Wireless battery management method and system based on computing power collaboration

Abstract

The invention belongs to the technical field of BMS, and discloses a wireless battery management method and system based on computing power collaboration, and the method comprises the steps that a main control battery system receives an abnormal battery parameter, a supervised battery parameter and a computing power collaboration request of an abnormal subsystem; the main control battery system determines each normal subsystem according to the abnormal subsystem and obtains the residual computing power of each normal subsystem; determining a first subsystem and a second subsystem according to the residual computing power; the main control battery system, the first subsystem and the second subsystem train an initial neural network corresponding to the computing power cooperation request based on the abnormal battery parameters and the supervised battery parameters to obtain a target neural network; the main control battery system sends the target neural network to an abnormal subsystem; and the abnormal subsystem inputs the abnormal battery parameters into the target neural network to obtain a request result. According to the method and the device, the problems that the training time is long and abnormal conditions cannot be handled in time due to insufficient computing power of the abnormal subsystem can be avoided, and the response speed of the abnormal subsystem is improved.

Description

Technical Field

[0001] The present application relates to the field of BMS technology, and in particular to a wireless battery management method and system based on computing power collaboration. Background Art

[0002] The essence of wireless BMS (battery management system) technology is to use wireless communication to reduce the wiring harnesses between slave MCUs and between slave and master BMS, while the sampling of voltage and temperature and the scheme of transmitting voltage and temperature to slave MCU are not affected; the main advantages of wireless BMS technology include: reducing system cost and production complexity by eliminating wiring harnesses; enabling new architectures to make better use of space and design battery packs with higher specific energy; improving automation by eliminating heavy, costly and frequently maintained wiring and reducing the number of connectors, thereby speeding up assembly and improving production efficiency. Each battery subsystem is matched with a wireless MCU, and the battery of the control subsystem responsible for sampling itself also has a certain computing power. Therefore, tasks such as battery data prediction or data analysis are often completed independently by the subsystem corresponding to the battery.

[0003] However, when training a neural network to predict or analyze battery data, the forward output and reverse optimization process needs to be repeated continuously. This requires a lot of training time for subsystems with limited computing power, especially when the subsystem itself has little remaining computing power, which is not conducive to the real-time response of the battery management system. Summary of the invention

[0004] The present application provides a wireless battery management method and system based on computing power collaboration, which can avoid the problem of insufficient computing power of the abnormal subsystem resulting in long training time and inability to respond to abnormal situations in a timely manner.

[0005] In a first aspect, an embodiment of the present application provides a wireless battery management method based on computing power collaboration, including:

[0006] The main control battery system receives abnormal battery parameters, supervision battery parameters and computing power coordination requests from abnormal subsystems;

[0007] The main control battery system determines each normal subsystem according to the abnormal subsystem, and obtains the remaining computing power of each normal subsystem; determines the first subsystem and the second subsystem according to the remaining computing power;

[0008] The main control battery system, the first subsystem and the second subsystem train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain the target neural network;

[0009] The main control battery system sends the target neural network to the abnormal subsystem;

[0010] The abnormal subsystem inputs the abnormal battery parameters into the target neural network and obtains the requested results.

[0011] Furthermore, the above-mentioned determining the first subsystem and the second subsystem according to the remaining computing power includes:

[0012] Sort the remaining computing power of each normal subsystem;

[0013] Select the normal subsystem with the largest remaining computing power as the first subsystem;

[0014] Select the normal subsystem with the second largest remaining computing power as the second subsystem.

[0015] Furthermore, the master control battery system, the first subsystem and the second subsystem train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain the target neural network, including:

[0016] Step 1: The main control battery system sends abnormal battery parameters and the initial neural network to the first subsystem;

[0017] The first subsystem inputs abnormal battery parameters into the initial neural network to obtain a forward output result;

[0018] The first subsystem sends the forward output result and the initial neural network to the second subsystem;

[0019] The main control battery system sends the monitoring battery parameters to the second subsystem;

[0020] The second subsystem updates the initial neural network according to the supervisory battery parameters and the forward output results, and sends the updated initial neural network to the main control battery system;

[0021] The main control battery system receives the initial neural network and increases the number of training times by 1; determines whether the number of training times reaches a first preset number; if so, uses the initial neural network as the target neural network; if not, returns to the first step.

[0022] Furthermore, the master control battery system, the first subsystem and the second subsystem train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain the target neural network, including:

[0023] Step 1: The main control battery system sends abnormal battery parameters and the initial neural network to the first subsystem;

[0024] The first subsystem inputs abnormal battery parameters into the initial neural network to obtain a forward output result;

[0025] The first subsystem sends the forward output result to the main control battery system;

[0026] When the second training number is equal to 0, the main control battery system updates the initial neural network according to the supervisory battery parameters and the forward output result, and increases the first training number by 1; and determines whether the first training number reaches the second preset threshold;

[0027] If or the second training number is not equal to 0, the forward output result, the supervision battery parameter and the initial neural network are sent to the second subsystem; if not, return to the first step;

[0028] The second subsystem updates the initial neural network according to the forward output result and the supervisory battery parameters, and sends the updated initial neural network to the main control battery system;

[0029] After receiving the updated initial neural network from the second subsystem, the main control battery system increases the second training times by 1;

[0030] Determine whether the second training reaches the third preset threshold at this time; if so, use the received initial neural network as the target neural network; if not, return to the first step.

[0031] Furthermore, the method further comprises:

[0032] The subsystem samples data of the target battery to obtain battery sampling data;

[0033] If a preset number of consecutive battery sampling data are all abnormal data, the subsystem is an abnormal subsystem, and the supervisory battery parameters are obtained based on the normal historical battery parameters, and the abnormal battery parameters are generated based on each abnormal data;

[0034] The supervision battery parameters, abnormal battery parameters and computing power coordination requests are sent to the main control battery system.

[0035] Furthermore, the method further comprises:

[0036] After obtaining the request result, the exception subsystem determines the exception type according to the request result;

[0037] The abnormal subsystem sends the abnormal type to the main control battery system.

[0038] Furthermore, the abnormal battery parameters include voltage data, current data, temperature data or remaining capacity.

[0039] In a second aspect, an embodiment of the present application provides a wireless battery management system based on computing power collaboration, including:

[0040] A main control battery system and a plurality of subsystems in wireless communication with the main control battery system;

[0041] The main control battery system is used to receive abnormal battery parameters, monitor battery parameters and computing power coordination requests of abnormal subsystems; determine each normal subsystem based on the abnormal subsystem, and obtain the remaining computing power of each normal subsystem;

[0042] and, determining the first subsystem and the second subsystem according to the remaining computing power, training the computing power coordination request corresponding to the first subsystem and the second subsystem based on the abnormal battery parameters and the supervisory battery parameters, obtaining the target neural network, and sending the target neural network to the abnormal subsystem;

[0043] The abnormal subsystem is used to input abnormal battery parameters into the target neural network to obtain the requested results.

[0044] Furthermore, the main control battery system is specifically used to sort the remaining computing power of each normal subsystem; select the normal subsystem with the largest remaining computing power as the first subsystem; and select the normal subsystem with the second largest remaining computing power as the second subsystem.

[0045] Furthermore, the main control battery system is specifically used to send the abnormal battery parameters and the initial neural network to the first subsystem; send the supervision battery parameters to the second subsystem; receive the initial neural network updated by the second subsystem, and increase the number of training times by 1; determine whether the number of training times reaches the first preset number; if so, use the initial neural network as the target neural network;

[0046] The first subsystem inputs the abnormal battery parameters into the initial neural network to obtain a forward output result; and sends the forward output result and the initial neural network to the second subsystem;

[0047] The second subsystem updates the initial neural network according to the supervisory battery parameters and the forward output results, and sends the updated initial neural network to the main control battery system.

[0048] In summary, compared with the prior art, the technical solution provided in the embodiment of the present application has at least the following beneficial effects:

[0049] The embodiment of the present application provides a wireless battery management method based on computing power coordination. When an abnormal subsystem occurs in the battery management system, the abnormal battery parameters, supervisory battery parameters and computing power coordination request of the abnormal subsystem are obtained through the main control battery system. The main control battery system and the first subsystem and the second subsystem with remaining computing power perform the initial neural network training process, and then send the trained target neural network to the abnormal subsystem for use. When an abnormality occurs in a subsystem, the above method makes full use of other subsystems with remaining computing power to train the initial neural network, thereby avoiding the problem of long training time due to insufficient computing power of the abnormal subsystem and failure to make corresponding processing for the abnormal situation in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flowchart of a wireless battery management method based on computing power collaboration is provided as an exemplary embodiment of the present application.

[0051] Figure 2 A structural diagram of a wireless battery management system based on computing power collaboration is provided as an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments.

[0053] Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of this application.

[0054] See also Figure 1 , the embodiment of the present application provides a wireless battery management method based on computing power collaboration, including:

[0055] Step S1, the main control battery system receives abnormal battery parameters, supervisory battery parameters and computing power coordination requests of abnormal subsystems.

[0056] The abnormal battery parameter includes one of voltage data, current data, temperature data or remaining capacity.

[0057] The data types of supervised battery parameters and abnormal battery parameters are consistent; the computing power coordination request includes the functions of the target neural network that needs to be trained, namely predicting future data, analyzing response strategies, etc.

[0058] Specifically, the subsystem samples data from the target battery to obtain battery sampling data; if a preset number of consecutive battery sampling data are all abnormal data, the subsystem is an abnormal subsystem. The target battery is the battery controlled by the subsystem.

[0059] The abnormal subsystem obtains the supervisory battery parameters based on the normal historical battery parameters, and generates abnormal battery parameters based on various abnormal data; and sends the supervisory battery parameters, abnormal battery parameters and computing power coordination requests to the main control battery system.

[0060] For example, assuming that the preset number is 2, under normal circumstances, the voltage change curve should be [V1 V2 V3 V4V5…], but the actual sampling is [V1 V2 V3 VS VSS…], where VS and VSS are abnormal data, then VS and VSS and the corresponding sampling time are sent to the main control battery system as abnormal battery parameters.

[0061] The supervisory battery parameters are obtained from the normal historical battery parameters of the target battery. Generally, the values ​​10 moments before the abnormal data and the corresponding sampling times are selected so that the target neural network can learn the normal voltage trend of the target battery.

[0062] Step S2, the main control battery system determines each normal subsystem according to the abnormal subsystem, and obtains the remaining computing power of each normal subsystem; and determines the first subsystem and the second subsystem according to the remaining computing power.

[0063] Specifically, see Figure 2 In the wireless battery management system, there is a main control MCU and multiple subsystem MCUs. The main control battery system here is the main control MCU. The subsystem periodically sends its own computing power application status (in percentage) to the main control battery system for decision-making.

[0064] The main control battery system can sort the remaining computing power of each normal subsystem; select the normal subsystem with the largest remaining computing power as the first subsystem; and select the normal subsystem with the second largest remaining computing power as the second subsystem.

[0065] Step S3: The main control battery system, the first subsystem and the second subsystem train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain the target neural network.

[0066] Step S4: the main control battery system sends the target neural network to the abnormal subsystem.

[0067] Step S5, the abnormal subsystem inputs the abnormal battery parameters into the target neural network to obtain the requested result.

[0068] The request result is the result corresponding to the computing power coordination request. For example, if the computing power coordination request is to predict the future voltage data of the target battery, the request result here is the voltage prediction data.

[0069] The above embodiment provides a wireless battery management method based on computing power coordination. When an abnormal subsystem occurs in the battery management system, the abnormal battery parameters, supervisory battery parameters and computing power coordination request of the abnormal subsystem are obtained through the main control battery system. The main control battery system and the first subsystem and the second subsystem with remaining computing power perform the initial neural network training process, and then send the trained target neural network to the abnormal subsystem for use. When an abnormality occurs in a subsystem, the above method makes full use of other subsystems with remaining computing power to train the initial neural network, thereby avoiding the problem of long training time due to insufficient computing power of the abnormal subsystem and the inability to deal with the abnormal situation in time.

[0070] In some embodiments, the master control battery system, the first subsystem, and the second subsystem train an initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain a target neural network, including:

[0071] Step S311, first step: the main control battery system sends the abnormal battery parameters and the initial neural network to the first subsystem.

[0072] The initial neural network here specifically refers to the model parameters of the initial neural network.

[0073] Step S312: the first subsystem inputs the abnormal battery parameters into the initial neural network to obtain a forward output result.

[0074] Step S313: The first subsystem sends the forward output result and the initial neural network to the second subsystem.

[0075] Step S314: the main control battery system sends the monitoring battery parameters to the second subsystem.

[0076] Step S315, the second subsystem updates the initial neural network according to the supervisory battery parameters and the forward output result, and sends the updated initial neural network to the main control battery system.

[0077] Here, the second subsystem updates the initial neural network based on the loss function of the initial neural network, analyzes the gap between the forward output results and the supervised battery parameters, and updates the model parameters of the initial neural network.

[0078] In step S316, the main control battery system receives the initial neural network and increases the number of training times by 1; determines whether the number of training times reaches a first preset number; if so, uses the initial neural network as the target neural network; if not, returns to the first step.

[0079] Specifically, the above steps can be considered as letting the first subsystem perform forward calculations in the neural network training process, and the second subsystem perform reverse optimization in the neural network training process. In this way, the two steps that occupy the most computing power in the neural network training process are respectively assigned to the two subsystems with the largest computing power to execute, which can speed up each forward calculation and reverse optimization operation.

[0080] In the specific implementation process, the number of selected subsystems can also be determined according to the steps of the neural network training process. If there are more training steps, more subsystems can be selected to share the execution of each step.

[0081] In some embodiments, the master control battery system, the first subsystem, and the second subsystem train an initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain a target neural network, including:

[0082] Step S321, first step: the main control battery system sends the abnormal battery parameters and the initial neural network to the first subsystem.

[0083] In step S322, the first subsystem inputs the abnormal battery parameters into the initial neural network to obtain a forward output result.

[0084] Step S323: the first subsystem sends the forward output result to the main control battery system.

[0085] Step S324, when the second training times is equal to 0, the main control battery system updates the initial neural network according to the supervisory battery parameters and the forward output result, and increases the first training times by 1; determines whether the first training times reaches the second preset threshold; if so or the second training times is not equal to 0, sends the forward output result, the supervisory battery parameters and the initial neural network to the second subsystem; if not, returns to the first step.

[0086] Step S325, the second subsystem updates the initial neural network according to the forward output result and the supervisory battery parameters, and sends the updated initial neural network to the main control battery system.

[0087] Step S326, after receiving the updated initial neural network from the second subsystem, the main control battery system increases the second training times by 1; determines whether the second training reaches the third preset threshold at this time; if so, uses the received initial neural network as the target neural network; if not, returns to the first step.

[0088] It can be understood that the second preset threshold + the third preset threshold in this embodiment is equal to the first preset threshold in the above embodiment. In this embodiment, the main control battery system is supported to first perform the reverse optimization step of the second preset threshold, and then the remaining third preset threshold number of the reverse optimization step is allocated to the second subsystem for completion; if the second training number is equal to 0, it means that the main control battery system completes the reverse optimization step at this time; if the first training number reaches the second preset threshold, or the second training number is not equal to 0, it means that the second subsystem completes the reverse optimization step at this time.

[0089] The above embodiment can support the main control battery system to share part of the reverse optimization operation when the computing power is sufficient or the computing power is greater than that of the second subsystem. Whether to use the main control battery system can be determined by the user.

[0090] In some embodiments, the method further comprises:

[0091] Step S6: After obtaining the request result, the exception subsystem determines the exception type according to the request result.

[0092] Step S7: The abnormal subsystem sends the abnormality type to the main control battery system.

[0093] Specifically, if the request result is a response strategy for an abnormal situation, the abnormal subsystem can infer the abnormality type based on the response strategy. For example, if the response strategy is to change the control of the current size or shut it off, the abnormality type is current abnormality.

[0094] After receiving the abnormal type, the main control battery system forwards it to the upper control system or remotely notifies the mobile terminal.

[0095] See also Figure 2 Another embodiment of the present application provides a wireless battery management system based on computing power collaboration, including a main control battery system and multiple subsystems that wirelessly communicate with the main control battery system.

[0096] The main control battery system is used to receive abnormal battery parameters, supervisory battery parameters and computing power coordination requests of the abnormal subsystem; determine each normal subsystem according to the abnormal subsystem, and obtain the remaining computing power of each normal subsystem; and determine the first subsystem and the second subsystem according to the remaining computing power, and train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters of the first subsystem and the second subsystem to obtain the target neural network, and send the target neural network to the abnormal subsystem.

[0097] The abnormal subsystem is used to input abnormal battery parameters into the target neural network to obtain the requested results.

[0098] Furthermore, the main control battery system is specifically used to sort the remaining computing power of each normal subsystem; select the normal subsystem with the largest remaining computing power as the first subsystem; and select the normal subsystem with the second largest remaining computing power as the second subsystem.

[0099] Furthermore, the main control battery system is specifically used to send abnormal battery parameters and the initial neural network to the first subsystem; send the supervision battery parameters to the second subsystem; receive the initial neural network updated by the second subsystem, and increase the number of training times by 1; determine whether the number of training times reaches a first preset number; if so, use the initial neural network as the target neural network.

[0100] The first subsystem inputs the abnormal battery parameters into the initial neural network to obtain a forward output result; and sends the forward output result and the initial neural network to the second subsystem.

[0101] The second subsystem updates the initial neural network according to the supervisory battery parameters and the forward output results, and sends the updated initial neural network to the main control battery system.

[0102] For the specific limitations of a wireless battery management system based on computing power collaboration provided in this embodiment, please refer to the above embodiment of a wireless battery management method based on computing power collaboration, which will not be repeated here. Each module in the above-mentioned wireless battery management system based on computing power collaboration can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0103] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0104] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A wireless battery management method based on computing power collaboration, characterized in that: include: The main control battery system receives abnormal battery parameters, supervision battery parameters and computing power coordination requests from abnormal subsystems; The main control battery system determines each normal subsystem according to the abnormal subsystem, and obtains the remaining computing power of each normal subsystem; determines the first subsystem and the second subsystem according to the remaining computing power; The master control battery system, the first subsystem, and the second subsystem train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain a target neural network; The main control battery system sends the target neural network to the abnormal subsystem; The abnormal subsystem inputs the abnormal battery parameters into the target neural network to obtain the requested result.

2. The wireless battery management method based on computing power collaboration according to claim 1 is characterized in that: The determining the first subsystem and the second subsystem according to the remaining computing power includes: Sorting the remaining computing power of each normal subsystem; Selecting the normal subsystem with the largest remaining computing power as the first subsystem; The normal subsystem with the second largest remaining computing power is selected as the second subsystem.

3. The wireless battery management method based on computing power collaboration according to claim 1 is characterized in that: The master control battery system, the first subsystem, and the second subsystem train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain a target neural network, including: Step 1: The main control battery system sends the abnormal battery parameters and the initial neural network to the first subsystem; The first subsystem inputs the abnormal battery parameter into the initial neural network to obtain a forward output result; The first subsystem sends the forward output result and the initial neural network to the second subsystem; The master control battery system sends the supervisory battery parameter to the second subsystem; The second subsystem updates the initial neural network according to the supervisory battery parameters and the forward output result, and sends the updated initial neural network to the main control battery system; The main control battery system receives the initial neural network, increases the number of training times by 1, determines whether the number of training times reaches a first preset number, if so, uses the initial neural network as the target neural network, if not, returns to the first step.

4. The wireless battery management method based on computing power collaboration according to claim 1 is characterized in that: The master control battery system, the first subsystem, and the second subsystem train the initial neural network corresponding to the computing power coordination request based on the abnormal battery parameters and the supervisory battery parameters to obtain a target neural network, including: Step 1: The main control battery system sends the abnormal battery parameters and the initial neural network to the first subsystem; The first subsystem inputs the abnormal battery parameter into the initial neural network to obtain a forward output result; The first subsystem sends the forward output result to the main control battery system; When the second training number is equal to 0, the master control battery system updates the initial neural network according to the supervisory battery parameter and the forward output result, and increases the first training number by 1; determines whether the first training number reaches a second preset threshold; if so or the second training number is not equal to 0, sends the forward output result, the supervisory battery parameter and the initial neural network to the second subsystem; if not, returns to the first step; The second subsystem updates the initial neural network according to the forward output result and the supervisory battery parameter, and sends the updated initial neural network to the main control battery system; After receiving the updated initial neural network from the second subsystem, the master control battery system increases the second training times by 1; determines whether the second training reaches a third preset threshold at this time; if so, uses the received initial neural network as the target neural network; if not, returns to the first step.

5. The wireless battery management method based on computing power collaboration according to claim 1, characterized in that: Also includes: The subsystem samples data of the target battery to obtain battery sampling data; If a preset number of consecutive battery sampling data are all abnormal data, the subsystem is an abnormal subsystem, the supervisory battery parameters are obtained according to normal historical battery parameters, and the abnormal battery parameters are generated based on each abnormal data; The supervisory battery parameter, the abnormal battery parameter and the computing power coordination request are sent to the master control battery system.

6. The wireless battery management method based on computing power collaboration according to claim 1, characterized in that: Also includes: After obtaining the request result, the exception subsystem determines the exception type according to the request result; The abnormal subsystem sends the abnormal type to the main control battery system.

7. The wireless battery management method based on computing power collaboration according to claim 1 is characterized in that: The abnormal battery parameters include voltage data, current data, temperature data or remaining capacity.

8. A wireless battery management system based on computing power collaboration, characterized in that: include: A main control battery system and a plurality of subsystems in wireless communication with the main control battery system; The master control battery system is used to receive abnormal battery parameters, supervisory battery parameters and computing power coordination requests from abnormal subsystems; Determine each normal subsystem according to the abnormal subsystem, and obtain the remaining computing power of each normal subsystem; and, determining a first subsystem and a second subsystem according to the remaining computing power, training an initial neural network corresponding to the computing power coordination request with the first subsystem and the second subsystem based on the abnormal battery parameters and the supervisory battery parameters to obtain a target neural network, and sending the target neural network to the abnormal subsystem; The abnormal subsystem is used to input the abnormal battery parameters into the target neural network to obtain the requested result.

9. The wireless battery management system based on computing power collaboration according to claim 8, characterized in that: The main control battery system is specifically used to sort the remaining computing power of each normal subsystem; select the normal subsystem with the largest remaining computing power as the first subsystem; and select the normal subsystem with the second largest remaining computing power as the second subsystem.

10. The wireless battery management system based on computing power collaboration according to claim 8, characterized in that: The main control battery system is specifically used to send the abnormal battery parameters and the initial neural network to the first subsystem; send the supervisory battery parameters to the second subsystem; receive the initial neural network updated by the second subsystem, increase the number of training times by 1; and determine whether the number of training times reaches a first preset number; If yes, then use the initial neural network as the target neural network; The first subsystem inputs the abnormal battery parameter into the initial neural network to obtain a forward output result; and sends the forward output result and the initial neural network to the second subsystem; The second subsystem updates the initial neural network according to the supervisory battery parameters and the forward output result, and sends the updated initial neural network to the main control battery system.

Images