A Fault Diagnosis Method for Energy Storage Grid Connected with Distributed Generation
Through multi-stage diagnostic methods, including electrochemical characteristics, electromagnetic characteristics diagnostic models, knowledge graphs, and principal component analysis algorithms, multi-source data is integrated to identify energy storage battery pack failures, and the problem of insufficient diagnostic accuracy in the existing technology is solved, and the Early detection and efficient maintenance have improved the stability and safety of the energy storage system.
Patent Information
- Application Number
- CN202411360636.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-27
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-27
AI Technical Summary
The existing energy storage battery pack fault diagnosis technology is difficult to accurately identify composite faults, especially in extreme environments, which lacks multi-source data fusion means, resulting in insufficient diagnostic accuracy and comprehensiveness.
Multi-stage diagnostic methods are adopted, including building an electrochemical characteristic diagnostic model and an electromagnetic characteristic diagnostic model for preliminary diagnosis, combining the knowledge graph for the second diagnosis, and using the principal component analysis algorithm to diagnose the measured signal for the third diagnosis, integrating all results to generate Final diagnostic report.
It realizes early detection, accurate identification and efficient maintenance of energy storage battery pack failures, significantly improves diagnostic efficiency and accuracy, reduces misdiagnosis rate and maintenance costs, and ensures the stable operation and safety of the energy storage system.
Smart Images

Figure CN118884257B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of power electronics, and in particular to a method for diagnosing faults in an energy storage grid to which a distributed power source is connected. Background Art
[0002] With the large-scale application of renewable energy and the development of smart grids, distributed power sources have become an important part of modern power systems. Energy storage technology plays a core role in balancing supply and demand and improving grid stability. However, the current fault diagnosis technology of energy storage battery packs mainly relies on a single traditional technical means, which makes it difficult to accurately identify complex faults.
[0003] Existing fault diagnosis technologies usually only focus on the status data of the battery pack, such as temperature, voltage and current, but do not fully consider the comprehensive analysis of measured signals such as acoustic signals, vibration signals and thermal imaging data, which limits the accuracy and comprehensiveness of diagnosis. Especially in extreme environments such as high temperature, low temperature, high humidity, and plateau hypoxia, the lack of effective multi-source data fusion means makes it difficult for existing methods to accurately determine the specific type of fault and its root cause. Summary of the invention
[0004] In view of the above existing problems, the present invention is proposed.
[0005] Therefore, the present invention provides a method for diagnosing faults in an energy storage grid with distributed power sources connected, which solves the problem of how to effectively integrate multi-dimensional heterogeneous data and achieve accurate diagnosis and positioning of complex faults in energy storage battery packs.
[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0007] In a first aspect, an embodiment of the present invention provides a method for diagnosing faults in an energy storage grid to which a distributed power source is connected, comprising obtaining status data and measured signals of an energy storage battery group;
[0008] Use state data to build electrochemical property diagnostic models and electromagnetic property diagnostic models;
[0009] Based on the electrochemical characteristic diagnosis model and the electromagnetic characteristic diagnosis model, the energy storage battery pack is preliminarily diagnosed to determine whether there is a fault in the energy storage battery pack;
[0010] When a fault occurs, a fault reasoning system based on knowledge graph is used for a second diagnosis to identify the fault type and cause of the energy storage battery pack;
[0011] If the second diagnosis fails to identify the fault type and cause of the energy storage battery pack, a principal component analysis algorithm is used to perform a third diagnosis on the measured signal to identify the fault type and cause of the energy storage battery pack;
[0012] Combine the preliminary diagnosis results, the second diagnosis results and the third diagnosis results, output the final diagnosis results of the energy storage grid and generate a diagnosis report.
[0013] As a preferred solution of the energy storage grid fault diagnosis method for distributed power sources connected to the present invention, the state data of the energy storage battery group is divided into data under different environmental conditions of normal environment, high temperature, low temperature, high humidity, and plateau hypoxia;
[0014] The state data includes the temperature, voltage, current, internal resistance and electromagnetic field strength of the battery pack;
[0015] The measured signals include acoustic signals, vibration signals, and thermal imaging data.
[0016] As a preferred solution of the method for diagnosing faults of energy storage grids connected to distributed power sources of the present invention, the electrochemical characteristic diagnosis model is constructed by using state data, which is expressed as:
[0017] ;
[0018] Among them, D C Indicates the electrochemical characteristic diagnostic value, V i represents the actual voltage value at the i-th time point, represents the open circuit voltage value at the i-th time point predicted by the OCV-SOC model, ΔR int Indicates the change in the internal resistance of the battery pack, ΔT i Indicates the change in battery pack temperature, represents the influence coefficient of voltage deviation, σ represents the influence coefficient of internal resistance change, τ represents the influence coefficient of temperature change, i represents the index of the time point, and N represents the number of data points in the time series.
[0019] As a preferred solution of the method for diagnosing faults of energy storage grids connected to distributed power sources according to the present invention, an electromagnetic characteristic diagnosis model is constructed using state data, which is expressed as:
[0020] ;
[0021] Among them, D E represents the electromagnetic characteristic diagnostic value, M represents the number of electromagnetic field intensity data points in the time series, and W i represents the wavelet coefficient obtained by wavelet transform at the i-th time point, θ i Indicates the set wavelet coefficient threshold, ΔH i represents the change in the electromagnetic field strength of the battery pack, μ represents the influence coefficient of the wavelet coefficient deviation, and ν represents the influence coefficient of the change in the electromagnetic field strength.
[0022] As a preferred solution of the energy storage grid fault diagnosis method for distributed power sources connected to the present invention, the energy storage battery pack is preliminarily diagnosed based on the electrochemical characteristic diagnosis model and the electromagnetic characteristic diagnosis model to determine whether the energy storage battery pack has a fault, including the following steps:
[0023] Based on historical status data, electrochemical characteristic diagnosis thresholds and electromagnetic characteristic diagnosis thresholds are set;
[0024] When the electrochemical characteristic diagnosis value is less than or equal to the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is less than or equal to the electromagnetic characteristic diagnosis threshold, it is determined that the energy storage battery pack has no fault;
[0025] When the electrochemical characteristic diagnosis value is less than or equal to the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is greater than the electromagnetic characteristic diagnosis threshold, it is judged that the electromagnetic field strength is abnormal and the energy storage battery pack has no fault;
[0026] When the electrochemical characteristic diagnosis value is greater than the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is greater than the electromagnetic characteristic diagnosis threshold, it is judged that multiple faults such as charge state estimation error, internal resistance increase, temperature abnormality and electromagnetic field strength abnormality occur;
[0027] When the electrochemical characteristic diagnostic value is greater than the electrochemical characteristic diagnostic threshold and the electromagnetic characteristic diagnostic value is less than or equal to the electromagnetic characteristic diagnostic threshold, it is judged that multiple faults such as charge state estimation error, increased internal resistance and abnormal temperature have occurred.
[0028] As a preferred solution of the energy storage grid fault diagnosis method for distributed power sources connected to the present invention, when a fault occurs, a second diagnosis is performed using a fault reasoning system based on a knowledge graph to identify the fault type and cause of the energy storage battery pack, including the following steps:
[0029] Based on historical fault types, fault causes, and fault manifestations, a knowledge graph containing nodes and edges is constructed using a graph database, where nodes represent fault types, fault causes, and fault manifestations, and edges represent fault modes, specifically the relationship between fault types, fault causes, and fault manifestations;
[0030] According to the preliminary diagnosis results, match the fault mode in the knowledge graph and calculate the matching degree;
[0031] Evaluate the correlation between the preliminary diagnosis results and the failure mode and calculate the correlation score;
[0032] The key features of the failure mode are extracted using the feature extraction function based on the knowledge graph, expressed as:
[0033] ;
[0034] in, represents the key features for extracting failure modes, represents the knowledge graph, represents the weight matrix from the input layer to the hidden layer, represents the weight matrix from the hidden layer to the output layer, represents the bias vector from the input layer to the hidden layer, b [2] represents the bias vector from the hidden layer to the output layer, represents the activation function of the hidden layer;
[0035] Combining the matching degree, correlation score and key features, the fault type diagnosis value is calculated and expressed as,
[0036] ;
[0037] Among them, D represents the fault type diagnosis value, B represents the knowledge graph matching degree, A represents the relevance score, and λ represents the adjustment factor for adjusting the relevance score;
[0038] Set fault type diagnosis thresholds based on historical fault types, fault causes, and fault manifestations;
[0039] When the fault type diagnosis value is greater than the fault type diagnosis threshold, it means that the fault type and cause are identified and the maintenance plan is determined;
[0040] When the fault type diagnosis value is less than or equal to the fault type diagnosis threshold, it means that the fault type and cause are not identified, and the third diagnosis is performed.
[0041] As a preferred solution of the energy storage grid fault diagnosis method for distributed power sources connected to the present invention, the principal component analysis algorithm is used to perform a third diagnosis on the measured signal to identify the fault type and cause of the energy storage battery group, including the following steps:
[0042] Extract frequency domain features, time domain features, and energy features from acoustic signals, vibration signals, and thermal imaging data, and fuse the features into a feature matrix;
[0043] Calculate the covariance matrix of the feature matrix, expressed as,
[0044] ;
[0045] Where n represents the number of samples, X represents the feature matrix, represents the covariance matrix, represents transpose;
[0046] Calculate the eigenvalues and corresponding eigenvectors of the covariance matrix, and select the eigenvectors corresponding to the first k largest eigenvalues as the principal components;
[0047] Project the feature matrix into the principal component space, expressed as,
[0048] ;
[0049] Among them, Y represents the data matrix after dimensionality reduction, V k represents the matrix containing the first k eigenvectors, k represents the eigenvector corresponding to the first k largest eigenvalues selected;
[0050] Calculate the similarity between the reduced-dimensional data matrix and the known fault mode, and determine the fault type and cause based on the similarity value;
[0051] If the fault type and cause are not identified in the third diagnosis, the collection of status data and real-time signal operations and the fault diagnosis operations are restarted.
[0052] As a preferred solution of the method for diagnosing faults of energy storage grids connected to distributed power sources of the present invention, the method comprises the following steps: combining the preliminary diagnosis result, the second diagnosis result and the third diagnosis result, outputting the final diagnosis result of the energy storage grid and generating a diagnosis report.
[0053] When a fault is found in the initial diagnosis, it is combined with the fault type and cause identified in the second diagnosis to output the final diagnosis result of the energy storage grid, design a maintenance plan, and generate a diagnosis report;
[0054] If the second diagnosis does not identify the fault type and cause, the fault type and cause identified in the third diagnosis are combined to output the final diagnosis result of the energy storage grid, design a maintenance plan, and generate a diagnosis report.
[0055] The beneficial effects of the present invention are as follows: by acquiring and preprocessing battery pack status data and measured signals, preliminary fault judgment is made in combination with the diagnosis model of electrochemical and electromagnetic characteristics, secondary diagnosis is performed using a fault reasoning system based on a knowledge graph to accurately identify the fault type and cause, and when necessary, a principal component analysis algorithm is used to further diagnose difficult-to-identify faults, and finally all diagnostic results are integrated to output a detailed diagnostic report and maintenance plan. The present invention realizes early detection, accurate identification and efficient maintenance of energy storage battery pack faults, significantly improves diagnostic efficiency and accuracy, reduces misdiagnosis rate and maintenance costs, and ensures stable operation and safety of the energy storage system. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0057] Figure 1 This is a flow chart of the method for diagnosing faults in an energy storage grid to which a distributed power source is connected in Example 1.
[0058] Figure 2 This is a decision diagram of the diagnostic process in Example 1. DETAILED DESCRIPTION
[0059] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.
[0060] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method for diagnosing faults in an energy storage grid to which a distributed power source is connected, comprising the following steps:
[0061] S1, obtaining the status data and measured signals of the energy storage battery pack, including the following steps:
[0062] The status data of the energy storage battery pack is divided into data under different environmental conditions: normal environment, high temperature, low temperature, high humidity, and plateau hypoxia. The normal environment usually refers to the standard conditions under which the battery pack can operate normally, such as the temperature range is generally between 20°C and 25°C, the humidity is moderate, and there are no extreme climate conditions. High temperature environment usually refers to conditions that exceed the optimal operating temperature range of the battery pack. For most lithium-ion battery packs, high temperature is defined as above 40°C or even higher, depending on the specific type and technical specifications of the battery pack. High temperature will have a significant impact on the chemical reaction rate of the battery pack, which may shorten the life of the battery pack. Low temperature environment usually refers to conditions below the optimal operating temperature range of the battery pack. For lithium-ion battery packs, low temperature may be defined as below 0°C or lower, and the specific value also depends on the design of the battery pack. Under low temperature conditions, the chemical reaction rate of the battery pack will decrease, which may affect the power output of the battery pack. High humidity environment usually refers to a high moisture content in the air, which may cause problems such as corrosion or short circuit of battery components. Although the specific humidity threshold may vary depending on the type of battery pack, in general, a relative humidity of more than 80% or even higher may be considered a high humidity environment. Plateau hypoxia refers to areas with high altitudes and low oxygen concentrations, usually at an altitude of 3,000 meters or above. Although the volume fraction of oxygen is still 21%, the oxygen partial pressure is reduced to about 157.5 mmHg (relative to 160 mmHg at sea level). This environment may affect the cooling effect of the battery system because the lower air pressure on the plateau may affect the heat dissipation efficiency. However, the specific impact of the plateau hypoxia environment on the performance of the battery pack depends on the specific situation, because many battery systems do not directly rely on oxygen in the atmosphere to operate.
[0063] Status data includes battery pack temperature, voltage, current, internal resistance, and electromagnetic field strength;
[0064] The measured signals include acoustic signals, vibration signals, and thermal imaging data.
[0065] It should be noted that by collecting the temperature, voltage, current, internal resistance and other status data of the battery pack, as well as the measured signals such as acoustic signals, vibration signals, and thermal imaging data, the comprehensive monitoring of the battery pack operation status is achieved. The status data is classified according to different environmental conditions to facilitate the subsequent diagnosis of faults in specific environments.
[0066] S2. Using the state data to construct an electrochemical characteristic diagnosis model and an electromagnetic characteristic diagnosis model, including the following steps:
[0067] The electrochemical characteristic diagnostic model is constructed using state data and is expressed as,
[0068] ;
[0069] Among them, D C Represents the electrochemical characteristic diagnostic value, which is used to represent the comprehensive evaluation result of the electrochemical state of the battery pack. i Indicates the actual voltage value at the i-th time point, reflecting the output voltage of the battery pack at that time point. It represents the open circuit voltage value at the i-th time point predicted by the OCV-SOC model (open circuit voltage-state of charge model), which is used as a reference for the battery pack voltage under ideal conditions. ΔR int Indicates the change in the internal resistance of the battery pack, reflecting the change in the internal resistance of the battery pack over time or under usage conditions, ΔT i Indicates the change in battery pack temperature, reflecting the temperature fluctuations of the battery pack at different time points. It represents the influence coefficient of voltage deviation, which is used to quantify the influence of the deviation between the actual voltage value and the predicted voltage value on the diagnostic value. σ represents the influence coefficient of internal resistance change, which represents the contribution of internal resistance change to the battery pack status assessment. τ represents the influence coefficient of temperature change, which reflects the influence of temperature change on the performance and status of the battery pack. i represents the index of the time point, and N represents the number of data points in the time series.
[0070] It should be noted that the basis for designing the above formula is based on the impact of changes in the electrochemical characteristics of the battery pack on its health status. By quantifying the impact of voltage deviation, internal resistance change and temperature change on the battery pack status, a comprehensive evaluation of the electrochemical status of the battery pack can be achieved.
[0071] The electromagnetic characteristic diagnosis model is constructed using the state data and is expressed as,
[0072] ;
[0073] Among them, D E It represents the electromagnetic characteristic diagnostic value, which is used to represent the comprehensive evaluation result of the electromagnetic state of the battery pack. M represents the number of electromagnetic field intensity data points in the time series. W i represents the wavelet coefficient obtained by wavelet transform at the i-th time point, which is used to analyze the time-frequency characteristics of the battery pack signal, θ i Represents the set wavelet coefficient threshold, which is used to analyze the time-frequency characteristics of the battery pack signal, ΔH i It represents the change of the electromagnetic field strength of the battery pack, reflecting the fluctuation of the electromagnetic field around the battery pack. μ represents the influence coefficient of the wavelet coefficient deviation, which represents the influence of the wavelet coefficient and the threshold deviation on the battery pack state assessment. ν represents the influence coefficient of the change of the electromagnetic field strength, which quantifies the influence of the change of the electromagnetic field strength on the battery pack state.
[0074] It should be noted that the basis for designing the above formula is based on the impact of changes in the electromagnetic field around the battery pack on its health status. By quantifying the impact of wavelet coefficient deviation and changes in electromagnetic field intensity on the battery pack status, a comprehensive evaluation of the electromagnetic status of the battery pack is achieved.
[0075] S3, based on the electrochemical characteristic diagnosis model and the electromagnetic characteristic diagnosis model, a preliminary diagnosis is performed on the energy storage battery pack to determine whether the energy storage battery pack has a fault, including the following steps:
[0076] Based on historical status data, electrochemical characteristic diagnosis thresholds and electromagnetic characteristic diagnosis thresholds are set;
[0077] When the electrochemical characteristic diagnosis value is less than or equal to the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is less than or equal to the electromagnetic characteristic diagnosis threshold, it is determined that the energy storage battery pack has no fault;
[0078] When the electrochemical characteristic diagnosis value is less than or equal to the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is greater than the electromagnetic characteristic diagnosis threshold, it is judged that the electromagnetic field strength is abnormal and the energy storage battery pack has no fault;
[0079] When the electrochemical characteristic diagnosis value is greater than the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is greater than the electromagnetic characteristic diagnosis threshold, it is judged that multiple faults such as charge state estimation error, internal resistance increase, temperature abnormality and electromagnetic field strength abnormality occur;
[0080] When the electrochemical characteristic diagnostic value is greater than the electrochemical characteristic diagnostic threshold and the electromagnetic characteristic diagnostic value is less than or equal to the electromagnetic characteristic diagnostic threshold, it is judged that multiple faults such as charge state estimation error, increased internal resistance and abnormal temperature have occurred.
[0081] It should be noted that by quantifying the impact of voltage deviation, internal resistance change and temperature change on the battery pack state, a comprehensive evaluation of the electrochemical state of the battery pack is achieved. By analyzing the wavelet coefficient deviation and electromagnetic field strength change, a comprehensive evaluation of the electromagnetic state of the battery pack is achieved. Setting a reasonable diagnostic threshold ensures that a fault is only determined when both the electrochemical and electromagnetic characteristics are abnormal, reducing the misdiagnosis rate.
[0082] S4. When a fault occurs, a second diagnosis is performed using a fault reasoning system based on a knowledge graph to identify the fault type and cause of the energy storage battery pack.
[0083] Collect relevant knowledge about energy storage battery pack failures from literature, expert experience, and failure cases. Specifically, obtain the failure mechanism, failure detection method, and preventive measures of energy storage battery packs by consulting scientific papers, technical reports, and industry standards. Interview or consult experts in the field of battery packs to understand the failure cases they encounter in practice and their treatment methods. Collect actual failure cases, including failure phenomena, treatment processes, and final results. Use the collected knowledge about energy storage battery pack failures to define the types of failures, such as overheating, short circuit, internal open circuit, capacity attenuation, etc., and the causes of failures, such as manufacturing defects, improper use, and environmental factors (temperature, humidity, Vibration) and fault manifestations, such as abnormal voltage, increased temperature, decreased charging and discharging performance, etc., use graph database to build knowledge graph, which includes nodes and edges, where nodes represent fault types (such as overheating, short circuit, etc.), components (such as battery cells, battery pack management system (BMS), connectors, etc.), symptoms (such as voltage reduction, temperature rise, etc.), and edges represent the relationship between entities, including the relationship between fault type and component, indicating which components a specific type of fault may occur on, the relationship between component and symptom, indicating the symptoms that may be manifested when a component has a problem, and the relationship between fault type and symptom, indicating the typical symptoms that may be caused by a certain fault;
[0084] According to the preliminary diagnosis results, match the fault mode in the knowledge graph and calculate the matching degree;
[0085] Evaluate the correlation between the diagnostic results and the failure mode and calculate the correlation score;
[0086] The key features of the failure mode are extracted using the feature extraction function based on the knowledge graph, expressed as:
[0087] ;
[0088] in, represents the key features for extracting failure modes, represents the knowledge graph, represents the weight matrix from the input layer to the hidden layer, represents the weight matrix from the hidden layer to the output layer, represents the bias vector from the input layer to the hidden layer, represents the bias vector from the hidden layer to the output layer, represents the activation function of the hidden layer;
[0089] Combining the matching degree, correlation score and extracted key features, the fault type diagnosis value is calculated and expressed as,
[0090] ;
[0091] Among them, D represents the fault type diagnosis value, B represents the knowledge graph matching degree, A represents the relevance score, and λ represents the adjustment factor for adjusting the relevance score;
[0092] Set fault type diagnosis thresholds based on fault diagnosis experience and historical data statistical analysis;
[0093] When the fault type diagnosis value is greater than the fault type diagnosis threshold, it means that the fault type and cause are identified and the maintenance plan is determined;
[0094] When the fault type diagnosis value is less than or equal to the fault type diagnosis threshold, it means that the fault type and cause are not identified, and the third diagnosis is performed.
[0095] It should be noted that the knowledge graph constructed based on literature, expert experience and fault cases provides a theoretical basis for fault diagnosis. The key features of the fault mode are extracted using neural networks, which improves the intelligent level of fault diagnosis. The fault type diagnostic value is calculated by combining the matching degree, correlation score and key features to ensure the reliability of the diagnosis results.
[0096] S5. If the second diagnosis fails to identify the fault type and cause of the energy storage battery pack, a third diagnosis is performed on the measured signal using a principal component analysis algorithm to identify the fault type and cause of the energy storage battery pack, including the following steps:
[0097] From acoustic signals, vibration signals and thermal imaging data, use Fourier transform to convert time domain signals into frequency domain signals, find the maximum value or several largest peaks in the spectrum, calculate the energy distribution of the signal, detect whether there are harmonic components and their relative strength, and measure the sharpness of the spectrum peak. The time domain features are directly extracted from the original signal, usually including the statistical properties of the signal. For thermal imaging data, the energy features are usually related to the brightness distribution of the image. The frequency domain features, time domain features, and energy features are obtained, and the features are fused into a feature matrix;
[0098] Calculate the covariance matrix of the feature matrix, expressed as,
[0099] ;
[0100] Where n represents the number of samples, X represents the feature matrix, represents the covariance matrix, represents transpose;
[0101] Calculate the eigenvalues and corresponding eigenvectors of the covariance matrix, and select the eigenvectors corresponding to the first k largest eigenvalues as the principal components;
[0102] Project the feature matrix into the principal component space, expressed as,
[0103] ;
[0104] Among them, Y represents the data matrix after dimensionality reduction, V k represents the matrix containing the first k eigenvectors, k represents the eigenvector corresponding to the first k largest eigenvalues selected;
[0105] Calculate the similarity between the reduced-dimensional data matrix and the known fault mode, and determine the fault type and cause based on the similarity value;
[0106] If the fault type and cause are not identified in the third diagnosis, the state data collection and measured signal operation and the fault diagnosis operation are restarted.
[0107] It should be noted that multiple features are extracted from the measured signals and fused into a feature matrix, which provides rich information for subsequent analysis. Through PCA dimensionality reduction processing, the data dimension is reduced while retaining the main information, which improves the calculation efficiency. The similarity between the reduced-dimensional data and the known fault mode is calculated to further improve the accuracy of diagnosis.
[0108] S6. Combining the preliminary diagnosis results, the second diagnosis results and the third diagnosis results, outputting the final diagnosis results of the energy storage grid and generating a diagnosis report, including the following steps:
[0109] When a fault is found in the initial diagnosis, it is combined with the fault type and cause identified in the second diagnosis to output the final diagnosis result of the energy storage grid, design a maintenance plan, and generate a diagnosis report;
[0110] If the second diagnosis does not identify the fault type and cause, the fault type and cause identified in the third diagnosis are combined to output the final diagnosis result of the energy storage grid, design a maintenance plan, and generate a diagnosis report.
[0111] The maintenance plan includes formulating a corresponding maintenance or parts replacement plan based on the final diagnosis results, taking into account the feasibility, cost and time of the maintenance;
[0112] The diagnostic report shall include a detailed description of the fault type and cause, the data and methods used in the diagnostic process, the interpretation of the comprehensive diagnostic indicators and results, the recommended repair or maintenance measures, preventive measures and future monitoring suggestions;
[0113] It should be noted that the diagnosis results of the three stages are combined to ensure the comprehensiveness and accuracy of the diagnosis conclusion. Reasonable maintenance plans are formulated based on the diagnosis results to improve the pertinence and efficiency of maintenance work. A diagnostic report containing fault type, cause, maintenance suggestions and other contents is generated to provide a basis for subsequent maintenance work. In summary, the present invention obtains and pre-processes the battery pack status data and measured signals, combines the diagnosis model of electrochemical and electromagnetic characteristics to perform preliminary fault judgment, uses the fault reasoning system based on the knowledge graph to perform secondary diagnosis to accurately identify the fault type and cause, and uses the principal component analysis algorithm to further diagnose difficult-to-identify faults when necessary, and finally integrates all diagnostic results to output detailed diagnostic reports and maintenance plans. The present invention realizes early detection, accurate identification and efficient maintenance of energy storage battery pack faults, significantly improves diagnostic efficiency and accuracy, reduces misdiagnosis rate and maintenance costs, and ensures the stable operation and safety of the energy storage system.
Claims
1. A method for diagnosing faults in an energy storage grid connected to a distributed power source, characterized in that: include, Obtain status data and measured signals of energy storage battery packs; Use state data to build electrochemical property diagnosis models and electromagnetic property diagnosis models; Based on the electrochemical characteristic diagnosis model and the electromagnetic characteristic diagnosis model, the energy storage battery pack is preliminarily diagnosed to determine whether there is a fault in the energy storage battery pack; When a fault occurs, a fault reasoning system based on knowledge graph is used for a second diagnosis to identify the fault type and cause of the energy storage battery pack; If the second diagnosis fails to identify the fault type and cause of the energy storage battery pack, a principal component analysis algorithm is used to perform a third diagnosis on the measured signal to identify the fault type and cause of the energy storage battery pack; Combine the preliminary diagnosis results, the second diagnosis results and the third diagnosis results to output the final diagnosis results of the energy storage grid and generate a diagnosis report; The electrochemical characteristic diagnostic model is constructed using state data and is expressed as, ; Among them, D C Indicates the electrochemical characteristic diagnostic value, V i represents the voltage of the battery pack at the i-th time point, represents the open circuit voltage value at the i-th time point predicted by the OCV-SOC model, ΔR int Indicates the change in the internal resistance of the battery pack, ΔT i Indicates the change in temperature of the battery pack, represents the influence coefficient of voltage deviation, σ represents the influence coefficient of internal resistance change, represents the influence coefficient of temperature change, i represents the index of the time point, and N represents the number of data points in the time series; The electromagnetic characteristic diagnosis model is constructed using the state data and is expressed as, ; Among them, D E represents the electromagnetic characteristic diagnostic value, M represents the number of electromagnetic field intensity data points in the time series, and W i represents the wavelet coefficient obtained by wavelet transform at the i-th time point, θ i Indicates the set wavelet coefficient threshold, ΔH i represents the change in the electromagnetic field strength of the battery pack, μ represents the influence coefficient of the wavelet coefficient deviation, and ν represents the influence coefficient of the change in the electromagnetic field strength.
2. The method for diagnosing faults in an energy storage grid connected to a distributed power source according to claim 1, characterized in that: The status data of the energy storage battery pack is divided into data under different environmental conditions: normal environment, high temperature, low temperature, high humidity, and plateau hypoxia; The state data includes the temperature, voltage, current, internal resistance and electromagnetic field strength of the battery pack; The measured signals include acoustic signals, vibration signals, and thermal imaging data.
3. The method for diagnosing faults in an energy storage grid connected to a distributed power source as claimed in claim 2, characterized in that: Based on the electrochemical characteristic diagnosis model and the electromagnetic characteristic diagnosis model, a preliminary diagnosis is performed on the energy storage battery pack to determine whether the energy storage battery pack has a fault, including the following steps: Based on historical status data, electrochemical characteristic diagnosis thresholds and electromagnetic characteristic diagnosis thresholds are set; When the electrochemical characteristic diagnosis value is less than or equal to the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is less than or equal to the electromagnetic characteristic diagnosis threshold, it is determined that the energy storage battery pack has no fault; When the electrochemical characteristic diagnosis value is less than or equal to the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is greater than the electromagnetic characteristic diagnosis threshold, it is judged that the electromagnetic field strength is abnormal and the energy storage battery pack has no fault; When the electrochemical characteristic diagnosis value is greater than the electrochemical characteristic diagnosis threshold, and the electromagnetic characteristic diagnosis value is greater than the electromagnetic characteristic diagnosis threshold, it is judged that multiple faults such as charge state estimation error, internal resistance increase, temperature abnormality and electromagnetic field strength abnormality occur; When the electrochemical characteristic diagnostic value is greater than the electrochemical characteristic diagnostic threshold and the electromagnetic characteristic diagnostic value is less than or equal to the electromagnetic characteristic diagnostic threshold, it is judged that multiple faults such as charge state estimation error, increased internal resistance and abnormal temperature have occurred.
4. The method for diagnosing faults in an energy storage grid connected to a distributed power source as claimed in claim 3, characterized in that: When a fault occurs, a second diagnosis is performed using a fault reasoning system based on a knowledge graph to identify the fault type and cause of the energy storage battery pack, including the following steps: Based on historical fault types, fault causes, and fault manifestations, a knowledge graph containing nodes and edges is constructed using a graph database, where nodes represent fault types, fault causes, and fault manifestations, and edges represent fault modes, specifically the relationship between fault types, fault causes, and fault manifestations; According to the preliminary diagnosis results, match the fault mode in the knowledge graph and calculate the matching degree; Evaluate the correlation between the preliminary diagnosis results and the failure mode and calculate the correlation score; The key features of the failure mode are extracted using the feature extraction function based on the knowledge graph, expressed as: ; in, represents the key features for extracting failure modes, represents the knowledge graph, represents the weight matrix from the input layer to the hidden layer, represents the weight matrix from the hidden layer to the output layer, represents the bias vector from the input layer to the hidden layer, represents the bias vector from the hidden layer to the output layer, represents the activation function of the hidden layer; Combining the matching degree, correlation score and key features, the fault type diagnosis value is calculated and expressed as, ; Among them, D represents the fault type diagnosis value, B represents the knowledge graph matching degree, A represents the relevance score, and λ represents the adjustment factor for adjusting the relevance score; Set fault type diagnosis thresholds based on historical fault types, fault causes, and fault manifestations; When the fault type diagnosis value is greater than the fault type diagnosis threshold, it means that the fault type and cause are identified and the maintenance plan is determined; When the fault type diagnosis value is less than or equal to the fault type diagnosis threshold, it means that the fault type and cause are not identified, and the third diagnosis is performed.
5. The method for diagnosing faults in an energy storage grid connected to a distributed power source as claimed in claim 4, characterized in that: The principal component analysis algorithm is used to perform a third diagnosis on the measured signal to identify the fault type and cause of the energy storage battery pack, including the following steps: Extract frequency domain features, time domain features, and energy features from acoustic signals, vibration signals, and thermal imaging data, and fuse the features into a feature matrix; Calculate the covariance matrix of the feature matrix, expressed as, ; Where n represents the number of samples, X represents the feature matrix, represents the covariance matrix, represents transpose; Calculate the eigenvalues and corresponding eigenvectors of the covariance matrix, and select the eigenvectors corresponding to the first k largest eigenvalues as the principal components; Project the feature matrix into the principal component space, expressed as, ; Among them, Y represents the data matrix after dimensionality reduction, V k represents the matrix containing the first k eigenvectors, k represents the eigenvector corresponding to the first k largest eigenvalues selected; Calculate the similarity between the reduced-dimensional data matrix and the known fault mode, and determine the fault type and cause based on the similarity value; If the fault type and cause are not identified in the third diagnosis, the state data collection and measured signal operation and the fault diagnosis operation are restarted.
6. The method for diagnosing faults in an energy storage grid connected to a distributed power source as claimed in claim 5, characterized in that: Combining the preliminary diagnosis results, the second diagnosis results and the third diagnosis results, the final diagnosis results of the energy storage grid are output and a diagnosis report is generated, including the following steps: When a fault is found in the initial diagnosis, it is combined with the fault type and cause identified in the second diagnosis to output the final diagnosis result of the energy storage grid, design a maintenance plan, and generate a diagnosis report; If the second diagnosis does not identify the fault type and cause, the fault type and cause identified in the third diagnosis are combined to output the final diagnosis result of the energy storage grid, design a maintenance plan, and generate a diagnosis report.
