the safety of power station operation.

[0045] In a further embodiment, the training submodule is also used to fine-tune the pre-trained AI large model using known abnormal battery feature data and known normal battery feature data to obtain a fine-tuned and trained AI large model.

[0046] In a further embodiment, the training submodule is also used to perform secondary battery feature quantization training on the trained AI big model to obtain a secondary trained AI big model.

[0047] In a further embodiment, the training submodule is also used to perform three battery feature pruning training on the AI ​​big model trained twice to obtain the AI ​​big model trained three times.

[0048] The present invention also proposes a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the energy storage battery safety warning method based on the AI ​​large model as described in any one of the above.

[0049] The present invention also proposes an electronic device, comprising: a processor and a memory, the memory being used to store one or more programs; when the one or more programs are executed by the processor, the energy storage battery safety warning method based on the AI ​​large model as described in any one of the above is implemented.

[0050] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A safety early warning method for energy storage batteries based on AI large model, characterized in that: include: Acquire real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system; wherein the real-time battery characteristic data includes the voltage, temperature and deformation force of each single battery at the current moment; wherein the historical battery characteristic data includes the voltage, temperature and deformation force of each single battery at each moment in the historical time period; Build a large AI model based on Transformer's self-attention mechanism; Obtain historical battery characteristic data of each single battery in the reference energy storage system; The AI ​​big model is trained using the historical battery characteristic data of each single battery in the reference energy storage system to obtain a trained AI big model; A pre-trained AI large model is used to provide safety warnings for each single battery in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system.

2. The energy storage battery safety early warning method based on AI large model according to claim 1 is characterized in that: The AI ​​big model is trained using the historical battery characteristic data of each single battery in the reference energy storage system to obtain a trained AI big model, which specifically includes: Preprocessing the historical battery characteristic data of each single battery in the reference energy storage system to obtain preprocessing results: wherein the preprocessing results include the voltage change rate of each single battery in each time period of the charging, discharging and resting period, and the voltage variance and voltage standard deviation between the single batteries at each moment, the temperature change rate of each single battery in each time period of the charging, discharging and resting period, the temperature variance and temperature standard deviation between the single batteries at each moment, the temperature difference change data of the temperature of adjacent single batteries in the charging and discharging period at each moment, and the deformation force change rate of each single battery in each time period of the charging, discharging and resting period, and the deformation force standard deviation between the single batteries at each moment; Setting safety warning control conditions and initial safety warning thresholds; wherein the initial safety warning thresholds include an initial temperature upper limit threshold, an initial temperature difference threshold, and an initial deformation force threshold; Obtaining warning results according to safety warning control conditions, initial safety warning thresholds and preprocessing results; The AI ​​big model is trained according to the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning results to obtain a trained AI big model.

3. The energy storage battery safety early warning method based on AI large model according to claim 2 is characterized in that: Safety early warning control conditions include minute-level advance safety early warning control conditions and hour-level advance safety early warning control conditions; The specific minute-level advance safety warning control conditions are as follows: If the temperature change rate of a certain i-cell battery exceeds the initial temperature upper limit threshold, and the temperature difference between the i-cell battery and the adjacent cell battery is higher than the initial temperature difference upper limit threshold, and the temperatures of the remaining cells except the i-cell battery are normal; at the same time, the deformation force of the i-cell battery at this time t is greater than the deformation force at the time before this time t, and is greater than the initial deformation force threshold, and the deformation force standard deviation of the i-cell battery at this time t is higher than the deformation force standard deviation at the time before this time t, and the deformation force standard deviation of the i-cell battery at this time t is the largest, then it is judged that the i-cell battery has a thermal runaway risk, and the warning result is abnormal; The hourly early warning conditions are as follows: If the voltage change rate of a certain i-cell battery changes more and more with time and exceeds the change rate threshold, and the difference between the voltage standard deviation of the i-cell battery at this moment t and the voltage standard deviation of the previous moment is large, and this difference continues to increase in the time series after this moment t, and the voltage variance and voltage change rate of the i-cell battery in the charge and discharge period are larger than the voltage variance and voltage change rate of the corresponding historical charge and discharge time; At the same time, the deformation force of the i-cell battery at this moment t is greater than the deformation force at the moment before this moment t, and the deformation force continues to increase in the time series after this moment t, and the deformation force change rate of the i-cell battery during the charging and discharging process is greater than the initial deformation force threshold, and the deformation force standard deviation of the i-cell battery at this moment t is greater than the deformation force standard deviation at the moment before this moment t, and the difference between the deformation force standard deviation of the i-cell battery at this moment t and the deformation force standard deviation at the previous moment is the largest, and the difference continues to increase with time t in the time series at and before this moment t; At the same time, the temperature change rate of the i-cell battery is increasing, while the temperature difference between the i-cell battery and the adjacent cell battery does not exceed the upper temperature difference threshold, and the difference between the temperature variance of the i-cell battery and the temperature variance at the previous moment is increasing with time t, and the temperature variance and temperature change rate of the i-cell battery in the charge and discharge period are larger than the temperature variance and temperature change rate of the corresponding historical charge and discharge time; It is judged that the i single battery has the risk of overcharge, overheating and runaway, and the warning result is abnormal.

4. The energy storage battery safety early warning method based on AI large model according to claim 2 is characterized in that: After the AI ​​big model is trained based on the historical battery characteristic data, initial safety warning threshold and warning results of each single battery in the reference energy storage system, the following steps are also included after the trained AI big model is obtained: The pre-trained AI big model is fine-tuned using known abnormal battery feature data and known normal battery feature data to obtain a fine-tuned AI big model.

5. The energy storage battery safety early warning method based on AI large model according to claim 4 is characterized in that: After fine-tuning the pre-trained AI big model using known abnormal battery feature data and known normal battery feature data to obtain the fine-tuned AI big model, the following steps are also included: The trained AI big model is subjected to secondary battery feature quantization training to obtain a secondary trained AI big model.

6. The energy storage battery safety early warning method based on AI large model according to claim 5 is characterized in that: After the trained AI model is trained for secondary battery feature quantification, it also includes: The AI ​​model trained twice is trained three times with battery feature pruning to obtain the AI ​​model trained three times.

7. A storage battery safety early warning system based on AI large model, characterized in that: include: An acquisition module, used to acquire real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system; wherein the real-time battery characteristic data includes the voltage, temperature and deformation force of each single battery at the current moment; wherein the historical battery characteristic data includes the voltage, temperature and deformation force of each single battery at each moment in the historical time period; Model building module, used to build a large AI model based on Transformer's self-attention mechanism; A model training module is used to obtain historical battery characteristic data of each single cell in the reference energy storage system, and use the historical battery characteristic data of each single cell in the reference energy storage system to train the AI ​​big model to obtain a trained AI big model; The early warning module is used to use a pre-trained AI large model to provide safety early warnings for each single battery in the target energy storage system based on the real-time battery characteristic data and historical battery characteristic data of each single battery in the target energy storage system.

8. The energy storage battery safety early warning system based on AI large model according to claim 7 is characterized in that: The model training module includes: A preprocessing submodule is used to preprocess the historical battery characteristic data of each single cell in the reference energy storage system to obtain a preprocessing result: wherein the preprocessing result includes the voltage change rate of each single cell in each time period of the charging, discharging and resting period, and the voltage variance and voltage standard deviation between the single cells at each moment, the temperature change rate of each single cell in each time period of the charging, discharging and resting period, the temperature variance and temperature standard deviation between the single cells at each moment, the temperature difference change data of the temperature of adjacent single cells in the charging and discharging period at each moment, and the deformation force change rate of each single cell in each time period of the charging, discharging and resting period, and the deformation force standard deviation between the single cells at each moment; The configuration submodule is used to set the safety warning control conditions and the initial safety warning threshold; wherein the initial safety warning threshold includes the initial temperature upper limit threshold, the initial temperature difference threshold, and the initial deformation force threshold; The prediction submodule is used to obtain the warning result according to the safety warning control conditions, the initial safety warning threshold and the preprocessing result; The training submodule is used to train the AI ​​big model based on the historical battery characteristic data of each single battery in the reference energy storage system, the initial safety warning threshold and the warning result to obtain a trained AI big model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the energy storage battery safety warning method based on the AI ​​large model as described in any one of claims 1 to 6.

10. An electronic device, characterized in that: include: A processor and a memory, the memory being used to store one or more programs; When one or more programs are executed by the processor, the energy storage battery safety early warning method based on the AI ​​large model as described in any one of claims 1 to 6 is implemented.

Images