Portable detection device, method and product of energy storage system

Abstract

The embodiment of the invention provides a portable detection device, method and product for an energy storage system. The device comprises an adapter and an operation state detection module, the operation state detection module comprises a state evaluation module and a fault type determination module; the adapter is used for being connected with energy storage systems of different communication protocols in a pluggable mode, adapting to the communication protocols of the connected energy storage systems to conduct detection data collection, converting the detection data into target data and sending the target data to the operation state detection module. The target data is detection data under a target communication protocol supported by the running state detection module; the state evaluation module is used for carrying out anomaly evaluation on the target data through a preset detection rule and carrying out corresponding fault warning based on an evaluation result; and the fault type determination module is used for processing the abnormal target data through the fault diagnosis model and determining the fault type corresponding to the fault alarm. The invention aims to improve the compatibility and accuracy of the operation state detection of the energy storage system.

Description

Technical Field

[0001] This application relates to the technical field of abnormal detection of energy storage systems, and specifically relates to a portable detection device, method and product for energy storage systems. Background Art

[0002] As an important part of new energy, electrochemical energy storage power stations are playing an increasingly important role in peak shaving, standby and emergency response of power systems. With the expansion of the scale of energy storage power stations, the demand for state detection and defect diagnosis of energy storage systems is increasing. The existing technologies mainly rely on the monitoring system software platform provided by EPC procurement or energy storage integrators for state evaluation and fault diagnosis, which is strictly bound to a single power station and has poor compatibility. Summary of the Invention

[0003] In view of this, embodiments of this application provide a portable detection device, method and product for energy storage systems. The aim is to improve the compatibility and accuracy of the operation state detection of energy storage systems.

[0004] The first aspect of this application provides a portable detection device for an energy storage system, and the device includes:

[0005] The device includes: an adapter and an operation state detection module; the operation state detection module includes: a state evaluation module and a fault type determination module;

[0006] The adapter is used to be pluggably connected to energy storage systems with different communication protocols, adapt to the communication protocol of the connected energy storage system for detecting data acquisition, and convert the detection data into target data and send it to the operation state detection module, where the target data is the detection data under the target communication protocol supported by the operation state detection module;

[0007] The state evaluation module is used to perform abnormal evaluation on the target data through preset detection rules and perform corresponding fault warnings based on the evaluation results;

[0008] The fault type determination module is used to process the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault warning.

[0009] Optionally, the adapter includes a data acquisition module, which is used to determine the communication protocol of the energy storage system, perform protocol configuration based on the determined communication protocol, and perform detection data acquisition on the energy storage system based on the configuration result;

[0010] The data acquisition module includes:

[0011] The basic data acquisition module is used to acquire the basic detection data at the cell level, module level, and container level;

[0012] The differential data acquisition module is used to acquire the first-level detection data at the cell level; and, to acquire the second-level detection data at the module level; and, to acquire the third-level detection data at the container level.

[0013] Optionally, the device further includes: a cross-level status analysis module and a data mining analysis module;

[0014] The status evaluation module is used to perform anomaly evaluation on the target data at different levels through a preset detection rule to obtain the corresponding evaluation result;

[0015] The status evaluation module is further used to determine the target level to which the target data with anomalies belongs according to the evaluation result, and perform a fault alarm for the target level;

[0016] The fault type determination module is used to process the target data at the target level through a fault diagnosis model to determine the fault type corresponding to the fault alarm at the target level;

[0017] The cross-level status analysis module is used to perform a correlation analysis on the target data at different levels, and determine the current status of the energy storage system at each level based on the correlation analysis result;

[0018] The data mining analysis module is used to analyze the target data at different levels for a historical preset duration through a fault prediction model to predict the status of the energy storage system at each level.

[0019] Optionally, the fault diagnosis model in the fault type determination module includes:

[0020] The first data input module is used to receive the target data at the target level, and perform standardization and time window alignment processing to obtain the corresponding three-dimensional input tensor;

[0021] The spatial feature extraction module is used to extract spatial features from the three-dimensional input tensor through a convolutional neural network;

[0022] The temporal feature extraction module is used to extract temporal features from the three-dimensional input tensor through a bidirectional long short-term memory network;

[0023] The multi-modal fusion module is used to perform dynamic weight allocation on the spatial features and the temporal features through a self-attention mechanism and calculate the cross-modal correlation to obtain a fused feature vector;

[0024] A diagnostic output module, which is used to process the fused feature vector through a fully connected layer and a classifier to determine the fault types at the abnormal target level.

[0025] Optionally, the fault diagnosis model in the fault type determination module includes:

[0026] A second data input module, which is used to receive the target data at the target level, perform normalization and time window alignment processing, and obtain the corresponding three-dimensional input tensor;

[0027] A supervised learning sub-model, which is used to process the three-dimensional input tensor to determine the probabilities corresponding to various fault types at the abnormal target level;

[0028] A time series sub-model, which is used to process the time series type tensor in the three-dimensional input tensor to determine the probabilities corresponding to various fault types at the abnormal target level;

[0029] A statistical analysis sub-model, which is used to perform distribution anomaly analysis on the target data at the abnormal target level to determine the probabilities corresponding to various fault types at the abnormal target level;

[0030] A first result fusion module, which is used to assign weights according to the confidence levels of each sub-model, fuse the output results of each sub-model, and determine the final fault type at the abnormal target level.

[0031] Optionally, the fault prediction model in the data mining and analysis module includes:

[0032] A third data input module, which is used to receive the target data at different levels in the historical preset duration, perform normalization and time window alignment processing, and obtain the corresponding three-dimensional input tensor;

[0033] A statistical sub-model, which is used to perform linear trend prediction on the target data at different levels in the historical preset duration to obtain the corresponding first prediction result;

[0034] A deep learning sub-model, which is used to perform non-linear feature extraction and prediction on the three-dimensional input tensor to obtain the corresponding second prediction result;

[0035] A second result fusion module, which is used to perform dynamic weight assignment to each sub-model according to the past prediction errors of each sub-model, and perform weighted fusion on the first prediction result and the second prediction result based on the weight assignment result to obtain the final prediction result.

[0036] Optionally, the fault prediction model in the data mining and analysis module includes:

[0037] The fourth data input module is used to receive target data at different levels for a historical preset duration, perform normalization and time window alignment processing, and obtain corresponding three-dimensional input tensors;

[0038] The clustering analysis module is used to perform clustering analysis on the three-dimensional input tensors through a clustering algorithm to determine the battery states at different levels;

[0039] The association module is used to determine the fault types corresponding to the data of abnormal battery states under the current environmental data based on pre-established abnormal association rules;

[0040] The health prediction module is used to predict the battery health state scores and remaining lifetimes at different levels according to the capacity attenuation rates determined by the three-dimensional input tensors.

[0041] Optionally, the adapter includes a communication module, which is used to convert the detection data into target data based on the communication protocol of the energy storage system and send it to the operation state detection module.

[0042] Optionally, the device further includes:

[0043] An algorithm library, which is used to store various algorithms required by the operation state detection module and connect to the operation state detection module through an API interface to provide various algorithms required for state detection.

[0044] Optionally, the device is connected to an algorithm management platform;

[0045] The algorithm management platform is used to obtain the detection data of each energy storage system for the effect evaluation of various algorithms, and optimize and update the various algorithms based on the evaluation results.

[0046] The second aspect of the present application provides an operation state detection method for an energy storage system. The method is applied to a portable detection device for an energy storage system according to the first aspect of the present application. The method includes:

[0047] Adapt the communication protocol of the energy storage system and collect detection data of the energy storage system based on the adaptation result;

[0048] Convert the detection data into target data, where the target data is the detection data under the target communication protocol supported by the operation state detection module;

[0049] Perform abnormal evaluation on the target data through preset detection rules, and perform corresponding fault alarms based on the evaluation results;

[0050] Process the abnormal target data through a fault diagnosis model to determine the fault types corresponding to the fault alarms.

[0051] In a third aspect of the present application, an electronic device is provided, including: a processor, a memory, and a computer program stored on the memory and running on the processor. When the computer program is executed by the processor, the steps in a method for detecting the operating state of an energy storage system as described in the second aspect of the present application are implemented.

[0052] In a fourth aspect of the present application, a computer-readable storage medium is provided. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps in a method for detecting the operating state of an energy storage system as described in the second aspect of the present application are implemented.

[0053] The portable detection device for an energy storage system provided by the present application has the following advantages:

[0054] A portable detection device for an energy storage system provided by an embodiment of the present application includes: an adapter and an operating state detection module; the operating state detection module includes: a state evaluation module and a fault type determination module; the adapter is used to be pluggably connected to energy storage systems with different communication protocols, adapt to the communication protocols of the connected energy storage systems to collect detection data, and convert the detection data into target data and send it to the operating state detection module. The target data is the detection data under the target communication protocol supported by the operating state detection module; the state evaluation module is used to perform anomaly evaluation on the target data through a preset detection rule and perform corresponding fault alarms based on the evaluation results; the fault type determination module is used to process the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault alarm. Thus, in this solution, an adapter is introduced to adapt to different communication protocols of different energy storage systems. Therefore, no matter how the communication protocol of the energy storage system changes, this detection device can collect and convert data through the adapter for detecting the operating state of the energy storage system, thereby solving the problem that the current monitoring system software platform is strictly bound to a single site and has poor compatibility, and improving the compatibility of detecting the operating state of the energy storage system. At the same time, while performing anomaly evaluation on the target data and performing fault alarms based on the evaluation results, a deep learning model is used to more accurately determine the fault type of the fault state based on the fault results, which helps the operation and maintenance personnel understand the cause and nature of the fault, so that the operation and maintenance personnel can propose solutions more efficiently and accurately. Description of the Drawings

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0056] Figure 1 Schematic diagram of a portable detection device for an energy storage system shown in an embodiment of the present application;

[0057] Figure 2 Another schematic diagram of a portable detection device for an energy storage system shown in an embodiment of the present application;

[0058] Figure 3 Flowchart of a method for detecting the operating state of an energy storage system shown in an embodiment of the present application. Detailed implementation manners

[0059] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0060] Figure 1 Schematic diagram of a portable detection device for an energy storage system shown in an embodiment of the present application. As Figure 1 shown, the device includes: an adapter 101 and an operating state detection module 102; the operating state detection module 102 includes: a state evaluation module 1021 and a fault type determination module 1022; the adapter 101 is used to be pluggably connected to energy storage systems with different communication protocols, adapt to the communication protocols of the connected energy storage systems to collect detection data, and convert the detection data into target data and send it to the operating state detection module, where the target data is the detection data under the target communication protocol supported by the operating state detection module; the state evaluation module 1021 is used to perform anomaly evaluation on the target data through a preset detection rule and perform corresponding fault alarms based on the evaluation results; the fault type determination module 1022 is used to process the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault alarm.

[0061] In this embodiment, a portable detection device for an energy storage system provided by the present application at least includes an adapter, a state evaluation module, and a fault type determination module. The adapter can be used to perform corresponding communication configurations based on different communication protocols adopted by different energy storage systems, so as to collect detection data of the energy storage system under various different communication protocols. A portable detection device for an energy storage system provided by the present application further includes an operating state detection model. The operating state detection module is used to analyze the collected detection data to determine the operating state of the energy storage system. However, since the communication protocols adopted by different energy storage systems may be different, the detection data collected by the adapter may be detection data under various communication protocols. In order for the operating state detection module to recognize and process the detection data collected by the adapter, the adapter in the present application is further used to convert the collected detection data into target data under the target communication protocol supported by the operating state detection module and send it to the operating state detection module, so that the operating state detection module can recognize and process the detection data collected by the adapter. Specifically: The adapter is used to connect to the energy storage system and automatically identify the communication protocol of the interconnected energy storage system, and perform a configuration corresponding to the communication protocol of the energy storage system based on the recognition result. After completing the configuration of the adapter, collect the detection data of the energy storage system based on the configuration result. After the adapter collects the detection data of the energy storage system, it converts the detection data into detection data under the target communication protocol supported by the operating state detection module, that is, target data. For example, if energy storage system 1 adopts communication protocol A, the adapter automatically identifies that energy storage system 1 adopts communication protocol A. At this time, the adapter performs a configuration corresponding to communication protocol A. After completing the configuration, the adapter will be able to collect the detection data of energy storage system 1. Then, based on communication protocol A, the adapter converts the detection data under communication protocol A collected into detection data under the target communication protocol supported by the operating state detection module and gives it to the operating state detection module for analysis and processing to determine the operating state of the energy storage system; if energy storage system 2 adopts communication protocol B, the adapter automatically identifies that energy storage system 2 adopts communication protocol B. At this time, the adapter performs a configuration corresponding to communication protocol B. After completing the configuration, the adapter will be able to collect the detection data of energy storage system 2. Then, based on communication protocol B, the adapter converts the detection data under communication protocol B collected into detection data under the target communication protocol supported by the operating state detection module and gives it to the operating state detection module for analysis and processing to determine the operating state of the energy storage system. Thus, the present application can detect the operating state of the energy storage system for any energy storage system adopting any communication protocol, thereby effectively improving the compatibility of the operating state detection of the energy storage system.The adapter of the portable detection device for the energy storage system features a quick connection / disconnection mechanism, facilitating the convenient connection of the portable detection device for the energy storage system to the energy storage system, as well as disconnecting from the current energy storage system and connecting to other energy storage systems. The adapter includes software algorithms for automatically identifying the communication protocol adopted by the connected energy storage system device. Meanwhile, based on the identification result, the adapter automatically adjusts the communication parameters to match the communication protocol of the energy storage system. The portable detection device for the energy storage system can be rapidly deployed and used among different energy storage power stations without being restricted by specific stations. Additionally, through an algorithm management platform wirelessly connected to the portable detection device for the energy storage system, cloud storage of data and algorithms for each energy storage power station can be achieved, enabling the sharing and access of data and analysis results among different stations.

[0062] In this embodiment, the status evaluation module in the operating status detection module receives target data from the adapter. Based on preset detection rules, this status evaluation module conducts anomaly evaluation on these target data, obtains corresponding evaluation results, and determines whether there is a fault in the energy storage system based on the evaluation results. In the case of a fault, an alarm is issued. For example, when it is detected that the temperature of the battery pack rises abnormally, the status evaluation module issues an alarm of "battery pack overheating". These detection rules include but are not limited to threshold judgment, trend analysis, etc. Meanwhile, when the detected data of the energy storage system collected includes multiple data types, the target data received by the status evaluation module from the adapter also includes multiple data types, and there are respective corresponding detection rules for each data type. The status evaluation module will determine the corresponding detection rule based on the data type of the target data, and then conduct anomaly evaluation on the target data using the corresponding detection rule. For example, when the detected data of energy storage system 1 includes detected data of type 1, type 2, and type 3, the target data received by the status evaluation module from the adapter also includes target data of type 1, type 2, and type 3. There is a corresponding detection rule 1 for the target data of type 1, a corresponding detection rule 2 for the target data of type 2, and a corresponding detection rule 3 for the target data of type 3.

[0063] In this embodiment, after the status evaluation module determines that there is a fault in the energy storage system and issues a corresponding alarm, the fault type determination module processes the target data for a period of time under the data type to which the abnormal target data belongs through a fault diagnosis model to determine the fault type corresponding to the fault alarm. This fault diagnosis model can be constructed based on algorithms such as machine learning and deep learning, and can identify and classify various potential faults. For example, when the status evaluation module issues an alarm of "battery overheating", the fault type determination module can determine whether the specific fault type belongs to an internal short circuit of the battery or a failure of the heat dissipation system based on the target data for a period of time under the data type to which the abnormal target data belongs.

[0064] In this embodiment, since the evaluation speed of the anomaly evaluation based on the detection rules is faster, when any anomaly occurs in the energy storage system in the solution of this application, the state evaluation module first performs an anomaly evaluation according to the preset detection rules and issues a corresponding fault warning. This fast response mechanism provides a valuable time window for the operation and maintenance personnel, enabling them to quickly realize that there are potential problems in the energy storage system and prepare for the necessary maintenance work. Once it is determined that there is a fault in the energy storage system, the fault type determination module will further use advanced means such as a deep learning model to conduct a more in-depth analysis of the abnormal target data to determine the specific type of the fault. This accurate diagnosis helps the operation and maintenance personnel understand the cause and nature of the fault, so as to enable the operation and maintenance personnel to propose solutions more efficiently and accurately.

[0065] A portable detection device for an energy storage system provided by an embodiment of this application includes: an adapter and an operating state detection module; the operating state detection module includes: a state evaluation module and a fault type determination module; the adapter is used to be pluggably connected to energy storage systems with different communication protocols, adapt to the communication protocols of the connected energy storage systems to collect detection data, and convert the detection data into target data and send it to the operating state detection module, where the target data is the detection data under the target communication protocol supported by the operating state detection module; the state evaluation module is used to perform an anomaly evaluation on the target data through preset detection rules and issue a corresponding fault warning based on the evaluation result; the fault type determination module is used to process the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault warning. Thus, this solution adapts to different communication protocols of different energy storage systems by introducing an adapter, so that no matter how the communication protocol of the energy storage system changes, this detection device can collect and convert data through the adapter for the operation state detection of the energy storage system, thereby solving the problem that the current monitoring system software platform is strictly bound to a single station and has poor compatibility, and improving the compatibility of the operation state detection of the energy storage system. At the same time, while performing an anomaly evaluation on the target data and issuing a fault warning based on the evaluation result, a more accurate fault type of the fault state is determined based on the fault result through a deep learning model, which helps the operation and maintenance personnel understand the cause and nature of the fault, so as to enable the operation and maintenance personnel to propose solutions more efficiently and accurately.

[0066] Combined with the above embodiments, in one implementation, the embodiments of the present application further provide a portable detection device for an energy storage system. In the portable detection device for the energy storage system, the adapter includes a data acquisition module, configured to determine the communication protocol of the energy storage system, perform protocol configuration based on the determined communication protocol, and acquire detection data of the energy storage system based on the configuration result; the data acquisition module includes: a basic data acquisition module, configured to acquire basic detection data at the cell level, module level, and container level; a differential data acquisition module, configured to acquire first-level detection data at the cell level; and, configured to acquire second-level detection data at the module level; and, configured to acquire third-level detection data at the container level.

[0067] In this embodiment, the adapter of the portable detection device for the energy storage system provided by the present application includes a data acquisition module. The data acquisition module is configured to automatically identify the communication protocol of the energy storage system, and automatically perform a configuration operation corresponding to the communication protocol based on the identified communication protocol of the energy storage system, so as to ensure that the data acquisition module can acquire the detection data of the energy storage system under this communication protocol. Among them, the configuration information configured by the configuration operation includes communication parameters (such as baud rate, data bits, etc.) and data formats. The adapter is connected to the energy storage system through the data acquisition module to realize the acquisition and transmission of detection data. The data acquisition module adapts to different communication protocols of various energy storage systems through flexible configuration, ensuring the versatility and scalability of the adapter.

[0068] In this embodiment, the current method for detecting the operating state of the energy storage system is to determine that the energy storage system has a fault and issue an alarm when abnormal detection data of certain data types is detected. For such an operating state detection method, it is not conducive to fault location of the energy storage system. To solve such a problem, the present application divides the detection data of the energy storage system into levels, and divides it into the smallest cell level in the energy storage system, the largest container level in the energy storage system, and the module level located between the cell level and the container level according to the scope covered by the detection data. The module level is a power unit formed by combining multiple single cells through conductive connectors, and this power unit is a component that is fixed in the designed position through processes and structures and collaborates to perform the functions of electric energy charging, discharging, and storage. At the same time, different types of detection data are acquired at different levels of the energy storage system to detect the operating state of the energy storage system. Through this method, not only can the operating state of the energy storage system be detected in a refined manner, improving the safety and reliability of the energy storage system, but also the state of the container level can be macroscopically controlled, and rapid responses can be made to abnormalities and faults at the system level of the energy storage system, thereby achieving a comprehensive coverage detection of the energy storage system.

[0069] Specifically: The energy storage system is divided into multiple container levels from large to small. For each container level, it is divided into multiple module levels, and for each module level, it is divided into multiple cell levels. The data acquisition module may include a basic data acquisition module and a differential data acquisition module. The data acquisition module is used to acquire the basic detection data at each cell level, the basic detection data at each module level, and the basic detection data at each container level. Among them, the basic detection data is the key basic data that reflects the operating state of various levels, and the basic detection data at least includes: voltage, current, and temperature. The differential data acquisition module is used to acquire the first-level detection data that is more concerned at each cell level; and, used to acquire the second-level detection data that is more concerned at each module level; and, used to acquire the third-level detection data that is more concerned at each container level. Among them, since the cell level pays more attention to the detailed performance parameters of individual cells, the first-level detection data at least includes the internal resistance and capacity of the cells. Since the module level pays more attention to the cooperation state among the cells at the module level, the second-level detection data at least includes the differences in parameters such as voltage, capacity, and internal resistance among the cells within the module level; the remaining power (or state of charge), temperature, and the voltage value differences during the charge and discharge process of each cell within the module level. Since the container level pays more attention to the overall performance of the system, the third-level detection data at least includes the energy conversion efficiency, which is the ratio of the output electric energy to the input electric energy within the container level; the charge and discharge efficiency; the internal temperature distribution within the container level.

[0070] Combined with the above embodiments, in one implementation, the embodiment of the present application further provides a portable detection device for an energy storage system. In the portable detection device for the energy storage system, the adapter includes a communication module, which is used to convert the detection data into target data based on the communication protocol of the energy storage system and send it to the operating state detection module.

[0071] In this embodiment, after the detection data of the energy storage system is acquired by the data acquisition module, the communication module of the adapter converts the acquired detection data into detection data under the target communication protocol supported by the operating state detection module, that is, the target data, based on the communication protocol used to acquire the detection data of the energy storage system, and then the communication module sends the target data to the operating state detection module. The communication module includes a multi-functional data access interface, which is used to receive the detection data acquired under the communication protocol of the energy storage system, and the multi-functional data access interface can at least support multiple communication protocols such as RS485, CAN, and Ethernet.

[0072] In this embodiment, the portable detection device of the energy storage system further includes an interaction interface. In the case of identification and / or configuration failure during the automatic identification and configuration of the communication protocol, manual configuration is performed through this interaction interface. A comprehensive system-level test is carried out on the portable detection device of the energy storage system. This system-level test includes aspects such as communication stability, data accuracy, and user interaction experience to ensure the stability and reliability of the entire portable detection device of the energy storage system.

[0073] Combined with the above embodiments, in one implementation, the embodiment of the present application further provides a portable detection device for an energy storage system. In this portable detection device of the energy storage system, the device further includes: a cross-level state analysis module and a data mining analysis module; the state evaluation module is used to perform anomaly evaluation on target data at different levels through a preset detection rule to obtain corresponding evaluation results; the state evaluation module is further used to determine the target level to which the target data with anomalies belongs according to the evaluation results and issue a fault alarm for the target level; the fault type determination module is used to process the target data at the target level through a fault diagnosis model to determine the fault type corresponding to the fault alarm at the target level; the cross-level state analysis module is used to perform correlation analysis on target data at different levels and determine the current state of the energy storage system at each level based on the correlation analysis results; the data mining analysis module is used to analyze the target data at different levels for a preset historical duration through a fault prediction model to predict the state of the energy storage system at each level.

[0074] In this embodiment, when the detection data of the energy storage system is collected by different hierarchical units, the portable detection device for the energy storage system provided by the present application further includes a cross-hierarchical state analysis module and a data mining and analysis module. When the detection data of the energy storage system is collected by different hierarchical units, since the abnormal determination criteria are different at different levels, for the target data corresponding to the basic detection data that belongs to the same data type and different levels, there will be corresponding detection rules respectively. For example, for the basic detection data of the data type of voltage, the voltage basic detection data at the cell level has a corresponding detection rule a1, the voltage basic detection data at the module level has a corresponding detection rule a2, and the voltage basic detection data at the container level has a corresponding detection rule a3. After receiving the target data converted from the detection data at different levels (including the cell level, the module level, and the container level) from the adapter, the state evaluation module determines the detection rule corresponding to both the data type and the level to which the target data belongs based on the data type and the level to which the target data belongs, and then performs an abnormal evaluation on the target data based on this detection rule to obtain a corresponding evaluation result. Based on this evaluation result, it is determined whether there is an abnormality in the level corresponding to the target data. In the case where an abnormality is determined, it is determined that the level corresponding to and pointed to by the target data is the target level. For example, when the internal resistance data of a certain cell level z1 in the energy storage system exceeds the threshold range, the state evaluation module will determine that there is an abnormality in this certain cell level z1 of the current energy storage system, and determine this abnormal cell level z1 as the target level. At this time, a fault alarm for this cell level will be triggered. It should be understood that this fault alarm will visually display the abnormal cell level z. Then, the fault type determination module will analyze and process the target data of all data types at the abnormal target level determined by the state evaluation module to determine the specific fault type corresponding to the fault alarm at the abnormal target level. For example, if the internal resistance abnormality of a certain cell level z1 is accompanied by a sudden increase in temperature, the fault diagnosis model in the fault type determination module may diagnose it as an internal short circuit of the cell; if the internal resistance of this certain cell level z1 is abnormal while the temperature is stable, the fault diagnosis model in the fault type determination module may diagnose it as cell aging deterioration.

[0075] In this embodiment, the cross - level state analysis module is used to perform joint analysis on multi - level target data through technologies such as time - series alignment and data correlation calculation. The correlation analysis of data is carried out on the target data at the container level, all the target data at the module level under the container level, and all the target data at the cell level under the container level within a container level. Before performing the correlation analysis on the target data at the container level, all the target data at the module level under the container level, and all the target data at the cell level under the container level within a container level, time alignment processing is first performed on these target data, and then the cross - level state analysis module performs correlation analysis on the target data at the container level, all the target data at the module level under the container level, and all the target data at the cell level under the container level after time alignment processing to determine the abnormal operating state of the energy storage system that is not recognized by the state evaluation module. An optional implementation method for determining that there is an abnormality in the energy storage system through correlation is: when there are deviations in a large amount of low - level target data and the deviation does not exceed the set threshold for determining that the target data is abnormal, and the deviation positively affects the deviation of the high - level target data including the large amount of low - level target data, it is determined that there is a hidden abnormality in the energy storage system, and corresponding alarms are sent to the operation and maintenance personnel. When the energy conversion efficiency at the container level decreases, the cross - level analysis module will synchronously retrieve the target data related to the cell cooperation parameters (such as the voltage difference between cells within the module level) in each module level included in the container level and the target data related to the single - cell performance data (such as internal resistance distribution) in each cell level included in the container level, and determine whether the efficiency decrease is caused by the deterioration of the cell consistency of a specific module by establishing a parameter correlation matrix. For example, if there is an obvious distribution difference in the temperature distribution of a certain container level y1, but the temperature has not reached the abnormal state, and at the same time, this distribution difference forms a spatio - temporal correlation with the temperature parameter difference at the module level included under the container level y1 (the difference has not reached the set threshold for determining that it is abnormal) and the local temperature sudden increase difference at the cell level (the difference has not reached the set threshold for determining that it is abnormal), it is predicted that the energy storage system is currently at a cross - level thermal runaway risk caused by the blockage of the heat dissipation system air duct. The correlation analysis result output by the cross - level state analysis module can predict in advance the possible abnormalities of the energy storage system, facilitating the operation and maintenance personnel to better maintain the normal operation of the energy storage system.

[0076] In this embodiment, the data mining and analysis module is used to analyze the target data at different levels obtained from a period of time from the current moment to the past (i.e., the historical preset duration) through a fault prediction model, and predict the current operating state of the energy storage system at each level. Among them, the historical preset duration can be set according to the actual application scenario and is not specifically limited here. For example, it is 1 hour from the current moment to the past. When the data mining and analysis module analyzes the target data at different levels, the different levels refer to the target data at the container level under a container level, all module-level target data under the container level, and all cell-level target data under the container level.

[0077] Combined with the above embodiments, in one implementation manner, the embodiment of the present application further provides a portable detection device for an energy storage system. In the portable detection device for the energy storage system, the fault diagnosis model in the fault type determination module includes: a first data input module, configured to receive the target data at the target level, and perform normalization and time window alignment processing to obtain a corresponding three-dimensional input tensor; a spatial feature extraction module, configured to extract spatial features from the three-dimensional input tensor through a convolutional neural network; a temporal feature extraction module, configured to extract temporal features from the three-dimensional input tensor through a bidirectional long short-term memory network; a multimodal fusion module, configured to perform dynamic weight allocation and calculate cross-modal correlations on the spatial features and the temporal features through a self-attention mechanism to obtain a fused feature vector; a diagnosis output module, configured to process the fused feature vector through a fully connected layer and a classifier to determine the fault type at the abnormal target level.

[0078] In this embodiment, an alternative implementation of the fault diagnosis model structure in the fault type determination module is as follows: The fault diagnosis model includes a first data input module, a spatial feature extraction module, a temporal feature extraction module, a multimodal fusion module, and a diagnosis output module. The first data input module is used to perform data standardization and time window alignment. The first data input module receives the target data at the target level with anomalies determined by the state evaluation module, and then performs data cleaning and standardization processing on the received target data. Data cleaning is a process of inspecting, correcting, and preprocessing the target data, aiming to remove noise, errors, and redundant information in the target data, improve the quality and usability of the data. Standardization processing is a process of converting target data of different data types into a unified scale or range, aiming to eliminate the dimensional differences between target data of different data types and improve the performance and accuracy of the algorithm. Then, time window alignment is performed on the standardized target data. This time window alignment operation extracts the time series data within a fixed time window for processing according to the fault diagnosis requirements. For example, for the temperature data at the cell level, the time series data from 30 seconds before the fault trigger moment to the fault trigger moment is extracted and divided into multiple time steps (such as one sampling point per second). Based on the target data after standardization and time window alignment processing, a three-dimensional input tensor corresponding to the target data is constructed. The dimension of the three-dimensional input tensor is (time step × channel × feature), where the channel corresponds to various data types of the target data at the abnormal target level.

[0079] In this embodiment, the spatial feature extraction module extracts spatial features from the obtained three-dimensional input tensor through a convolutional neural network (CNN). Specifically, by sliding the convolutional kernel in the hierarchical channel dimension, the spatial correlation between different features within the abnormal target level is captured. For example, for the voltage data and temperature data at the cell level, the co-variation in space between the two is extracted through convolution operations. Thus, spatial features are extracted from the three-dimensional input tensor by the convolutional kernel. The temporal feature extraction module extracts temporal features from the obtained three-dimensional input tensor through a bidirectional long short-term memory network (Bi-LSTM). Specifically, the LSTM unit is used to capture the temporal evolution law before the fault occurs. For example, whether the cell temperature shows a gradual upward trend before the fault trigger. The reverse LSTM unit is used to analyze the data changes after the fault trigger. For example, whether the cell voltage drops suddenly after the fault trigger. The forward and reverse temporal features are concatenated to form a complete temporal feature vector, which is the temporal feature extracted by the bidirectional long short-term memory network (Bi-LSTM) from the obtained three-dimensional input tensor.

[0080] In this embodiment, the multimodal fusion module dynamically assigns weights to the extracted spatial features and temporal features through the self-attention mechanism, and performs cross-modal correlation analysis calculation after weight assignment to obtain the final fused feature vector. Specifically: Based on the obtained spatial features and temporal features, the importance weights of the spatial features and temporal features are calculated through the self-attention mechanism. For example, if the fault is mainly caused by abnormal spatial distribution, a higher weight is assigned to the spatial features. Then, after completing the dynamic weight assignment of the spatial features and temporal features, cross-attention calculation is performed on the spatial features and temporal features to capture the implicit correlation between the two, so as to obtain the final fused feature vector after weighted fusion. For example, whether the abnormal spatial distribution of the cell temperature is related to the current fluctuation at a specific time point.

[0081] In this embodiment, the diagnosis output module processes the obtained fused feature vector through a fully connected layer and a Softmax classifier to determine the specific fault type under the abnormal target level. Specifically: The fused feature vector is mapped to a low-dimensional space through the fully connected layer, and high-order abstract features are extracted. The Softmax classifier processes the high-order abstract features extracted by the fully connected layer and outputs the probability distribution of various fault types. For example, the output fault types are "internal short circuit of the cell" (probability 85%) and "aging deterioration of the cell" (probability 15%). When the probability of a certain fault type exceeds the set threshold, it is determined that the fault type under the abnormal target level belongs to the fault type that exceeds the set threshold.

[0082] Combined with the above embodiments, in one implementation manner, the embodiment of the present application further provides a portable detection device for an energy storage system. In the portable detection device for the energy storage system, the fault diagnosis model in the fault type determination module includes: a second data input module, configured to receive the target data under the target level, and perform normalization and time window alignment processing to obtain a corresponding three-dimensional input tensor; a supervised learning sub-model, configured to process the three-dimensional input tensor to determine the probabilities corresponding to various fault types under the abnormal target level; a time series sub-model, configured to process the time series type tensor in the three-dimensional input tensor to determine the probabilities corresponding to various fault types under the abnormal target level; a statistical analysis sub-model, configured to perform distribution anomaly analysis on the target data under the abnormal target level to determine the probabilities corresponding to various fault types under the abnormal target level; a first result fusion module, configured to assign weights according to the confidence levels of each sub-model, and fuse the output results of each sub-model to determine the final fault type under the abnormal target level.

[0083] In this embodiment, another alternative implementation of the fault diagnosis model structure in the fault type determination module is: a second data input module, a supervised learning sub-model, a time series sub-model, a statistical analysis sub-model, and a first result fusion module. The second data input module is similar in type to the first data input module in the previous alternative implementation. This second data input module is used for data standardization and time window alignment. The second data input module receives the target data at the target level determined by the status evaluation module to be abnormal, and then performs data cleaning and standardization processing on the received target data. Data cleaning is a process of checking, correcting, and preprocessing the target data, aiming to remove noise, errors, and redundant information in the target data, and improve the quality and usability of the data. Standardization processing is a process of converting target data of different data types into a unified scale or range, aiming to eliminate the dimensionality differences between target data of different data types and improve the performance and accuracy of the algorithm. Then, time window alignment is performed on the standardized target data. This time window alignment operation extracts time series data within a fixed time window for processing according to the fault diagnosis requirements. For example, for the temperature data at the cell level, the time series data from 30 seconds before the fault trigger moment to the fault trigger moment is extracted and divided into multiple time steps (such as one sampling point per second). Based on the target data after standardization and time window alignment processing, a three-dimensional input tensor corresponding to the target data is constructed. The dimensions of this three-dimensional input tensor are (time step × channel × feature), where the channel corresponds to various data types of the target data at the abnormal target level.

[0084] In this embodiment, the supervised learning sub-model processes the obtained three-dimensional input tensor to determine the probabilities corresponding to various fault types at the abnormal target level. Specifically: spatial features of the three-dimensional input tensor are extracted through a convolutional neural network (CNN) or a residual network (ResNet). For example, the voltage distribution pattern between battery cell levels or the temperature thermal map features at the container level are identified. The probability distributions of various fault types are output through a fully connected layer and a Softmax classifier. The time series sub-model analyzes based on the time series pattern of LSTM / Transformer to determine the probabilities corresponding to various fault types at the abnormal target level. Specifically: the time series features of the three-dimensional input tensor are extracted through an LSTM or a Transformer network, and the probability distributions of various fault types are output through a time series classifier. The statistical analysis sub-model performs anomaly detection based on distribution and hypothesis testing. Specifically: the deviation of the data distribution is identified through a clustering algorithm (such as K-means) or a hypothesis testing (such as T-test). For example, it is detected whether the internal resistance of the battery cell deviates from the historical normal distribution. The probability distribution of the fault type is output based on a predefined statistical rule library (such as the "3σ principle"). For example, the probability of "battery cell aging and deterioration" is output as 82%. The first result fusion module is used to assign weights to the historical accuracies of the supervised learning sub-model, the time series sub-model, and the statistical analysis sub-model (such as supervised learning: 0.5, time series model: 0.3, statistical analysis: 0.2), and then based on the assigned weights, the probability distributions of each sub-model are weighted and summed to generate the final probabilities of each fault type. Then, the fault type with the maximum probability and exceeding the set threshold is determined as the most serious fault type at the abnormal target level.

[0085] In this embodiment, the fault type determination tolerance can be improved by various model fusion fault type determination methods. For example, the misjudgment of a single sub-model can be corrected by other sub-models.

[0086] Combined with the above embodiments, in one implementation, the embodiments of the present application further provide a portable detection device for an energy storage system. In the portable detection device for the energy storage system, the fault prediction model in the data mining and analysis module includes: a third data input module, configured to receive target data at different levels for a historical preset duration, and perform normalization and time window alignment processing to obtain a corresponding three-dimensional input tensor; a statistical sub-model, configured to perform a linear trend prediction on the target data at different levels for the historical preset duration to obtain a corresponding first prediction result; a deep learning sub-model, configured to perform non-linear feature extraction and prediction on the three-dimensional input tensor to obtain a corresponding second prediction result; a second result fusion module, configured to perform dynamic weight assignment on each sub-model according to the past prediction errors of each sub-model, and perform weighted fusion on the first prediction result and the second prediction result based on the weight assignment result to obtain a final prediction result.

[0087] In this embodiment, an optional implementation of the structure composition of the fault prediction model in the data mining and analysis module is: a third data input module, a statistical sub-model, a deep learning sub-model, and a second result fusion module. The third data input module is used to perform data standardization and time window alignment. The third data input module receives target data at different levels from the current moment to a past period of time (i.e., the historical preset period) of the energy storage system. The different levels refer to the target data at the container level under a container level, all module-level target data under the container level, and all cell-level target data under the container level. Then, the received target data is subjected to data cleaning and standardization processing. Data cleaning is a process of checking, correcting, and preprocessing the target data, aiming to remove noise, errors, and redundant information in the target data, improve the quality and usability of the data. Standardization processing is a process of converting target data of different data types into a unified scale or range, aiming to eliminate the dimensional difference between target data of different data types and improve the performance and accuracy of the algorithm. Then, time window alignment is performed on the standardized target data. This time window alignment operation extracts time series data within a fixed time window for processing according to fault diagnosis requirements. Based on the target data after standardization and time window alignment processing, a three-dimensional input tensor corresponding to the target data is constructed. The dimension of the three-dimensional input tensor is (time step × channel × feature), where the channel corresponds to various data types of the target data at different levels.

[0088] In this embodiment, the statistical sub-model performs fault prediction based on a linear trend. Specifically: Input target data at different levels for the historical preset period, and perform trend fitting analysis on the target data of the time series type through linear regression. For example, fit the linear growth curve of the internal resistance of the cell by the least squares method. Generate predicted values for the future time period based on the fitting result, and calculate the prediction confidence interval. For example, predict that the cell voltage will decrease at a rate of 0.1V / hour within the next 1 hour. Output the fault probability in combination with a predefined threshold (such as a voltage lower than 3.2V is abnormal). For example, if it is predicted that the voltage will be lower than the threshold after 2 hours, output the probability of "insufficient cell voltage" as 80%.

[0089] In this embodiment, the deep learning sub-model performs fault prediction based on non-linear time series features. Specifically: Input the obtained three-dimensional input tensor, capture the time series dependence and hierarchical correlation of the data through an LSTM or Transformer network to obtain corresponding non-linear features, and process the obtained non-linear features through a fully connected layer and a Sigmoid function to output the fault probability. For example, output the probability of "thermal runaway of the cell" as 65%.

[0090] In this embodiment, the second result fusion module predicts whether the energy storage system may fail in the future through dynamic weight allocation and result fusion. Specifically: it inputs the prediction results of the statistical sub-model and the deep learning sub-model, calculates the historical prediction errors of each sub-model (such as the mean absolute error MAE). For example, the average error of the statistical sub-model in the past 10 predictions is 5%, and that of the deep learning sub-model is 8%. The weights of each sub-model are dynamically adjusted according to the errors (the smaller the error, the higher the weight). For example, the weight of the statistical sub-model is 0.6, and that of the deep learning sub-model is 0.4. Based on the weight allocation, the first prediction result obtained by the statistical sub-model and the second prediction result obtained by the deep learning sub-model are weighted and fused to obtain the final prediction result, which includes the probabilities of various faults. For example, the statistical sub-model predicts that the probability of "insufficient cell voltage" is 80%, and the deep learning sub-model predicts it as 70%. The weighted result is 0.6×80% + 0.4×70% = 76%. If the fusion probability exceeds the preset threshold (such as 75%), the corresponding fault warning is triggered.

[0091] Combined with the above embodiments, in one implementation, the embodiment of the present application also provides a portable detection device for an energy storage system. In the portable detection device for the energy storage system, the fault prediction model in the data mining and analysis module includes: a fourth data input module, configured to receive target data at different levels for a historical preset duration, and perform standardization and time window alignment processing to obtain corresponding three-dimensional input tensors; a clustering analysis module, configured to perform clustering analysis on the three-dimensional input tensors through a clustering algorithm to determine the battery states at different levels; an association module, configured to determine the fault types corresponding to the data of the abnormal battery states under the current environmental data based on the pre-established abnormal association rules; a health prediction module, configured to predict the battery health state score and remaining life at different levels according to the capacity attenuation rate determined by the three-dimensional input tensors.

[0092] In this embodiment, another alternative implementation of the structure composition of the fault prediction model in the data mining and analysis module is: a fourth data input module, a clustering analysis module, an association module, and a health prediction module. The fourth data input module is used to perform data standardization and time window alignment. The fourth data input module receives target data at different levels from the current moment to a certain duration in the past (i.e., the historical preset duration) of the energy storage system. The different levels refer to the target data at the container level under one container level, all module-level target data under this container level, and all cell-level target data under this container level. Then, the received target data is subjected to data cleaning and standardization processing. Data cleaning is a process of checking, correcting, and preprocessing the target data, aiming to remove noise, errors, and redundant information in the target data, improve the quality and usability of the data. Standardization processing is a process of converting target data of different data types into a unified scale or range, aiming to eliminate the dimensionality differences between target data of different data types and improve the performance and accuracy of the algorithm. Then, time window alignment is performed on the standardized target data. This time window alignment operation extracts time series data within a fixed time window for processing according to fault diagnosis requirements. Based on the target data after standardization and time window alignment processing, a three-dimensional input tensor corresponding to the target data is constructed. The dimension of the three-dimensional input tensor is (time step × channel × feature), where the channel corresponds to various data types of the target data at different levels.

[0093] In this embodiment, the clustering analysis module classifies the states of hierarchical batteries. Specifically: the obtained three-dimensional input tensor is input, and clustering algorithms (such as K-means, DBSCAN, etc.) are respectively applied to the three-dimensional input tensors at each level. For the three-dimensional input tensor at the cell level, based on features such as the capacity attenuation rate and internal resistance change, the cells are divided into three categories: "healthy", "slightly aged", and "severely aged". For the three-dimensional input tensor at the module level, based on features such as the voltage difference between cells and the standard deviation of temperature distribution, the modules are divided into three categories: "well balanced", "partially imbalanced", and "severely imbalanced". For the three-dimensional input tensor at the container level, based on features such as energy efficiency and environmental temperature fluctuation, the containers are divided into three categories: "stable operation", "potential risk", and "performance degradation". At the same time, the state correlation between different levels is determined through association rule mining (such as the Apriori algorithm) to determine the state of the energy storage system. For example, there is a 90% probability that the module level with the proportion of severely aged cells greater than 20% is in a severely imbalanced state.

[0094] In this embodiment, the association module performs fault type mapping through environmental perception. Specifically: it inputs the clustering analysis results obtained by the clustering analysis module (the clustering analysis results record the abnormal battery states at each level) and the current environmental data (such as temperature, humidity, charge and discharge rate), and based on the pre-established abnormal association rules, determines the fault type corresponding to the data of the abnormal battery state under the current environmental data. For example, the abnormal association rules include Rule 1: If the environmental temperature is greater than 40°C and the proportion of "severely aged" battery cells is greater than 15%, then determine the fault type as the risk of thermal runaway; Rule 2: If the charge and discharge rate is greater than 2C and the number of severely imbalanced modules is greater than 5, then determine the fault type as overload damage.

[0095] In this embodiment, the health prediction module performs life assessment based on capacity attenuation. Specifically: it inputs the obtained three-dimensional input tensor, and determines the capacity attenuation rate at each level through linear regression or exponential fitting. For example, the capacity of a certain battery cell drops from 100% to 80% in 500 charge and discharge cycles, and the attenuation rate is 0.04% per cycle. For the calculation of the health state score at the battery cell level: SOH = current capacity / initial capacity × 100%; for the calculation of the health state score at the module level: SOH = (average battery cell SOH) × (1 - voltage difference coefficient); for the calculation of the health state score at the container level: SOH = energy efficiency × average module SOH. Preset the failure thresholds at each level, that is, when the health degree at each level is lower than its corresponding failure threshold, it is determined that the life of this level ends. Then, based on the determined capacity attenuation rate and health degree, determine the remaining life at different levels. For example, the current SOH is 85%, the attenuation rate is 0.04% per cycle, and the remaining life is (85 - 70) / 0.04 = 375 charge and discharge cycles.

[0096] Combined with the above embodiments, in one implementation manner, the embodiments of the present application further provide a portable detection device for an energy storage system. In this portable detection device for the energy storage system, the device further includes: an algorithm library, which is used to store various algorithms required by the operation state detection module, and is used to connect to the operation state detection module through an API interface to provide various algorithms required for state detection.

[0097] In this embodiment, the portable detection device of the energy storage system provided by the present application further includes an algorithm library for centrally managing various algorithms required for state detection of the energy storage system. The algorithm library stores various algorithms required by the operating state detection module. The algorithm library is connected to the operating state detection module through an API interface to provide it with various algorithms required for state detection of the energy storage system. The algorithm library supports dynamic loading and updating. The various algorithms include, but are not limited to, anomaly detection, fault diagnosis, health prediction, data preprocessing and other algorithms. The operating state detection module does not need to build in all algorithms, reducing resource occupation. New algorithms can be seamlessly integrated through the API without downtime for upgrade.

[0098] Combined with the above embodiments, in one implementation, the embodiment of the present application further provides a portable detection device for an energy storage system. In the portable detection device for the energy storage system, the device is connected to an algorithm management platform; the algorithm management platform is used to obtain the detection data of each energy storage system for evaluating the effects of various algorithms, and based on the evaluation results, optimize and update the various algorithms.

[0099] In this embodiment, as Figure 2 shown, the portable detection device for the energy storage system provided by the present application is connected to an algorithm management platform. The algorithm management platform is used to obtain the detection data of each energy storage system for evaluating the processing effects of various algorithms in the device, and based on the evaluation results, perform corresponding optimization and update on the various algorithms, so as to further improve the accuracy of subsequent state detection of the energy storage system. Specifically: perform data cleaning on the obtained data to remove outliers and fill in missing data, and perform manual or automated annotation on the fault warning results (such as confirming whether "cell short circuit" is a false alarm), so as to obtain a standardized detection data set for algorithm effect evaluation. Perform performance evaluation on various algorithms based on the obtained standardized detection data set, including the accuracy rate, false alarm rate, response delay, cross-system comparison (i.e., the performance difference of the same algorithm in different energy storage systems), etc., to obtain the evaluation results of each algorithm. Based on the evaluation results, perform targeted optimization and update on the algorithms with poor performance in the aspects with poor performance, so as to improve the accuracy of subsequent state detection of the energy storage system and the response speed of operating state detection.

[0100] In this embodiment, the portable detection device of the energy storage system provided by the present application includes an automated report generation module, which is used to provide formatted and detailed analysis results and suggestions, facilitating the quick understanding and action-taking by operation and maintenance personnel. The portable detection device of the energy storage system includes a feedback receiving module, which is used to collect the feedback of operation and maintenance personnel on the diagnostic results and the on-site performance data of the device performance. Through these feedback data, the performance of the algorithm is continuously monitored, and adjustments and optimizations are made as needed. The portable detection device of the energy storage system also includes a power supply module, which is used to provide energy during the detection process for the portable detection device of the energy storage system. Each module in the portable detection device of the energy storage system provided by the present application adopts a modular design, and can be quickly replaced or module-upgraded according to different detection requirements, select a suitable embedded processor or FPGA to meet the requirements of data analysis and algorithm processing, and configure sufficient RAM and storage space to ensure the smoothness of data processing and the requirements of data storage.

[0101] Based on the same inventive concept, an embodiment of the present application provides a method for detecting the operating state of an energy storage system. The method is applied to a portable detection device of an energy storage system described in the first aspect of the present application. The method includes:

[0102] Step S31: Adapt the communication protocol of the energy storage system, and collect detection data of the energy storage system based on the adaptation result;

[0103] Step S32: Convert the detection data into target data, where the target data is detection data under the target communication protocol supported by the operating state detection module;

[0104] Step S33: Evaluate the abnormality of the target data through a preset detection rule, and perform corresponding fault warnings based on the evaluation result;

[0105] Step S34: Process the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault warning.

[0106] Optionally, determine the communication protocol of the energy storage system through the data acquisition module included in the adapter, perform protocol configuration based on the determined communication protocol, and collect detection data of the energy storage system based on the configuration result;

[0107] Optionally, the method further includes:

[0108] Collect basic detection data at the cell level, module level, and container level through the basic data acquisition module in the data acquisition module; and collect first-level detection data at the cell level through the differential data acquisition module in the data acquisition module; and collect second-level detection data at the module level; and collect third-level detection data at the container level.

[0109] Optionally, perform anomaly assessment on the target data through a preset detection rule, and perform corresponding fault alarms based on the assessment results, including: performing anomaly assessment on the target data at different levels through a preset detection rule to obtain corresponding assessment results; determining the target level to which the target data with anomalies belongs according to the assessment results, and performing a fault alarm for the target level;

[0110] Processing the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault alarm, including: processing the target data at the target level through a fault diagnosis model to determine the fault type corresponding to the fault alarm at the target level;

[0111] Perform correlation analysis on the target data at different levels, and determine the current state of the energy storage system at each level based on the correlation analysis results;

[0112] Analyze the target data at different levels for a historical preset duration through a fault prediction model to predict the state of the energy storage system at each level.

[0113] Optionally, processing the target data at the target level through a fault diagnosis model to determine the fault type corresponding to the fault alarm at the target level includes: receiving the target data at the target level, and performing normalization and time window alignment processing to obtain a corresponding three-dimensional input tensor; extracting spatial features from the three-dimensional input tensor through a convolutional neural network; extracting temporal features from the three-dimensional input tensor through a bidirectional long short-term memory network; performing dynamic weight allocation and calculating cross-modal correlation on the spatial features and the temporal features through a self-attention mechanism to obtain a fused feature vector; processing the fused feature vector through a fully connected layer and a classifier to determine the fault type at the abnormal target level.

[0114] Optionally, processing the target data at the target level through a fault diagnosis model to determine the fault type corresponding to the fault alarm at the target level includes: receiving the target data at the target level, and performing normalization and time window alignment processing to obtain a corresponding three-dimensional input tensor; processing the three-dimensional input tensor to determine the probabilities corresponding to various fault types at the abnormal target level; processing the temporal type tensor in the three-dimensional input tensor to determine the probabilities corresponding to various fault types at the abnormal target level; performing distribution anomaly analysis on the target data at the abnormal target level to determine the probabilities corresponding to various fault types at the abnormal target level; fusing the output results of each sub-model according to the confidence levels of each sub-model to determine the final fault type at the abnormal target level.

[0115] Optionally, analyzing the target data at different levels of the historical preset duration through the fault prediction model to predict the state of the energy storage system at each level includes: receiving the target data at different levels of the historical preset duration, performing normalization and time window alignment processing to obtain the corresponding three-dimensional input tensor; performing linear trend prediction on the target data at different levels of the historical preset duration to obtain the corresponding first prediction result; performing non-linear feature extraction and prediction on the three-dimensional input tensor to obtain the corresponding second prediction result; dynamically allocating weights to each sub-model according to the past prediction errors of each sub-model, and performing weighted fusion on the first prediction result and the second prediction result based on the weight allocation result to obtain the final prediction result.

[0116] Optionally, analyzing the target data at different levels of the historical preset duration through the fault prediction model to predict the state of the energy storage system at each level includes: receiving the target data at different levels of the historical preset duration, performing normalization and time window alignment processing to obtain the corresponding three-dimensional input tensor; performing clustering analysis on the three-dimensional input tensor through a clustering algorithm to determine the battery state at different levels; determining the fault type corresponding to the data of the abnormal battery state under the current environmental data based on the pre-established abnormal association rules; predicting the battery health state score and remaining life at different levels according to the capacity attenuation rate determined by the three-dimensional input tensor.

[0117] Optionally, the method further includes: storing various algorithms required by the operating state detection module through an algorithm library, and connecting to the operating state detection module through an API interface to provide various algorithms required for state detection.

[0118] Optionally, the method further includes: obtaining the detection data of each energy storage system through an algorithm management platform to evaluate the effects of various algorithms, and optimizing and updating the various algorithms based on the evaluation results.

[0119] For the method embodiments, since they are basically similar to the system embodiments, the description is relatively simple. For the relevant parts, refer to the partial description of the method embodiments.

[0120] It should be noted that for the method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the embodiments of the present application are not limited by the described action sequence, because according to the embodiments of the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0121] The embodiments in this specification are all described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other.

[0122] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the embodiments of the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0124] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0125] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0126] Although the preferred embodiments of the embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present application.

[0127] Finally, it should also be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, article or terminal device comprising the said element.

[0128] The above has introduced in detail a portable detection device, method and product of an energy storage system provided by the present application. Specific examples are used in this text to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A portable detection device for an energy storage system, characterized in that, The device includes: an adapter and an operating status detection module; the operating status detection module includes: a status evaluation module and a fault type determination module; The adapter is used to be pluggably connected to energy storage systems with different communication protocols, adapt to the communication protocols of the connected energy storage systems to collect detection data, and convert the detection data into target data and send it to the operating status detection module. The target data is the detection data under the target communication protocol supported by the operating status detection module; The status evaluation module is used to perform anomaly evaluation on the target data through a preset detection rule and perform corresponding fault alarms based on the evaluation results; The fault type determination module is used to process the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault alarm.

2. The portable detection device for an energy storage system according to claim 1, characterized in that The adapter includes a data collection module, which is used to determine the communication protocol of the energy storage system, perform protocol configuration based on the determined communication protocol, and collect detection data from the energy storage system based on the configuration result; The data collection module includes: A basic data collection module, which is used to collect basic detection data at the cell level, module level, and container level; A differential data collection module, which is used to collect first-level detection data at the cell level; and collect second-level detection data at the module level; and collect third-level detection data at the container level.

3. The portable detection device for an energy storage system according to claim 2, characterized in that, The device further includes: a cross-level status analysis module and a data mining analysis module; The status evaluation module is used to perform anomaly evaluation on the target data at different levels through a preset detection rule to obtain corresponding evaluation results; The status evaluation module is further used to determine the target level to which the abnormal target data belongs according to the evaluation results and perform a fault alarm for the target level; The fault type determination module is used to process the target data at the target level through a fault diagnosis model to determine the fault type corresponding to the fault alarm at the target level; The cross-level status analysis module is used to perform a correlation analysis on the target data at different levels and determine the current status of the energy storage system at each level based on the correlation analysis results; The data mining analysis module is used to analyze the target data at different levels for a historical preset duration through a fault prediction model to predict the status of the energy storage system at each level.

4. The portable detection device for an energy storage system according to claim 3, characterized in that, The fault diagnosis model in the fault type determination module includes: A first data input module, which is used to receive the target data at the target level, perform standardization and time window alignment processing to obtain a corresponding three-dimensional input tensor; A spatial feature extraction module, which is used to extract spatial features from the three-dimensional input tensor through a convolutional neural network; A temporal feature extraction module, which is used to extract temporal features from the three-dimensional input tensor through a bidirectional long short-term memory network; A multimodal fusion module, which is used to perform dynamic weight allocation on the spatial features and the temporal features through a self-attention mechanism and calculate the cross-modal correlation to obtain a fused feature vector; The diagnostic output module is used to process the fused feature vector through a fully connected layer and a classifier to determine the fault types at the abnormal target level.

5. The portable detection device for an energy storage system according to claim 3, characterized in that, The fault diagnosis model in the fault type determination module includes: The second data input module is used to receive the target data at the target level, perform normalization and time window alignment processing, and obtain the corresponding three-dimensional input tensor. The supervised learning sub-model is used to process the three-dimensional input tensor to determine the probabilities corresponding to various fault types at the abnormal target level. The time series sub-model is used to process the time series type tensor in the three-dimensional input tensor to determine the probabilities corresponding to various fault types at the abnormal target level. The statistical analysis sub-model is used to perform distribution anomaly analysis on the target data at the abnormal target level to determine the probabilities corresponding to various fault types at the abnormal target level. The first result fusion module is used to assign weights according to the confidence levels of each sub-model, fuse the output results of each sub-model, and determine the final fault type at the abnormal target level.

6. The portable detection device for an energy storage system according to claim 3, characterized in that, The fault prediction model in the data mining and analysis module includes: The third data input module is used to receive the target data at different levels for a historical preset duration, perform normalization and time window alignment processing, and obtain the corresponding three-dimensional input tensor. The statistical sub-model is used to perform linear trend prediction on the target data at different levels for a historical preset duration to obtain the corresponding first prediction result. The deep learning sub-model is used to perform non-linear feature extraction and prediction on the three-dimensional input tensor to obtain the corresponding second prediction result. The second result fusion module is used to dynamically assign weights to each sub-model according to the past prediction errors of each sub-model, and perform weighted fusion on the first prediction result and the second prediction result based on the weight assignment result to obtain the final prediction result.

7. The portable detection device for an energy storage system according to claim 3, wherein The fault prediction model in the data mining and analysis module includes: The fourth data input module is used to receive the target data at different levels for a historical preset duration, perform normalization and time window alignment processing, and obtain the corresponding three-dimensional input tensor. The clustering analysis module is used to perform clustering analysis on the three-dimensional input tensor through a clustering algorithm to determine the battery states at different levels. The association module is used to determine the fault types corresponding to the data of the abnormal battery state under the current environmental data based on the pre-established abnormal association rules. The health prediction module is used to predict the battery health state score and remaining life at different levels according to the capacity attenuation rate determined by the three-dimensional input tensor.

8. A portable detection device for an energy storage system according to claim 1, characterized in that, The adapter includes a communication module, which is used to convert the detection data into target data based on the communication protocol of the energy storage system and send it to the operation state detection module.

9. The portable detection device for an energy storage system according to claim 1, characterized in that, The device further includes: The algorithm library is used to store various algorithms required by the operation state detection module, and is used to connect to the operation state detection module through an API interface to provide various algorithms required for state detection.

10. The portable detection device for an energy storage system according to claim 1, characterized in that, The device is connected to the algorithm management platform; The algorithm management platform is used to obtain the detection data of each energy storage system for the effect evaluation of various algorithms, and optimize and update the various algorithms based on the evaluation results.

11. A method for detecting the operating state of an energy storage system according to claim 1, characterized in that, The method is applied to a portable detection device for an energy storage system according to any one of claims 1 to 10, and the method includes: Adapting the communication protocol of the energy storage system, and collecting detection data of the energy storage system based on the adaptation result; Converting the detection data into target data, where the target data is the detection data under the target communication protocol supported by the operation status detection module; Evaluating the abnormality of the target data through a preset detection rule, and performing corresponding fault alarms based on the evaluation result; Processing the abnormal target data through a fault diagnosis model to determine the fault type corresponding to the fault alarm.

12. An electronic device, characterized in that, Including: A processor, a memory, and a computer program stored on the memory and running on the processor. When the computer program is executed by the processor, it implements the steps in a method for detecting the operation status of an energy storage system according to claim 11.

13. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements the steps in a method for detecting the operation status of an energy storage system according to claim 11.

Images