Flexible DC converter valve submodule fault positioning method and system

Abstract

The invention discloses a flexible direct current converter valve submodule fault positioning method and system, and the method comprises the steps: obtaining operation data of a multi-level flexible direct current power transmission MMC converter valve after open-circuit faults occur at different submodule positions, carrying out the multi-mode decomposition and reconstruction, and outputting a de-noised multi-mode decomposition and reconstruction signal data set; mapping and converting the multi-mode decomposition reconstruction signal data from an electric signal into a pseudo-color two-dimensional image; training an MMC converter valve submodule fault positioning model according to the pseudo-color two-dimensional image; the method comprises the following steps: acquiring DC side voltage value data of an MMC converter valve in a set period in real time, and carrying out preliminary fault positioning; and according to the MMC converter valve sub-module fault positioning model, obtaining the position of the fault sub-module, and if the position of the sub-module is different from the initial fault positioning, correcting the position of the fault sub-module according to the initial fault positioning and the confidence coefficient thereof. According to the invention, accurate positioning of the open-circuit fault of the IGBT sub-module in the flexible DC power transmission converter valve is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of fault positioning, and more particularly to a flexible direct current (DC) converter valve sub-module fault positioning method and system. BACKGROUND

[0002] DC power transmission technology is an important way to implement the West-to-East power transmission and achieve the "double carbon" target. DC power transmission has the ability of point-to-point, ultra-long distance and large capacity power transmission, and has become one of the effective ways to solve the problem of new energy grid connection and power consumption in China. Modular multilevel converter (MMC) is a very important device in high-voltage DC power transmission system, which plays an important role in AC / DC conversion, maintaining power system stability and improving power quality. It is composed of a plurality of sub-modules with the same structure in cascade, has the advantages of easy expansion in voltage and current, low switching frequency, low cost of redundancy and fault tolerance operation, excellent output waveform, flexible output voltage level, strong fault tolerance, etc. The rapid diagnosis of sub-module fault and the realization of rapid protection play a very important role in the stable operation of MMC-HVDC. Accurate and typical flexible DC converter valve sub-module fault diagnosis or identification detection method is beneficial to reduce the risk of safe and stable operation of flexible DC converter valve fault and power grid.

[0003] CN120009656A discloses a MMC flexible DC converter valve sub-module fault positioning method and device, which belongs to the technical field of DC power transmission. The method consists of two links of fault monitoring and fault positioning. The fault monitoring link relies on the MMC system bridge arm voltage deviation value to monitor whether the system has a fault and locates the fault to the bridge arm level. The fault positioning link constructs a sliding mode observer for each sub-module of the fault bridge arm, estimates the sub-module capacitor voltage value, and compares it with the measured capacitor voltage value to generate a residual sequence, which is then compared with an adaptive threshold to locate the fault to a specific sub-module. However, the comparative document directly locates the sub-module based on the residual threshold, which lacks fault tolerance mechanism for complex working conditions. SUMMARY

[0004] To solve the problems in the prior art, the application provides a flexible DC converter valve sub-module fault positioning method and system.

[0005] The application adopts the following technical solutions.

[0006] The beneficial effects of the present application are that, compared with the prior art, the present application transforms the MMC converter valve sub-module fault de-energizing signal into an image for fault positioning by denoising multi-modal decomposition reconstruction data, taking into account the speed while improving the positioning accuracy and noise immunity of the model, and realizing accurate positioning of the IGBT sub-module open circuit fault inside the flexible DC transmission converter valve; the voltage value based preliminary screening criterion realizes rapid fault detection and preliminary positioning, the results of the model positioning are corrected through preliminary positioning, the detection error under complex working conditions such as harmonics, bridge arm circulating current distortion caused by inrush current and switching frequency is avoided, and the correction combined with confidence evaluation effectively guarantees the reliability of the correction results. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 A flow chart of the flexible DC converter valve sub-module fault positioning method;

[0008] Figure 2 A structure diagram of the fault positioning model.

[0009] Figure 3 A graph of the fault positioning test diagnosis result accuracy rate of the method of the present application under a specific implementation example.

[0010] Figure 4 A comparison of fault positioning accuracy rates under different fault positioning algorithms. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. The embodiments described in the present application are only a part of the embodiments of the present application, not all the embodiments. All other embodiments obtained by those skilled in the art without creative labor based on the spirit of the present application are within the protection scope of the present application.

[0012] As Figure 1 shown, embodiment 1 of the present application provides a flexible DC converter valve sub-module fault positioning method, characterized in that, comprising:

[0013] Obtaining the operating data in a set period after open circuit fault of different sub-modules of the multi-level flexible DC transmission MMC converter valve, performing multi-modal decomposition reconstruction, and outputting a denoised multi-modal decomposition reconstruction signal data set; mapping and converting the multi-modal decomposition reconstruction signal data from an electrical signal to a pseudo-color two-dimensional image;

[0014] Preferably in the present embodiment, the operating data comprises three-phase upper and lower bridge arm current data and three-phase bridge arm voltage data of the MMC converter valve; and the set period is the average duration of the open circuit fault of the MMC converter valve sub-module in the historical data.

[0015] The embodiment preferably, the multi-modal decomposition reconstruction is carried out, and the denoised multi-modal decomposition reconstruction signal data set is output, specifically:

[0016] After the running data of each MMC converter valve when the open circuit fault occurs in different sub-module positions is denoised, it is decomposed into a plurality of modal functions, and a constraint variational model of the original input data is constructed, which is converted into a corresponding augmented Lagrange expression; the update formula of the modal frequency spectrum, the center frequency and the Lagrange multiplier corresponding to each modal function is established to perform repeated cyclic iteration, and when the set convergence criterion is met, the reconstruction result of all the modal functions at present is output as the denoised multi-modal decomposition reconstruction signal data set.

[0017] Specifically, the constraint variational model is constructed:

[0018]

[0019] In the formula, f(t) represents a sequence signal composed of the original input data, t is the time, and the decomposition is K modal functions, wherein the kth modal function is u k (t), and the corresponding center frequency is w k . is the estimated center frequency; δ(t) is an impulse function; represents the two-norm calculation, represents the time derivative of the function behind; j is an imaginary number; the corresponding augmented Lagrange expression L({u k (t)},{w k},λ(t)) is established:

[0020]

[0021] In the formula, λ(t) represents the Lagrange operator; β is a quadratic penalty factor, and the embodiment takes 0.1; <·> represents the inner product operation of the Lagrange operator and the constraint function; and the update formula of the modal frequency spectrum, the center frequency and the Lagrange multiplier is further established:

[0022]

[0023]

[0024] In the formula, f(w) is the frequency spectrum function corresponding to the sequence signal composed of the original input data; is the spectrum function corresponding to the kth modal function of the n+1th iteration, and w is the frequency; are the spectrum functions corresponding to the i th modal function of the n+1th and n th iterations, respectively; are the center frequencies corresponding to the kth modal function of the n+1th and n th iterations, respectively; is the Lagrange operator of the n+1, n iteration; epsilon is a noise tolerance parameter, and 0.01 is taken in the embodiment; is the estimation of the spectrum function corresponding to the sequence signal composed of the original input data, and f(w) is taken in the embodiment;

[0025] In the iterative calculation process, the above calculation formula is repeatedly updated through iteration, until the following convergence criterion is met:

[0026]

[0027] In the formula, is the spectrum function corresponding to the kth modal function of the n+1 iteration; psi is a convergence criterion factor, which is set according to requirements, and 0.01 is taken in the embodiment; the alternating direction multiplier method is used to iteratively solve the minimization problem in the formula, and if the iteration process meets the criterion condition in the formula, the denoised multi-modal decomposition reconstructed signal data set is finally output.

[0028] In the embodiment, the multi-modal decomposition reconstructed signal data is preferably converted from the electrical signal mapping to the pseudo-color two-dimensional image, specifically:

[0029] According to the sampling time corresponding to each sampling point of the operation data when the open-circuit fault occurs at different sub-module positions of the MMC converter valve, the value of the multi-modal decomposition reconstructed signal data corresponding to each sampling point is obtained, and the formula is:

[0030] x = {x1, x2, …, x t …, x T},

[0031] wherein x t is the value of the sampling point corresponding to t, T is the average duration of the open-circuit fault of the sub-module of the MMC converter valve, and x is the set of sampling points.

[0032] The sampling points are divided into a plurality of corresponding non-overlapping number intervals through quantile division, and the formula is:

[0033] y = {y1, y2, …, y l …, y Q}

[0034] wherein y l is the lth non-overlapping number interval, each interval contains a plurality of sampling points, and Q is the total number of non-overlapping number intervals; y is the set of all non-overlapping number intervals.

[0035] A transition probability matrix of Q*Q is constructed, and the element of the qth row and the pth column in the transition probability matrix represents the number of times that a sampling point changes from the qth interval to the pth interval to its adjacent sampling point.

[0036] The transition probability matrix is normalized after being weighted according to the set weight to obtain a normalized weighted transition matrix, which is converted into a Markov transition matrix W of Q*Q; non-overlapping number of interval q g and q h The element of the q g row and the q h column in the normalized weighted transition matrix is the element of the g row and the h column in the Markov transition matrix, the value of the element of the Markov transition matrix is taken as a gray value to form a gray image, and the range of the gray value is divided into multiple sections, and different sections are converted into different colors.

[0037] The MMC converter valve sub-module fault positioning model is trained according to the pseudo-color two-dimensional image;

[0038] Preferably, the MMC converter valve sub-module fault positioning model specifically comprises:

[0039] As shown in Figure 2 , the structure of the MMC converter valve sub-module fault positioning model comprises an input convolutional layer, a channel attention mechanism layer, a grouped convolutional-residual module, a flattening layer, a multi-head attention mechanism layer, a fully connected layer module, a splicing layer and a classification layer, the classification layer outputs the probability of each sub-module position being faulty, and the sub-module position with the maximum probability is the fault position.

[0040] Real-time acquisition of the MMC converter valve DC side voltage value data in a set period, based on the voltage value data and the constructed MMC converter valve sub-module fault identification and positioning preliminary screening criterion formula, to determine whether the MMC converter valve sub-module is faulty and to perform preliminary positioning of the fault, and to calculate the confidence of the preliminary positioning;

[0041] Preferably, the determination of whether the MMC converter valve sub-module is faulty and the preliminary positioning of the fault specifically comprises:

[0042] The formula for determining whether the MMC converter valve sub-module is faulty and the preliminary positioning of the fault is:

[0043]

[0044] In the formula, u φp (t) and u φn (t) are the voltages of the upper bridge arm and the lower bridge arm of the φ phase at time t, φ includes any phase A, B or C; ξ1 and ξ2 are weight factors, both of which are 0.5 in this embodiment; u ci (t) and G irepresents the DC side capacitor voltage value of the i-th converter valve sub-module and the conduction state value of the corresponding IGBT; T is the average duration of the open circuit fault of the MMC converter valve sub-module; N is the number of sub-module units of the bridge arm; and Δu is the deviation between the theoretical value and the actual measured value of the bridge arm voltage.

[0045] The embodiment preferably calculates the confidence of the preliminary fault location, in particular:

[0046] When it is determined that a fault occurs, the confidence C of the preliminary fault location is calculated according to the following formula:

[0047]

[0048] In the formula, δ t is the judgment consistency index at the t-th moment, and is calculated according to the following formula: If the positive and negative of the calculation result are the same as those of Δu, then 1, otherwise 0; α is a set time decay coefficient, and in the embodiment, 0.2 is taken; Φ(·) is a normal cumulative distribution function; and σ is the standard deviation of Δu calculated under normal conditions in historical data. are respectively a set time confidence weight and a fluctuation confidence weight, and in the embodiment, 0.4 and 0.6 are respectively taken.

[0049] The running data of the MMC converter valve in a set period is collected in real time, converted into a pseudo-color two-dimensional image, and then input into the MMC converter valve sub-module fault location model to obtain the sub-module position of the fault. If the sub-module position is different from the preliminary fault location, the sub-module position of the fault is corrected according to the preliminary fault location and the confidence thereof.

[0050] The embodiment preferably corrects the sub-module position of the fault according to the preliminary fault location and the confidence thereof, in particular:

[0051] The MMC converter valve sub-module fault location model outputs the probability of the occurrence of the fault in each sub-module position. When the sub-module position with the maximum probability is not within the range of the preliminary fault location, the maximum value of the probability of the occurrence of the fault in all the sub-module positions in the preliminary fault location is extracted. If the difference between the maximum value of the probability of the occurrence of the fault in all the sub-module positions and the maximum value of the probability of the occurrence of the fault in all the sub-module positions in the preliminary fault location is less than or equal to a set difference threshold value, and the confidence of the preliminary fault location is greater than a set confidence threshold value, then the corrected sub-module position of the fault is the sub-module position with the maximum probability of the occurrence of the fault in the preliminary fault location. Otherwise, the position corresponding to the maximum value of the probability of the occurrence of the fault in all the sub-module positions is still taken as the sub-module position of the fault.

[0052] Preferably, the difference threshold is 0.1 times the maximum value of the probability of all sub-module position sending faults, and the confidence threshold is 0.75.

[0053] Embodiment 2 of the present application proposes a flexible direct current valve sub-module fault positioning system based on the method described in Embodiment 1 of the present application, which comprises an electrical signal mapping module, a fault positioning model construction module, a preliminary positioning module and a final positioning module, specifically:

[0054] The electrical signal mapping module: acquires the operating data in a set period after an open circuit fault occurs at different sub-module positions of a multi-level flexible direct current transmission MMC converter valve, performs multi-modal decomposition and reconstruction, and outputs a denoised multi-modal decomposition and reconstruction signal data set; and converts the multi-modal decomposition and reconstruction signal data from electrical signal mapping to a pseudo-color two-dimensional image.

[0055] The fault positioning model construction module: trains an MMC converter valve sub-module fault positioning model according to the pseudo-color two-dimensional image.

[0056] The preliminary positioning module: acquires the MMC converter valve DC side voltage value data in a set period in real time, judges whether the MMC converter valve sub-module has a fault and performs preliminary positioning of the fault based on the voltage value data and the constructed MMC converter valve sub-module fault identification and positioning preliminary screening criterion, and calculates the confidence of the preliminary positioning.

[0057] The final positioning module: acquires the operating data of the MMC converter valve in a set period in real time, converts the data into a pseudo-color two-dimensional image, inputs the image into the MMC converter valve sub-module fault positioning model, obtains the fault sub-module position, and corrects the fault sub-module position according to the preliminary positioning of the fault and the confidence of the preliminary positioning if the fault sub-module position is different from the preliminary positioning.

[0058] In this embodiment, the historical data of multiple faults of the converter valve sub-module in a certain three-phase nine-level flexible direct current transmission system are collected, and the first 80% of the data are used as training sample data, and the remaining 20% of the data are used as running test data. The test diagnosis result is shown in Figure 3 The figure is a schematic diagram of whether the fault sub-module position is actually faulty or not and the fault sub-module position judged by the present application, wherein the vertical coordinate of the two-dimensional matrix is whether the fault sub-module position is actually faulty or not, and the horizontal coordinate is the fault sub-module position judged by the present application; it can be found that different sub-module fault samples can be correctly judged, the number of misjudgment samples is very small, the accuracy is high, the generalization ability of the model is strong, and it is proved that the method provided by the present application can achieve the expected prediction effect, and the embodiment verifies the effectiveness of the method.

[0059] At the same time, different algorithms VMD-MTF-CNN, VMD-CNN, VMD-Adaboost, VMD-XGBoost, SVM are selected as the algorithm of fault location learning and training link respectively, and the test accuracy of the present application is compared with them. The precision in the machine learning measurement index is taken as the measurement index to judge the model performance, such as Figure 4 As shown in the figure, the effect comparison of the method of the present application and other fault diagnosis algorithms is shown. From Figure 4 It can be seen that the accuracy of the soft direct current valve sub-module fault location based on the method of the present application is generally better than that of other algorithms.

[0060] The present disclosure can be a system, a method and / or a computer program product. The computer program product can include a computer readable storage medium having computer readable program instructions stored therein, which are used to enable a processor to implement various aspects of the present disclosure.

[0061] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application, but not to limit it. Although the present application has been described in detail with reference to the above examples, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered within the protection scope of the claims of the present application.

Claims

1. A fault location method for a flexible DC converter valve submodule, characterized in that, include: The system acquires operational data within a set period after an open-circuit fault occurs at different sub-module positions of the MMC converter valve in a multi-level flexible DC transmission system, performs multi-modal decomposition and reconstruction, and outputs a set of denoised multi-modal decomposition and reconstruction signal data; the system then converts the multi-modal decomposition and reconstruction signal data from electrical signals into pseudo-color two-dimensional images. A fault location model for the MMC converter valve submodule was trained based on pseudo-color two-dimensional images. Real-time acquisition of DC side voltage data of MMC converter valve within a set period; based on the voltage data and the constructed initial screening criterion for fault identification and location of MMC converter valve submodule, determining whether the MMC converter valve submodule has failed and performing preliminary fault location, and calculating the confidence level of the preliminary location. The operating data of the MMC converter valve is collected in real time within a set period. After being converted into a pseudo-color two-dimensional image, it is input into the fault location model of the MMC converter valve submodule to obtain the location of the faulty submodule. If the location of the submodule is different from the initial fault location, the location of the faulty submodule is corrected according to the initial fault location and its confidence level.

2. The method for fault location and identification of a flexible DC converter valve submodule according to claim 1, characterized in that: The operational data includes the three-phase upper and lower arm current data and the three-phase arm voltage data of the MMC converter valve; the set period is the average duration of open-circuit faults in the MMC converter valve submodule in historical data.

3. The method for fault location and identification of a flexible DC converter valve submodule according to claim 1, characterized in that: The process of performing multimodal decomposition and reconstruction, and outputting a denoised multimodal decomposition and reconstruction signal data set, specifically involves: After denoising the operating data of different sub-modules of each MMC converter valve when an open circuit fault occurs, the data is decomposed into multiple modal functions. A constrained variational model of the original input data is constructed and transformed into the corresponding augmented Lagrange expression. An update formula for the modal spectrum, center frequency and Lagrange multiplier corresponding to each modal function is established and iterated repeatedly. When the set convergence criterion is met, the reconstruction results of all current modal functions are output as the set of denoised multimodal decomposition and reconstruction signal data.

4. The method for fault location and identification of a flexible DC converter valve submodule according to claim 1, characterized in that: The process of converting multimodal decomposition and reconstruction signal data from electrical signals into pseudo-color two-dimensional images specifically involves: Based on the sampling time corresponding to each sampling point of the operating data when an open-circuit fault occurs at different submodule locations of the MMC converter valve, the value of each sampling point corresponding to the multimodal decomposition and reconstruction signal data is obtained. These sampling points are divided into multiple non-overlapping intervals using quantiles, with a total of Q non-overlapping intervals. A Q*Q transition probability matrix is ​​constructed, where the element in the q-th row and p-th column of the transition probability matrix represents the number of times a sampling point transitions from the q-th interval to its adjacent sampling point. The transition probability matrix is ​​then weighted according to set weights and normalized to obtain a normalized weighted transition matrix, which is then converted into a Q*Q Markov transition matrix. The non-overlapping intervals q corresponding to the g-th and h-th sampling points of the multimodal decomposition and reconstruction signal data are obtained respectively. g and q h In the normalized weighted transition matrix, the q-th g line q h The elements of the column are the elements in the g-th row and h-th column of the Markov transition matrix. The values ​​of the elements of the Markov transition matrix are used as gray values ​​to form a grayscale image. The range of gray values ​​is divided into multiple segments, and different segments are converted into different colors.

5. The method for fault location and identification of a flexible DC converter valve submodule according to claim 1, characterized in that: The fault location model for the MMC converter valve submodule is as follows: The fault location model structure of the MMC converter valve submodule includes an input convolutional layer, a channel attention mechanism layer, a grouped convolutional-residual module, a flattening layer, a multi-head attention mechanism layer, a fully connected layer module, a splicing layer, and a classification layer. The classification layer outputs the probability of a fault occurring at each submodule location, and the submodule location with the highest probability is the fault location.

6. The method for fault location and identification of a flexible DC converter valve submodule according to claim 2, characterized in that: The process of determining whether the MMC converter valve submodule has malfunctioned and performing preliminary fault location is as follows: The formula for determining whether the MMC converter valve submodule has malfunctioned and for preliminary fault location is as follows: In the formula, u φp (t), u φn (t) represents the voltages at time t of the upper and lower bridge arms of phase φ, respectively, where φ includes any phases A, B, and C; ξ1 and ξ2 are weighting factors; u ci (t) and G i This represents the DC-side capacitor voltage value of the i-th converter valve submodule and the corresponding IGBT conduction state value; T represents the average duration of an open-circuit fault in the MMC converter valve submodule; N represents the number of bridge arm submodule units; Δu is the deviation between the theoretical value and the actual measured value of the bridge arm voltage.

7. The method for fault location and identification of a flexible DC converter valve submodule according to claim 6, characterized in that: The calculation of the confidence level for this preliminary location is specifically as follows: When a fault is determined to have occurred, the confidence level C for preliminary fault location is calculated using the following formula: Where: δ t To calculate the consistency index at time t, the value at time t is... If the sign of the calculated result is the same as the sign of Δu, then it is 1; otherwise, it is 0. α is the set time decay coefficient; Φ(·) is the normal cumulative distribution function; σ is the standard deviation of Δu calculated under normal conditions in historical data. These are the set time confidence weights and volatility confidence weights, respectively.

8. The method for fault location and identification of a flexible DC converter valve submodule according to claim 5, characterized in that: The step of correcting the location of the faulty submodule based on the initial fault location and its confidence level is as follows: The MMC converter valve submodule fault location model outputs the probability of a fault occurring at each submodule location. If the submodule location with the highest probability is not within the initial fault location range, the maximum value among all submodule locations in the initial fault location is extracted. If the difference between the maximum value among all submodule locations and the extracted maximum value among all submodule locations in the initial fault location is less than or equal to a set difference threshold, and the confidence level of the initial fault location is greater than the set confidence level threshold, then the corrected faulty submodule location is the submodule location with the highest probability of a fault occurring in the initial fault location; otherwise, the location corresponding to the maximum value among all submodule locations is still used as the faulty submodule location.

9. The method for fault location and identification of a flexible DC converter valve submodule according to claim 8, characterized in that: The difference threshold is 0.1 times the maximum probability of transmission failure at all submodule locations, and the confidence threshold is 0.

75.

10. A fault location system for a flexible DC-DC converter valve submodule based on the method of any one of claims 1-9, comprising an electrical signal mapping module, a fault location model construction module, a preliminary location module, and a final location module, characterized in that: Electrical signal mapping module: acquires the operating data within a set period after an open circuit fault occurs at different sub-module positions of the MMC converter valve of the multi-level flexible DC transmission, performs multi-modal decomposition and reconstruction, and outputs a set of denoised multi-modal decomposition and reconstruction signal data; converts the multi-modal decomposition and reconstruction signal data from electrical signal mapping into a pseudo-color two-dimensional image; Fault location model construction module: Train the fault location model of the MMC converter valve submodule based on pseudo-color two-dimensional images; Preliminary location module: Real-time acquisition of DC side voltage data of MMC converter valve within a set period; Based on the voltage data and the constructed fault identification and location preliminary screening criteria of MMC converter valve submodule, determine whether the MMC converter valve submodule has failed and perform preliminary fault location, and calculate the confidence level of the preliminary location. Final positioning module: Real-time acquisition of the operating data of the MMC converter valve within a set period, conversion of the data into a pseudo-color two-dimensional image, input into the MMC converter valve sub-module fault positioning model to obtain the location of the faulty sub-module. If the location of the sub-module differs from the initial fault location, the location of the faulty sub-module is corrected based on the initial fault location and its confidence level.

Images