A fault intelligent detection method and system for energy storage system operation and maintenance

Abstract

The present application relates to the technical field of system voltage detection, and specifically to a fault intelligent detection method and system for energy storage system operation and maintenance, the method comprising: obtaining the high-frequency components of the voltage signals of each phase of the energy storage system after performing a layer of wavelet decomposition; determining the relative index of the frequency distribution at each maximum point on the spectrum curve of the high-frequency components; determining the distribution uniformity of the spectrum curve and screening the fault phase according to the distribution range of all maximum points in the spectrum curve and the concentrated distribution of the relative index of the frequency distribution at all maximum points; completing the wavelet decomposition of the voltage signal of the three phases according to the difference in frequency distribution and energy distribution between the fault phase and other phases; obtaining the fault detection result of the energy storage system according to the spectrum curves of all levels of the voltage signal of the three phases in the wavelet decomposition. The present application can improve the fault detection accuracy of the energy storage system operation and maintenance.

Description

Technical Field

[0001] The present application relates to the technical field of system voltage detection, and in particular to a fault intelligent detection method and system for operation and maintenance of an energy storage system. Background Art

[0002] In energy storage systems, voltage detection is a key technology for identifying and diagnosing faults. Generally, voltage faults are mostly caused by short circuits and grounding. When detecting short circuit faults and grounding faults, the voltage signal is usually decomposed step by step using wavelets. Short circuit faults usually introduce high-frequency components into the voltage signal, while grounding faults may appear as changes in low-frequency components, thereby detecting voltage faults.

[0003] The fault characteristics of the voltage signal of the energy storage system are usually weak, and multi-layer decomposition may be required to capture these subtle differences. Although this multi-layer decomposition can improve the accuracy of detection, it also increases the amount of calculation; if the wavelet decomposition level of the voltage is small, the fault characteristics have not been decomposed, and the characteristic difference characteristics of each layer after the current wavelet decomposition cannot be determined, resulting in low detection accuracy. Therefore, there is uncertainty in the level of wavelet decomposition, resulting in low fault detection accuracy in energy storage system operation and maintenance. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a fault intelligent detection method and system for energy storage system operation and maintenance. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a fault intelligent detection method for energy storage system operation and maintenance, the method comprising the following steps:

[0006] S1, obtaining the three-phase voltage signals of the energy storage system;

[0007] S2, obtaining the high-frequency components of the voltage signals of each phase after performing a layer of wavelet decomposition; determining the relative index of the frequency distribution at each maximum point according to the frequency distribution at each maximum point on the spectrum curve of the high-frequency components;

[0008] S3, according to the distribution range of all maximum value points in the spectrum curve, calculate the maximum value frequency ratio of the spectrum curve; according to the maximum value frequency ratio of the spectrum curve and the centralized distribution of the frequency distribution relative index at all maximum value points, determine the distribution uniformity of the spectrum curve; perform threshold analysis on the distribution uniformity of the spectrum curve of all phase voltage signals to screen the fault phase;

[0009] S4, according to the frequency distribution and energy distribution differences between the fault phase and other phases, complete the wavelet decomposition of the three-phase voltage signal; the details are as follows:

[0010] If there is no fault phase, proceed directly to the next decomposition until the preset decomposition level is reached;

[0011] If there is a faulty phase, do the following:

[0012] According to the frequency range difference between the spectrum curve of each fault phase and the spectrum curve of other phases, the spectrum deviation of each fault phase is obtained; according to the energy distribution difference and spectrum deviation of the spectrum curve of each fault phase and the spectrum curve of other phases, the characteristic abnormality index of each fault phase is obtained; according to the threshold analysis results of the characteristic abnormality index of all fault phases, the level of wavelet decomposition is obtained;

[0013] S5, obtaining the fault detection result of the energy storage system according to the frequency spectrum curves of all levels of the three-phase voltage signal in the wavelet decomposition.

[0014] Furthermore, the method of determining the relative index of frequency distribution at each maximum point according to the frequency distribution at each maximum point on the spectrum curve of the high-frequency component includes:

[0015] Obtain each maximum point on the spectrum curve of the high-frequency component; obtain the difference between the frequency at each maximum point and the minimum frequency of the spectrum curve, recorded as the first difference; obtain the difference between the maximum frequency and the minimum frequency of the spectrum curve, recorded as the second difference; and use the ratio of the first difference to the second difference as a relative index of the frequency distribution at each maximum point.

[0016] Furthermore, the method for calculating the maximum frequency ratio of the frequency spectrum curve includes:

[0017] The difference between the maximum frequency and the minimum frequency corresponding to all the maximum points on each spectrum curve is obtained, which is recorded as the third difference; the ratio of the third difference to the frequency range value of each spectrum curve is taken as the maximum frequency ratio of each spectrum curve; the frequency range value of each spectrum curve is the difference between the maximum frequency and the minimum frequency of the spectrum curve.

[0018] Furthermore, the method for obtaining the distribution uniformity of the frequency spectrum curve includes:

[0019]

[0020] in, is the distribution uniformity of the vth spectrum curve; Relative index of frequency distribution of the ith maximum point on the vth spectrum curve; Represents the average value of the relative index of the frequency distribution of all maximum points on the vth spectrum curve; Indicates the number of maximum points on the vth spectrum curve; is the maximum frequency ratio of the vth spectrum curve; softmax() is the softmax normalization function.

[0021] Furthermore, the method for screening fault phases includes: recording the phase corresponding to the frequency spectrum curve whose distribution uniformity is less than a preset fault threshold as the fault phase.

[0022] Furthermore, the method for obtaining the spectrum deviation of each fault phase includes:

[0023] Calculate the intersection of the frequency range of each fault phase and the frequency spectrum curves of other phases; the frequency range of the frequency spectrum curve is an interval with the minimum frequency of the frequency spectrum curve as the lower line and the maximum frequency as the upper line;

[0024] The union of the intersections of each fault phase and all other phases is recorded as the cross-spectrum interval of each fault phase; the difference between the upper limit and the lower line of the cross-spectrum interval is recorded as the range value of the cross-spectrum interval; the ratio between the range value of the cross-spectrum interval and the range value of the union interval of the frequency ranges of all phase spectrum curves is recorded as the proportion of the cross-spectrum interval; the difference between the value 1 and the proportion is used as the spectrum deviation degree of each fault phase.

[0025] Furthermore, the method for obtaining the characteristic abnormality index of each fault phase includes: marking the characteristic abnormality index of the vth fault phase as , ;in, represents the energy entropy of all frequencies in the spectrum curve of the vth fault phase, E represents the mean energy entropy of all frequencies in the spectrum curves of all phases except all fault phases, Indicates the spectrum deviation of the vth fault phase; softmax() is the softmax normalization function.

[0026] Furthermore, the wavelet decomposition level is obtained according to the threshold analysis results of the characteristic abnormal indicators of all fault phases, specifically including:

[0027] If the characteristic abnormal index of any fault phase is greater than the preset fault threshold, the wavelet decomposition is stopped; otherwise, the next decomposition is performed until the preset decomposition level is reached.

[0028] Furthermore, the fault detection result of the energy storage system is obtained according to the spectrum curves of all levels of the three-phase voltage signal in the wavelet decomposition, including:

[0029] The three-phase voltage signals of the energy storage system collected a preset number of times are obtained; the fault conditions of the voltage signals are manually identified as fault signals and non-fault signals as a training set; the SVM model is trained using the training set and the frequency spectrum curves of all levels of all voltage signals in the training set in the wavelet decomposition to obtain a trained SVM model; the trained SVM model is used to detect the three-phase voltage signals to be detected to obtain the fault conditions of the voltage signals.

[0030] In a second aspect, an embodiment of the present application further provides a fault intelligent detection system for operation and maintenance of an energy storage system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of any one of the above-mentioned fault intelligent detection methods for operation and maintenance of an energy storage system are implemented.

[0031] This application has at least the following beneficial effects:

[0032] This application analyzes the high-frequency and low-frequency components of the voltage signal of the energy storage system by using wavelet changes, and then obtains the situation that the level of wavelet decomposition is uncertain during fault detection, resulting in the inability to accurately extract fault features; firstly, the frequency distribution of the high-frequency components of the voltage signal during the wavelet decomposition process is analyzed, and the distribution uniformity of the spectrum curve is calculated according to the fluctuation of the spectrum curve of the high-frequency components, reflecting the fault conditions of each phase, and then screening the fault phase; further analyzes the balanced relationship of the spectrum distribution between the three phases under normal circumstances, and judges whether to extract fault features according to the frequency distribution and energy distribution differences between the fault phase and other phases, and then decides whether to stop the wavelet decomposition to obtain the level of wavelet decomposition. The spectrum curves of each level obtained by decomposition contain the fault characteristics of the voltage signal, which are used to train the sub-model and identify the fault conditions of the voltage signal, which can improve the fault detection accuracy of the energy storage system operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 A flowchart of a method for intelligent fault detection for energy storage system operation and maintenance provided by an embodiment of the present application;

[0035] Figure 2 A block diagram for obtaining the wavelet decomposition levels of a three-phase voltage signal provided by one embodiment of the present application. DETAILED DESCRIPTION

[0036] 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, not all of the embodiments. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0038] The following is a detailed description of a specific scheme of a fault intelligent detection method and system for energy storage system operation and maintenance provided by the present application in conjunction with the accompanying drawings.

[0039] See also Figure 1 , which shows a flowchart of a method for intelligent fault detection for energy storage system operation and maintenance provided by an embodiment of the present application, the method comprising the following steps:

[0040] S1, obtain the three-phase voltage signals of the energy storage system.

[0041] The specific implementation scenario of this application is the fault detection scenario of the energy storage system; when the wavelet decomposition method is usually used for fault processing, a fixed level is determined, and each voltage signal is segmented and decomposed at multiple levels. When the difference between the voltage signals of the short circuit fault and the ground fault is obvious, multi-level wavelet decomposition will increase the amount of calculation; if the difference between the voltage signals of the two is not obvious, the fixed-level wavelet decomposition cannot extract a large feature difference, so it is necessary to extract different levels of wavelet decomposition according to different signal differences, which can reduce the amount of calculation while obtaining signal features.

[0042] Most power grid systems are designed based on three-phase alternating current, and the energy storage system needs to be compatible with the existing power grid. Therefore, the energy storage system converts direct current (DC) into three-phase alternating current through an inverter to achieve grid-connected operation. The real-time monitoring system equipped with the energy storage system records and stores the three-phase voltage signals of the current energy storage system. This embodiment performs wavelet decomposition on the three-phase voltage signals, and determines the current wavelet decomposition level according to the frequency distribution of the signal after wavelet decomposition.

[0043] When the three phases of voltage are decomposed by wavelet, the frequency spectrum distribution of the three phases of voltage is different when they are faulty and when they are not faulty. In case of bidirectional grounding fault, the fault phase may be manifested as a change in low-frequency components, and short circuit usually introduces high-frequency components into the voltage signal. Therefore, the fault phase of the circuit with the current fault can be determined according to the frequency distribution of the frequency spectrum curve of the detail coefficient after each layer of wavelet decomposition, and the abnormal characteristics of the fault phase of the circuit can be determined. If the abnormal characteristics meet the wavelet decomposition conditions, the wavelet needs to be further decomposed to determine the final decomposition level of the wavelet.

[0044] S2, obtaining the high-frequency components of the voltage signals of each phase after performing a layer of wavelet decomposition; determining the relative index of the frequency distribution at each maximum point according to the frequency distribution at each maximum point on the spectrum curve of the high-frequency components.

[0045] Wavelet decomposition can decompose the voltage signal into signals at different scales, each scale corresponding to a different frequency range. This process is carried out step by step, and each step involves filtering and downsampling operations on the signal. Therefore, this embodiment processes the wavelet frequency situation after each layer of decomposition, and determines that the current decomposed wavelet signal meets the extraction requirements of the wavelet signal features of fault detection. Then the wavelet decomposition of the voltage curve can be stopped, and the maximum level of the wavelet decomposition of the current voltage is determined, and the characteristics of the wavelet decomposition are obtained.

[0046] First, a wavelet hierarchical decomposition tree of the voltage curve is established. In this embodiment, the Moret wavelet function is selected as the wavelet basis of the wavelet change. The voltage signal of each phase is decomposed by wavelet to obtain the high-frequency component of the signal, namely the detail coefficient and the low-frequency signal, namely the approximate coefficient.

[0047] In case of a bidirectional ground fault, the fault phase may be manifested as a change in low-frequency components, while a short-circuit fault causes a sharp drop in circuit impedance, resulting in a sharp increase in current, which generates high-frequency components that usually introduce high-frequency components into the voltage signal. The fault phase and the non-fault phase can be distinguished based on the frequency distribution of the high-frequency components of the voltage signal of different phases, and the abnormal characteristics of the current voltage signal can be determined based on the energy change difference between the fault phase and the non-fault phase.

[0048] Specifically, a spectrum curve is obtained after wavelet transformation of the high-frequency components of the voltage signal of each phase; the horizontal axis of the spectrum curve is frequency, and the vertical axis is energy intensity; each maximum point on the spectrum curve is obtained, if the frequency of the maximum point is at a relatively high frequency, it means that a short circuit fault may occur in the current circuit, if the frequency of the maximum point is in a relatively low frequency range, it means that a ground fault may occur in the current circuit.

[0049] For any maximum point, determine the relative index of frequency distribution according to the frequency distribution at the maximum point. The relative index of frequency distribution at the i-th maximum point is The calculation formula is:

[0050]

[0051] in, Indicates the maximum frequency of the spectrum curve, represents the minimum frequency of the frequency curve, Represents the frequency at the i-th maximum point.

[0052] in, Represents the frequency range of the spectrum curve, It represents the relative distribution index of the frequency of the current i-th maximum value in the current frequency range. If the value is close to 0, it means that a short circuit fault has occurred in the current circuit and the local maximum value is in a relatively low frequency range. If the value is close to 1, it means that a ground fault has occurred in the current circuit and the local maximum value is in a relatively large frequency range.

[0053] As another embodiment of the present application, the relative index of the frequency distribution at the i-th maximum point is Can be used directly It is used to analyze the relative size of the frequency at the maximum point.

[0054] S3, calculate the maximum frequency ratio of the spectrum curve according to the distribution range of all maximum points in the spectrum curve; determine the distribution uniformity of the spectrum curve according to the maximum frequency ratio of the spectrum curve and the centralized distribution of the frequency distribution relative index at all maximum points; perform threshold analysis on the distribution uniformity of the spectrum curve of all phase voltage signals to screen the fault phase.

[0055] When there is a fault, in the frequency spectrum curve of the high-frequency component of the voltage signal, the frequency distribution of the fault phase is relatively concentrated in certain specific frequency bands, and the frequency distribution of the non-fault phase is relatively uniform. Therefore, according to the distribution range of all maximum points in the frequency spectrum curve, the maximum frequency ratio of the frequency spectrum curve is calculated; specifically: the maximum frequency ratio of the vth frequency spectrum curve is recorded as , ;in, Represents the maximum frequency corresponding to all maximum points on the vth spectrum curve, Indicates the minimum frequency corresponding to all maximum points on the vth spectrum curve, Represents the frequency range of the vth spectrum curve. Among them, Indicates the maximum spectrum range of the spectrum curve, Indicates the proportion of the spectrum range with the maximum value in the current spectrum range. The larger the value, the more uniform the frequency distribution.

[0056] Further, according to the maximum frequency ratio of the spectrum curve and the centralized distribution of the frequency distribution relative index at all maximum points, the distribution uniformity of the spectrum curve is determined. The calculation formula is:

[0057]

[0058] in, is the distribution uniformity of the vth spectrum curve; Relative index of frequency distribution of the ith maximum point on the vth spectrum curve; Represents the average value of the relative index of the frequency distribution of all maximum points on the vth spectrum curve; Indicates the number of maximum points on the vth spectrum curve; is the maximum frequency ratio of the vth spectrum curve; softmax() is the softmax normalization function.

[0059] In the above formula, It indicates the concentration distribution of the frequency distribution at all maximum points relative to the index. The smaller the value is, the more stable the frequency distribution of the spectrum curve is. The larger the proportion of the maximum frequency of the spectrum curve is, the more uniform the frequency distribution is, and the greater the distribution uniformity of the spectrum curve is.

[0060] The fault threshold is determined to be 0.4 according to the distribution uniformity of the spectrum curve. If the distribution uniformity of the spectrum curve of the current phase is less than the fault threshold, it means that the current phase is a faulty phase.

[0061] S4, according to the frequency distribution and energy distribution differences between the fault phase and other phases, complete the wavelet decomposition of the three-phase voltage signals.

[0062] Since the spectrum distribution between the three phases is balanced under normal circumstances, the distribution of the spectrum after wavelet transformation is also close. If there is a fault in a phase, the distribution of its spectrum will change. After determining the fault phase, the spectrum deviation of the fault phase can be determined according to the distribution of the spectrum of the current three phases.

[0063] If there is no fault phase, proceed directly to the next decomposition until the preset decomposition level is reached.

[0064] If there is a faulty phase, do the following:

[0065] According to the frequency range difference between the spectrum curve of each fault phase and the spectrum curve of other phases, the spectrum deviation of each fault phase is obtained; according to the energy distribution difference and spectrum deviation of the spectrum curve of each fault phase and the spectrum curve of other phases, the characteristic abnormality index of each fault phase is obtained; according to the threshold analysis results of the characteristic abnormality indexes of all fault phases, the level of wavelet decomposition is obtained.

[0066] Specifically, the intersection of the frequency range of each fault phase and the frequency spectrum curve of other phases is calculated; the frequency range of the spectrum curve is an interval with the minimum frequency of the spectrum curve as the lower line and the maximum frequency as the upper limit; the union of the intersection of each fault phase and all other phases is recorded as the cross-spectrum interval of each fault phase; the difference between the upper limit and the lower line of the cross-spectrum interval is recorded as the range value of the cross-spectrum interval; the spectrum deviation calculation formula of the fault phase is:

[0067]

[0068] in, Indicates the spectrum deviation of the vth fault phase; Indicates the range of the cross-spectrum interval of the vth fault phase, Represents the range value of the union interval of the frequency range of all phase spectrum curves, is the proportion of the cross-spectrum interval. The closer the value is to 1, the higher the overlap of the spectrum distribution between the non-fault phase and the spectrum interval of the current fault phase is, and the smaller the deviation is.

[0069] If the maximum value of the modulus of the spectrum of the fault phase is on both sides, and the normal spectrum distribution in the middle is relatively low, it cannot be determined by using the offset of the fault phase, so the distribution of the current wavelet spectrum in each spectrum interval can be determined based on the energy entropy of the current fault.

[0070] The energy entropy of the fault phase will increase significantly, while the energy entropy of other non-fault phases will decrease relatively or remain unchanged. The greater the difference in energy entropy between the non-fault phase and the fault phase, the more obvious the current voltage abnormality. Therefore, this embodiment obtains the characteristic abnormality index of each fault phase based on the energy distribution difference and spectrum deviation between the spectrum curve of each fault phase and the spectrum curve of other phases:

[0071]

[0072] in, is the characteristic abnormality indicator of the vth fault phase; represents the energy entropy of all frequencies in the spectrum curve of the vth fault phase, E represents the mean energy entropy of all frequencies in the spectrum curves of all phases except all fault phases, Indicates the spectrum deviation of the vth fault phase; softmax() is the softmax normalization function.

[0073] To determine the level of the current wavelet decomposition, the initial value of the level is 0; if there is no fault phase in the current wavelet decomposition, the next wavelet decomposition is directly performed, and the level is increased by one. If the preset decomposition level 10 is reached, the wavelet decomposition is stopped; if the level is less than or equal to 10, the wavelet decomposition is continued and the process goes to step S2;

[0074] If there is a fault phase in the current wavelet decomposition, determine whether the characteristic abnormality index of each fault phase is greater than the fault threshold 0.7. If it is greater than the threshold, stop the wavelet decomposition; otherwise, perform the next decomposition until the preset decomposition level is reached. After completing the wavelet decomposition of the three-phase circuit of the current voltage, the wavelet hierarchical decomposition tree is obtained.

[0075] The block diagram of the acquisition of the wavelet decomposition level of the three-phase voltage signal is as follows: Figure 2 shown.

[0076] S5, obtaining the fault detection result of the energy storage system according to the frequency spectrum curves of all levels of the three-phase voltage signal in the wavelet decomposition.

[0077] According to the above, determine the voltage of the current three-phase circuit, and determine the spectrum curve of the wavelet change of the three-phase circuit voltage at each level. Use manual identification of the three-phase circuit of the current voltage and the voltage fault condition as fault signals and non-fault signals, and establish a sample set containing 2000 samples in total. 80% is used as a training set and 20% is used as a test set. Use the training set and the spectrum curves of all voltage signals in the training set at all levels in the wavelet decomposition to train the SVM model to obtain a trained SVM model; use the trained SVM model to detect the voltage signal of the three phases to be detected and obtain the fault condition of the voltage signal.

[0078] Based on the same inventive concept as the above method, an embodiment of the present application also provides a fault intelligent detection system for operation and maintenance of an energy storage system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of any one of the above-mentioned fault intelligent detection methods for operation and maintenance of an energy storage system are implemented.

[0079] Through the above description of the implementation method in combination with the accompanying drawings, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0080] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A fault intelligent detection method for energy storage system operation and maintenance, characterized in that: The method comprises the following steps: S1, obtaining the three-phase voltage signals of the energy storage system; S2, obtaining the high-frequency components of the voltage signals of each phase after performing a layer of wavelet decomposition; determining the relative index of the frequency distribution at each maximum point according to the frequency distribution at each maximum point on the spectrum curve of the high-frequency components; S3, according to the distribution range of all maximum value points in the spectrum curve, calculate the maximum value frequency ratio of the spectrum curve; according to the maximum value frequency ratio of the spectrum curve and the centralized distribution of the frequency distribution relative index at all maximum value points, determine the distribution uniformity of the spectrum curve; perform threshold analysis on the distribution uniformity of the spectrum curve of all phase voltage signals to screen the fault phase; S4, according to the frequency distribution and energy distribution differences between the fault phase and other phases, complete the wavelet decomposition of the three-phase voltage signal; the details are as follows: If there is no fault phase, proceed directly to the next decomposition until the preset decomposition level is reached; If there is a faulty phase, do the following: According to the frequency range difference between the spectrum curve of each fault phase and the spectrum curve of other phases, the spectrum deviation of each fault phase is obtained; according to the energy distribution difference and spectrum deviation of the spectrum curve of each fault phase and the spectrum curve of other phases, the characteristic abnormality index of each fault phase is obtained; according to the threshold analysis results of the characteristic abnormality index of all fault phases, the level of wavelet decomposition is obtained; S5, obtaining the fault detection result of the energy storage system according to the frequency spectrum curves of all levels of the three-phase voltage signals in the wavelet decomposition; Determining the relative index of the frequency distribution at each maximum point according to the frequency distribution at each maximum point on the spectrum curve of the high-frequency component includes: Obtain each maximum point on the spectrum curve of the high-frequency component; obtain the difference between the frequency at each maximum point and the minimum frequency of the spectrum curve, recorded as the first difference; obtain the difference between the maximum frequency and the minimum frequency of the spectrum curve, recorded as the second difference; and use the ratio of the first difference to the second difference as a relative index of the frequency distribution at each maximum point.

2. A fault intelligent detection method for energy storage system operation and maintenance according to claim 1, characterized in that: The method for calculating the maximum frequency ratio of the spectrum curve includes: The difference between the maximum frequency and the minimum frequency corresponding to all the maximum points on each spectrum curve is obtained, which is recorded as the third difference; the ratio of the third difference to the frequency range value of each spectrum curve is taken as the maximum frequency ratio of each spectrum curve; the frequency range value of each spectrum curve is the difference between the maximum frequency and the minimum frequency of the spectrum curve.

3. A fault intelligent detection method for energy storage system operation and maintenance according to claim 1, characterized in that: The method for obtaining the distribution uniformity of the frequency spectrum curve comprises: in, is the distribution uniformity of the vth spectrum curve; Relative index of frequency distribution of the ith maximum point on the vth spectrum curve; Represents the average value of the relative index of the frequency distribution of all maximum points on the vth spectrum curve; Indicates the number of maximum points on the vth spectrum curve; is the maximum frequency ratio of the vth spectrum curve; softmax() is the softmax normalization function.

4. A fault intelligent detection method for energy storage system operation and maintenance according to claim 1, characterized in that: The method for screening fault phases includes: recording the phase corresponding to the frequency spectrum curve whose distribution uniformity is less than a preset fault threshold as the fault phase.

5. A fault intelligent detection method for energy storage system operation and maintenance as claimed in claim 2, characterized in that: The method for obtaining the spectrum deviation of each fault phase includes: Calculate the intersection of the frequency range of each fault phase and the frequency spectrum curves of other phases; the frequency range of the frequency spectrum curve is an interval with the minimum frequency of the frequency spectrum curve as the lower line and the maximum frequency as the upper line; The union of the intersections of each fault phase and all other phases is recorded as the cross-spectrum interval of each fault phase; the difference between the upper limit and the lower line of the cross-spectrum interval is recorded as the range value of the cross-spectrum interval; the ratio between the range value of the cross-spectrum interval and the range value of the union interval of the frequency ranges of all phase spectrum curves is recorded as the proportion of the cross-spectrum interval; the difference between the value 1 and the proportion is used as the spectrum deviation degree of each fault phase.

6. A fault intelligent detection method for energy storage system operation and maintenance according to claim 1, characterized in that: The method for obtaining the characteristic abnormality index of each fault phase comprises: marking the characteristic abnormality index of the vth fault phase as , ;in, represents the energy entropy of all frequencies in the spectrum curve of the vth fault phase, E represents the mean energy entropy of all frequencies in the spectrum curves of all phases except all fault phases, Indicates the spectrum deviation of the vth fault phase; softmax() is the softmax normalization function.

7. The intelligent fault detection method for energy storage system operation and maintenance according to claim 1, characterized in that: The wavelet decomposition level is obtained according to the threshold analysis results of the characteristic abnormal indicators of all fault phases, specifically including: If the characteristic abnormal index of any fault phase is greater than the preset fault threshold, the wavelet decomposition is stopped; otherwise, the next decomposition is performed until the preset decomposition level is reached.

8. The intelligent fault detection method for energy storage system operation and maintenance according to claim 1, characterized in that: The fault detection result of the energy storage system is obtained according to the frequency spectrum curves of all levels of the three-phase voltage signal in the wavelet decomposition, including: The three-phase voltage signals of the energy storage system collected a preset number of times are obtained; the fault conditions of the voltage signals are manually identified as fault signals and non-fault signals as a training set; the SVM model is trained using the training set and the frequency spectrum curves of all levels of all voltage signals in the training set in the wavelet decomposition to obtain a trained SVM model; the trained SVM model is used to detect the three-phase voltage signals to be detected to obtain the fault conditions of the voltage signals.

9. A fault intelligent detection system for operation and maintenance of an energy storage system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the intelligent fault detection method for energy storage system operation and maintenance as described in any one of claims 1-8 are implemented.

Images