A harmonic fault diagnosis method, medium and system for a distributed battery energy storage power station

Abstract

The present invention provides a harmonic fault diagnosis method, medium and system for a distributed battery energy storage power station, belonging to the technical field of distributed battery energy storage power stations, comprising: acquiring electrical parameters output by each inverter of the distributed battery energy storage power station in real time, and establishing an inverter tree; traversing the inverter tree to obtain the sequence number of each inverter; establishing an electrical two-dimensional grid map, converting the electrical parameters collected in real time into RGB color values, and filling the two-dimensional grid map to obtain a two-dimensional color map; acquiring the electrical parameters output by each inverter of the historical harmonic fault record, and the fault cause corresponding to the historical harmonic fault record, establishing a two-dimensional color map corresponding to each harmonic fault, recorded as a comparison two-dimensional map; establishing a sliding window, using the comparison two-dimensional map corresponding to each harmonic fault to perform similarity matching on the two-dimensional color map, selecting the harmonic fault corresponding to the comparison two-dimensional map with the highest similarity as the harmonic fault cause of the distributed battery energy storage power station, and outputting it.

Description

Technical Field

[0001] The present invention belongs to the technical field of distributed battery energy storage power stations, and in particular, relates to a harmonic fault diagnosis method, medium and system for distributed battery energy storage power stations. Background Art

[0002] In recent years, large-scale distributed battery energy storage power stations have been widely used and promoted around the world. However, distributed battery energy storage power stations also face many technical challenges in the process of connecting to the power grid, such as power fluctuations, harmonic defects, and power quality degradation, which brings new problems to the stable operation of the power system. Harmonics refer to other frequency components other than the fundamental wave in the power system, which are usually generated by nonlinear electrical equipment, resulting in additional energy loss and equipment heating. In distributed battery energy storage power stations, various harmonic defects are easily generated due to the use of a large number of inverters to convert DC power into AC power and connect to the grid for output. These harmonics not only affect the safety of distributed battery energy storage power stations, but also spread to the entire power grid through transmission lines, interfering with other electrical equipment and causing a larger range of energy waste. Therefore, timely detection and diagnosis of harmonic faults in distributed battery energy storage power stations is of great significance for improving system operation efficiency and reducing operation and maintenance costs.

[0003] Traditional harmonic detection and diagnosis methods mainly rely on manual analysis of monitoring data, which has the following prominent problems:

[0004] 1. Low work efficiency. Distributed battery energy storage power stations have many devices, and the massive monitoring data is difficult to process manually. Manual analysis is time-consuming and labor-intensive, with slow response speed, and it is difficult to achieve real-time online diagnosis.

[0005] 2. Poor reliability and high misdiagnosis rate. Harmonics are diverse and complex, and many hidden defects are not easy to detect. Manual analysis often has subjective biases and inconsistent judgment standards, which often lead to missed diagnoses and misdiagnoses. Summary of the invention

[0006] In view of this, the present invention provides a distributed battery energy storage power station harmonic fault diagnosis method, medium and system, which can solve the technical problems of low work efficiency, poor reliability and high misdiagnosis rate in the prior art when diagnosing harmonic faults in distributed battery energy storage power stations.

[0007] The present invention is achieved in that:

[0008] A first aspect of the present invention provides a method for diagnosing harmonic faults in a distributed battery energy storage power station, comprising the following steps:

[0009] S10, obtaining in real time the electrical parameters output by each inverter of the distributed battery energy storage power station, including current, voltage, and frequency;

[0010] S20, establishing an inverter tree according to the inverter topology diagram of the energy storage power station;

[0011] S30, traversing the inverter tree, and obtaining a sequence number of each inverter according to the traversal order;

[0012] S40, establishing an electrical two-dimensional grid diagram, wherein the abscissa of the electrical two-dimensional grid diagram is time, the ordinate is the inverter number, and the duration of each grid is one acquisition cycle;

[0013] S50, converting the electrical parameters collected in real time into RGB color values, and filling the two-dimensional grid map to obtain a two-dimensional color map;

[0014] S60, obtaining electrical parameters output by each inverter in the historical harmonic fault records and the fault causes corresponding to the historical harmonic fault records, and establishing a two-dimensional color map corresponding to each harmonic fault, recorded as a comparative two-dimensional map;

[0015] S70, establishing a sliding window, using the comparative two-dimensional graph corresponding to each harmonic fault to perform similarity matching on the two-dimensional color graph, selecting the harmonic fault corresponding to the comparative two-dimensional graph with the highest similarity as the harmonic fault cause of the distributed battery energy storage power station, and outputting it to the operator.

[0016] On the basis of the above technical solution, the harmonic fault diagnosis method of a distributed battery energy storage power station of the present invention can also be improved as follows:

[0017] The step of obtaining the electrical parameters output by each inverter of the distributed battery energy storage power station in real time specifically includes:

[0018] First, the original sampling data of the electrical parameters at the output end of each inverter is obtained;

[0019] Then, the original sampled data is preprocessed, including format unification, removal of outliers and noise, and interpolation of missing values.

[0020] The method of traversing the inverter tree is pre-order traversal, in-order traversal or post-order traversal.

[0021] Wherein, the acquisition cycle is 0.001 seconds to 0.005 seconds.

[0022] Among them, the step of converting the electrical parameters collected in real time into RGB color values ​​is specifically: normalizing the voltage value, current value and frequency value of the electrical parameters to the range of 0-1 respectively, and then using the three normalized values ​​as the RGB three primary color values.

[0023] Among them, the steps of obtaining the electrical parameters of each inverter output of the historical harmonic fault records, and the fault causes corresponding to the historical harmonic fault records, and establishing a two-dimensional color map corresponding to each harmonic fault, specifically include: establishing a harmonic fault sample library, extracting relevant electrical parameter data and fault cause descriptions of various harmonic faults from the historical fault data; then preprocessing the relevant electrical parameter data, and constructing a two-dimensional color image based on the preprocessed relevant electrical parameter data as a comparative two-dimensional map corresponding to various harmonic faults.

[0024] Among them, the step of establishing a sliding window and using the comparative two-dimensional image corresponding to each harmonic fault to perform similarity matching on the two-dimensional color image is specifically: constructing a sliding window, and the sliding window will slide on the two-dimensional color image with a fixed step size; extracting the image to be matched in the sliding window; traversing the fault sample library established in step S60, and calculating the similarity score between each comparative image and the image to be matched, and the similarity score is calculated based on the pixel color distance and weight.

[0025] Furthermore, in the step of calculating the similarity score between each comparison image and the image to be matched, the fault delay phenomenon is taken into account, and the image to be matched is matched with the sample library comparison image at different time offsets; the one with the highest similarity score with the image to be matched is selected from all the comparison images, and the corresponding fault cause is output as the diagnosis result.

[0026] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned method for diagnosing harmonic faults in a distributed battery energy storage power station.

[0027] A third aspect of the present invention provides a distributed battery energy storage power station harmonic fault diagnosis system, which includes the above-mentioned computer-readable storage medium.

[0028] Compared with the prior art, the beneficial effects of the harmonic fault diagnosis method, medium and system of a distributed battery energy storage power station provided by the present invention are:

[0029] 1. Convert electrical parameter data into intuitive two-dimensional color images, and use image processing and computer vision technology to identify faults, greatly improving diagnostic efficiency and accuracy.

[0030] 2. For similar faults, clustering and image merging strategies are used to generate the corresponding total image, effectively extracting the common features of the faults. Then, based on the segmentation algorithm, the total image is further disassembled according to the potential texture differences to mine more hidden detail features.

[0031] 3. For each type of fault image, a corresponding "sample library comparison image" is constructed. Based on the idea of ​​image similarity matching, the real-time detection data is matched and compared with the comparison image in the sample library, so as to quickly and accurately identify the fault type and its cause.

[0032] 4. Considering that there is a certain time delay in faults, a sliding window and time series alignment mechanism are introduced to enable the diagnosis system to tolerate and find deviations in the time of fault occurrence.

[0033] 5. An innovative similarity measurement method is adopted, which integrates the two dimensions of pixel color distance and position weight, and fully considers image differences when scoring.

[0034] In summary, the solution of the present invention solves the technical problems of low working efficiency, poor reliability and high misdiagnosis rate in the prior art when diagnosing harmonic faults in distributed battery energy storage power stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0036] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0038] like Figure 1 As shown, it is a flow chart of a method for diagnosing harmonic faults in a distributed battery energy storage power station provided by the present invention. The method comprises the following steps:

[0039] S10, obtaining in real time the electrical parameters output by each inverter of the distributed battery energy storage power station, including current, voltage, and frequency;

[0040] S20, establishing an inverter tree according to the inverter topology diagram of the energy storage power station;

[0041] S30, traversing the inverter tree, and obtaining the sequence number of each inverter according to the traversal order;

[0042] S40, establishing an electrical two-dimensional grid diagram, wherein the abscissa of the electrical two-dimensional grid diagram is time, and the ordinate is the inverter number, wherein the duration of each grid is one acquisition cycle;

[0043] S50, converting the electrical parameters collected in real time into RGB color values, and filling the two-dimensional grid map to obtain a two-dimensional color map;

[0044] S60, obtaining electrical parameters of each inverter output in the historical harmonic fault record and the fault cause corresponding to the historical harmonic fault record, and establishing a two-dimensional color map corresponding to each harmonic fault, recorded as a comparative two-dimensional map;

[0045] S70, establish a sliding window, use the comparative two-dimensional graph corresponding to each harmonic fault to perform similarity matching on the two-dimensional color graph, select the harmonic fault corresponding to the comparative two-dimensional graph with the highest similarity as the harmonic fault cause of the distributed battery energy storage power station, and output it to the operator.

[0046] The specific implementation methods of the above steps are described in detail below:

[0047] The specific implementation of step S10 is as follows:

[0048] First, it is necessary to install current, voltage, frequency and other sensors on all inverters in the distributed battery energy storage power station. These sensors can collect the electrical parameters output by the inverter in real time. There are many types of sensors. For example, the sensors used to measure current mainly include current transformers, current vector converters, open-loop and closed-loop Hall current sensors, etc. The sensors for measuring voltage include resistor dividers, capacitor dividers, and mutual inductance dividers. These sensors have different measurement ranges, accuracy and prices. When selecting sensors, you need to weigh them based on actual needs and cost budgets.

[0049] Secondly, all sensors need to be connected to the computer system of the power station, and the collected data should be transmitted to the computer in real time by wired or wireless means. The main wired connection methods are RS-485, CAN, Ethernet, etc.; wireless connection methods include Bluetooth, WiFi, 4G / 5G, etc. Each connection method has advantages and disadvantages, and the most appropriate one needs to be selected according to the specific situation. For example, if the installation location is close and does not need to be moved frequently, the wired connection is more reliable and stable; if the installation location is relatively scattered and moved frequently, the wireless connection is more flexible and convenient.

[0050] After receiving the sensor data, the computer system needs to preprocess it, including removing outliers and interpolating missing data. The outliers can be processed by algorithms such as median filtering and Kalman filtering. If the data is missing, linear interpolation, polynomial interpolation, spline interpolation and other methods can be used for interpolation. The preprocessed data is then stored for subsequent analysis.

[0051] The specific implementation of step S20 is:

[0052] The inverter tree is established according to the actual wiring topology of the energy storage power station. The busbar or collector line can be used as the root node of the tree, and each inverter string can be used as a child node. The structure of the tree can be drawn according to the actual series and parallel relationship. Each node contains the basic information of the string / group inverter, such as model, rated power, installation location, etc.

[0053] The inverter tree can be drawn manually or generated with the help of special software tools. If you use software that draws logic trees, you only need to enter the basic information of the inverter and the connection relationship to quickly generate it. You can also use general drawing software or programming language code to draw it yourself. For example, using Python's NetworkX package, you can quickly create and visualize a graphical network.

[0054] The inverter tree is used to visually display the topology of the entire power station, making it easy to retrieve and process the electrical parameters of each inverter. The tree structure is easy to traverse and access according to the hierarchical relationship, so as to analyze and locate the origin of harmonic faults.

[0055] The specific implementation of step S30 is:

[0056] The established inverter tree is traversed, and a sequence number is assigned to each traversed inverter node.

[0057] The algorithms for traversing the tree include pre-order traversal, in-order traversal, post-order traversal, and level-order traversal. Different traversal orders will affect the final inverter numbering, but have no effect on the effectiveness of the method of the present invention. Generally, pre-order traversal or level-order traversal can be used. Both start traversal from the root node, and the resulting sequence numbers are more intuitive and continuous. For example, for a complete binary tree, the pre-order traversal sequence numbers are 1, 2, 4, 5, 3, 6, 7; the level-order traversal sequence numbers are 1, 2, 3, 4, 5, 6, 7.

[0058] Take the pre-order traversal algorithm in Python as an example:

[0059] def preorder(root):

[0060] global node_seq

[0061] if root:

[0062] node_seq+=1

[0063] root.no=node_seq

[0064] preorder(root.left)

[0065] preorder(root.right)

[0066] Where root is the root node of the tree, node_seq is the sequence counter, and root.no is the sequence number of the current node.

[0067] The reason for encoding the inverter serial number in this step is mainly to facilitate its mapping to the ordinate of the subsequent two-dimensional grid diagram. With the inverter serial number as the ordinate and time as the abscissa, the electrical parameters of each inverter can be clearly visualized in the time dimension.

[0068] The specific implementation of step S40 is:

[0069] Create a two-dimensional grid diagram with time and inverter number as coordinate axes.

[0070] The horizontal axis time is the time axis. Set a reasonable time range, such as the last 24 hours. Divide the time range into several equal time periods as the scale marks of the time axis. For example, divide 24 hours into 96 time periods, each lasting 15 minutes. The time period corresponding to the current time is the rightmost end of the horizontal axis on the two-dimensional grid chart.

[0071] The vertical axis no is the inverter serial number obtained in the previous step. The serial number ranges from 1 to n, where n is the total number of inverters in the power station. According to the actual situation, the serial number range can be divided into multiple intervals, and the vertical axis scale spacing can be reduced to make the image display effect better.

[0072] In the two-dimensional plane formed by the time axis and the serial number axis, horizontal and vertical grid lines are drawn at equal intervals, and the intersection of the grid lines is the grid unit. Each grid unit represents the change of the electrical parameters of a certain inverter in a time period.

[0073] The size of the grid needs to be set according to the actual situation. If the time axis scale is too large or too small, it will affect the resolution. Generally, 15 minutes is a more reasonable period. The same is true for the vertical axis scale. If there are too many inverters, the scale value needs to be appropriately reduced according to the actual situation.

[0074] The two-dimensional grid diagram facilitates the intuitive display of two important dimensions, time and equipment serial number, at the same time, laying the foundation for visual analysis of electrical parameters.

[0075] The specific implementation of step S50 is:

[0076] First, the real-time collected electrical parameters such as voltage, current, and frequency need to be converted into RGB color values. One conversion method is to normalize the voltage value U, current value I, and frequency value F to the range of 0-1, that is:

[0077] u=(U-Umin) / (Umax-Umin)

[0078] i=(I-Imin) / (Imax-Imin)

[0079] f=(F-Fmin) / (Fmax-Fmin)

[0080] Among them, Umin, Umax, Imin, Imax, Fmin, and Fmax are the theoretical minimum and maximum values ​​of voltage, current, and frequency respectively.

[0081] Then the three normalized quantities (u, i, f) are used as the RGB primary color values ​​to get the corresponding colors. The changes in the electrical parameters can be directly reflected through the different changes in color. Auxiliary tools such as color scale bars can be used to mark the specific parameter values ​​corresponding to various colors.

[0082] Secondly, according to the color value of each inverter in the current time period, the corresponding grid unit is filled with the color. For example, in a certain period of time, the voltage of inverter 1 is 220V, the current is 50A, and the frequency is 49Hz. According to the above method, the RGB value is (0.8, 0.6, 0.7), and the corresponding grid unit is filled with this color.

[0083] Finally, by filling all the grid cells, an intuitive two-dimensional color map can be obtained. The color change of each pixel in the map corresponds to the change of the electrical parameters of the inverter at that location during that period. This visualization method can quickly locate abnormalities and provide strong support for fault diagnosis.

[0084] The purpose of this step is to integrate multiple electrical parameter information output by the inverter into a single color, thereby using a unified and intuitive visual representation. Through the change of color brightness, abnormal points can be quickly identified, and combined with location information, the specific inverter device and the time of occurrence can be located.

[0085] This step uses the pseudo-color coding technology commonly used in image processing, which can map one-dimensional data to color space, so that even small changes in data can be clearly presented in the form of color differences. Compared with previous display methods such as curve graphs, two-dimensional color maps have the advantages of large amount of information, novel form, strong visibility, etc., and are easier to be automatically analyzed by artificial intelligence algorithms, so they are very suitable for the method of the present invention.

[0086] The specific implementation of step S60 is:

[0087] First, a harmonic fault sample library needs to be established to collect and store relevant data of various historical harmonic faults. These data include: the voltage, current, frequency curves or discrete data points output by each inverter within 10 seconds before the fault occurs; and the corresponding fault cause description, such as inverter internal fault, grid-connected side fault, etc.

[0088] The data in the sample library can come from the actual operation and maintenance site records, or can be obtained through theoretical modeling and simulation. The data collected on site is authentic and reliable, but is limited by the low frequency of failures; while the simulation data lacks authenticity, it can quickly and efficiently generate data under various failure conditions, so the combination of the two is more ideal.

[0089] Secondly, the raw data of the harmonic fault samples are preprocessed, including format unification, removal of outliers and noise, and interpolation of missing values. The data preprocessing method mentioned in step S10 can be used as a reference.

[0090] Then, based on the preprocessed data, for each harmonic fault condition, a corresponding two-dimensional color map is constructed according to steps S40-S50, and stored in the sample library together with the fault cause. These color maps are the "comparison two-dimensional maps" that match the harmonic fault.

[0091] The sample library will eventually contain the following contents: explanations of the causes of various harmonic faults, the corresponding raw data of electrical parameters, and the corresponding two-dimensional color "comparison chart". When a new harmonic fault occurs, the most similar comparison chart and its corresponding fault cause can be found.

[0092] The purpose of this step is to establish a knowledge base containing historical fault samples to provide a reference for subsequent fault diagnosis. Similar to the practice of establishing a pathology sample library in medicine, the symptom data of known diseases are collected for comparison and analysis in the future.

[0093] The specific implementation of step S70 is:

[0094] First, a sliding window is constructed. The window size is set according to actual needs, for example, it can be set to represent the time range of the past 4 hours. The window will slide to the right on the two-dimensional color map at a rate of 1 time period.

[0095] In the sliding window, the two-dimensional image data of the electrical parameters of all inverters within this time range are extracted and recorded as the image to be matched.

[0096] Then, the fault sample library established in S60 is traversed to calculate the similarity score between the comparison image and the image to be matched in each sample library. The similarity score calculation formula is as follows:

[0097] Similarity = sum(color distance weight * color distance score)

[0098] The color distance score is calculated based on the Euclidean distance of the pixel color in the RGB space. The smaller the distance, the higher the score. The weight is set according to the importance of the fault location. The inverter of the power station trunk line is given a higher weight to make it more influential.

[0099] In the process of similarity calculation, the delay of fault occurrence also needs to be taken into account. That is, after the main fault source occurs, other inverters may show abnormal performance after a certain time delay. For this reason, the matching graph can be matched with the sample library comparison graph according to different time offsets to obtain the best matching score.

[0100] Finally, the one with the highest similarity score with the image to be matched is selected from all sample library comparison images, and the corresponding fault cause is output, which is the diagnosis result of this time.

[0101] If the highest score is lower than the preset threshold, it means that there is no matching fault in the library, and the system will issue a prompt to the operator, and it cannot be effectively diagnosed. The threshold can be set based on experience, and generally 0.8 points or more can be considered a basic match.

[0102] In the above process, various computer vision and signal processing algorithms are involved, such as image similarity calculation, pattern matching, time series alignment, threshold filtering, etc. The key is:

[0103] 1) Similarity calculation algorithm:

[0104] Using weighted Euclidean distance as a similarity metric can comprehensively consider the effects of both spatial position and color differences.

[0105] 2) Time series alignment algorithm:

[0106] In view of the time delay of the inverter response after a fault occurs, it is necessary to perform time offset alignment on the matching graph and the sample graph to improve the matching accuracy.

[0107] 3)Threshold filtering:

[0108] The similarity score threshold is introduced to filter out those cases that are too different from known fault samples to avoid misdiagnosis. The threshold can also be set to adaptive adjustment.

[0109] The purpose of this step is to use the existing historical fault sample knowledge to quickly diagnose the newly occurring real-time fault situation, give the judgment result and the corresponding fault cause explanation, and provide a reference basis for the subsequent precise positioning of the fault source and maintenance.

[0110] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned method for diagnosing harmonic faults in a distributed battery energy storage power station.

[0111] A third aspect of the present invention provides a distributed battery energy storage power station harmonic fault diagnosis system, which includes the above-mentioned computer-readable storage medium.

[0112] Specifically, the method of the present invention can efficiently and accurately diagnose harmonic faults and their causes in distributed battery energy storage power stations. The main principles are as follows:

[0113] 1. Strong image representation capabilities

[0114] Compared with directly processing the data stream itself, after converting the data into an intuitive two-dimensional color image, the human visual system and computer vision algorithms have a strong ability to perceive and extract the various patterns, features and laws contained therein. Given a complex data sequence, it is difficult for the human brain to discover its internal connections, but after rendering it into an image through color, the visual effect immediately highlights the relevant information, such as regularity, symmetry, texture, etc. This is the fundamental reason why this method adopts the image representation strategy.

[0115] 2. Clustering and merging to extract common features

[0116] For the same type of fault, although the specific values ​​may be different, there are some common features in general. The image clustering algorithm can automatically discover these internal connections and find out the potential essential factors of the fault. And merging the clustered images to generate the total image is to extract and condense their common regular features and abstract the typical "fingerprint" of the fault. This idea greatly reduces the dependence on a single sample data, which helps the generalization ability and universality of the method.

[0117] 3. Segmentation and comparison to discover hidden rules

[0118] The total image after simple merging may still confuse some details. This method re-segmented the total image, starting from its inherent potential texture, edge and other features, disassembled it into local images of multiple sub-regions, and rearranged them in a matrix. The key role of this step is that the segmentation algorithm can automatically dig out the deeper implicit laws between the influencing factors and the fault, and visualize them in the local pattern of the image, so that the entire algorithm can describe the cause of the fault more comprehensively and finely.

[0119] 4. Similarity matching to find the best diagnosis result

[0120] For the image data monitored in real time, this method does not analyze the meaning of each pixel, but extracts a relevant image through a sliding window, and matches and compares it with the pre-built fault standard "comparison map" in the sample library as a whole. The similarity matching idea is derived from the classic pattern recognition algorithm, but when calculating the similarity score, this method not only considers the pixel color distance, but also incorporates the spatial position weight, which can reflect the differences in both color and texture.

[0121] 5. Sliding windows and time alignment to cope with failure delays

[0122] In actual situations, the occurrence of faults is often not immediate, but there is a certain time delay or dislocation. In order to improve the robustness of the method and deal with this kind of timing deviation, the present invention introduces sliding window and time series alignment technology. The sliding window extracts local fragments from the monitoring data at a fixed step size, while the time series alignment takes into account a certain time offset when calculating the distance between each window and the comparison graph, thereby including all possible time points of the fault to the greatest extent, significantly improving the time fault tolerance of the diagnosis.

[0123] 6. Preprocessing to ensure image quality

[0124] A high-quality input image is crucial for subsequent feature analysis and matching. In the preprocessing stage, this method automatically applies a number of image enhancement, denoising, correction and other technologies, such as Gaussian smoothing filtering, color space transformation, etc., to ensure that the image data input to the system has good clarity, contrast and consistency, laying the foundation for accurate matching.

[0125] In summary, the method of the present invention integrates multiple cutting-edge theories and technologies such as computer vision, pattern recognition, signal processing, image processing, etc., and combines the characteristics of power system operation to form a new set of effective harmonic fault intelligent diagnosis ideas. The innovation of the method is that it cleverly combines various algorithm strategies, gives full play to their respective advantages, complements each other, and finally achieves the goals of automation, efficiency, and intelligence.

[0126] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for diagnosing harmonic faults in a distributed battery energy storage power station, characterized in that: The following steps are involved: S10, obtaining in real time the electrical parameters output by each inverter of the distributed battery energy storage power station, including current, voltage, and frequency; S20, establishing an inverter tree according to the inverter topology diagram of the energy storage power station; S30, traversing the inverter tree, and obtaining a sequence number of each inverter according to the traversal order; S40, establishing an electrical two-dimensional grid diagram, wherein the abscissa of the electrical two-dimensional grid diagram is time, the ordinate is the inverter number, and the duration of each grid is one acquisition cycle; S50, converting the electrical parameters collected in real time into RGB color values, and filling the two-dimensional grid map to obtain a two-dimensional color map; S60, obtaining electrical parameters output by each inverter in the historical harmonic fault records and the fault causes corresponding to the historical harmonic fault records, and establishing a two-dimensional color map corresponding to each harmonic fault, recorded as a comparative two-dimensional map; S70, establishing a sliding window, using the comparative two-dimensional graph corresponding to each harmonic fault to perform similarity matching on the two-dimensional color graph, selecting the harmonic fault corresponding to the comparative two-dimensional graph with the highest similarity as the harmonic fault cause of the distributed battery energy storage power station, and outputting it to the operator.

2. A method for diagnosing harmonic faults in a distributed battery energy storage power station according to claim 1, characterized in that: The step of obtaining in real time the electrical parameters output by each inverter of the distributed battery energy storage power station specifically includes: First, the original sampling data of the electrical parameters at the output end of each inverter is obtained; Then, the original sampled data is preprocessed, including format unification, removal of outliers and noise, and interpolation of missing values.

3. A method for diagnosing harmonic faults in a distributed battery energy storage power station according to claim 1, characterized in that: The method of traversing the inverter tree is pre-order traversal, in-order traversal, or post-order traversal.

4. A method for diagnosing harmonic faults in a distributed battery energy storage power station according to claim 1, characterized in that: The acquisition cycle is 0.001 seconds to 0.005 seconds.

5. A method for diagnosing harmonic faults in a distributed battery energy storage power station according to claim 1, characterized in that: The step of converting the electrical parameters collected in real time into RGB color values ​​is specifically: normalizing the voltage value, current value and frequency value of the electrical parameters to the range of 0-1 respectively, and then using the three normalized values ​​as the RGB three primary color values.

6. A method for diagnosing harmonic faults in a distributed battery energy storage power station according to claim 1, characterized in that: The step of obtaining the electrical parameters of each inverter output of the historical harmonic fault record, and the fault cause corresponding to the historical harmonic fault record, and establishing a two-dimensional color map corresponding to each harmonic fault, specifically comprises: establishing a harmonic fault sample library, extracting relevant electrical parameter data and fault cause descriptions of various harmonic faults from the historical fault data; Then, the relevant electrical parameter data are preprocessed, and a two-dimensional color image is constructed based on the preprocessed relevant electrical parameter data as a comparative two-dimensional image corresponding to various types of harmonic faults.

7. A method for diagnosing harmonic faults in a distributed battery energy storage power station according to claim 1, characterized in that: The step of establishing a sliding window and using the comparative two-dimensional image corresponding to each harmonic fault to perform similarity matching on the two-dimensional color image is specifically: constructing a sliding window, which will slide on the two-dimensional color image with a fixed step size; extracting the image to be matched in the sliding window; traversing the fault sample library established in step S60, and calculating the similarity score between each comparative image and the image to be matched, wherein the similarity score is calculated based on the pixel color distance and the weight.

8. A method for diagnosing harmonic faults in a distributed battery energy storage power station according to claim 7, characterized in that: In the step of calculating the similarity score between each comparison image and the image to be matched, the fault delay phenomenon is taken into account, and the image to be matched is matched with the sample library comparison image at different time offsets; the one with the highest similarity score with the image to be matched is selected from all the comparison images, and the corresponding fault cause is output as the diagnosis result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, which, when executed, are used to execute a harmonic fault diagnosis method for a distributed battery energy storage power station according to any one of claims 1 to 8.

10. A distributed battery energy storage power station harmonic fault diagnosis system, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images