Energy storage system fault detection method and energy storage system

Abstract

The application relates to a fault detection method of an energy storage system and the energy storage system. The energy storage system is composed of multiple device components and is applied to an energy storage edge. The fault detection method of the energy storage system comprises the following steps: obtaining basic operation data generated by the multiple device components in an operation process, and performing fault identification on the basic operation data to obtain fault event data, wherein the fault event data represents fault types of the multiple device components; extracting target fault event data in a preset time window from a preset fault event time sequence database; performing fault evaluation on the multiple device components in the preset time window according to the target fault event data to obtain fault evaluation feature data, wherein the fault evaluation feature data represents fault occurrence conditions of the multiple device components in the preset time window; and sending the target basic operation data and the fault evaluation feature data in the preset time window to an energy storage cloud. The method can improve the fault detection effect of the energy storage system.

Description

Technical Field

[0001] This application relates to the field of energy storage battery management system technology, and in particular to a fault detection method for an energy storage system and an energy storage system. Background Technology

[0002] With the continuous development of technology, energy storage systems have been widely used in many fields. An energy storage system is usually composed of multiple equipment components such as a power conversion system (PCS), a battery management system (BMS), and an energy management system (EMS). The various equipment components work together to stably complete the entire system of energy storage and release. Therefore, the operational stability of the equipment components is directly related to the safety and efficiency of the energy storage system.

[0003] Currently, fault detection in energy storage systems typically relies on edge devices or cloud platforms for real-time monitoring. However, this often results in issues such as detection lag, identification errors, and excessive time consumption, making it impossible to balance real-time performance and accuracy in the fault detection process. Consequently, the current methods for fault detection in energy storage systems are ineffective. Summary of the Invention

[0004] Therefore, it is necessary to provide a fault detection method and energy storage system for energy storage systems that can improve the effectiveness of fault detection in energy storage systems, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a fault detection method for an energy storage system, applied at the edge of the energy storage system, wherein the energy storage system comprises multiple equipment components; the fault detection method for the energy storage system includes:

[0006] Acquire basic operational data generated by the multiple device components during operation, and perform fault identification on the basic operational data to obtain fault event data, wherein the fault event data characterizes the fault type of the multiple device components;

[0007] Extract target fault event data within a preset time window from a preset fault event time series database, wherein the preset fault event time series database stores fault event data recorded over time;

[0008] Based on the target fault event data, fault assessment is performed on the multiple device components within the preset time window to obtain fault assessment feature data, wherein the fault assessment feature data characterizes the fault occurrence of the multiple device components within the preset time window;

[0009] The system sends the target basic operating data and the fault assessment feature data within the preset time window to the energy storage cloud. The energy storage cloud then inputs the target basic operating data and the fault assessment feature data into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when the multiple equipment components fail, obtaining the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, the system generates a cloud-based fault detection result for the energy storage system. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system.

[0010] Secondly, this application also provides a fault detection method for an energy storage system, applied to an energy storage cloud, wherein the energy storage system is composed of multiple device components; the fault detection method for the energy storage system includes:

[0011] The system acquires target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge device. The fault assessment feature data is obtained by the energy storage edge device based on target fault event data extracted from a preset fault event time series database to assess the faults of the multiple device components. The preset fault event time series database stores fault event data recorded over time. The fault event data is obtained by fault identification of the basic operating data generated by the multiple device components during operation.

[0012] Based on the system type of the energy storage system, a fault correlation analysis is performed on the fault assessment feature data to obtain the fault correlation analysis results, wherein the fault correlation analysis results include the fault feature weights corresponding to the fault assessment feature data.

[0013] The target basic operating data and the fault assessment feature data are input into a preset fault risk prediction model. The preset fault risk prediction model is used to predict the impact risk on the energy storage system when the multiple equipment components fail, and the fault risk prediction results corresponding to each of the multiple equipment components are obtained.

[0014] Based on all fault risk prediction results and the fault feature weights, cloud-based fault detection results for the energy storage system are generated.

[0015] Thirdly, this application also provides a fault detection device for an energy storage system, applied at the edge of the energy storage system, wherein the energy storage system is composed of multiple equipment components; the fault detection device for the energy storage system includes:

[0016] The first acquisition module is used to acquire basic operating data generated by the multiple device components during operation, and to perform fault identification on the basic operating data to obtain fault event data, wherein the fault event data characterizes the fault type of the multiple device components;

[0017] The extraction module is used to extract target fault event data within a preset time window from a preset fault event time series database, wherein the preset fault event time series database stores fault event data recorded over time.

[0018] The fault assessment module is used to assess the faults of the multiple equipment components within the preset time window based on the target fault event data, and obtain fault assessment feature data, wherein the fault assessment feature data characterizes the fault occurrence of the multiple equipment components within the preset time window.

[0019] The sending module is used to send the target basic operating data and the fault assessment feature data within the preset time window to the energy storage cloud, so that the energy storage cloud can input the target basic operating data and the fault assessment feature data into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk of the energy storage system when the multiple equipment components fail, obtains the fault risk prediction results corresponding to each of the multiple equipment components, and generates the cloud-based fault detection results of the energy storage system based on all fault risk prediction results and fault feature weights. The fault feature weights are obtained by the energy storage cloud from fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system.

[0020] Fourthly, this application also provides a fault detection device for an energy storage system, applied in an energy storage cloud, wherein the energy storage system comprises multiple equipment components; the fault detection device for the energy storage system includes:

[0021] The second acquisition module is used to acquire target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge device. The fault assessment feature data is obtained by the energy storage edge device based on target fault event data extracted from a preset fault event time series database to perform fault assessment on the multiple device components. The preset fault event time series database stores fault event data recorded over time. The fault event data is obtained by fault identification of the basic operating data generated by the multiple device components during operation.

[0022] The fault analysis module is used to perform fault correlation analysis on the fault assessment feature data according to the system type of the energy storage system, and obtain fault correlation analysis results, wherein the fault correlation analysis results include fault feature weights corresponding to the fault assessment feature data.

[0023] The prediction module is used to input the target basic operating data and the fault assessment feature data into a preset fault risk prediction model, and to predict the impact risk of the energy storage system when the multiple equipment components fail through the preset fault risk prediction model, so as to obtain the fault risk prediction results corresponding to each of the multiple equipment components.

[0024] The generation module is used to generate cloud-based fault detection results for the energy storage system based on all fault risk prediction results and the fault feature weights.

[0025] Fifthly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0026] The system acquires basic operational data generated by the multiple device components during operation and performs fault identification on the basic operational data to obtain fault event data, wherein the fault event data characterizes the fault type of the multiple device components; it extracts target fault event data within a preset time window from a preset fault event time series database, wherein the preset fault event time series database stores fault event data recorded over time; based on the target fault event data, it performs fault assessment on the multiple device components within the preset time window to obtain fault assessment feature data, wherein the fault assessment feature data characterizes the fault occurrence status of the multiple device components within the preset time window; and it sends the data to the energy storage cloud. The target basic operating data and the fault assessment feature data within the preset time window are input together by the energy storage cloud into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when the multiple equipment components fail, obtaining the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, a cloud-based fault detection result for the energy storage system is generated. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system; or...

[0027] The system acquires target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge. The fault assessment feature data is obtained by the energy storage edge based on target fault event data extracted from a preset fault event time-series database, which stores fault event data recorded over time. This fault event data is obtained by identifying faults in the basic operating data generated by the multiple device components during operation. Based on the system type of the energy storage system, a fault correlation analysis is performed on the fault assessment feature data to obtain a fault correlation analysis result, which includes the fault feature weights corresponding to the fault assessment feature data. The target basic operating data and the fault assessment feature data are jointly input into a preset fault risk prediction model. This model predicts the impact risk on the energy storage system when the multiple device components fail, obtaining fault risk prediction results for each of the multiple device components. Finally, based on all fault risk prediction results and the fault feature weights, a cloud-based fault detection result for the energy storage system is generated.

[0028] Sixthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0029] The system acquires basic operational data generated by the multiple device components during operation and performs fault identification on the basic operational data to obtain fault event data, wherein the fault event data characterizes the fault type of the multiple device components; it extracts target fault event data within a preset time window from a preset fault event time series database, wherein the preset fault event time series database stores fault event data recorded over time; based on the target fault event data, it performs fault assessment on the multiple device components within the preset time window to obtain fault assessment feature data, wherein the fault assessment feature data characterizes the fault occurrence status of the multiple device components within the preset time window; and it sends the data to the energy storage cloud. The target basic operating data and the fault assessment feature data within the preset time window are input together by the energy storage cloud into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when the multiple equipment components fail, obtaining the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, a cloud-based fault detection result for the energy storage system is generated. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system; or...

[0030] The system acquires target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge. The fault assessment feature data is obtained by the energy storage edge based on target fault event data extracted from a preset fault event time-series database, which stores fault event data recorded over time. This fault event data is obtained by identifying faults in the basic operating data generated by the multiple device components during operation. Based on the system type of the energy storage system, a fault correlation analysis is performed on the fault assessment feature data to obtain a fault correlation analysis result, which includes the fault feature weights corresponding to the fault assessment feature data. The target basic operating data and the fault assessment feature data are jointly input into a preset fault risk prediction model. This model predicts the impact risk on the energy storage system when the multiple device components fail, obtaining fault risk prediction results for each of the multiple device components. Finally, based on all fault risk prediction results and the fault feature weights, a cloud-based fault detection result for the energy storage system is generated.

[0031] In a seventh aspect, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0032] The system acquires basic operational data generated by the multiple device components during operation and performs fault identification on the basic operational data to obtain fault event data, wherein the fault event data characterizes the fault type of the multiple device components; it extracts target fault event data within a preset time window from a preset fault event time series database, wherein the preset fault event time series database stores fault event data recorded over time; based on the target fault event data, it performs fault assessment on the multiple device components within the preset time window to obtain fault assessment feature data, wherein the fault assessment feature data characterizes the fault occurrence status of the multiple device components within the preset time window; and it sends the data to the energy storage cloud. The target basic operating data and the fault assessment feature data within the preset time window are input together by the energy storage cloud into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when the multiple equipment components fail, obtaining the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, a cloud-based fault detection result for the energy storage system is generated. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system; or...

[0033] The system acquires target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge. The fault assessment feature data is obtained by the energy storage edge based on target fault event data extracted from a preset fault event time-series database, which stores fault event data recorded over time. This fault event data is obtained by identifying faults in the basic operating data generated by the multiple device components during operation. Based on the system type of the energy storage system, a fault correlation analysis is performed on the fault assessment feature data to obtain a fault correlation analysis result, which includes the fault feature weights corresponding to the fault assessment feature data. The target basic operating data and the fault assessment feature data are jointly input into a preset fault risk prediction model. This model predicts the impact risk on the energy storage system when the multiple device components fail, obtaining fault risk prediction results for each of the multiple device components. Finally, based on all fault risk prediction results and the fault feature weights, a cloud-based fault detection result for the energy storage system is generated.

[0034] The aforementioned fault detection method and energy storage system are applied at the energy storage edge, where the system consists of multiple equipment components. The energy storage edge first collects basic operational data generated by these components during operation and performs fault identification on this data to obtain fault event data. This fault event data characterizes the fault types of the multiple equipment components. Then, target fault event data within a preset time window is extracted from a preset fault event time-series database, which stores fault event data recorded over time. Based on the target fault event data, fault assessment is performed on the multiple equipment components within the preset time window to obtain fault assessment feature data. This fault assessment feature data characterizes the fault occurrence status of the multiple equipment components within the preset time window. This allows for rapid processing and analysis of basic operational data at the energy storage edge, enabling preliminary fault detection of the energy storage system. The objective is to further provide the target basic operation data and fault assessment feature data within a preset time window to the energy storage cloud for inputting together the target basic operation data and fault assessment feature data into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk of multiple equipment components failing on the energy storage system, obtaining the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, the cloud-based fault detection results of the energy storage system are generated. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the type of the energy storage system. This enables the energy storage cloud to perform in-depth analysis and accurate prediction of the operation status related data uploaded by the energy storage edge, and, combined with the correlation analysis logic adapted to the system type, to analyze and obtain the overall impact assessment and risk level of each equipment component failure.Because the energy storage edge is deployed close to the equipment components, it can capture basic operational data in real time and perform rapid feature analysis. This allows for rapid early warning of fault conditions in each equipment component without relying on remote communication transmission, effectively ensuring timely fault response. Meanwhile, the energy storage cloud possesses powerful computing capabilities and massive data processing capabilities. It can utilize the complete operational status parameters and fault assessment feature data uploaded from the energy storage edge, combined with system type, to perform targeted complex correlation analysis and model prediction. This enables accurate detection of fault conditions and impact risks for each equipment component. Finally, the final fault diagnosis of the energy storage system is obtained by comparing the results from both the energy storage edge and the energy storage cloud. The detection results demonstrate that the system achieves automated, end-to-end fault detection and risk assessment of energy storage systems through collaborative efforts between edge and cloud-based energy storage, rather than relying solely on the limited computing power of edge devices for simple detection or manual inspection by operators. Therefore, it overcomes the technical shortcomings of insufficient edge device computing power to support complex diagnostics, delayed cloud response due to communication latency, and limited manual identification capabilities and efficiency that cannot match the needs of large-scale deployment of energy storage systems. These shortcomings often lead to detection delays, errors, misjudgments of fault risks, or excessively long detection times. Thus, the system improves the effectiveness of fault detection in energy storage systems. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart illustrating a fault detection method for an energy storage system in one embodiment;

[0037] Figure 2 This is a flowchart illustrating a fault detection method for an energy storage system in another embodiment;

[0038] Figure 3 The following is a control flowchart of a fault detection method for an energy storage system in another embodiment, which performs fault detection on the energy storage system by coordinating the energy storage edge and the energy storage cloud.

[0039] Figure 4 This is a structural block diagram of a fault detection device for an energy storage system applied at the energy storage edge in one embodiment;

[0040] Figure 5 This is a structural block diagram of a fault detection device for an energy storage system applied to an energy storage cloud in one embodiment;

[0041] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0043] First, it should be understood that with the continuous development of technology, energy storage systems have been widely used in scenarios such as grid peak shaving and frequency regulation, and renewable energy grid integration and consumption. However, energy storage systems are prone to faults during operation, such as cell consistency degradation, abnormal temperature, sensor failure, and communication delays. Therefore, fault detection of energy storage systems is essential. Currently, fault detection in energy storage systems often relies on edge devices such as BMS or PCS for real-time monitoring, but their computing power is limited, making it difficult to diagnose and predict complex models. Alternatively, manual troubleshooting methods may be used, but these methods are limited by human identification capabilities. The detection efficiency is difficult to adapt to the needs of large-scale deployment of energy storage systems, and problems such as detection lag, identification errors and excessive time consumption are likely to occur. On the other hand, although cloud platforms have powerful data analysis and modeling capabilities, they are difficult to respond to abnormal situations in a timely manner due to communication latency and bandwidth limitations. Consequently, they cannot balance real-time performance and accuracy in the fault detection process. Therefore, how to fully leverage the real-time capabilities of the edge and the computing power of the cloud to achieve efficient and accurate fault detection of energy storage systems has become an urgent problem to be solved. In other words, there is an urgent need for a fault detection method for energy storage systems that can improve the effectiveness of fault detection.

[0044] In one embodiment, such as Figure 1As shown, a fault detection method for an energy storage system is provided. This embodiment uses the application of this method to the energy storage edge as an example. The energy storage edge refers to the computing, control, and communication units deployed locally or nearby to the energy storage device in the energy storage system. It is the layer closest to the physical device in the entire energy storage system architecture and can also be called the "energy storage edge". Specifically, it may include embedded edge controllers, industrial-grade edge computing gateways, smart sensors, or actuators, etc. The energy storage edge consists of multiple device components. Device components refer to independent hardware or hardware / software combined units that support the core functions of the energy storage edge, such as data acquisition, local computing, control scheduling, and communication transmission. Specifically, it may be an energy storage converter, a battery management system, and an energy storage device. Furthermore, in management systems, multiple device components can be heterogeneous. This allows different components to collaborate based on their respective functional advantages to achieve their functions. Heterogeneous components refer to different categories of equipment with fundamental differences in functional positioning, hardware form, processing logic, or communication methods. In contrast, homogeneous components refer to units of the same type, specifications, or functions that are completely identical. For example, in one feasible implementation, battery cells / modules, BMUs, CMUs, PCSs, EMSs, and NTC temperature sensors are considered heterogeneous device components, while multiple battery cells of the same specification and multiple sets of BMU acquisition units of the same model are considered homogeneous components. The energy storage edge is deployed with a first acquisition module, an extraction module, and... The system includes a fault assessment module and a transmission module. The first acquisition module acquires basic operational data generated by multiple device components during operation and performs fault identification on this data to obtain fault event data, where the fault event data represents the fault type of the multiple device components. The extraction module extracts target fault event data within a preset time window from a preset fault event time-series database, which stores fault event data recorded over time. The fault assessment module performs fault assessment on multiple device components within the preset time window based on the target fault event data, obtaining fault assessment feature data, where the fault assessment feature data represents the fault type of the multiple device components within the preset time window. The system displays the fault occurrence status within a preset time window. The sending module is used to send the target basic operation data and fault assessment feature data within a preset time window to the energy storage cloud. The energy storage cloud then inputs the target basic operation data and fault assessment feature data into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when multiple equipment components fail, obtaining the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, the cloud-based fault detection results of the energy storage system are generated. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system.For example, in one feasible approach, the primary function of the first acquisition module is to collect basic operational data from energy storage devices. This module can be deployed in smart sensors or data acquisition units at the energy storage edge. The primary function of the extraction module is to filter and extract target fault event data within a preset time window from a preset fault event time series database, completing data filtering and preliminary processing. This module can be deployed in an industrial-grade edge computing gateway at the energy storage edge, leveraging the gateway's local data processing capabilities to quickly access historical and real-time fault event data from the time series database, improving data extraction efficiency. The primary function of the fault assessment module is to analyze the frequency of fault occurrence, duration of fault occurrence, and fault correlation of each energy storage device component within a preset time window based on the extracted target fault event data, generating fault assessment feature data characterizing the fault occurrence. This module can be deployed in an embedded edge controller at the energy storage edge, utilizing the controller's efficient computing power to complete local fault assessment analysis, reducing reliance on cloud computing power. The primary function of the transmission module is to establish a communication connection between the energy storage edge and the energy storage cloud, securely, stably, and in real-time transmitting the target basic operational data and fault assessment feature data to the cloud, ensuring... The integrity and timeliness of data transmission can be ensured by deploying communication modules (such as 4G / 5G communication modules, fiber optic communication modules) at the energy storage edge or in the communication units of industrial-grade edge computing gateways. Supporting multiple communication protocols, it adapts to the cloud communication needs of different energy storage scenarios. Furthermore, because the energy storage edge is deployed close to the equipment components, it can capture basic operating data of the equipment in real time and perform rapid feature analysis. This allows for rapid early warning of fault conditions in each equipment component without relying on remote communication transmission, effectively ensuring timely fault response. Meanwhile, the energy storage cloud possesses powerful computing capabilities and massive data processing capabilities. It can then use the complete operating status parameters and fault assessment feature data uploaded by the energy storage edge, combined with the system type, to perform complex correlation analysis and model prediction, thereby accurately detecting the fault conditions and impact risks of each equipment component. Finally, the final fault detection result of the energy storage system is obtained by comparing and verifying the results from both the energy storage edge and the energy storage cloud. This achieves the goal of automating and conducting full-process fault detection and risk assessment of the energy storage system through collaboration between the energy storage edge and the energy storage cloud, thus improving the effectiveness of fault detection in energy storage systems. In this embodiment, the method includes the following steps:

[0045] Step 202: Obtain basic operating data generated by multiple device components during operation, and perform fault identification on the basic operating data to obtain fault event data, wherein the fault event data represents the fault type of multiple device components.

[0046] It should be noted that the equipment component is an independent hardware or hardware / software combined unit that supports the core functions of energy storage edge, such as data acquisition, local computing, control scheduling, and communication transmission. During operation, the equipment component generates corresponding basic operating data, which includes, but is not limited to, current, voltage, active power, reactive power, and liquid cooling rate. For example, in one feasible implementation, assuming the equipment component is an energy storage converter, the basic operating data that can be collected by sensors includes input and output current, input and output voltage, active power, reactive power, converter efficiency, module operating temperature, and liquid cooling rate. Assuming the equipment component is a battery management system, the basic operating data that can be collected by sensors includes individual battery voltage, total battery module voltage, battery pack charging and discharging current, individual battery and module temperature, remaining battery capacity, battery health status, and balancing current.

[0047] It should be noted that after obtaining the basic operational data, lightweight diagnostic algorithms can be used to quickly identify faults in the energy storage system to obtain fault event data. Specifically, these lightweight diagnostic algorithms can be threshold detection algorithms or Kalman filter residual analysis algorithms. The fault event data characterizes the fault types of multiple device components, including but not limited to communication interruptions, exceeding operating parameter thresholds, and abnormal switch operation. After obtaining the fault event data, the energy storage edge can store it locally. For example, based on different device components of the energy storage system and their device identification information, the obtained fault event data can be classified and cached. Specifically, the local cache uses both key-value pair structures and time-series data. The structure is as follows: 1) A key-value pair structure is used to store data related to fault statistics. Its key includes at least device identification information and fault type identification. The value corresponding to the key is used to record the statistical information of the device component under the corresponding fault type, which can be the number of communication interruptions, the frequency of communication interruptions, or the level of communication interruptions; 2) A time-series data structure is used to store the occurrence time of basic operating data and fault event data to support subsequent time window statistical analysis. In one feasible approach, the energy storage edge can be an SCU controller. Its local storage can be implemented by combining a ring buffer and a lightweight time-series key-value storage engine to support continuous data writing and breakpoint resume.

[0048] As an example, step 202 includes: collecting basic operating data generated by multiple device components during operation through sensors, and using a Kalman filter residual analysis algorithm to identify the basic operating data to obtain fault event data.

[0049] Step 204: Extract target fault event data within a preset time window from the preset fault event time sequence database, wherein the preset fault event time sequence database stores fault event data recorded over time.

[0050] Understandably, to facilitate data management and subsequent fault analysis, after obtaining fault event data, a preset fault event time-series database can be constructed based on the fault event data. The preset fault event time-series database stores fault event data recorded over time. For example, in one feasible approach, the fault event data in the preset fault event time-series database can be stored using a key-value pair structure and a time-series data structure, thus enabling the fault event data to possess time-series attributes.

[0051] It should be noted that the preset time window refers to a specific time interval pre-set for filtering target fault event data. It can be flexibly configured according to actual fault analysis needs, such as setting it to a fixed duration of the most recent 1 hour, 6 hours, 24 hours, or 7 days, or dynamically adjusting the time span according to changes in the energy storage system's operating load and periods of high fault incidence. By setting specific start and end time nodes, and matching the time sequence attributes of fault event data in the preset fault event time sequence database, the target fault event data can be extracted from the preset fault event time sequence database. This ensures that the extracted data accurately corresponds to the time range to be analyzed, providing targeted data support for the subsequent fault assessment module to conduct fault occurrence analysis.

[0052] As an example, step 204 includes: using the start and end times of a preset time window as indexes, querying fault event data within the preset time window in a preset fault event time series database, and extracting the fault event data within the preset time window as target fault event data.

[0053] Step 206: Based on the target fault event data, perform fault assessment on multiple equipment components within a preset time window to obtain fault assessment feature data, wherein the fault assessment feature data characterizes the fault occurrence of multiple equipment components within the preset time window.

[0054] It should be noted that after obtaining the target fault event data, fault assessment feature data can be obtained, thereby enabling the determination of the faults in multiple equipment components within a preset time window. Specifically, the fault assessment feature data can represent the occurrence of fault events, the frequency of occurrence of fault events, and the level of fault events.

[0055] As an example, step 206 includes: performing statistical analysis on the target fault event data within a preset time window to obtain fault assessment feature data.

[0056] Step 208: Send the target basic operation data and fault assessment feature data within a preset time window to the energy storage cloud so that the energy storage cloud can input the target basic operation data and fault assessment feature data into the preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when multiple equipment components fail, and obtains the fault risk prediction results corresponding to each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, the cloud-based fault detection results of the energy storage system are generated. The fault feature weights are obtained by the energy storage cloud from fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system.

[0057] It should be noted that the detection of faults in energy storage systems by the energy storage edge has certain limitations, specifically in terms of limited computing power, making it difficult to diagnose and predict complex models. Therefore, after generating fault assessment feature data, it is possible to further collaborate with the energy storage cloud to complete the fault detection of the energy storage system.

[0058] It should be noted that the energy storage cloud is a remote server cluster or platform built on cloud computing technology for the management and data analysis of energy storage systems. Specifically, it can be a centralized management platform for large-scale energy storage power stations or an energy storage cloud platform, etc. The cloud fault detection results are obtained by the energy storage cloud using machine learning models to perform multi-dimensional feature analysis. Specifically, it can be the qualitative fault detection results or the quantitative fault detection results of the energy storage system. The qualitative fault detection results of the energy storage system can indicate that the energy storage system is faulty or that the energy storage system is fault-free, while the quantitative fault detection results of the energy storage system can be the health score of the energy storage system.

[0059] As an example, step 208 includes: sending target basic operation data and fault characteristic data within a preset time window to the energy storage cloud, so that the energy storage cloud can input the target basic operation data and fault assessment characteristic data into a preset fault risk prediction model, predict the impact risk of multiple equipment components failing on the energy storage system through the preset fault risk prediction model, obtain the fault risk prediction results corresponding to each of the multiple equipment components, and generate cloud-based fault detection results of the energy storage system based on all fault risk prediction results and fault characteristic weights. The fault characteristic weights are obtained by the energy storage cloud from fault correlation analysis of the fault assessment characteristic data according to the system type of the energy storage system.

[0060] The aforementioned fault detection method for energy storage systems is applied at the energy storage edge, where the system consists of multiple equipment components. The energy storage edge first collects basic operational data generated by these components during operation and performs fault identification on this data to obtain fault event data. This fault event data characterizes the fault types of the multiple equipment components. Then, target fault event data within a preset time window is extracted from a preset fault event time-series database, which stores fault event data recorded over time. Based on the target fault event data, fault assessment is performed on the multiple equipment components within the preset time window to obtain fault assessment feature data. This fault assessment feature data characterizes the fault occurrence status of the multiple equipment components within the preset time window. This allows for the rapid processing and analysis of basic operational data at the energy storage edge, achieving the goal of preliminary fault detection for the energy storage system. Furthermore, the energy storage edge will transmit the target basic operation data and fault assessment feature data within a preset time window to the energy storage cloud. The target basic operation data and fault assessment feature data will be input into the preset fault risk prediction model. The preset fault risk prediction model will predict the impact risk of multiple equipment components failing on the energy storage system, and obtain the fault risk prediction results corresponding to each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, the cloud fault detection results of the energy storage system will be generated. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the type of energy storage system. This enables the energy storage cloud to perform in-depth analysis and accurate prediction of the operation status related data uploaded by the energy storage edge, and combine the correlation analysis logic adapted to the system type to analyze and obtain the overall impact assessment and risk level of each equipment component failure.Because the energy storage edge is deployed close to the equipment components, it can capture basic operational data in real time and perform rapid feature analysis. This allows for rapid early warning of fault conditions in each equipment component without relying on remote communication transmission, effectively ensuring timely fault response. Meanwhile, the energy storage cloud possesses powerful computing capabilities and massive data processing capabilities. It can utilize the complete operational status parameters and fault assessment feature data uploaded from the energy storage edge, combined with system type, to perform targeted complex correlation analysis and model prediction. This enables accurate detection of fault conditions and impact risks for each equipment component. Finally, the final fault diagnosis of the energy storage system is obtained by comparing the results from both the energy storage edge and the energy storage cloud. The detection results demonstrate that the system achieves automated, end-to-end fault detection and risk assessment of energy storage systems through collaborative efforts between edge and cloud-based energy storage, rather than relying solely on the limited computing power of edge devices for simple detection or manual inspection by operators. Therefore, it overcomes the technical shortcomings of insufficient edge device computing power to support complex diagnostics, delayed cloud response due to communication latency, and limited manual identification capabilities and efficiency that cannot match the needs of large-scale deployment of energy storage systems. These shortcomings often lead to detection delays, errors, misjudgments of fault risks, or excessively long detection times. Thus, the system improves the effectiveness of fault detection in energy storage systems.

[0061] In one embodiment, the fault assessment feature data includes fault assessment frequency data; based on target fault event data, fault assessments are performed on multiple device components within a preset time window to obtain fault assessment feature data, including:

[0062] For any target device component in the energy storage system, identify the target fault type corresponding to the target fault event data and extract the fault occurrence time of the target device component; based on the fault occurrence time, update the cumulative total number of faults corresponding to the target fault type to obtain the updated cumulative total number of faults; based on the fault occurrence time and the updated cumulative total number of faults, count the occurrence frequency of the target fault type within a preset time window to obtain fault assessment frequency data.

[0063] It should be noted that, in order to accurately assess the occurrence of faults in multiple equipment components within a preset time window from the perspective of fault frequency quantification, the frequency of different fault types of the target equipment components can be assessed by sequentially executing the steps of fault type classification and identification, fault timestamp matching and statistics, and frequency calculation and quantification. The fault assessment frequency data represents the frequency of fault occurrence of multiple equipment components within the preset time window, which can be 0.25 times / hour or 3 times / day, etc. Since the target fault event data is stored in key-value pairs at the energy storage edge, the corresponding fault occurrence time can be extracted for any target equipment component in the energy storage system, and then the relevant parameters can be updated to obtain the fault assessment frequency data.

[0064] As an example, for any target device component in an energy storage system, the corresponding target fault type is queried using the target fault event data as an index, and the fault occurrence time of the target device component is extracted from the target fault event data; the cumulative total number of faults corresponding to the target fault type is updated based on the fault occurrence time to obtain the updated cumulative total number of faults; based on the fault occurrence time and the updated cumulative total number of faults, the occurrence frequency of the target fault type within a preset time window is counted to obtain fault assessment frequency data.

[0065] In one feasible approach, at the energy storage edge, for each device component and its corresponding fault type in a centralized energy storage system, a fault frequency statistical analysis based on a preset time window is performed to quantify the intensity of different faults at different time scales. Specifically, for any target device component, the energy storage edge maintains an independent fault statistics record for each fault type during the caching phase. The fault statistics record is stored using a key-value structure, where the key is the fault type (fault_type), and the value includes at least the cumulative occurrence count (total_count). The occurrence count within the preset time window can be represented as last_window_count, and the fault occurrence timestamp sequence can be represented as timestamps. In the implementation process, the energy storage edge configures at least one preset time window for each fault type, which can specifically be a sliding window. Alternatively, a rolling window can be used, with different time windows corresponding to different time scales, such as hourly, daily, and weekly scales. When a fault event is detected in a target device component at the energy storage edge, the following frequency statistics steps are performed based on the target fault event data: 1) Timestamp recording: The occurrence time of the fault event is recorded in the timestamp sequence timestamps corresponding to the fault_type; 2) Cumulative count update: The cumulative occurrence count total_count corresponding to the fault_type is incremented by one; 3) Count calculation within the window: Based on the current time and the start and end times of the preset time window, the number of timestamps falling within the preset time window is filtered from timestamps, and the filtered count is used as the last_window_count of the occurrence count of the fault_type within the preset time window.

[0066] This embodiment constructs a sophisticated fault frequency statistics mechanism at the energy storage edge. Relying on the key-value storage structure of local caching and independent fault statistics records at the energy storage edge, it avoids the bandwidth consumption caused by frequent uploading of fault event data to the energy storage cloud, thereby reducing the cloud computing load and improving the response speed of fault frequency statistics. At the same time, by setting a frequency statistics process of real-time recording of timestamp sequences, dynamic updating of cumulative counts, and precise filtering of counts within a window, the accuracy and real-time nature of fault frequency data can be guaranteed, providing a reliable quantitative basis for the generation of fault assessment feature data. Therefore, it further improves the effect of fault detection in energy storage systems.

[0067] In one embodiment, the fault assessment feature data includes fault assessment probability data; based on the target fault event data, fault assessments are performed on multiple device components within a preset time window to obtain fault assessment feature data, including:

[0068] Obtain the statistical characteristics of the number of failures, the intensity of use, and the frequency of use of the target equipment components during the historical operating cycle; standardize the failure assessment frequency data based on the statistical characteristics of the number of failures to obtain standard failure assessment frequency data; and normalize and map the standard failure assessment frequency data based on the intensity of use and the frequency of use to obtain failure assessment probability data.

[0069] It should be noted that, in order to accurately assess the failure occurrence of multiple equipment components within a preset time window from the perspective of failure probability quantification, the following steps can be performed sequentially: feature value statistics, standardization processing of failure assessment frequency data, and normalization processing of failure assessment frequency standard data. This allows for the assessment of the probability of different failure types of the target equipment component. The failure assessment probability data represents the probability of failure of multiple equipment components within the preset time window, specifically 0.5 or 0.6, etc. The failure assessment frequency data can be represented as count_i, representing the number of times a certain target equipment component i has a corresponding failure type within the preset time window, i.e., last_window_count. The failure count feature value can be represented as u_i, n_i, where u_i refers to the target equipment component i. The mean number of occurrences of the corresponding fault type for component i within the historical operating cycle, and n_i refers to the standard deviation of the number of occurrences of the corresponding fault type for target equipment component i within the historical operating cycle; the usage intensity characteristic value can be represented as U_i, which is the usage intensity index of the target equipment component, used to characterize the load level or workload of the target equipment component within a preset time window; the usage frequency characteristic value can be represented as F_i, which is the usage frequency index of the target equipment component, used to characterize the number of start-stops, schedulings, or work cycles of the target equipment component within a preset time window; it can be understood that the usage intensity characteristic value and the usage frequency characteristic value can be calculated based on the basic operating data of the target equipment component, which at least includes power, current, operating duration, and control command trigger records.

[0070] As an example, based on the basic operational data of the target equipment components, statistical characteristic values ​​of the number of failures, usage intensity, and usage frequency of the target equipment components within their historical operating cycles are collected and calculated. The failure assessment frequency data is then standardized using a preset standardization formula to obtain standard failure assessment frequency data. The preset standardization formula can be as follows:

[0071]

[0072] in, For fault assessment frequency standard data, count_i represents the fault assessment frequency data, u_i refers to the mean number of occurrences of the corresponding fault type of target equipment component i within the historical operating cycle, and n_i refers to the standard deviation of the number of occurrences of the corresponding fault type of target equipment component i within the historical operating cycle. Based on the usage intensity feature value and usage frequency feature value, a preset normalization mapping processing formula is used to normalize and map the fault assessment frequency standard data to obtain fault assessment probability data. The specific preset normalization mapping processing formula is as follows:

[0073]

[0074] in, This is fault assessment probability data, where U_i is the usage intensity feature value and F_i is the usage frequency feature value. This is the first weighting factor, used to balance the impact of usage intensity on failure frequency. This is the second weighting factor, used to balance the impact of usage frequency on failure frequency. To prevent the stabilization factor from having a denominator of zero, its value can be a preset positive number. It is understandable that by introducing usage intensity and frequency corrections, fault events occurring under high load or high frequency operating conditions are not excessively amplified, thereby improving the accuracy of fault assessment. Thus, by introducing statistical characteristic values ​​of the number of faults, usage intensity, and usage frequency within the historical operating cycle of the target equipment component as correction criteria, and by standardizing the fault assessment frequency data based on the statistical characteristic values ​​of the number of faults, the fault frequency distributions of different equipment components can be made comparable. Furthermore, by combining the usage intensity and usage frequency characteristic values ​​of the actual operating conditions of the equipment, a normalized mapping correction is completed for the standardized frequency data, offsetting the frequency deviation caused by differences in equipment service life, load, and operating environment. This achieves the goal of transforming discrete fault frequency data into a representation of the probability of fault occurrence of the target equipment component under the current operating state, thus further improving the effectiveness of fault detection in energy storage systems.

[0075] In one embodiment, the fault assessment feature data includes fault assessment level data; based on the target fault event data, fault assessments are performed on multiple device components within a preset time window to obtain fault assessment feature data, including:

[0076] Based on the target fault type, the corresponding target fault assessment level is queried in the preset mapping table. The preset mapping table stores the mapping relationship between the fault type and the fault assessment level of the target equipment component. Based on the target fault assessment level, the target equipment component fault level record table is updated to obtain the fault assessment level data.

[0077] It should be noted that, in order to accurately assess the failure occurrence of multiple equipment components within a preset time window from the perspective of failure probability quantification, a mapping relationship between the failure type and assessment level of the equipment component can be established to achieve the assessment of the failure level of different failure types of the target equipment component. For example, in one feasible approach, during the assessment of the failure level of the equipment component, the failure level information corresponding to the failure type stored according to KV can be queried by querying a preset mapping table. The preset mapping table can be constructed based on a rule table or an experience base.

[0078] As an example, the target fault type is used as an index to query the corresponding target fault assessment level in a pre-defined mapping table. Based on the target fault assessment level, the historical fault assessment level data in the target equipment component fault level record table is updated to obtain the fault assessment level data. In this way, by querying the pre-defined mapping table using the target fault type as an index, a standardized target fault assessment level can be quickly matched, avoiding the subjectivity and error of manual judgment, thereby improving the efficiency and consistency of fault level determination. Simultaneously, dynamically updating the historical fault assessment level data based on the target fault assessment level can form a time-series record of fault levels throughout the entire lifecycle of the target equipment component, completely preserving the evolution trajectory of fault levels. This provides a basis for subsequent analysis of the correlation between fault levels and equipment operating status, thus further improving the effectiveness of fault detection in energy storage systems.

[0079] In one embodiment, such as Figure 2As shown, a fault detection method for an energy storage system is provided. This embodiment uses the application of this method to an energy storage cloud as an example. The energy storage cloud refers to a cloud platform based on cloud computing technology for remote monitoring, data management, analysis, decision-making, and intelligent operation and maintenance of the energy storage system. The energy storage cloud includes, but is not limited to, personal computers and laptops. The energy storage cloud deploys a faulty energy storage system for the energy storage system. The faulty energy storage system includes a second acquisition module, a fault analysis module, a prediction module, and a generation module. The second acquisition module is used to acquire target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge. The fault assessment feature data is obtained by the energy storage edge based on target fault event data extracted from a preset fault event time-series database, which stores fault event data recorded over time. The fault event data is obtained by fault identification of basic operating data generated by multiple equipment components during operation. The fault analysis module performs fault correlation analysis on the fault assessment feature data according to the system type of the energy storage system, obtaining the fault correlation analysis results, which include the fault feature weights corresponding to the fault assessment feature data. The prediction module inputs the target basic operating data and fault assessment feature data into a preset fault risk prediction model, predicting the impact risk of multiple equipment component failures on the energy storage system, obtaining the fault risk prediction results for each equipment component. The generation module generates cloud-based fault detection results for the energy storage system based on all fault risk prediction results and fault feature weights. Through information interaction between the second acquisition module, fault analysis module, prediction module, and generation module, the goal of predicting the impact risk of equipment component failures on the energy storage system and obtaining cloud-based fault detection results for the energy storage system can be achieved in the energy storage cloud. In other words, it achieves the goal of automating and conducting full-process fault detection and risk assessment of energy storage systems through collaborative energy storage edge devices and energy storage cloud, rather than relying solely on the limited computing power of edge devices for simple detection or manual inspection by operators. Therefore, it overcomes the technical shortcomings of insufficient computing power of edge devices to support complex diagnostics, delayed response of the cloud due to communication latency, and limited manual identification capabilities and detection efficiency that cannot match the needs of large-scale deployment of energy storage systems, which can lead to detection delays, errors, misjudgments of fault risks, or excessively long detection times. Thus, it improves the effectiveness of fault detection in energy storage systems. In this embodiment, the method includes the following steps:

[0080] Step 302: Obtain the target basic operation data and fault assessment feature data within the preset time window sent by the energy storage edge. The fault assessment feature data is obtained by the energy storage edge based on the target fault event data extracted from the preset fault event time series database to perform fault assessment on multiple equipment components. The preset fault event time series database stores fault event data recorded over time. The fault event data is obtained by fault identification of the basic operation data generated by multiple equipment components during operation.

[0081] It should be noted that the steps of obtaining fault event data by identifying faults in the basic operating data generated by multiple equipment components during operation, and obtaining fault assessment feature data by assessing faults in multiple equipment components based on the fault event data, can refer to the above embodiment, and will not be repeated here.

[0082] As an example, step 302 includes acquiring target basic operating data within a preset time window transmitted by the energy storage edge, and acquiring fault assessment feature data transmitted by the energy storage edge.

[0083] In one feasible approach, the target basic operational data uploaded from the energy storage edge to the energy storage cloud characterizes the operating status of equipment components within a corresponding preset time window. Specifically, this may include, but is not limited to, time-series data such as current, voltage, temperature, active power, reactive power, and liquid cooling flow rate. The target fault event data uploaded from the energy storage edge to the energy storage cloud can be encapsulated using a key-value structure. For each target equipment component and its corresponding fault type, it includes at least the following three types of fault characteristic indicators: 1) Fault occurrence frequency data: used to characterize the scale of fault occurrence within the preset time window; 2) Fault occurrence probability data: used to characterize the relative degree of abnormality after correction for usage intensity and usage frequency; 3) Fault level data: used to characterize the impact level of the corresponding fault type on equipment safety and system reliability.

[0084] Step 304: Based on the system type of the energy storage system, perform fault correlation analysis on the fault assessment feature data to obtain the fault correlation analysis results, wherein the fault correlation analysis results include the fault feature weights corresponding to the fault assessment feature data.

[0085] It should be noted that the system type of the energy storage system refers to the system category classified by core attributes such as the application scenario, energy storage medium type, and grid-connected operation mode; the fault correlation analysis results are obtained by multi-dimensional analysis of the correlation between fault assessment feature data and overall fault risk, combined with the specific type of the energy storage system; the fault feature weight refers to the quantitative coefficient of the impact of a single fault assessment feature data on the overall fault risk of the energy storage system. For example, the fault frequency weight of the energy storage converter is 0.35, the single-cell voltage imbalance weight of the battery management system is 0.28, the abnormal liquid cooling speed weight of the temperature control system is 0.15, the communication interruption weight of the smart sensor is 0.08, and the current fluctuation weight of the combiner cabinet is 0.14.

[0086] Before performing fault correlation analysis on fault assessment characteristic data based on the system type of the energy storage system to obtain the fault correlation analysis results, the method also includes:

[0087] Based on the device metadata of the energy storage system, the system type of the energy storage system is detected. The device metadata includes at least one of the following: device attribute data of multiple device components and data on the connection relationships between devices.

[0088] It should be noted that the energy storage cloud can perform system type detection based on the device metadata of the energy storage system. Device metadata refers to structured data describing the basic attributes, functional characteristics, and topological relationships of each device component in the energy storage system. It is the core basis for the energy storage cloud to identify the system type. Device metadata includes internal attribute data of multiple device components and inter-device connection relationship data. Internal attribute data refers to basic data characterizing the inherent characteristics of a single device component, specifically including device identification, device model, energy storage medium type, rated power, rated voltage, device function positioning, and manufacturer. Inter-device connection relationship data refers to data describing the network topology and interaction relationships of each device component at the physical and logical levels. Specifically, it includes communication link type, signal transmission direction, control command flow, and physical wiring topology between devices. Examples include bidirectional data communication connection between battery clusters and battery management systems, control command transmission connection between BMS and energy storage converters, grid connection between PCS and the public power grid, and energy input connection between photovoltaic arrays and energy storage batteries. In short, the device metadata of the energy storage system can automatically identify the architecture type of the energy storage system to distinguish between different types of energy storage systems.

[0089] As an example, the system type of an energy storage system can be queried using its device metadata as an index. By using standardized device metadata as a unified index to query the system type, automated and unbiased accurate determination of system type can be achieved, avoiding the subjective errors and experience reliance of manual classification. Simultaneously, relying on dual-dimensional verification of device attribute data and inter-device connection relationship data significantly improves the accuracy and reliability of system type identification, effectively avoiding deviations in subsequent fault correlation analysis and fault feature weight allocation due to misjudgment of system type. Furthermore, the above index query method has good compatibility and scalability, adapting to metadata formats of energy storage devices from different manufacturers and with different specifications. It eliminates the need to develop separate identification logic for specific system types, significantly reducing the operation and maintenance costs and development complexity of cloud systems. Therefore, it further lays the foundation for improving the effectiveness of fault detection in energy storage systems.

[0090] Step 306: Input the target basic operation data and fault assessment feature data into the preset fault risk prediction model. The preset fault risk prediction model is used to predict the impact risk on the energy storage system when multiple equipment components fail, and obtain the fault risk prediction results for each of the multiple equipment components.

[0091] It should be noted that the preset fault risk prediction model is used to predict the impact risk on the energy storage system when multiple equipment components fail. Specifically, the preset fault risk prediction model can be a deep neural network model or a graph neural network model; the fault risk prediction result can be a fault risk score.

[0092] As an example, step 306 includes: inputting the target basic operating data and fault assessment feature data into a preset fault risk prediction model, predicting the impact risk on the energy storage system when multiple equipment components fail through the preset fault risk prediction model, and obtaining the fault risk prediction results corresponding to each of the multiple equipment components.

[0093] In one feasible approach, after system architecture identification, fault correlation analysis, and fault feature weight allocation are completed based on system type, the energy storage cloud will construct a pre-defined fault risk prediction model. Specifically, the pre-defined fault risk prediction model can be constructed based on the following two routes: 1) Model input feature construction: The energy storage cloud will construct multi-dimensional input features based on the data uploaded from the energy storage edge and the cloud analysis results for deep learning model training and inference. The input features include at least one or more of the following: multi-channel time-series operation features, used to characterize the dynamic behavior of equipment components in the time dimension, including at least voltage, current, temperature, active power, reactive power, and liquid cooling flow rate; fault statistical features, including the number of fault occurrences, normalized fault frequency, and fault severity, used to reflect the degree of abnormality of the fault in the statistical dimension; operating condition and environmental features, including load level, operating mode, ambient temperature, and altitude. The system includes features such as: equipment model, commissioning years, capacity, and maintenance strategy parameters; similar group features, including similar equipment or similar fault cluster features obtained from the aforementioned horizontal comparative analysis or cluster analysis; and then, by fusing time-series features, statistical features, and equipment profile features, a multimodal input feature space that can comprehensively characterize the equipment's operating status is constructed. 2) Model structure and dependency modeling: In one embodiment, the cloud selects a corresponding family of deep learning models to construct based on the type and structure of the input features: For multi-channel time-series operating features, recurrent neural networks, time-series convolutional networks, or attention-based time-series models are used to learn the time dependencies of the equipment's operating status; For dependencies across devices and clusters in a centralized energy storage system, a graph structure is introduced to represent the topological relationships of the devices, and a graph neural network is used to model the association features between different devices.

[0094] For anomaly pattern recognition, an autoencoder or density-based anomaly detection model is introduced to identify abnormal behaviors that deviate from normal operating modes. By combining different model structures, joint modeling of time dependence, structural dependence, and anomaly features is achieved. Then, a pre-set fault risk prediction model outputs fault risk prediction results. The model output and the output of the risk assessment deep learning model include at least one or more of the following results: multi-label fault type prediction results, used to output the set of possible fault types; fault location results, used to determine the location of the equipment component, equipment cluster, or rack where the fault occurs; and confidence scores or severity scores for each predicted fault type.

[0095] Step 308: Generate cloud-based fault detection results for the energy storage system based on all fault risk prediction results and fault feature weights.

[0096] As an example, step 308 includes: calculating the cloud-based fault detection result of the energy storage system by inputting all fault risk prediction results and fault feature weights into a preset fault detection result formula, wherein the preset fault detection result formula may be as follows:

[0097]

[0098] in, This represents the prediction score or severity score (fault risk prediction result) of the preset fault risk prediction model for fault type i. This indicates the overall risk level of the current energy storage system operation (cloud-based fault detection results). This represents the weight of the fault characteristics.

[0099] The aforementioned fault detection method and energy storage system are applied to an energy storage cloud platform. The energy storage system consists of multiple device components. The energy storage cloud platform first acquires target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge. The fault assessment feature data is obtained by the energy storage edge platform based on target fault event data extracted from a preset fault event time-series database, which stores fault event data recorded over time. The fault event data is obtained by fault identification of the basic operating data generated by the multiple device components during operation. Then, according to the system type of the energy storage system, fault correlation analysis is performed on the fault assessment feature data to obtain the fault correlation analysis results. The fault correlation analysis results include the fault feature weights corresponding to the fault assessment feature data.

[0100] The system inputs target basic operational data and fault assessment feature data into a preset fault risk prediction model. This model then predicts the impact risk on the energy storage system when multiple equipment components fail, yielding fault risk prediction results for each component. Finally, based on all fault risk prediction results and fault feature weights, a cloud-based fault detection result for the energy storage system is generated. Because the energy storage edge is deployed close to the equipment components, it can capture basic operational data in real time and perform rapid feature analysis, enabling rapid early warning of fault conditions for each component without relying on remote communication transmission. This effectively ensures timely fault response. Furthermore, the energy storage cloud possesses powerful computing capabilities and massive data processing capabilities, enabling it to utilize complete operational status parameters and fault assessment feature data uploaded from the energy storage edge, combined with the system type, to perform targeted and complex fault detection. Correlation analysis and model prediction are used to accurately detect the fault conditions and impact risks of each equipment component. Finally, the final fault detection result of the energy storage system is obtained by comparing and verifying the results from both the energy storage edge and the energy storage cloud. In other words, it achieves the goal of automating and conducting full-process fault detection and risk assessment of the energy storage system in collaboration between the energy storage edge and the energy storage cloud, rather than relying solely on the limited computing power of edge devices for simple detection or manual inspection by operators. Therefore, it overcomes the technical defects of insufficient computing power of edge devices to support complex diagnosis, delayed response of the cloud due to communication latency, and limited manual identification capabilities and detection efficiency that cannot meet the needs of large-scale deployment of energy storage systems, which can lead to detection delays, detection errors, misjudgments of fault risks, or excessively long detection times. Thus, it improves the effectiveness of fault detection in energy storage systems.

[0101] In one embodiment, detecting the system type of the energy storage system based on device metadata of multiple device components includes:

[0102] Based on the device metadata, generate the type characteristic determination conditions for the energy storage system; if the energy storage system meets the type characteristic determination conditions, determine the system type of the energy storage system.

[0103] It should be noted that the device metadata can be integrated to obtain type feature determination conditions. Through the type feature determination conditions, the type characteristics of the energy storage system can be clearly analyzed. Then, based on the matching relationship between the energy storage system and the type feature determination conditions, the system type of the energy storage system can be determined. Specifically, the system type can include centralized or distributed.

[0104] As an example, based on device metadata, type characteristic determination conditions for energy storage systems are generated; if the energy storage system is found to meet the type characteristic determination conditions, the system type of the energy storage system is determined.

[0105] In one embodiment, the type characteristic determination conditions of the energy storage system are generated based on the device metadata, including:

[0106] Obtain topology power correlation data between multiple device components. The topology power correlation data includes at least one of the following: power aggregation node distribution data, electrical connection level data, number of power aggregation devices connected data, and topology aggregation density data. Generate type feature determination conditions based on the topology power correlation data and device metadata.

[0107] It is understandable that, in the process of determining the system type, energy storage systems can also combine multiple conditions for architecture identification. Specifically, they can also combine topology power correlation data between multiple device components. Among them, topology power correlation data includes at least one of the following: power aggregation node distribution data, electrical connection level data, power aggregation device connection quantity data, and topology aggregation density data.

[0108] As an example, topology power correlation data between multiple device components is obtained, wherein the topology power correlation data includes at least one of power aggregation node distribution data, electrical connection level data, power aggregation device connection quantity data, and topology aggregation density data; the topology power correlation data and device metadata are integrated into type feature determination data, and the type feature determination data is encapsulated into type feature determination conditions.

[0109] In one embodiment, the type feature determination condition includes a centralized feature determination condition, which includes at least one of the following:

[0110] Multiple device components aggregate power through the same power aggregation device; multiple device components are located at the same electrical grid-connected node; multiple device components share preset common conditions; the topology among multiple device components is a preset aggregation structure.

[0111] It is understandable that, for centralized features, one or more of the above centralized feature determination conditions can be used for determination.

[0112] In one embodiment, the type feature determination condition includes a distributed feature determination condition, which includes at least one of the following:

[0113] Each device component is independently configured with a corresponding power aggregation device, and multiple device components are connected to different electrical grid-connected nodes; there is no centralized power aggregation relationship between multiple device components; the operating conditions of multiple device components are not consistent; the topology of multiple device components is a preset independent structure.

[0114] It is understandable that, for distributed features, one or more of the above distributed feature determination conditions can be used for determination.

[0115] In one embodiment, based on the system type, a fault correlation analysis is performed on the fault assessment feature data to obtain the fault correlation analysis results, including:

[0116] In the case of a centralized system, a corresponding fault assessment feature vector is constructed based on the fault assessment feature data; a fault correlation matrix of the energy storage system is generated based on the correlation degree of different fault types in the fault assessment feature vector; and the fault correlation matrix is ​​quantitatively analyzed to obtain the fault correlation analysis results.

[0117] It should be noted that different fault correlation analysis methods are used to detect faults in energy storage systems of different types. Specifically, after identifying the architecture of the energy storage system, for sites identified as centralized energy storage systems, the energy storage cloud can perform correlation analysis and dimensionality reduction on fault data from multiple devices at the same site, in the same operating environment, and under similar operating conditions to identify system-level fault modes and key risk factors. For example, in one feasible approach, for centralized energy storage sites, the cloud uses equipment components as the analysis object and constructs a fault feature vector based on fault statistics uploaded from the edge side. The fault feature vector includes at least: the normalized fault frequency (Norm) of different fault types over a long time scale. Freq; the severity level corresponding to the fault type; optionally, including the count of fault occurrences or its temporal distribution characteristics; the long time scale can be daily, weekly, or monthly, used to reflect the stable correlation characteristics of faults within the system's operating cycle; then, a fault correlation matrix is ​​calculated: in centralized energy storage systems, because multiple equipment components are under the same environmental conditions and similar operating conditions, their fault behaviors often show strong correlations in both the time and type dimensions; therefore, the cloud calculates the correlation between different fault types over a long time scale based on the above fault characteristics, constructing a fault correlation matrix R; the element R_ij of the correlation matrix R represents the degree of correlation between fault type i and fault type j.

[0118] As an example, in the case of a centralized system, a corresponding fault assessment feature vector is constructed based on the fault assessment feature data; based on the correlation between different fault types in the fault assessment feature vector, a fault correlation matrix of the energy storage system is generated, wherein the fault correlation matrix is ​​generated based on the following formula:

[0119]

[0120] in, This is the fault correlation matrix. , The fault characteristic sequences of fault types i and j over a long time scale are represented respectively. Furthermore, by constructing a correlation matrix R, the linkage relationships of different fault types in centralized energy storage stations can be quantified, providing a foundation for system-level risk assessment. The fault correlation matrix is ​​then quantitatively analyzed to obtain the fault correlation analysis results. Thus, considering the characteristics of centralized energy storage systems—dense equipment and the ease with which faults can cause cascading propagation—the construction of standardized fault assessment feature vectors enables the structured integration and quantitative representation of multiple fault types, solving the problem of difficult collaborative analysis of dispersed fault data. The fault correlation matrix generated based on the fault characteristic sequences can accurately quantify the linkage strength and propagation probability between different fault types, further laying the foundation for improving the effectiveness of fault detection in energy storage systems.

[0121] In one embodiment, a corresponding fault assessment feature vector is constructed based on the fault assessment feature data, including:

[0122] The fault assessment feature data is dimensionality reduced to obtain a low-dimensional fault feature representation; the low-dimensional fault feature representation is then clustered to obtain a fault assessment feature vector.

[0123] It should be noted that, in order to reduce the redundancy and computational complexity caused by high-dimensional fault features, the energy storage cloud can also perform dimensionality reduction processing on fault features or correlation matrices.

[0124] As an example, the fault assessment feature data is dimensionality reduced to obtain a low-dimensional fault feature representation; the low-dimensional fault feature representation is then clustered to obtain a fault assessment feature vector.

[0125] In one feasible approach, the energy storage system can employ principal component analysis or manifold learning methods to reduce the dimensionality of fault features, extracting a low-dimensional feature space representing the system's main fault modes. Furthermore, methods such as unified manifold approximation and projection can be used to preserve the local structural relationships between fault features. After obtaining the low-dimensional fault feature representation, clustering analysis is performed in the cloud based on these features. The clustering method includes at least hierarchical clustering or density-based clustering. Through clustering analysis, fault types similar in temporal behavior and statistical characteristics are grouped into the same fault cluster, and representative fault types are identified within each cluster to characterize the main risk features of that cluster, thus obtaining a fault assessment feature vector.

[0126] In one embodiment, based on the system type, a fault correlation analysis is performed on the fault assessment feature data to obtain the fault correlation analysis results, including:

[0127] In the case of a distributed system, target fault assessment feature data with consistent adaptability is selected from the fault assessment feature data based on preset matching conditions, and a corresponding fault assessment feature vector is constructed based on the target fault assessment feature data; a fault correlation matrix of the energy storage system is generated based on the correlation degree of different fault types in the fault assessment feature vector; the fault correlation matrix is ​​quantitatively analyzed to obtain the fault correlation analysis results.

[0128] It should be noted that if the energy storage system is a distributed system, then it is necessary to construct preset matching conditions based on information such as the difference in commissioning time, model and scale, and select target fault assessment feature data with consistent adaptability. Other processing procedures are the same as those for centralized processing procedures, and will not be elaborated here.

[0129] As an example, in the case of a distributed system, based on preset matching conditions, target fault assessment feature data with consistent adaptability is selected from the fault assessment feature data, and a corresponding fault assessment feature vector is constructed based on the target fault assessment feature data. A fault correlation matrix for the energy storage system is generated based on the correlation degree of different fault types in the fault assessment feature vector. The fault correlation matrix is ​​then quantitatively analyzed to obtain the fault correlation analysis results. In this way, considering the characteristics of distributed energy storage systems—distributed deployment, independent operation on the user side, and limited fault impact range—selecting target fault assessment feature data with consistent adaptability through preset matching conditions can accurately eliminate redundant features irrelevant to the distributed scenario, effectively improving the data targeting and effectiveness of fault analysis. Constructing fault assessment feature vectors and generating correlation matrices based on the selected feature data can focus on the fault correlation relationships of equipment components within a single user site, rather than large-scale cascading fault propagation across sites. Quantitative analysis of the correlation matrix can output fault feature weights adapted to the distributed scenario and uncover user-side-specific fault occurrence patterns, providing a precise basis for the formulation of personalized operation and maintenance strategies for distributed energy storage systems.

[0130] In one embodiment, the fault feature weights are obtained by fusing prior weights, correlation weights, and statistical significance weights.

[0131] Specifically, the prior weights are preset for different fault types. Prior weights are determined based on industry safety standards, equipment manufacturer technical documents, or maintenance experience, and are used to reflect the inherent risk level of the fault type; correlation weights are assigned to fault type i based on the correlation matrix R. The correlation weight is used to reflect the importance of the fault in system-level fault linkage, and its calculation method is as follows:

[0132]

[0133] Here, the function g(·) is a monotonically increasing function used to map the correlation strength to weight values; the statistical significance weight is assigned to fault type i based on the statistical results of fault occurrence. The statistical significance weight is related to the statistical significance of the fault occurrence, and its calculation method satisfies: ,in, This represents the statistical significance level of fault type i; the higher the significance, the greater the corresponding weight. Finally, the energy storage cloud will integrate the above-mentioned prior weights, correlation weights, and statistical significance weights to calculate the comprehensive weight of fault type i. , , where γ, δ, and ζ are coefficients used to adjust the contribution of each weight factor, and these coefficients can be adjusted through historical data analysis or operation and maintenance feedback; normalize(·) represents the normalization operation, which is used to ensure that each fault weight is within a uniform dimension range.

[0134] In one feasible approach, refer to Figure 3 , Figure 3 This diagram illustrates the process of fault detection in an energy storage system through interaction between the energy storage edge and the energy storage cloud. First, at the energy storage edge, basic operational data generated by multiple device components during operation is acquired, and fault identification is performed on this data to obtain fault event data. This fault event data represents the fault type of multiple device components. Next, target fault event data within a preset time window is extracted from a preset fault event time-series database, which stores fault event data recorded over time. Based on the target fault event data, fault assessment is performed on multiple device components within the preset time window to obtain fault assessment feature data. This fault assessment feature data represents the fault type of multiple device components. The system detects the occurrence of faults in components within a preset time window; it sends target basic operating data and fault assessment feature data within the preset time window to the energy storage cloud, which then inputs the target basic operating data and fault assessment feature data into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when multiple equipment components fail, obtaining the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, the system generates cloud-based fault detection results for the energy storage system. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system.

[0135] Furthermore, to improve the stability and interpretability of alarm determination, in addition to making alarm decisions based on the single Risk indicator, the cloud can also fuse Risk with the three fault factors calculated by the edge. The fusion process includes at least the following steps: using Risk as a system-level risk indicator; using the number of fault occurrences, normalized fault frequency, and fault severity as local fault feature indicators; and combining the fault weight W_i to perform weighted fusion of the system-level indicator and the local indicator to obtain the final risk score Risk_final.

[0136] In one embodiment, Risk_final is calculated according to the following formula:

[0137]

[0138] Among them, λ, Here, is the weighting coefficient, and f(·) is the function that maps the three factors to risk scores. Furthermore, the cloud platform classifies the current system operation status into different alarm levels based on the value of Risk_final. The alarm levels include at least the following: Low risk level: Risk_final is within the first preset threshold range, indicating that the system is operating basically normally; Medium risk level: Risk_final is within the second preset threshold range, indicating that the system has potential anomalies and requires attention; High risk level: Risk_final... Exceeding the third preset threshold indicates a high security risk in the system, requiring intervention. The threshold range can be set based on historical operational data statistics, equipment safety specifications, or operational experience, and supports dynamic adjustment. After determining the alarm level, the cloud sends corresponding alarm information and operational decision suggestions to the edge. The sent content includes at least one or more of the following: alarm information, including fault type, fault location, alarm level, and risk value; operational suggestions, including inspection suggestions, maintenance priorities, or operational adjustment suggestions; optionally, updates to operational thresholds or policy parameters are sent to the edge for adjusting alarm thresholds or operational control policies. Policy parameter updates are sent using a canary release mechanism and support rollback operations in abnormal situations to ensure the security and stability of system operation. In addition, alarm and maintenance closed-loop mechanisms are also implemented.

[0139] In one feasible approach, refer to Figure 3 , Figure 3The control flowchart for fault detection in energy storage systems through collaborative energy storage edge and energy storage cloud is as follows: First, data acquisition and preprocessing are performed at the energy storage edge. Specifically, sensors deployed at the energy storage edge collect basic operational data generated by multiple device components during operation in real time, as well as processed fault event data. This fault event data is then locally cached, and a pre-defined fault event time-series database is constructed using a circular queue. Next, edge preprocessing is performed on the target fault event data, which is then used to generate the target fault event data. The energy storage edge then sends the target fault event data to the energy storage cloud for storage and intelligent analysis. Specifically, the energy storage cloud receives target basic operation data and fault assessment feature data within a preset time window, and then completes the system type determination of the energy storage system to distinguish whether the energy storage system is a centralized energy storage system or a distributed energy storage system. Then, it adopts different methods to conduct fault correlation analysis and assign fault feature weights. Finally, it uses a deep learning model to complete the reasoning and generate cloud fault detection results. In addition, the edge side and application side receive cloud fault detection results and participate in decision-making and action. Different action modes are adopted under different risk levels. Finally, the operation and maintenance feedback closed loop results are sent back to the energy storage cloud.

[0140] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0141] Based on the same inventive concept, this application also provides a fault detection device for an energy storage system to implement the fault detection method for the energy storage system described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the fault detection device for an energy storage system provided below can be found in the limitations of the fault detection method for the energy storage system described above, and will not be repeated here.

[0142] In one exemplary embodiment, such as Figure 4As shown, a fault detection device for an energy storage system is provided, which is applied at the edge of the energy storage system. The energy storage system is composed of multiple equipment components. The fault detection device for the energy storage system includes: a first acquisition module 401, an extraction module 402, a fault assessment module 403, and a transmission module 404, wherein;

[0143] The first acquisition module 401 is used to acquire basic operating data generated by multiple device components during operation, and to identify faults in the basic operating data to obtain fault event data, wherein the fault event data represents the fault type of multiple device components.

[0144] Extraction module 402 is used to extract target fault event data within a preset time window from a preset fault event time sequence database, wherein the preset fault event time sequence database stores fault event data recorded over time.

[0145] The fault assessment module 403 is used to assess the faults of multiple equipment components within a preset time window based on the target fault event data, and obtain fault assessment feature data, wherein the fault assessment feature data characterizes the fault occurrence of multiple equipment components within the preset time window.

[0146] The sending module 404 is used to send target basic operating data and fault assessment feature data within a preset time window to the energy storage cloud. The energy storage cloud then inputs the target basic operating data and fault assessment feature data into a preset fault risk prediction model. The preset fault risk prediction model predicts the impact risk on the energy storage system when multiple equipment components fail, and obtains the fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, the cloud-based fault detection results of the energy storage system are generated. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data according to the system type of the energy storage system.

[0147] In one exemplary embodiment, such as Figure 5 As shown, a fault detection device for an energy storage system is provided, applied in an energy storage cloud. The energy storage system consists of multiple equipment components. The fault detection device includes: a second acquisition module 501, a fault analysis module 502, a prediction module 503, and a generation module 504, wherein:

[0148] The second acquisition module 501 is used to acquire target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge. The fault assessment feature data is obtained by the energy storage edge based on target fault event data extracted from a preset fault event time series database to perform fault assessment on multiple equipment components. The preset fault event time series database stores fault event data recorded over time. The fault event data is obtained by fault identification of the basic operating data generated by multiple equipment components during operation.

[0149] The fault analysis module 502 is used to perform fault correlation analysis on fault assessment feature data according to the system type of the energy storage system, and obtain fault correlation analysis results, wherein the fault correlation analysis results include fault feature weights corresponding to the fault assessment feature data.

[0150] Prediction module 503 is used to input the target basic operation data and fault assessment feature data into the preset fault risk prediction model. The preset fault risk prediction model is used to predict the impact risk of multiple equipment components on the energy storage system when they fail, and to obtain the fault risk prediction results corresponding to each of the multiple equipment components.

[0151] The generation module 504 is used to generate cloud-based fault detection results for the energy storage system based on all fault risk prediction results and fault feature weights.

[0152] Each module in the fault detection device of the aforementioned energy storage system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0153] In one exemplary embodiment, an energy storage system is provided, which includes a management terminal for the energy storage system. The internal structure diagram of the management terminal can be as follows: Figure 6As shown. The control terminal of the energy storage system includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the control terminal provides computing and control capabilities. The memory of the control terminal includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the control terminal is used for exchanging information between the processor and external devices. The communication interface of the control terminal is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a fault detection method for the energy storage system. Those skilled in the art will understand that... Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the control terminal of the energy storage system to which the present application is applied. The control terminal of a specific energy storage system may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0154] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0155] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0156] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A fault detection method for an energy storage system, characterized in that, Applied to the edge of energy storage, the energy storage system is composed of multiple equipment components; the fault detection method of the energy storage system includes: Acquire basic operational data generated by the multiple device components during operation, and perform fault identification on the basic operational data to obtain fault event data, wherein the fault event data characterizes the fault type of the multiple device components; Extract target fault event data within a preset time window from a preset fault event time series database, wherein the preset fault event time series database stores fault event data recorded over time; Based on the target fault event data, fault assessment is performed on the multiple device components within the preset time window to obtain fault assessment feature data. The fault assessment feature data characterizes the fault occurrence of the multiple device components within the preset time window. The fault assessment feature data includes fault assessment frequency data, fault assessment probability data, and fault assessment level data. The fault assessment frequency data characterizes the frequency of fault occurrence of the multiple device components within the preset time window. The fault assessment probability data characterizes the probability of fault occurrence of the multiple device components within the preset time window. The fault assessment level data characterizes the different fault types and fault levels of the multiple device components. The system sends target basic operating data and fault assessment feature data within the preset time window to the energy storage cloud. The energy storage cloud then inputs both the target basic operating data and the fault assessment feature data into a preset fault risk prediction model. This model predicts the impact risk on the energy storage system when multiple equipment components fail, obtaining fault risk prediction results for each of the multiple equipment components. Based on all fault risk prediction results and fault feature weights, a cloud-based fault detection result for the energy storage system is generated. The target basic operating data is the basic operating data within the preset time window. The fault feature weights are obtained by the energy storage cloud through fault correlation analysis of the fault assessment feature data based on the system type of the energy storage system. The fault correlation analysis based on the system type, yielding the fault correlation analysis results, includes: In the case of a centralized system, a corresponding fault assessment feature vector is constructed based on the fault assessment feature data; a fault correlation matrix of the energy storage system is generated based on the correlation degree of different fault types in the fault assessment feature vector; the fault correlation matrix is ​​quantitatively analyzed to obtain the fault correlation analysis result. Alternatively, in the case of a distributed system, target fault assessment feature data with consistent adaptability is selected from the fault assessment feature data based on preset matching conditions, and a corresponding fault assessment feature vector is constructed based on the target fault assessment feature data; a fault correlation matrix of the energy storage system is generated based on the correlation degree of different fault types in the fault assessment feature vector; the fault correlation matrix is ​​quantitatively analyzed to obtain the fault correlation analysis result.

2. The fault detection method for an energy storage system according to claim 1, characterized in that, The step of performing fault assessment on the multiple device components within the preset time window based on the target fault event data to obtain fault assessment feature data includes: For any target device component of the energy storage system, identify the target fault type corresponding to the target fault event data, and extract the fault occurrence time of the target device component; Based on the time of the fault occurrence, the total number of faults corresponding to the target fault type is updated to obtain the updated total number of faults. Based on the fault occurrence time and the updated cumulative total number of faults, the frequency of occurrence of the target fault type within the preset time window is counted to obtain the fault assessment frequency data.

3. The fault detection method for the energy storage system according to claim 2, characterized in that, The step of performing fault assessment on the multiple device components within the preset time window based on the target fault event data to obtain fault assessment feature data includes: Obtain statistical characteristic values ​​of the number of failures, usage intensity, and usage frequency of the target device component within its historical operating cycle; Based on the statistical characteristic value of the number of faults, the fault assessment frequency data is standardized to obtain standard fault assessment frequency data. Based on the usage intensity feature value and the usage frequency feature value, the fault assessment frequency standard data is normalized and mapped to obtain the fault assessment probability data.

4. The fault detection method for the energy storage system according to claim 3, characterized in that, The step of performing fault assessment on the multiple device components within the preset time window based on the target fault event data to obtain fault assessment feature data includes: Based on the target fault type, the corresponding target fault assessment level is queried in a preset mapping table, wherein the preset mapping table stores the mapping relationship between the fault type and the fault assessment level of the target device component; Based on the target fault assessment level, the target equipment component fault level record table is updated to obtain the fault assessment level data.

5. A fault detection method for an energy storage system, characterized in that, Applied to energy storage cloud computing, the energy storage system consists of multiple device components; the fault detection method of the energy storage system includes: The system acquires target basic operating data and fault assessment feature data within a preset time window sent by the energy storage edge device. The target basic operating data refers to the basic operating data within the preset time window. The fault assessment feature data is obtained by the energy storage edge device based on target fault event data extracted from a preset fault event time-series database, which performs fault assessments on the multiple device components. The fault assessment feature data includes fault assessment frequency data, fault assessment probability data, and fault assessment level data. The fault assessment frequency data represents the frequency of fault occurrence of the multiple device components within the preset time window. The fault assessment probability data represents the probability of fault occurrence of the multiple device components within the preset time window. The fault assessment level data represents the different fault types and fault levels of the multiple device components. The preset fault event time-series database stores fault event data recorded over time. The fault event data is obtained by fault identification of the basic operating data generated by the multiple device components during operation. Based on the system type of the energy storage system, a fault correlation analysis is performed on the fault assessment feature data to obtain the fault correlation analysis results, wherein the fault correlation analysis results include the fault feature weights corresponding to the fault assessment feature data. The target basic operating data and the fault assessment feature data are input into a preset fault risk prediction model. The preset fault risk prediction model is used to predict the impact risk on the energy storage system when the multiple equipment components fail, and the fault risk prediction results corresponding to each of the multiple equipment components are obtained. Based on all fault risk prediction results and the fault feature weights, cloud-based fault detection results for the energy storage system are generated; fault correlation analysis is performed on the fault assessment feature data according to the system type to obtain fault correlation analysis results, including: In the case of a centralized system, a corresponding fault assessment feature vector is constructed based on the fault assessment feature data; a fault correlation matrix of the energy storage system is generated based on the correlation degree of different fault types in the fault assessment feature vector; the fault correlation matrix is ​​quantitatively analyzed to obtain the fault correlation analysis result. Alternatively, in the case of a distributed system, target fault assessment feature data with consistent adaptability is selected from the fault assessment feature data based on preset matching conditions, and a corresponding fault assessment feature vector is constructed based on the target fault assessment feature data; a fault correlation matrix of the energy storage system is generated based on the correlation degree of different fault types in the fault assessment feature vector; the fault correlation matrix is ​​quantitatively analyzed to obtain the fault correlation analysis result.

6. The fault detection method for an energy storage system according to claim 5, characterized in that, Before performing fault correlation analysis on the fault assessment feature data according to the system type of the energy storage system to obtain the fault correlation analysis results, the method further includes: Based on the device metadata of the energy storage system, the system type of the energy storage system is detected, wherein the device metadata includes at least one of the device attribute data of the plurality of device components and the device connection relationship data.

7. The fault detection method for an energy storage system according to claim 6, characterized in that, The step of detecting the system type of the energy storage system based on the device metadata of the multiple device components includes: Based on the device metadata, generate the type characteristic determination conditions for the energy storage system; If the energy storage system is found to meet the type characteristic determination conditions, the system type of the energy storage system is determined.

8. The fault detection method for the energy storage system according to claim 7, characterized in that, The step of generating type characteristic determination conditions for the energy storage system based on the device metadata includes: Obtain topology power correlation data among the plurality of device components, wherein the topology power correlation data includes at least one of power aggregation node distribution data, electrical connection level data, power aggregation device connection quantity data, and topology aggregation density data; Based on the topology power correlation data and the device metadata, the type feature determination conditions are generated.

9. The fault detection method for an energy storage system according to claim 8, characterized in that, The type feature determination conditions include centralized feature determination conditions, which include at least one of the following: The multiple device components converge power through the same power convergence device; The multiple equipment components are located at the same electrical grid connection node; The multiple device components share preset common conditions; The topology among the multiple device components is a preset aggregation structure.

10. The fault detection method for an energy storage system according to claim 8, characterized in that, The type feature determination conditions include distributed feature determination conditions, and the distributed feature determination conditions include at least one of the following: Each device component is independently configured with a corresponding power aggregation device, and the multiple device components are respectively connected to different electrical grid-connected nodes; There is no centralized power convergence relationship among the multiple device components; The operating conditions of the multiple device components are not consistent with each other; The topology among the multiple device components is a preset independent structure.

11. The fault detection method for an energy storage system according to claim 6, characterized in that, The step of constructing a corresponding fault assessment feature vector based on the fault assessment feature data includes: The fault assessment feature data is dimensionality reduced to obtain a low-dimensional fault feature representation. Clustering is performed on the low-dimensional fault feature representation to obtain the fault assessment feature vector.

12. The fault detection method for an energy storage system according to claim 5, characterized in that, The fault feature weights are obtained by fusing prior weights, correlation weights, and statistical significance weights.

13. An energy storage system comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the fault detection method for the energy storage system according to any one of claims 1 to 4 or 5 to 12.

Images