Fault diagnosis method and system for energy storage system under multi-level converter topological structure

Abstract

The invention relates to the technical field of energy storage systems, and provides an energy storage system fault diagnosis method and system under a multi-level converter topological structure. The method comprises the following steps: identifying a cascade series-parallel topology of a target converter; traversing the cascade series-parallel topology, and setting diagnosis point cloud distribution; constructing a swing function; based on the diagnosis point cloud distribution and swing function, network construction control risk map mining in multiple working modes is executed, and a fault diagnosis module is constructed; the topology electric signals are interacted, a fault diagnosis module is combined to make a point cloud data diagnosis decision in a target working mode, and a fault diagnosis list is determined; and the terminal displays and pre-warns the fault diagnosis list, and triggers an emergency plan, so as to solve the technical problem that when the complexity of the topological structure of the multi-level converter is increased, the fault diagnosis cannot be quickly positioned and the fault-tolerant capability is insufficient, so that the operation stability of the energy storage system is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of energy storage systems, and in particular to a method and system for diagnosing energy storage system faults under a multi-level converter topology. Background Art

[0002] In modern power systems, with the widespread application of renewable energy and energy storage systems, power converter technology has gradually become the core of system operation. However, with the expansion of the scale of energy storage systems and the complexity of multi-level converter topology, how to ensure the reliability of converters and the stability of the system has become a technical problem that needs to be solved urgently. Traditional energy storage system fault diagnosis methods mainly rely on the detection of single fault characteristics, with low diagnostic accuracy and difficulty in adapting to multi-module systems with complex topologies. In addition, factors such as system frequency fluctuations, topological structures between modules, and dynamic changes in power further increase the difficulty of fault location, resulting in delays in fault diagnosis and affecting the safety and stability of system operation. Summary of the invention

[0003] The present application provides a method and system for diagnosing energy storage system faults under a multi-level converter topology structure, aiming to solve the technical problem that when the complexity of the multi-level converter topology structure increases, the fault diagnosis cannot be quickly located and the fault tolerance is insufficient, resulting in reduced operating stability of the energy storage system. The application achieves the technical effect of quickly identifying faults in multiple working modes by combining the topology-based fault diagnosis point cloud distribution with the swing function, thereby improving the diagnostic efficiency and system fault tolerance.

[0004] In view of the above problems, the present application provides a method and system for diagnosing faults of an energy storage system under a multi-level converter topology.

[0005] The first aspect disclosed in the present application provides a method for diagnosing faults in an energy storage system under a multi-level converter topology structure, the method comprising: identifying a cascaded series-parallel topology of a target converter, wherein the series-parallel topology includes a series-parallel structure between modules and a series-parallel structure of components within a module; traversing the cascaded series-parallel topology, setting a diagnostic point cloud distribution, and the diagnostic point cloud distribution is determined based on series-parallel nodes; constructing a swing function, wherein the swing function includes a first swing function constructed based on a topology fault-tolerant configuration, and a second swing function constructed based on system frequency fluctuations; based on the diagnostic point cloud distribution and the swing function, performing network control risk map mining under multiple working modes, and constructing a fault diagnosis module; interacting with topological electrical signals, combining the fault diagnosis module to make a point cloud data diagnosis decision under the target working mode, and determining a fault diagnosis sheet; a terminal displays and warns the fault diagnosis sheet, and triggers an emergency plan, and the emergency plan meets a preset risk level.

[0006] Another aspect disclosed in the present application provides a fault diagnosis system for an energy storage system under a multi-level converter topology structure, the system comprising: a topology identification unit, the topology identification unit being used to identify a cascaded series-parallel topology of a target converter, wherein the series-parallel topology comprises an inter-module series-parallel structure and an intra-module component series-parallel structure; a point cloud distribution setting unit, the point cloud distribution setting unit being used to traverse the cascaded series-parallel topology and set a diagnostic point cloud distribution, wherein the diagnostic point cloud distribution is determined based on series-parallel nodes; a swing function construction unit, the swing function construction unit being used to construct a swing function, wherein the swing function comprises a point cloud distribution based on a series-parallel node; A first swing function constructed by a topological fault-tolerant configuration, and a second swing function constructed based on system frequency fluctuations; a graph mining unit, the graph mining unit is used to perform network control risk graph mining under multiple working modes based on the diagnosis point cloud distribution and the swing function, and to construct a fault diagnosis module; a diagnosis decision unit, the diagnosis decision unit is used to interact with topological electrical signals, and to make point cloud data diagnosis decisions under a target working mode in combination with the fault diagnosis module to determine a fault diagnosis sheet; an early warning unit, the early warning unit is used to display and warn the fault diagnosis sheet on a terminal, and to trigger an emergency plan, and the emergency plan meets a preset risk level.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The energy storage system fault diagnosis method under the multi-level converter topology structure is to identify and traverse the cascaded series-parallel topology structure of the converter, clarify the series-parallel relationship between modules and within the module, and then set the diagnosis point cloud distribution, and use the node-based series-parallel structure as the basis for fault detection to ensure accurate coverage of the diagnosis range. At the same time, a swing function is constructed, and the first swing function based on the topology fault tolerance configuration and the second swing function based on the system frequency fluctuation are combined to perform fault diagnosis in multiple working modes. On this basis, the network control risk map mining is performed to build a fault diagnosis module. Through the interaction with the topological electrical signal, the fault diagnosis module can perform point cloud data analysis in the target working mode, and finally generate a fault diagnosis list, and trigger the emergency plan according to the preset risk level to ensure the accuracy of fault warning and processing. This method effectively improves the real-time and accuracy of fault diagnosis, especially in the complex structure of multiple modules and multiple working modes, it can quickly identify the fault point and enhance the stability and fault tolerance of the energy storage system.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 The figure is a flow chart of a method for diagnosing a fault of an energy storage system in a multi-level converter topology structure in one embodiment.

[0012] Figure 2 1 is a diagram of an energy storage system fault diagnosis system architecture under a multi-level converter topology in one embodiment.

[0013] Explanation of the accompanying drawings: topology recognition unit 1, point cloud distribution setting unit 2, swing function construction unit 3, graph mining unit 4, diagnosis decision unit 5, early warning unit 6. DETAILED DESCRIPTION

[0014] The embodiments of the present application provide a method and system for diagnosing a fault of an energy storage system under a multi-level converter topology structure, thereby solving the technical problem that when the complexity of the multi-level converter topology structure increases, the fault diagnosis cannot be quickly located and the fault tolerance is insufficient, resulting in reduced operating stability of the energy storage system. The embodiment of the present application achieves the technical effect of quickly identifying faults in multiple working modes and improving the diagnostic efficiency and system fault tolerance by combining the topology-based fault diagnosis point cloud distribution with the swing function.

[0015] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0016] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0017] Embodiment 1, as Figure 1 As shown, the present application provides a method for diagnosing faults in an energy storage system under a multi-level converter topology structure, the method comprising:

[0018] A cascaded series-parallel topology of a target converter is identified, wherein the series-parallel topology includes an inter-module series-parallel structure and an intra-module component series-parallel structure.

[0019] In the embodiment of the present application, identifying the cascaded series-parallel topology of the target converter refers to analyzing and identifying the overall structure of the converter in the energy storage system. The structure includes the series and parallel connection modes between the converter modules, and the series and parallel relationship between the components inside each module. The series-parallel structure between modules refers to the connection of multiple converter modules in series or in parallel, which directly affects the distribution of voltage and current and the operating characteristics of the system. The series-parallel structure of the components in the module refers to the series-parallel relationship between the components (such as capacitors, switch tubes, etc.) inside each individual converter module, which determines the electrical performance and working mode inside the module. By identifying these topological structures, the interaction and connection mode of each module and its internal components can be clarified, thereby providing a basis for subsequent fault diagnosis and fault-tolerant mechanisms. This identification helps to locate the specific location where the fault may occur, thereby improving the accuracy and response speed of system diagnosis.

[0020] The cascaded series-parallel topology is traversed to set a diagnosis point cloud distribution, wherein the diagnosis point cloud distribution is determined based on the series-parallel nodes.

[0021] In one embodiment, when traversing the cascaded series-parallel topology, the connection mode between each module in the converter is checked one by one, including the series and parallel connections between modules, and the series and parallel structures of the components inside each module. By analyzing these connections, all electrical nodes and the specific roles of these nodes in the system operation are determined. This step not only covers the external connections of the converter modules, but also goes deep into the electrical connections between the components inside each module, ensuring that the overall topological structure of the system is fully presented. When setting the diagnostic point cloud distribution, based on the series and parallel nodes determined during the traversal process, a point cloud distribution for monitoring and diagnosis is generated. Each point cloud represents a diagnostic monitoring point, and the distribution positions of these points are strictly determined according to the key series and parallel nodes in the system. The design of the diagnostic point cloud distribution ensures that all possible fault locations are covered, especially the connection nodes between converter modules and the key nodes of the internal components of the modules. In this way, real-time monitoring and data collection can be performed for important nodes in the entire system to form a comprehensive diagnostic network, providing accurate information basis for subsequent fault detection and fault source location.

[0022] Furthermore, the present application provides the setting of the diagnosis point cloud distribution, including:

[0023] With rectification-filtering-inversion-modulation as functions, the functional response nodes of the cascaded series-parallel topology are determined; based on the functional response nodes, a first diagnostic point cloud is constructed, wherein the first diagnostic point cloud is located in the series-parallel structure of components within the module, and the first diagnostic point cloud corresponds one-to-one to the cascade module; based on the series-parallel connection nodes between modules, a second diagnostic point cloud is determined; the first diagnostic point cloud and the second diagnostic point cloud are aggregated to determine the distribution of the diagnostic point cloud.

[0024] Preferably, with rectification, filtering, inversion and modulation as the four main functions, each functional module in the cascaded series-parallel topology is identified and decomposed to determine the functional response node of the cascaded series-parallel topology. Among them, rectification is the process of converting the input alternating current into direct current. The rectifier circuit representing the rectification function is usually located near the power input terminal. By determining the nodes where the rectifier bridge, diode or other rectifier elements are located, and marking these nodes with rectification, the rectification function response node can be determined. Filtering is used to smooth the rectified direct current and remove excess voltage and current fluctuations. The filter circuit representing the filtering function is usually located after the rectification. By determining the nodes where the capacitor, inductor or other filter elements are located, and marking these nodes with filtering, the filter function response node can be determined. Inversion is to convert direct current into alternating current. The inverter circuit representing the inverter function is usually located after the filter. By determining the nodes where the IGBT, MOSFET or other power electronic components are located, and marking these nodes with inversion, the inverter function response node can be determined. Modulation is to control the output of the inverter so that it works according to a specific modulation strategy (such as PWM modulation). The modulation circuit representing the modulation function is usually located on the control signal transmission path, or at the junction of the control circuit responsible for modulation and the power electronic component. By determining the nodes where the PWM controller or other modulation components are located and marking these nodes for modulation, the modulation function response nodes can be determined. After clarifying these function response nodes, the corresponding first diagnostic point cloud is generated by analyzing the electrical connection of each function response node. The first diagnostic point cloud is the diagnostic point marked in the series-parallel structure of the components in the module. Each diagnostic point cloud represents a possible fault monitoring point, monitoring the voltage, current, signal status and other parameters of these points, so as to detect whether the components inside the module are working normally. Each cascade module contains multiple function response nodes, representing different functions (such as rectification, filtering, inversion, modulation). When constructing the first diagnostic point cloud, it is necessary to ensure that each diagnostic point cloud corresponds to a specific cascade module, that is, the distribution of the diagnostic point cloud inside each module is based on the series-parallel structure of the components in the module, ensuring that all key components and their connection nodes are covered in each module. After the construction of the first diagnostic point cloud is completed, the second diagnostic point cloud is determined based on the series-parallel connection nodes between the modules. Unlike the first diagnostic point cloud, the second diagnostic point cloud mainly focuses on the connection nodes between the modules. These connection nodes reflect the electrical connection between the modules, which are usually series or parallel circuit nodes, such as the connection points at the output of the module. By monitoring these nodes, the electrical interaction between the modules can be understood, thereby assisting in fault location. Finally, the first diagnostic point cloud and the second diagnostic point cloud are aggregated, that is, the first diagnostic point cloud and the second diagnostic point cloud are summarized to form an overall diagnostic point cloud distribution.This distribution diagram integrates the diagnostic points of key components inside the module (first diagnostic point cloud) and connection nodes between modules (second diagnostic point cloud), ensuring that all key locations in the entire system are monitored. This diagnostic point cloud distribution provides a comprehensive monitoring perspective that can capture possible faults in the system in a timely manner and provide an accurate basis for fault diagnosis.

[0025] A swing function is constructed, wherein the swing function includes a first swing function constructed based on a topology fault-tolerant configuration and a second swing function constructed based on a system frequency fluctuation.

[0026] In one embodiment, in the specific process of constructing the swing function, it is divided into two main parts, namely, the first swing function based on the topological fault-tolerant configuration and the second swing function based on the system frequency fluctuation. First, the topological structure of the converter is analyzed to identify the series-parallel relationship between the modules and the specific implementation method of the fault-tolerant mechanism. In the energy storage system, some abnormal conditions may not directly affect the overall operation of the system because the system has a certain fault-tolerant capability. Based on this, a first swing function is constructed to measure the response limit of the system under this fault-tolerant mechanism. The first swing function defines a normal working limit range. Even if there are slight abnormal fluctuations in the system within this range, such as slight deviations in the electrical parameters of the components between the modules, the system can still automatically adjust through the fault-tolerant configuration so that the overall operation is not affected. For example, the current fluctuation of a component in the module slightly exceeds the set value, but the fault-tolerant mechanism of the component can allow electrical deviations within a certain range. The swing function can cover these deviation ranges, thereby correcting these minor anomalies. The first swing function is as follows: Among them, S1(P) = 1 means that the key electrical parameter P is within the allowable fault tolerance range and the system can operate normally. S1(P) = 0 means that the key electrical parameter P exceeds the fault tolerance range and fault diagnosis is required. min It is the minimum allowable value of the key electrical parameter P, which is determined based on the rated electrical parameters of the equipment. maxIt is the maximum allowable value of the key electrical parameter P, which is determined according to the rated electrical parameters of the equipment. ΔP is the allowable deviation range of the key electrical parameter P of the system under fault-tolerant configuration, which is determined according to design specifications and industry standards. For example, in power converters, the deviation of voltage and current usually cannot exceed ±5% or ±10% of the rated value. P is a key electrical parameter and is the main monitoring object in the diagnosis process, including but not limited to voltage and current. In addition, the normal fluctuation of system frequency is also considered. In actual operation, due to the influence of factors such as magnetic field, working mode, load changes, etc., the electrical parameters of the system, such as frequency, voltage and current, are usually not constant. Even in a stable operating state, the system frequency will fluctuate slightly. Based on this, a second swing function is constructed, which is used to capture the frequency fluctuation of the system under different factors. The second swing function can describe the frequency fluctuation range of the system to ensure that these fluctuations are within the normal working range. Even if there are slight fluctuations, as long as the fluctuation amplitude does not exceed the range defined by the function, the system can still maintain normal response. The second swing function is as follows: Among them, S1(f) = 1 means that the system operating frequency f is within the allowable fluctuation range and the system can respond normally. S1(f) = 0 means that the system operating frequency f fluctuates beyond the allowable range and fault diagnosis is required. min It is the minimum permissible value of the system operating frequency f, which is determined based on the standard grid frequency. max It is the maximum permissible value of the system operating frequency f, which is determined according to the standard grid frequency. Δf is the allowable fluctuation tolerance of the system operating frequency f, which is determined according to the grid operation regulations and national standards. For example, the deviation of the operating frequency usually cannot exceed ±0.5Hz or ±1Hz of the standard value. In order to achieve overall monitoring of fault tolerance and frequency fluctuations, the first swing function is combined with the second swing function to form a complete swing function to ensure that the system can operate stably under the influence of multiple factors, whether it is a slight electrical deviation caused by component fault tolerance or an electrical change caused by the frequency fluctuation of the system itself. When slight abnormalities or frequency fluctuations occur, the system can still work normally, providing a stable basis for subsequent fault diagnosis.

[0027] Based on the diagnosis point cloud distribution and the swing function, network control risk map mining in multiple working modes is performed to build a fault diagnosis module.

[0028] In one embodiment, the monitoring points of each key node in the system are determined by the generated diagnostic point cloud distribution. These diagnostic point clouds cover the internal components of the module and the connection nodes between the modules, and can monitor the operating status of the system in real time. The diagnostic point cloud distribution provides all possible fault locations in the system. By monitoring these points, the key electrical parameters and operating status data of the system can be obtained. At the same time, the constructed swing function is used to determine whether the system is operating within the normal fluctuation range, identify which data fluctuations exceed the allowable tolerance range, and mark them. In actual operation, it may be in different working modes (such as different loads, different power supply conditions, etc.). In each working mode, the operating parameters and frequency fluctuation characteristics will be different, so it is necessary to mine the control risk map of the system for different working modes. By analyzing the operating data under multiple working modes, the diagnostic point cloud states under different modes are compared with the corresponding swing functions, and the risks that may exist in each working mode are mined, thereby constructing a network control risk map. This network control risk map can provide the operating conditions under multiple working modes and identify high-risk working modes and nodes. After completing the risk map mining, the obtained network control risk map is encapsulated to construct a fault diagnosis module. The fault diagnosis module will monitor all diagnostic point clouds in real time according to the corresponding relationship in the network control risk map, and judge the status of the node in combination with the swing function. Whenever the system operation exceeds the allowable fluctuation range, the diagnosis module will trigger fault detection, mark the fault location and identify the abnormal mode. In this way, the diagnosis module can provide timely fault warnings and diagnostic suggestions for each stage of system operation.

[0029] Furthermore, this application provides a network control risk map, including:

[0030] The operation record data of the interactive target converter is clustered based on the working mode to determine N groups of records; a group of records is identified and the spectrum energy distribution and envelope characteristics under the diagnosis point cloud distribution are mined to determine the first point cloud relationship, the first point cloud relationship includes a first electrical data relationship and a first fault relationship, the first fault relationship is identified with a fault feature, including a fault tracing element and a fault type; the N groups of records are traversed to determine the Nth point cloud relationship, the first point cloud relationship is integrated until the Nth point cloud relationship, and the network control risk map is constructed.

[0031] Preferably, the operation record data of the target converter is obtained according to the monitoring points arranged in the target converter, and these data cover the operation of the system in different time periods and working modes. By collecting electrical parameters such as voltage, current, and frequency, the working state of the converter can be understood. These data provide a basis for subsequent analysis. Subsequently, the operation record data of the target converter is preliminarily divided according to the constructed swing function, and each operation record data is marked, that is, marked with 1 or 0, so that the impact of the operation record data on the operation of the system can be distinguished. Afterwards, according to different working modes during operation, cluster analysis is performed on the operation record data belonging to the same mark. In the clustering process, taking the operation record data marked as 1 as an example, when performing feature clustering based on the working mode, the Gaussian mixture model (GMM) method is used to process the data distribution under different working modes. According to the working mode of the system, it is assumed that the data distribution under these modes can be described by multiple Gaussian distributions, and each working mode corresponds to one or more Gaussian distributions. Then initialize the model, set the initial mean, variance, and mixing coefficient, and these initial parameters will affect the convergence effect of the model. Then use the EM algorithm (Expectation-Maximization) for parameter estimation. The EM algorithm is divided into the expectation step and the maximization step. The expectation step (E step) is to calculate the probability of each data point belonging to a specific working mode based on the existing parameter estimation, and then calculate the relationship between each point and the corresponding Gaussian distribution. Then, enter the maximization step (M step), in which the mean, variance and mixing coefficient of each Gaussian distribution are re-estimated through the probability results calculated in the E step, and these model parameters are updated. This process needs to be iterated repeatedly, executing the E step and the M step until the model converges, that is, the parameters no longer change significantly, ensuring that each working mode can be described by a Gaussian distribution or a combination of multiple Gaussian distributions. In this way, the operating data under different working modes can be effectively classified, providing a clear classification basis for subsequent fault analysis and diagnosis. Through clustering, all data are divided into N groups of records, representing electrical characteristics under different working conditions. For example, some records may represent light load conditions, while others represent high load conditions. The purpose of clustering is to identify typical data groups under different modes and provide a classification basis for subsequent fault analysis.

[0032] After obtaining N groups of records, a group of records is randomly selected, and the spectral energy distribution and envelope characteristics under the diagnostic point cloud distribution are analyzed for the data in the group. Spectral energy distribution is to perform spectral analysis on electrical parameters through fast Fourier transform (FFT) or other transformation techniques to determine the frequency components of signals such as voltage and current, and identify the spectral energy range when the system is working normally. This helps to determine whether the electrical parameters are operating within the expected frequency range when the system is working normally. For example, if the spectral energy of the voltage is concentrated in a specific frequency range and the energy is at a stable level, it means that the system is operating normally in this working mode. The envelope feature is to extract the envelope curve of the signal using Hilbert transform or other methods. Hilbert transform can capture the instantaneous amplitude of the signal and calculate its envelope curve to show the trend of signal strength over time, reflecting the amplitude characteristics of the signal. By analyzing the envelope curve of the signal waveform, the amplitude characteristics of the signal fluctuation can be understood, especially when a fault occurs, the envelope may show abnormal fluctuations. On the basis of determining the spectral energy distribution and envelope characteristics, the electrical data relationship is analyzed in combination with the diagnostic point cloud distribution. The point cloud distribution reflects the electrical characteristics of each node of the system, such as the changes in data such as voltage and current under different working conditions. Based on the spectrum energy distribution and envelope characteristics, the changes in electrical data of each node in the diagnostic point cloud are analyzed. For example, the characteristic signals of voltage or current will be different at different nodes before and after rectification. Through spectrum analysis, the frequency components of the AC signal before rectification and the stability of the DC signal after rectification can be identified, and the electrical characteristic change points of the key nodes can be found. This can help analyze the voltage and current change rules under different working conditions and determine the first electrical data relationship between nodes. At the same time, according to the spectrum energy distribution and envelope characteristics, it is also determined which nodes may have faults. For example, if the spectrum energy distribution of a node is abnormal, or the envelope line fluctuates violently, it indicates that there is a problem with the electrical components of the node. By mining the relationship between these features, the first fault relationship can be determined. The first fault relationship contains fault characteristics in the system, such as fault type and fault tracing components. By analyzing the electrical data and spectrum characteristics, it is possible to identify which components may fail and the type of fault (such as overcurrent, overvoltage, short circuit, etc.). Subsequently, the first electrical data relationship and the first fault relationship are summarized to determine the first point cloud relationship. After that, repeat the above analysis process for each set of data, and gradually determine the point cloud relationship of each set of records. For each set of data, it is necessary to identify the corresponding electrical data relationship and fault characteristics, and find out the key nodes and fault characteristics under different working modes. When the point cloud relationships of N sets of records are identified, they are integrated together to form a complete network control risk map. This map shows the distribution of fault risks under different working modes, clarifies which nodes are prone to failure under different conditions, and the typical electrical data characteristics and fault types under each working mode.The risk map can help monitor the risks in the working mode in real time and provide early warning and emergency response suggestions when anomalies occur. Through this multi-level analysis method, the fault characteristics of the converter under different working conditions can be fully understood, and a complete risk map can be constructed to provide a basis for accurate fault diagnosis.

[0033] Further, the present application provides determining a first fault relationship, including:

[0034] The spectrum energy distribution and envelope characteristics are traversed to determine linear fault classes and nonlinear fault classes; for the linear fault classes, fault linear relationships are mined; for the nonlinear fault classes, nonlinear fault feature dimensionality reduction screening under kernel principal component analysis is performed to mine fault cost relationships, wherein the fault cost relationship is an approximate linear relationship of nonlinear features; based on the fault linear relationship and the fault cost relationship, the first fault relationship is determined.

[0035] Optionally, the spectrum energy distribution is analyzed to identify the energy fluctuations under different frequency components. If there is an abnormal increase in energy or significant frequency fluctuation in the spectrum, it means that there is a fault in a certain part. By traversing the change trend of the spectrum energy, the type of fault can be preliminarily determined. Then analyze the envelope characteristics of the signal. If the envelope curve shows abnormal fluctuations, such as a sudden increase in amplitude or a drastic non-steady-state change, it may indicate the presence of a fault. By combining the comprehensive analysis of the spectrum energy distribution and the envelope characteristics, linear fault classes and nonlinear fault classes can be distinguished. The linear fault class is a fault feature that changes relatively smoothly over time and can be described by a linear relationship. For example, voltage or current exceeding the preset range is a typical linear fault class. The nonlinear fault class is a fault feature that is more complex and shows nonlinear characteristics over time, which is difficult to directly describe with a simple linear relationship. For example, rapid changes in load in a short period of time, or fluctuations under the combined action of multiple factors, belong to the nonlinear fault class. For the identified linear fault class, the electrical parameters belonging to the linear fault class are linearly regressed and fitted, that is, these electrical parameters and the corresponding acquisition time are input into the linear regression model, and the regression function P(t)=at+b in the linear regression model is fitted to determine the specific slope a and intercept b. After the fitting is completed, the fault linear relationship is determined, which clearly describes the change law of the electrical parameters (such as current or voltage) over time. If the slope a is positive and large, it means that the electrical parameters increase rapidly over time, such as the rapid increase in current, indicating that the system is in a linear state of overcurrent fault. If the slope a is negative, it means that the parameter gradually decreases over time, which may reflect the gradual attenuation of a certain electrical parameter, resulting in system function failure. For nonlinear fault classes, due to the complex fault characteristics, traditional linear analysis methods are not enough to capture their changes. Therefore, kernel principal component analysis (KPCA) is used to process nonlinear data. KPCA is a dimensionality reduction method based on kernel functions, which can effectively map high-dimensional nonlinear data to a lower-dimensional linear space, thereby better identifying potential fault characteristics. Specifically, first, the nonlinear fault data is mapped to a high-dimensional space through a selected kernel function (such as a Gaussian kernel, RBF kernel, etc.). The kernel function is determined by the distribution of the fault data and the dimension of the data. For example, if the fault data presents a local burst of nonlinear characteristics, a Gaussian kernel or an RBF kernel is used. If the fault data presents a relatively smooth nonlinear trend, a polynomial kernel is used. Subsequently, the nonlinear fault data is mapped to a high-dimensional space through a kernel function. In a high-dimensional space, the nonlinear relationship of the original data is converted into a linear relationship, which is convenient for subsequent feature extraction. The kernel matrix is ​​then calculated through the selected kernel function, and this kernel matrix represents the similarity of each data point with other data points. Afterwards, the kernel matrix is ​​subjected to eigenvalue decomposition to obtain a set of eigenvectors and corresponding eigenvalues. The larger the eigenvalue, the greater the contribution of the eigenvector to the principal component of the data.By sorting the eigenvalues, the first few principal components (i.e., those with larger eigenvalues) are selected. These principal components are the components that contribute most to the data change and can retain the main features of the data. Through the selected principal components, the data in the high-dimensional space is mapped back to the linear space of lower dimensions. After this step is completed, the original nonlinear fault data is converted into low-dimensional data in the linear space. After dimensionality reduction, multiple principal components are obtained, each of which represents the one-dimensional data after dimensionality reduction. By analyzing the contribution of each principal component (i.e., the size of the eigenvalue), the main components that best represent the fault characteristics can be screened out. Then, the same linear regression fitting as above is performed using the screened principal components to determine the fault cost relationship, that is, how these characteristics cause certain specific types of faults in the system. Finally, the linear fault relationship is merged with the nonlinear fault relationship to form a complete first fault relationship. This includes the type and source of the fault, as well as information associated with the electrical characteristics of each node in the diagnostic point cloud, thereby helping to accurately identify and handle faults.

[0036] The interactive topological electrical signal is combined with the fault diagnosis module to make a point cloud data diagnosis decision under the target working mode to determine the fault diagnosis list.

[0037] In one embodiment, during the fault diagnosis process, the current topological electrical signal data is obtained through the monitoring points in the target converter. These electrical signals reflect the electrical connection and working status between different modules and components inside the target converter. Subsequently, these topological electrical signal data are combined with the constructed fault diagnosis module. The fault diagnosis module has previously defined possible fault modes and characteristics by analyzing the working mode and fault characteristics of the system. Therefore, the currently acquired topological electrical signal data will be input into the fault diagnosis module and compared and analyzed with the fault mode defined in the diagnosis module. In this step, the current working mode and point cloud data will be combined for diagnostic decision making, that is, the matching degree between the current signal and the known fault mode in the diagnosis module is analyzed to identify whether there is an abnormality. Afterwards, a fault diagnosis sheet is generated based on the diagnosis results. This fault diagnosis sheet contains detailed information about the fault, such as the type of fault, the affected components, etc., to provide a basis for subsequent early warnings.

[0038] Furthermore, the present application provides the method of determining a fault diagnosis sheet, including:

[0039] Identify the topological electrical signal, perform the conversion under the diagnosis point cloud distribution, and determine the point cloud distribution map; determine the target working mode, and trigger the target networking control risk map mapped by the mode; use the swing function as a constraint, combine the target networking control risk map, make a fault diagnosis decision on the point cloud distribution map, and determine the fault diagnosis list.

[0040] Preferably, real-time electrical signal data is obtained from multiple monitoring points. These data represent the operating status of each part, such as voltage, current fluctuation, frequency, etc. Subsequently, these electrical signal data are converted, that is, the topological nodes corresponding to these electrical signal data are matched one by one with the diagnostic point cloud distribution, so as to generate a point cloud distribution map, which reflects the operating status and mutual relationship of each node. Since the operating mode of the target converter may change according to different operating conditions or load conditions, after determining the point cloud distribution map, the current operating mode of the target converter will be identified to determine the target operating mode of the target converter. Based on the determined target operating mode, the mode mapping process is triggered, that is, the current point cloud distribution is compared with the network control risk map. The network control risk map describes the faults that may be faced under different operating modes, which can help predict potential problems under different working conditions. By comparing the data in the current point cloud distribution with the corresponding data type in the network control risk map, it is determined whether the data is abnormal, and marked according to the judgment result to form a target network control risk map. Afterwards, the swing function is used as a constraint to ensure that the electrical signal is judged as abnormal only when the fluctuation exceeds the normal swing range. Based on the target network control risk map, the electrical signal characteristics with abnormal identification in the point cloud distribution map are compared with the swing function to make a fault diagnosis decision. If certain electrical parameters exceed the tolerance range of the swing function and match the fault characteristics in the risk map, the location will be determined as the fault point. Then, relevant information is collected for these determined fault points, including fault type, fault location, etc., and the collected information is stored one by one to form a fault diagnosis sheet, which details the type of fault, location of occurrence, and the impact on the overall system.

[0041] Furthermore, the present application provides a method for making a fault diagnosis decision on the point cloud distribution map, including:

[0042] Based on the cascaded series-parallel topology, the active inertia is located, where the active inertia is the converter influencing inertia that affects the dynamic response characteristics and stability; the active inertia vector is determined to assist in fault diagnosis decision-making.

[0043] Optionally, in the cascaded series-parallel topology of the target converter, taking the energy storage element as an example, the energy storage element (such as capacitor and inductor) of each power module and the control strategy jointly determine the dynamic response characteristics and stability. The energy storage element affects the speed of change of current and voltage by storing energy. For example, the capacitor can store electrical energy and mitigate voltage fluctuations, while the inductor stores magnetic energy and suppresses rapid changes in current. The control strategy affects the response speed to load changes or external interference by adjusting the working state of these energy storage elements, thereby ensuring the stable operation of the system. On this basis, active inertia is used to describe the influence of these energy storage elements and control strategies on the dynamic response characteristics. Active inertia not only determines the response speed to electrical disturbances, but also affects the stability of the system. For example, a large capacitor makes the voltage change of the system relatively smooth, while a large inductor can suppress rapid fluctuations in current. Therefore, active inertia is an important parameter affecting the electrical characteristics of the system. In order to more accurately describe this inertia effect, an inertia model is introduced to quantify the influence of energy storage elements with a formula. The form of the model is: j = C·V 2 +L·I 2 ; where j represents active inertia, C is the capacitance of the capacitor, V represents voltage, L is the inductance of the inductor, and I is the current. In this way, the role of the energy storage element can be quantified as the inertia effect of the system. Subsequently, these influencing factors are integrated by establishing an active inertia vector, which provides a basis for fault diagnosis of the target converter. The active inertia vector can be expressed as In this vector, j C is the contribution of the capacitor to the converter inertia, representing the buffering effect of capacitor energy storage on voltage changes; j L is the contribution of inductance to the converter inertia, reflecting the inhibitory effect of inductor energy storage on current changes; and j control This indicates the role of the control strategy in the dynamic response of the converter. During the operation of the converter, the components of the active inertia vector will be dynamically adjusted as the working state of the converter changes. When a storage element (such as a capacitor or inductor) has a problem, the related components will change abnormally. For example, the performance degradation of the capacitor may cause j C The fluctuation of the component and the problem of control strategy will cause control By monitoring the changes in these inertia components, the fault of the converter can be quickly identified and the fault location and type can be determined. When performing fault detection, when a component of the active inertia vector fluctuates abnormally, the fault diagnosis mechanism will be triggered. For example, if j C A large fluctuation in the component may indicate a capacitor failure; L The change of may represent an inductor fault. Combining these diagnostic results can assist the fault diagnosis decision-making process and ensure the accuracy of the generated fault diagnosis decision.

[0044] The terminal displays and warns the fault diagnosis sheet and triggers an emergency plan, which meets the preset risk level.

[0045] In one embodiment, during the fault diagnosis process of the converter, when a fault is detected and a fault diagnosis sheet is generated, the terminal will display the diagnosis results in real time and issue an early warning signal. This fault diagnosis sheet contains detailed information about the fault, including the location, type, severity, etc. of the fault. These key information are provided to operators through the terminal to ensure that they can understand the current fault situation in a timely manner. When the fault diagnosis result shows that the fault reaches or exceeds the preset risk level, the corresponding emergency plan will be automatically triggered. The content of the emergency plan varies according to different risk levels. For lower-level faults, a warning will be issued and an inspection or adjustment will be recommended. For higher-risk faults, more urgent measures will be taken, such as automatically shutting down some equipment, switching to a backup system, or performing other safety protection operations to prevent the fault from further expanding or causing more serious effects on the entire converter system. Throughout the process, not only is timely fault information display provided, but the safe operation of the converter is also guaranteed through the preset emergency plan, ensuring that appropriate countermeasures can be taken quickly when a fault occurs, reducing losses and improving system reliability.

[0046] Furthermore, the present application provides on-off management and control complementary regulation, including:

[0047] The modules are relatively independent and cooperate with each other to determine the faulty module and perform on-off management and control complementary adjustment of the topology switch tube; wherein, the on-off management and control complementary adjustment of the topology switch tube include: locating the faulty module based on the fault diagnosis sheet; interrupting the topology switch tube for the faulty module, and performing balanced complementary adjustment based on the collaborative module, wherein the topology switch tube is the response hardware, and the power conversion system and battery management system are the adjustment response software.

[0048] Preferably, when a module is identified as faulty through a fault diagnosis sheet, the specific module where the fault occurs must first be determined. This process locates the module causing the problem by analyzing the diagnostic data. The module may cause abnormal current or voltage due to a fault in an internal circuit element (such as a capacitor, inductor or other power element). Once the faulty module is determined, the topological switch tube in the module is managed, specifically by interrupting the on-off state of the topological switch tube of the module. The topological switch tube is a key hardware component responsible for controlling the on-off of the current. By interrupting the switch tube of the module, the electrical connection between the module and the entire converter system can be cut off to prevent the fault from spreading to other modules. This operation isolates the faulty module from the normal module to avoid instability in other parts of the system caused by the faulty module. After interrupting the topological switch tube of the faulty module, compensation adjustment is performed based on the remaining normal modules to maintain the overall balance and stability of the system. This compensation process relies on collaborative modules, i.e., other normally operating modules in the converter system, which balance the electrical changes caused by the exit of the faulty module through appropriate adjustments. Specifically, the collaborative module balances the power distribution of the entire system by adjusting its output voltage, current and other parameters to ensure that the shutdown of the faulty module will not have a significant impact on the overall performance of the system. In order to achieve this adjustment, it is necessary to rely on the coordinated management of the software control layer. The topological switch tube, as a hardware response element, is responsible for physical on-off control, while the power conversion system and the battery management system, as the adjustment response software layer, jointly adjust the control strategy. The power conversion system is responsible for managing the overall power distribution and conversion of the converter. When a module fails, it ensures the power balance of the system by adjusting the power output of other modules. The battery management system is responsible for managing the charging and discharging process of the energy storage device. When a module fails to work properly, the battery management system adjusts the energy storage management strategy according to the load demand to provide stable power output. The mutual cooperation between the power conversion system and the battery management system, by adjusting the parameters of the response software layer, promptly responds to the hardware operation of the topological switch tube, ensuring that the system can be quickly adjusted and maintain normal operation when a fault occurs.

[0049] In summary, the embodiments of the present application have at least the following technical effects:

[0050] The embodiment of the present application identifies the cascaded series-parallel topology of the converter, sets the diagnosis point cloud distribution based on the series-parallel nodes, and constructs a swing function to perform network control risk map mining in multiple working modes. Then, by analyzing the electrical signal data, combined with the fault diagnosis module, the point cloud data diagnosis decision is made, a fault diagnosis list is generated and a warning is displayed, and an emergency plan is triggered. The diagnosis process involves identifying the series-parallel structure between modules, using kernel principal component analysis to reduce the dimension of nonlinear fault characteristics, and assisting diagnosis through active inertia vector. In the event of a fault, the isolation of the fault module and the guarantee of system stability can be achieved by interrupting the topological switch tube and the balanced complementary adjustment of the collaborative module. These technical effects jointly solve the technical problem that when the complexity of the multi-level converter topology increases, the fault diagnosis cannot be quickly located and the fault tolerance is insufficient, resulting in reduced operational stability of the energy storage system, and realize the technical effect of quickly identifying faults in multiple working modes and improving diagnostic efficiency and system fault tolerance by combining the fault diagnosis point cloud distribution based on the topology structure with the swing function.

[0051] Embodiment 2 is based on the same inventive concept as the energy storage system fault diagnosis method under the multi-level converter topology structure in the above embodiment. Figure 2 As shown, the present application provides a fault diagnosis system for an energy storage system under a multi-level converter topology structure, the system comprising: a topology identification unit 1: the topology identification unit 1 is used to identify the cascaded series-parallel topology of the target converter, wherein the series-parallel topology includes a series-parallel structure between modules and a series-parallel structure of components within the module; a point cloud distribution setting unit 2: the point cloud distribution setting unit 2 is used to traverse the cascaded series-parallel topology and set the diagnosis point cloud distribution, wherein the diagnosis point cloud distribution is determined based on the series-parallel nodes; a swing function construction unit 3: the swing function construction unit 3 is used to construct a swing function, wherein the swing function includes a point cloud distribution based on the topology capacitance. A first swing function constructed by misconfiguration, and a second swing function constructed based on system frequency fluctuations; a graph mining unit 4: the graph mining unit 4 is used to perform network control risk graph mining under multiple working modes based on the diagnosis point cloud distribution and the swing function, and to construct a fault diagnosis module; a diagnosis decision unit 5: the diagnosis decision unit 5 is used to interact with topological electrical signals, and to make point cloud data diagnosis decisions under the target working mode in combination with the fault diagnosis module, to determine a fault diagnosis sheet; an early warning unit 6: the early warning unit 6 is used to display and warn the fault diagnosis sheet on the terminal, and to trigger an emergency plan, which meets the preset risk level.

[0052] Furthermore, the point cloud distribution setting unit 2 is also used to perform the following method:

[0053] With rectification-filtering-inversion-modulation as functions, the functional response nodes of the cascaded series-parallel topology are determined; based on the functional response nodes, a first diagnostic point cloud is constructed, wherein the first diagnostic point cloud is located in the series-parallel structure of components within the module, and the first diagnostic point cloud corresponds one-to-one to the cascade module; based on the series-parallel connection nodes between modules, a second diagnostic point cloud is determined; the first diagnostic point cloud and the second diagnostic point cloud are aggregated to determine the distribution of the diagnostic point cloud.

[0054] Furthermore, the graph mining unit 4 is also used to perform the following method:

[0055] The operation record data of the interactive target converter is clustered based on the working mode to determine N groups of records; a group of records is identified and the spectrum energy distribution and envelope characteristics under the diagnosis point cloud distribution are mined to determine the first point cloud relationship, the first point cloud relationship includes a first electrical data relationship and a first fault relationship, the first fault relationship is identified with a fault feature, including a fault tracing element and a fault type; 4 traverses the N groups of records to determine the Nth point cloud relationship, integrates the first point cloud relationship until the Nth point cloud relationship, and constructs the network control risk map.

[0056] Furthermore, the graph mining unit 4 is also used to perform the following method:

[0057] Traversing the spectrum energy distribution and envelope characteristics, determining linear fault classes and nonlinear fault classes; for the linear fault class, mining fault linear relationships; for the nonlinear fault class, performing nonlinear fault feature dimensionality reduction screening under kernel principal component analysis, mining fault cost relationships, wherein the fault cost relationship is an approximate linear relationship of nonlinear features;

[0058] The first fault relationship is determined based on the fault linear relationship and the fault cost relationship.

[0059] Furthermore, the diagnosis decision unit 5 is also used to execute the following method:

[0060] Identify the topological electrical signal, perform the conversion under the diagnosis point cloud distribution, and determine the point cloud distribution map; determine the target working mode, and trigger the target networking control risk map mapped by the mode; use the swing function as a constraint, combine the target networking control risk map, make a fault diagnosis decision on the point cloud distribution map, and determine the fault diagnosis list.

[0061] Furthermore, the diagnosis decision unit 5 is also used to execute the following method:

[0062] Based on the cascaded series-parallel topology, the active inertia is located, where the active inertia is the converter influencing inertia that affects the dynamic response characteristics and stability; the active inertia vector is determined to assist in fault diagnosis decision-making.

[0063] Furthermore, the early warning unit 6 is also used to execute the following method:

[0064] The modules are relatively independent and cooperate with each other to determine the faulty module and perform on-off management and control complementary adjustment of the topology switch tube; wherein, the on-off management and control complementary adjustment of the topology switch tube include: locating the faulty module based on the fault diagnosis sheet; interrupting the topology switch tube for the faulty module, and performing balanced complementary adjustment based on the collaborative module, wherein the topology switch tube is the response hardware, and the power conversion system and battery management system are the adjustment response software.

[0065] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0066] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0067] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for diagnosing faults in an energy storage system under a multi-level converter topology, characterized in that: Applied to a modular multilevel converter, the method comprises: Identify a cascaded series-parallel topology of a target converter, wherein the series-parallel topology includes an inter-module series-parallel structure and an intra-module component series-parallel structure; Traversing the cascaded series-parallel topology, setting a diagnosis point cloud distribution, wherein the diagnosis point cloud distribution is determined based on the series-parallel nodes; Constructing a swing function, wherein the swing function comprises a first swing function constructed based on a topological fault-tolerant configuration and a second swing function constructed based on a system frequency fluctuation; Based on the diagnosis point cloud distribution and the swing function, network control risk map mining in multiple working modes is performed to build a fault diagnosis module; Interactive topological electrical signals, combined with the fault diagnosis module, make point cloud data diagnosis decisions under the target working mode to determine the fault diagnosis list; The terminal displays and warns the fault diagnosis sheet and triggers an emergency plan, which meets the preset risk level.

2. The method for diagnosing faults of an energy storage system under a multi-level converter topology structure according to claim 1, characterized in that: The setting of the diagnosis point cloud distribution includes: Determine the functional response node of the cascaded series-parallel topology with rectification-filtering-inversion-modulation as functions; Based on the functional response node, construct a first diagnostic point cloud, wherein the first diagnostic point cloud is located in a series-parallel structure of components within the module, and the first diagnostic point cloud corresponds one-to-one to the cascade module; Determine a second diagnostic point cloud based on the serial-parallel connection nodes between the modules; The first diagnosis point cloud and the second diagnosis point cloud are aggregated to determine the distribution of the diagnosis point cloud.

3. The method for diagnosing faults of an energy storage system under a multi-level converter topology structure according to claim 1, characterized in that: Perform network control risk graph mining in multiple working modes, including: The operation record data of the interactive target converter is clustered based on the working mode to determine N groups of records; Identify a group of records and mine the spectrum energy distribution and envelope characteristics under the diagnosis point cloud distribution, determine a first point cloud relationship, the first point cloud relationship includes a first electrical data relationship and a first fault relationship, the first fault relationship is identified with a fault feature, including a fault tracing element and a fault type; The N groups of records are traversed to determine the Nth point cloud relationship, the first point cloud relationship is integrated until the Nth point cloud relationship, and the network control risk map is constructed.

4. The method for diagnosing faults of an energy storage system under a multi-level converter topology structure according to claim 3, characterized in that: Determining a first fault relationship includes: Traversing the spectrum energy distribution and envelope characteristics to determine linear fault types and nonlinear fault types; For the linear fault class, mining the linear relationship of the fault; For the nonlinear fault class, perform dimensionality reduction screening of nonlinear fault features under kernel principal component analysis to mine fault cost relationships, wherein the fault cost relationship is an approximate linear relationship of nonlinear features; The first fault relationship is determined based on the fault linear relationship and the fault cost relationship.

5. The method for diagnosing faults of an energy storage system under a multi-level converter topology structure according to claim 1, characterized in that: The step of determining a fault diagnosis sheet includes: Identify the topological electrical signal, perform conversion under the diagnosis point cloud distribution, and determine a point cloud distribution map; Determine the target working mode and trigger the target network control risk map of the mode mapping; Taking the swing function as a constraint and combining it with the target network control risk map, a fault diagnosis decision is made on the point cloud distribution map to determine the fault diagnosis list.

6. The method for diagnosing faults of an energy storage system under a multi-level converter topology structure according to claim 5, characterized in that: Performing fault diagnosis decision on the point cloud distribution map includes: Based on the cascaded series-parallel topology, positioning active inertia, the active inertia is the converter influencing inertia that affects dynamic response characteristics and stability; Determine the active inertia vector to assist in fault diagnosis decision making.

7. The method for diagnosing faults of an energy storage system under a multi-level converter topology structure according to claim 1, characterized in that: The modules are relatively independent and work together to identify faulty modules and perform on-off management and control complementary regulation of topological switches. Among them, the on-off management and control complementary regulation of the topological switch tube include: Based on the fault diagnosis sheet, locate the faulty module; For the fault module, the topology switch tube is interrupted, and balanced complementary adjustment is performed based on the collaborative module, wherein the topology switch tube is used as the response hardware, and the power conversion system and the battery management system are used as the adjustment response software.

8. A fault diagnosis system for an energy storage system under a multi-level converter topology, characterized in that: The steps for implementing the method for diagnosing a fault in an energy storage system under a multi-level converter topology structure according to any one of claims 1 to 7 include: A topology identification unit: identifying a cascaded series-parallel topology of a target converter, wherein the series-parallel topology includes a series-parallel structure between modules and a series-parallel structure of components within a module; Point cloud distribution setting unit: traverses the cascaded series-parallel topology and sets the diagnosis point cloud distribution, wherein the diagnosis point cloud distribution is determined based on the series-parallel nodes; A swing function construction unit: constructs a swing function, wherein the swing function includes a first swing function constructed based on a topological fault-tolerant configuration and a second swing function constructed based on a system frequency fluctuation; A graph mining unit: based on the diagnosis point cloud distribution and the swing function, performs network control risk graph mining in multiple working modes to build a fault diagnosis module; Diagnosis decision unit: interactive topological electrical signals, combined with the fault diagnosis module to make point cloud data diagnosis decisions under the target working mode, and determine the fault diagnosis list; Early warning unit: The terminal displays and warns the fault diagnosis sheet and triggers an emergency plan, which meets the preset risk level.

Images