Energy storage wireless bms system multi-modal fault prediction method, device and storage medium

Abstract

A method, apparatus, device, and storage medium for multimodal fault prediction in an energy storage wireless BMS system are disclosed. The method involves collecting preset modal data at preset time intervals, extracting spatial and local temporal features from battery operating parameters, and extracting globally dependent feature sequences from communication signals and environmental information. These spatial, local temporal, and globally dependent feature sequences are then fused to obtain a fused feature vector. This fused feature vector is input into a multipath topology model for analysis to obtain the probability of different fault types for each battery. By fusing battery operating parameters, communication signals, and environmental data and performing multimodal fault prediction based on a multipath topology model, the method exhibits high sensitivity, low false alarm rate, and significantly improves fault prediction accuracy.

Description

Technical Field

[0001] This invention belongs to the field of intelligent energy storage operation and maintenance technology, and in particular relates to a multimodal fault prediction method, device and storage medium for an energy storage wireless BMS system. Background Technology

[0002] With the development of the new energy industry and the widespread application of energy storage systems, battery management systems (BMS) play a crucial role in ensuring battery safety, extending battery life, and improving system reliability. Especially in large-scale energy storage power stations, electric vehicles, and industrial-grade energy storage systems, the performance of the BMS directly affects the stable operation of the entire system.

[0003] Existing BMS systems generally employ threshold-based fault detection and alarm mechanisms using single parameters such as voltage and temperature. This means that when the battery voltage exceeds the normal range or the temperature rises abnormally, the system triggers a protection mechanism or alarm response. While this method is relatively simple to implement and has a fast response time, it has significant limitations: Firstly, many actual faults do not immediately cause abnormal abrupt changes in voltage or temperature; for example, latent faults such as poor contact, micro-short circuits, and increased internal resistance are difficult to detect in a timely manner. Secondly, this type of diagnostic method based on single-variable thresholds is ill-suited to address the nonlinear evolution trends that occur during system operation and cannot identify potential risks from multi-source information. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method, device and storage medium for multimodal fault prediction in an energy storage wireless BMS system. By integrating battery operating parameters, communication signals and environmental data and performing multimodal fault prediction based on a multipath topology model, it has high sensitivity and low false alarm rate, and can significantly improve fault prediction accuracy.

[0005] This application provides a method for multimodal fault prediction in an energy storage wireless BMS system, including:

[0006] Preset modal data is collected at preset time intervals, including battery operating parameters, communication signals, and environmental information;

[0007] Spatial and local temporal features are extracted from the battery operating parameters, and global dependency feature sequences are extracted from the communication signals and environmental information.

[0008] The spatial features, the local temporal features, and the global dependency feature sequence are fused to obtain a fused feature vector.

[0009] The fused feature vectors are input into a multipath topology model for analysis to obtain the different types of failure probabilities for each battery.

[0010] In one embodiment, extracting spatial features and local temporal features from the battery operating parameters includes:

[0011] The battery operating parameters are input into a convolutional neural network, and analyzed using sliding convolution kernels of different lengths and strides to obtain the spatial features and the local temporal features. The spatial features refer to the differences and correlations in the states of different batteries at the same time. The local temporal features refer to the trend of state changes of the same battery within a certain time period.

[0012] In one embodiment, extracting the globally dependent feature sequence from the communication signal and the environmental information includes:

[0013] The communication signal and the environmental information are input into a deep learning model based on a self-attention mechanism for analysis to obtain the global dependency feature sequence; the global dependency feature sequence includes the cross-modal correlation between the communication signal and the environmental information within a preset time period.

[0014] In one embodiment, the step of inputting the communication signal and the environmental information into a deep learning model based on a self-attention mechanism for analysis to obtain the globally dependent feature sequence includes:

[0015] The communication signal and the environmental information are input into a deep learning model based on a self-attention mechanism, and then encoded by a modality-specific encoder to obtain a communication modality feature sequence and an environmental modality feature sequence.

[0016] The communication mode feature sequence and the environment mode feature sequence are aligned in time to form an input feature sequence;

[0017] Based on the dynamic time adaptive rule, the input feature sequence is analyzed for same-modal and different-modal attention responses to obtain the global dependent feature sequence.

[0018] In one embodiment, the step of performing same-modal and different-modal attention response analysis on the input feature sequence based on dynamic time adaptive rules to obtain the globally dependent feature sequence includes:

[0019] By fusing time decay factor and modality bias, and performing long-term correlation analysis between dynamic same-modality and different-modality features on the input feature sequence based on attention weight, a globally dependent feature sequence with context awareness is obtained.

[0020] In one embodiment, the multipath topology model includes: a multipath topology network model composed of three adjacency matrices, wherein the three adjacency matrices represent different fault propagation channels.

[0021] In one embodiment, the step of inputting the fused feature vector into a multi-path topology model for analysis to obtain the different types of failure probabilities for each battery includes:

[0022] The fused feature vector is input into the multi-path topology model, and analysis is performed in the three adjacency matrices based on the weights of different propagation paths.

[0023] The analysis results of the three adjacency matrices are aggregated, and the different types of failure probabilities of each battery are obtained based on the aggregated results.

[0024] A second aspect of this application provides a multimodal fault prediction device for an energy storage wireless BMS system, comprising:

[0025] The acquisition module is used to acquire preset mode data at preset time intervals. The preset mode data includes battery operating parameters, communication signals, and environmental information.

[0026] The extraction module is used to extract spatial features and local temporal features from the battery operating parameters, and to extract global dependency feature sequences from the communication signals and the environmental information;

[0027] The fusion module is used to fuse the spatial features, the local temporal features, and the global dependency feature sequence to obtain a fused feature vector;

[0028] The analysis module is used to input the fused feature vector into the multipath topology model for analysis to obtain the different types of failure probabilities of each battery.

[0029] In one embodiment, the extraction module is specifically used for:

[0030] The battery operating parameters are input into a convolutional neural network, and analyzed using sliding convolution kernels of different lengths and strides to obtain the spatial features and the local temporal features. The spatial features refer to the differences and correlations in the states of different batteries at the same time. The local temporal features refer to the trend of state changes of the same battery within a certain time period.

[0031] In one embodiment, the extraction module is specifically used for:

[0032] The communication signal and the environmental information are input into a deep learning model based on a self-attention mechanism for analysis to obtain the global dependency feature sequence; the global dependency feature sequence includes the cross-modal correlation between the communication signal and the environmental information within a preset time period.

[0033] In one embodiment, the extraction module is specifically used for:

[0034] The communication signal and the environmental information are input into a deep learning model based on a self-attention mechanism, and then encoded by a modality-specific encoder to obtain a communication modality feature sequence and an environmental modality feature sequence.

[0035] The communication mode feature sequence and the environment mode feature sequence are aligned in time to form an input feature sequence;

[0036] Based on the dynamic time adaptive rule, the input feature sequence is analyzed for same-modal and different-modal attention responses to obtain the global dependent feature sequence.

[0037] In one embodiment, the extraction module is specifically used for:

[0038] By fusing time decay factor and modality bias, and performing long-term correlation analysis between dynamic same-modality and different-modality features on the input feature sequence based on attention weight, a globally dependent feature sequence with context awareness is obtained.

[0039] In one embodiment, the multipath topology model includes: a multipath topology network model composed of three adjacency matrices, wherein the three adjacency matrices represent different fault propagation channels.

[0040] In one embodiment, the analysis module is specifically used for:

[0041] The fused feature vector is input into the multi-path topology model, and analysis is performed in the three adjacency matrices based on the weights of different propagation paths.

[0042] The analysis results of the three adjacency matrices are aggregated, and the different types of failure probabilities of each battery are obtained based on the aggregated results.

[0043] The third aspect of this application provides a multimodal fault prediction device for an energy storage wireless BMS system, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor; the processor executes the computer program to implement the steps of the method described in the first aspect above.

[0044] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect above.

[0045] The beneficial effects of this application's embodiments are as follows: By collecting preset modal data at preset time intervals, spatial features and local temporal features are extracted from battery operating parameters, and globally dependent feature sequences are extracted from communication signals and environmental information; the spatial features, local temporal features, and globally dependent feature sequences are fused to obtain a fused feature vector; the fused feature vector is input into a multipath topology model for analysis to obtain the probability of different types of faults for each battery. By fusing battery operating parameters, communication signals, and environmental data and performing multimodal fault prediction based on a multipath topology model, high sensitivity and low false alarm rate are achieved, significantly improving fault prediction accuracy. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A schematic diagram illustrating the implementation process of a multimodal fault prediction method for an energy storage wireless BMS system provided in an embodiment of this application;

[0048] Figure 2 A schematic diagram of a multimodal fault prediction device for an energy storage wireless BMS system provided in an embodiment of this application;

[0049] Figure 3 This is a schematic diagram of a multimodal fault prediction device for an energy storage wireless BMS system provided in an embodiment of this application. Detailed Implementation

[0050] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0052] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0053] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0054] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0055] In the description of the embodiments of this application, the term "multiple frames" refers to two or more (including two).

[0056] In the description of the embodiments of this application, the technical terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this application and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0057] This invention provides a multimodal fault prediction method for an energy storage wireless BMS system, which can predict potential faults of the energy storage system in real time based on multi-dimensional operational data analysis, so as to take maintenance measures in advance and improve the safety and reliability of system operation.

[0058] Please see Figure 1 As shown, Figure 1 This is a schematic diagram illustrating the implementation flow of a multimodal fault prediction method for an energy storage wireless BMS system according to an embodiment of this application. This multimodal fault prediction method for an energy storage wireless BMS system is applicable to energy storage wireless BMS systems and is implemented by a multimodal fault prediction device for the energy storage wireless BMS system. Details are as follows:

[0059] S110: Collect preset mode data at preset time intervals. The preset mode data includes battery operating parameters, communication signals, and environmental information.

[0060] For example, battery operating parameters include basic electrical indicators such as voltage, current, temperature, and state of charge. Communication signals include indicators such as wireless signal strength, packet loss rate, and retransmission count. Environmental information includes vibration level, ambient temperature, external temperature, and EMI interference intensity.

[0061] S120: Extract spatial and local temporal features from battery operating parameters, and extract global dependency feature sequences from communication signals and environmental information.

[0062] It should be noted that the battery operating parameters, communication signals, and environmental information can first undergo time synchronization processing, missing value imputation, noise filtering, and normalization processing to unify the data scale and ensure the accuracy and stability of feature extraction. The specific processes of time synchronization processing, missing value imputation, noise filtering, and normalization are not explained in detail here.

[0063] In one embodiment, extracting spatial features and local temporal features from battery operating parameters includes: inputting the battery operating parameters into a convolutional neural network, analyzing them in the convolutional neural network using sliding convolution kernels of different lengths and strides to obtain spatial features and local temporal features; spatial features refer to the differences and correlations in the states of different batteries at the same time; local temporal features refer to the trend of state changes of the same battery within a certain time period.

[0064] For example, suppose a wireless BMS node in an energy storage system collects the following battery operating parameters once per second: voltage, current, temperature, and state of charge (SOC). A 60-row × 4-column time-series data matrix is ​​generated every 10 minutes. This 60-row × 4-column (time × channel) data is input into a convolutional neural network (CNN) such as a 1D-CNN. In this CNN, sliding convolution kernels of different lengths and strides are used to extract state change trends under different time windows, such as rapid voltage drops (short circuit signs), abnormal SOC oscillations (sampling distortion), and sharp temperature increases (abnormal heat dissipation). These state change trends under different time windows are used as corresponding local time-series features. The state parameters of different batteries at each time point where local trend changes occur are extracted, and the differences and correlations of the state parameters are analyzed to obtain spatial features. This embodiment uses CNN, which can effectively detect small fluctuation patterns within a short time period and is more suitable than RNN for extracting local time-series features. This embodiment, by combining spatial features with local time-series features, not only helps to locate local anomalies and determine structural problems (spatial features), but also helps to identify fault trends and predict future states (local time-series features).

[0065] In one embodiment, extracting a globally dependent feature sequence from communication signals and environmental information includes: inputting the communication signals and environmental information into a deep learning model based on a self-attention mechanism for analysis to obtain a globally dependent feature sequence; the globally dependent feature sequence includes the cross-modal correlation between the communication signals and environmental information within a preset time period.

[0066] For example, the communication signal and the environmental information are input into a deep learning model based on a self-attention mechanism for analysis to obtain the globally dependent feature sequence, including: inputting the communication signal and environmental information into the deep learning model based on a self-attention mechanism, encoding them respectively through a modality-specific encoder to obtain a communication modality feature sequence and an environmental modality feature sequence; aligning the communication modality feature sequence and the environmental modality feature sequence by time to form an input feature sequence; and performing same-modality and different-modality attention response analysis on the input feature sequence based on dynamic time adaptive rules to obtain the globally dependent feature sequence.

[0067] It's important to note that conventional self-attention mechanisms, when processing time-series data, assume by default that the influence between each time point is equivalent or statically defined. This step, however, introduces dynamic time-adaptive rules, allowing the model to autonomously determine "which time period to focus on" and "the intensity of attention" based on the structure of historical data or the structure between data points. Furthermore, traditional self-attention mechanisms are prone to "mixed-modality" issues when processing multimodal data, meaning they cannot distinguish the differences in the contributions of different modalities to the attention distribution. This embodiment, however, utilizes same-modal and different-modal attention response analysis to explicitly guide the model in analyzing the internal correlations between same-modalities (such as between communication features) and the interactive effects between different modalities (such as communication and the environment); this has significant value in improving the accuracy of cross-modal dependency analysis.

[0068] In one embodiment, the input feature sequence is subjected to attention response analysis for both same-modality and different-modality based on dynamic time adaptive rules to obtain a globally dependent feature sequence, including: fusing time decay factor and modality bias, and performing long-term correlation analysis between dynamic same-modality and different-modality based on attention weights to obtain a globally dependent feature sequence with context awareness.

[0069] For example, the time decay factor is expressed as: ;in, For learnable time-aware parameters, The interval between the i-th time point and the j-th time point; This is a time decay factor that controls the effect of time span on attention distribution.

[0070] Attention weights are represented as: ;in, These are modal tags used to distinguish between the same modality and different modalities. =1 indicates the same mode. =0 indicates a heteromodal state; The modal bias coefficient is... The time decay coefficient, and These are used to dynamically adjust attention weights and guide information fusion; , For attention key-value pairs, To query the distance, This represents the fusion attention weight between the i-th time point and the j-th time point.

[0071] Finally, by fusing the time decay factor and modality bias, and performing a dynamic long-term correlation analysis between the same and different modalities of the input feature sequence based on attention weights, the resulting context-aware global dependency feature sequence at time point i is expressed as:

[0072] ;in, Let be the context-aware feature representation at the i-th time point. The feature vector is generated by linear mapping from the feature vector corresponding to the j-th time point or mode in the input feature sequence.

[0073] By aligning communication signals (such as RSSI, packet loss rate, and retransmission count) with environmental information (such as temperature, EMI, and humidity) in time, an input feature sequence is formed. Then, learnable time-aware parameters are introduced to model the time-distance impact of input features at different time points using a time decay factor, enhancing the flexibility of long-term dependency modeling. Furthermore, a modality label matrix Mᵢⱼ is constructed to distinguish between same-mode and different-mode features. Attention weight allocation is guided by the modality bias coefficient β and the time decay coefficient γ, and the context-aware features at each time point are output through an attention mechanism. That is, a sequence of global feature representations that includes cross-modal relationships and dynamic time dependencies. = { ,..., }

[0074] Through the above process, the strength of dependencies across time dimensions and the complementary information of cross-modal features can be perceived simultaneously, thereby generating a global dependency feature sequence with context awareness capabilities, providing multi-dimensional feature support for subsequent fault analysis and state assessment.

[0075] S130: The spatial features, local temporal features, and global dependent feature sequences are fused to obtain a fused feature vector.

[0076] Spatial features, local temporal features, and global dependency feature sequences can be fused using concatenation or weighted summation to obtain a fused feature vector. Specifically, this embodiment does not limit the fusion method.

[0077] S140: Input the fused feature vector into the multipath topology model for analysis to obtain the different types of failure probabilities for each battery.

[0078] In one embodiment, the multipath topology model includes: a multipath topology network model composed of three adjacency matrices, the three adjacency matrices representing different fault propagation channels.

[0079] Specifically, the three adjacency matrices include the physical connection adjacency matrix, which represents the electrical connection relationship between batteries, such as series and parallel structures; the thermal coupling adjacency matrix, which represents the thermal conduction or temperature rise relationship between batteries; and the communication link adjacency matrix, which represents the communication structure and relay path between wireless communication nodes.

[0080] For example, the fused feature vector is input into a multi-path topology model for analysis to obtain the different types of failure probabilities for each battery, including: inputting the fused feature vector into the multi-path topology model and analyzing it based on the weights of different propagation paths in three adjacency matrices; aggregating the analysis results of the three adjacency matrices and obtaining the different types of failure probabilities for each battery based on the aggregation results.

[0081] Specifically, by inputting the fused feature vector into the multi-path topology model, graph neural network propagation and feature updates are performed under each path corresponding to the adjacency matrix to capture the impact of different structural relationships on fault evolution. Subsequently, the node feature update results from the three paths are aggregated (such as weighted fusion or attention fusion) to obtain node representations containing multi-source dependency information. Finally, based on the fused node feature vector, it is input into the fault classification sub-network or probabilistic inference module to output the predicted probability of each battery experiencing different types of faults (such as overheating, overcharging, communication interruption, etc.) at the current time.

[0082] As can be seen from the above analysis, the multimodal fault prediction method for energy storage wireless BMS systems provided in this application collects preset modal data at preset time intervals, extracts spatial and local temporal features from battery operating parameters, and extracts globally dependent feature sequences from communication signals and environmental information; it then fuses the spatial features, local temporal features, and globally dependent feature sequences to obtain a fused feature vector; and finally, it inputs the fused feature vector into a multipath topology model for analysis to obtain the different types of fault probabilities for each battery. By fusing battery operating parameters, communication signals, and environmental data and performing multimodal fault prediction based on a multipath topology model, it exhibits high sensitivity, low false alarm rate, and can significantly improve fault prediction accuracy.

[0083] Please see Figure 2 , Figure 2 This is a schematic diagram of a multimodal fault prediction device for an energy storage wireless BMS system according to an embodiment of this application. The modules or units included in the multimodal fault prediction device for the energy storage wireless BMS system are used to perform... Figure 1 The steps in the corresponding embodiments. Please refer to the details. Figure 1 The relevant descriptions in the corresponding embodiments are shown below. For ease of explanation, only the parts relevant to this embodiment are shown. See also... Figure 2 A multimodal fault prediction device 200 for an energy storage wireless BMS system includes:

[0084] The acquisition module 210 is used to acquire preset mode data at preset time intervals. The preset mode data includes battery operating parameters, communication signals and environmental information.

[0085] Extraction module 220 is used to extract spatial features and local temporal features from the battery operating parameters, and to extract global dependency feature sequences from the communication signals and the environmental information;

[0086] The fusion module 230 is used to fuse the spatial features, the local temporal features, and the global dependency feature sequence to obtain a fused feature vector;

[0087] Analysis module 240 is used to input the fused feature vector into the multipath topology model for analysis to obtain the different types of failure probabilities of each battery.

[0088] In one embodiment, the extraction module 220 is specifically used for:

[0089] The battery operating parameters are input into a convolutional neural network, and analyzed using sliding convolution kernels of different lengths and strides to obtain the spatial features and the local temporal features. The spatial features refer to the differences and correlations in the states of different batteries at the same time. The local temporal features refer to the trend of state changes of the same battery within a certain time period.

[0090] In one embodiment, the extraction module 220 is specifically used for:

[0091] The communication signal and the environmental information are input into a deep learning model based on a self-attention mechanism for analysis to obtain the global dependency feature sequence; the global dependency feature sequence includes the cross-modal correlation between the communication signal and the environmental information within a preset time period.

[0092] In one embodiment, the extraction module 220 is specifically used for:

[0093] The communication signal and the environmental information are input into a deep learning model based on a self-attention mechanism, and then encoded by a modality-specific encoder to obtain a communication modality feature sequence and an environmental modality feature sequence.

[0094] The communication mode feature sequence and the environment mode feature sequence are aligned in time to form an input feature sequence;

[0095] Based on the dynamic time adaptive rule, the input feature sequence is analyzed for same-modal and different-modal attention responses to obtain the global dependent feature sequence.

[0096] In one embodiment, the extraction module 220 is specifically used for:

[0097] By fusing time decay factor and modality bias, and performing long-term correlation analysis between dynamic same-modality and different-modality features on the input feature sequence based on attention weight, a globally dependent feature sequence with context awareness is obtained.

[0098] In one embodiment, the multipath topology model includes: a multipath topology network model composed of three adjacency matrices, wherein the three adjacency matrices represent different fault propagation channels.

[0099] In one embodiment, the analysis module 240 is specifically used for:

[0100] The fused feature vector is input into the multi-path topology model, and analysis is performed in the three adjacency matrices based on the weights of different propagation paths.

[0101] The analysis results of the three adjacency matrices are aggregated, and the different types of failure probabilities of each battery are obtained based on the aggregated results.

[0102] Please see Figure 3 , Figure 3 This is a schematic diagram of a multimodal fault prediction device for an energy storage wireless BMS system provided in an embodiment of this application. Figure 3 It is understood that the multimodal fault prediction device 300 for an energy storage wireless BMS system includes: a processor 310, a memory 320, and a computer program 330 stored in the memory 320 and executable on the processor 310; when the processor 310 executes the computer program 330, it implements the steps in the above-described embodiments of the multimodal fault prediction method for an energy storage wireless BMS system, for example... Figure 1 The steps S110 to S140 are shown. Alternatively, when the processor 310 executes the computer program 330, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 2 The functions of modules 210 to 240 are shown.

[0103] For example, the computer program 330 can be divided into one or more modules / units, one or more of which are stored in the memory 320 and executed by the processor 310 to complete this application. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 330 in the multimodal fault prediction device of the energy storage wireless BMS system. For example, the computer program 330 can be divided into an acquisition module, an extraction module, a fusion module, and an analysis module.

[0104] The multimodal fault prediction device for the energy storage wireless BMS system provided in this embodiment may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that... Figure 3 This is merely an example of a multimodal fault prediction device for an energy storage wireless BMS system and does not constitute a limitation on such devices. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, a multimodal fault prediction device for an energy storage wireless BMS system may also include input / output devices, network access devices, buses, etc.

[0105] The processor 310 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0106] The memory 320 can be an internal storage unit of the multimodal fault prediction device for the energy storage wireless BMS system, such as a hard drive or memory. The memory 320 can also be an external storage device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, or flash card. Furthermore, the multimodal fault prediction device for the energy storage wireless BMS system can include both internal and external storage units. The memory 320 is used to store computer programs and other programs and data required by the energy storage wireless BMS system. The memory 320 can also be used to temporarily store data that has been output or will be output.

[0107] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0108] This application also provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above method embodiments.

[0109] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0110] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the above-described method embodiments.

[0111] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0112] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0113] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0114] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0115] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0116] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A multimodal fault prediction method for an energy storage wireless BMS system, characterized in that, include: Preset modal data is collected at preset time intervals, including battery operating parameters, communication signals, and environmental information; The battery operating parameters are input into a convolutional neural network, where they are analyzed using sliding convolutional kernels of different lengths and strides to obtain spatial features and local temporal features. Spatial features refer to the state differences and correlations of different batteries at the same time. Local temporal features refer to the state change trends of the same battery within a certain time period. The communication signal and environmental information are input into a deep learning model based on a self-attention mechanism, and encoded by a modality-specific encoder to obtain a communication modality feature sequence and an environmental modality feature sequence. The communication modality feature sequence and the environmental modality feature sequence are aligned by time to form an input feature sequence. By fusing time decay factor and modality bias, and performing long-term correlation analysis between dynamic same-modality and different-modality features on the input feature sequence based on attention weight, a global dependency feature sequence with context awareness is obtained. The global dependency feature sequence includes the cross-modal correlation between the communication signal and the environmental information within a preset time period; A multipath topology model, comprising: a multipath topology network model composed of three adjacency matrices, wherein the three adjacency matrices represent different fault propagation channels; The spatial features, the local temporal features, and the global dependency feature sequence are fused to obtain a fused feature vector. The fused feature vectors are input into a multipath topology model for analysis to obtain the different types of failure probabilities for each battery.

2. The multimodal fault prediction method for an energy storage wireless BMS system as described in claim 1, characterized in that, The step of inputting the fused feature vector into a multi-path topology model for analysis to obtain the different types of failure probabilities for each battery includes: The fused feature vector is input into the multi-path topology model, and analysis is performed in the three adjacency matrices based on the weights of different propagation paths. The analysis results of the three adjacency matrices are aggregated, and the different types of failure probabilities of each battery are obtained based on the aggregated results.

3. A multimodal fault prediction device for an energy storage wireless BMS system, characterized in that, include: The acquisition module is used to acquire preset mode data at preset time intervals. The preset mode data includes battery operating parameters, communication signals, and environmental information. An extraction module is used to input the battery operating parameters into a convolutional neural network, and analyze them through sliding convolutional kernels of different lengths and strides to obtain spatial features and local temporal features. The spatial features refer to the state differences and correlations of different batteries at the same time. The local temporal features refer to the state change trends of the same battery within a certain time period. The communication signals and environmental information are input into a deep learning model based on a self-attention mechanism, and encoded by a modality-specific encoder to obtain communication modality feature sequences and environmental modality feature sequences. The communication modality feature sequences and the environmental modality feature sequences are aligned by time to form an input feature sequence. By fusing time decay factor and modality bias, and performing long-term correlation analysis between dynamic same-modality and different-modality features on the input feature sequence based on attention weight, a global dependency feature sequence with context awareness is obtained. The global dependency feature sequence includes the cross-modal correlation between the communication signal and the environmental information within a preset time period; A multipath topology model, comprising: a multipath topology network model composed of three adjacency matrices, wherein the three adjacency matrices represent different fault propagation channels; The fusion module is used to fuse the spatial features, the local temporal features, and the global dependency feature sequence to obtain a fused feature vector; The analysis module is used to input the fused feature vector into the multipath topology model for analysis to obtain the different types of failure probabilities of each battery.

4. A multimodal fault prediction device for an energy storage wireless BMS system, characterized in that, include: Processor, memory, and computer programs stored in said memory and executable on said processor; When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 2.

5. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 2.

Images