Photovoltaic power generation anomaly diagnosis method and system

Abstract

The application discloses a photovoltaic power generation anomaly diagnosis method and system, the method comprises the following steps: collecting basic topology data, and constructing an initial topology graph based on the basic topology data; extracting the space-time characteristics of the graph nodes, and constructing a space-time characteristic pool; screening the steady-state space-time characteristic data in a preset period, inputting the initial topology graph, and generating a reference feature matrix under normal working conditions; calculating the real-time graph correlation degree according to the reference feature matrix, and positioning the abnormal node area according to the calculation result; performing topology backtracking on the abnormal area, combing the fault conduction path, and confirming the fault type of the abnormal node area. The application breaks through the dependence on the number of fault samples, is suitable for new and old photovoltaic power stations, simultaneously fuses physical topology and space-time two-dimensional characteristics, excavates subtle abnormalities through the correlation deviation between nodes, can accurately identify implicit, concurrent and coupled faults that cannot be perceived by various traditional technologies, and has excellent diagnosis depth and coverage range.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, and specifically to a method and system for diagnosing photovoltaic power generation anomalies. Background Technology

[0002] Currently, the trend of large-scale and clustered construction of photovoltaic power plants is significant. The installed capacity of power plants continues to grow, and array layouts are becoming increasingly complex. Photovoltaic modules operate under complex outdoor conditions for extended periods, making them susceptible to multiple factors such as temperature, radiation, rain, snow, wind, sand, and aging degradation. This makes them prone to various faults, including module hot spots, branch mismatch, performance degradation, and abnormal coupling of multiple branches. The power generation efficiency and equipment health status of photovoltaic power plants directly determine the operation and maintenance revenue of the power plant and the stability of the power supply system. Accurate, comprehensive, and intelligent fault diagnosis technology is the key to ensuring the long-term efficient, safe, and economical operation of photovoltaic power plants.

[0003] Currently, the main methods for diagnosing photovoltaic (PV) power generation faults are traditional threshold analysis and supervised learning-based intelligent diagnosis. Traditional threshold analysis determines equipment anomalies by setting fixed power, voltage, and current thresholds, relying solely on single-node time-series operational data for fault screening. Its simple structure and ease of implementation have led to its widespread application in conventional PV monitoring systems. Supervised learning-based diagnostic techniques, on the other hand, train diagnostic models using massive amounts of labeled fault samples, enabling basic fault classification and identification. Compared to traditional threshold algorithms, this method offers improved intelligence and represents the mainstream research and application direction for intelligent diagnosis at present.

[0004] However, traditional thresholding algorithms and conventional supervised learning models only make anomaly judgments based on single-device, single-dimensional time-series data, ignoring the inherent series-parallel topological electrical coupling relationships and spatial correlation characteristics of photovoltaic power plants. Existing intelligent diagnostic technologies rely heavily on manually labeled fault sample datasets. However, in the real operation scenarios of photovoltaic power plants, the number of latent faults and multi-branch coupled concurrent fault samples is scarce, the collection is difficult, and the labeling cost is extremely high. This results in insufficient training samples for the model, poor generalization ability, and inability to adapt to the diagnostic scenarios of newly built fault-free power plants and old and complex power plants. Summary of the Invention

[0005] Based on this, the purpose of this invention is to provide a photovoltaic power generation anomaly diagnosis method and system, which aims to solve the problems of current photovoltaic power generation fault diagnosis methods having single anomaly judgment results, high dependence on fault data samples, and poor generalization ability.

[0006] To achieve the above objectives, this invention proposes a method for diagnosing photovoltaic power generation anomalies, the method comprising: Collect basic topology data and construct an initial topology map based on the basic topology data. The initial topology map contains several map nodes, and any map node is an independent monitoring unit of the photovoltaic power station. Extract the spatiotemporal features of the graph nodes, construct a spatiotemporal feature pool, filter steady-state spatiotemporal feature data within a preset period, input the initial topological graph, and generate a baseline feature matrix under normal operating conditions. Calculate the real-time map correlation degree based on the benchmark feature matrix, and locate abnormal node regions based on the calculation results; Perform topological backtracking on the abnormal region to trace the fault propagation path and confirm the fault type of the abnormal node region.

[0007] According to one aspect of the above technical solution, the step of collecting basic topology data and constructing an initial topology map based on the basic topology data includes: Batch collection of original design data and equipment ledger data of photovoltaic power plants; definition of all independent monitoring units of photovoltaic power plants as map nodes; unified map node coding; and establishment of a mapping relationship between the unique identifier of map nodes and actual equipment. Based on the physical location and electrical relationship between graph nodes, a physical and electrical dual-weight association mechanism is constructed, and the topological edge weights of any two graph nodes are assigned. Based on the device locations of the graph nodes, the connection relationships between the graph nodes, and the topological edge weights between any two graph nodes, an initial topological graph is constructed.

[0008] According to one aspect of the above technical solution, the steps of extracting the spatiotemporal features of the graph nodes, constructing a spatiotemporal feature pool, filtering steady-state spatiotemporal feature data within a preset period, inputting the initial topological graph, and generating a baseline feature matrix under normal operating conditions include: The device operation data of each map node is collected at a preset frequency, and the device operation data is preprocessed. A time-domain window is set to extract the time dimension features of the graph nodes. Specifically, a short-term time-domain window and a long-term time-domain window are defined and used to extract the short-term fluctuation features and long-term decay features of the graph nodes, respectively. Using adjacent or same branch nodes in the initial topological graph as comparison objects, the spatial correlation deviation between different graph nodes is calculated to extract spatial dimension features. The time-dimensional features and the spatial-dimensional features are normalized and then weighted and fused based on feature importance to construct a spatiotemporal feature pool.

[0009] According to one aspect of the above technical solution, the steps following the construction of the spatiotemporal feature pool are as follows: Steady-state spatiotemporal feature data within a preset period are selected as the training set. A pre-constructed initial topological map is input, and a baseline feature matrix under normal operating conditions is constructed through iterative learning via a graph attention mechanism.

[0010] According to one aspect of the above technical solution, the step of calculating the real-time map correlation degree based on the benchmark feature matrix and locating abnormal node regions based on the calculation results includes: Based on the spatiotemporal features and the baseline feature matrix in the spatiotemporal feature pool, calculate the node feature deviation of the current map node. :

[0011] in, The normalized spatiotemporal feature vector within the spatiotemporal feature pool. The baseline feature matrix; Node coupling consistency of synchronous computation graph nodes :

[0012] In the formula, This represents the number of adjacent nodes. For topological edge weights, For average topological weights, The characteristic Pearson correlation coefficient is given.

[0013] According to one aspect of the above technical solution, after calculating and obtaining the node feature deviation degree and node coupling consistency, when the node feature deviation degree, the node coupling consistency, and the duration all reach a preset threshold, the current map node is determined to be an abnormal node. Based on the physical location of each node in the initial topology graph, obtain the physical spatial coordinates of all abnormal nodes, retrieve the number of abnormal nodes within the neighborhood radius of any abnormal node, and complete the classification of abnormal node types. Aggregate all mutually density-reachable core nodes and boundary nodes to form a continuous independent anomalous region cluster, and locate the independent anomalous region cluster as an anomalous node region.

[0014] According to one aspect of the above technical solution, the step of performing topological backtracking on the abnormal region, tracing the fault propagation path, and confirming the fault type of the abnormal node region includes: Starting from the abnormal node region, traverse the topological edges in reverse to sort out the fault propagation path and record the topological edge weights and propagation delays of the fault propagation path. The fault impact factor of each graph node on the fault propagation path is calculated based on the topological edge weight and the propagation delay. Based on the parameter value of the fault impact factor of each graph node, the level of coupled fault is determined and located to the specific photovoltaic string number, branch number, and inverter port.

[0015] The present invention also provides a photovoltaic power generation anomaly diagnosis system, which is used to implement the above-mentioned photovoltaic power generation anomaly diagnosis method, the system comprising: The graph construction module is used to collect basic topology data and construct an initial topology graph based on the basic topology data. The initial topology graph contains several graph nodes, and any graph node is an independent monitoring unit of the photovoltaic power station. The feature extraction module is used to extract the spatiotemporal features of the graph nodes, construct a spatiotemporal feature pool, filter steady-state spatiotemporal feature data within a preset period, input the initial topological graph, and generate a baseline feature matrix under normal operating conditions. The anomaly localization module is used to calculate the real-time map correlation degree based on the benchmark feature matrix and locate the abnormal node region based on the calculation result. The type confirmation module is used to perform topological backtracking on the abnormal region, trace the fault propagation path, and confirm the fault type of the abnormal node region.

[0016] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the photovoltaic power generation anomaly diagnosis method as described above.

[0017] The present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the photovoltaic power generation anomaly diagnosis method as described above.

[0018] In summary, the photovoltaic power generation anomaly diagnosis method provided by this invention overcomes the technical limitations of traditional supervised learning and single time-series analysis, and innovatively proposes an unsupervised spatiotemporal coupling graph autonomous modeling mechanism. This mechanism integrates the inherent physical topology of photovoltaic power plants with the characteristics of time-series operational data, requiring no manual fault labeling samples. It autonomously learns normal operating condition benchmarks based on the normal spatiotemporal correlation patterns between equipment nodes, and accurately uncovers latent anomalies through operating condition deviation. Simultaneously, it utilizes topological weights and reverse tracing to achieve hierarchical decomposition, source point location, and quantitative assessment of multi-coupled faults. This fundamentally solves the industry's technical bottlenecks of difficult identification of latent photovoltaic faults, difficulty in tracing concurrent faults, and difficulty in training due to scarce samples. Compared to existing technologies, this invention represents substantial innovation and significant technological progress. This invention overcomes the dependence on the number of fault samples, is adaptable to both newly built and old photovoltaic power plants, and integrates physical topology and spatiotemporal dual-dimensional features. By uncovering subtle anomalies through inter-node correlation deviations, it can accurately identify various latent, concurrent, and coupled faults that are imperceptible by traditional technologies, exhibiting excellent diagnostic depth and coverage.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] Figure 1 This is a flowchart of the photovoltaic power generation anomaly diagnosis method in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the photovoltaic power generation anomaly diagnosis method in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the photovoltaic power generation anomaly diagnosis system in Embodiment 3 of the present invention; Figure 4 This is a structural block diagram of the electronic device in Embodiment 5 of the present invention. Detailed Implementation

[0021] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0022] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0023] Unless otherwise defined, 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 invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0024] Example 1 like Figure 1 The diagram shows a flowchart of a photovoltaic power generation anomaly diagnosis method according to Embodiment 1 of the present invention. The photovoltaic power generation anomaly diagnosis method includes the following steps S11-S14, wherein: S11. Collect basic topology data and construct an initial topology map based on the basic topology data, wherein the initial topology map contains several map nodes, and any map node is an independent monitoring unit of the photovoltaic power station.

[0025] Import original design data and equipment ledger data of photovoltaic power plants in batches, including the layout coordinates of photovoltaic module arrays, the string-parallel connection relationship, the branch allocation rules of combiner boxes, and the corresponding relationship of inverter access ports, clarifying the physical location and electrical subordination of each photovoltaic module, each branch, and each inverter.

[0026] All independent monitoring units of the power station are defined as graph nodes, with the smallest unit refined to a single photovoltaic string. A unified graph node code is established, and a mapping relationship between the unique identifier of the graph node and the actual equipment is established. Based on the physical location and electrical relationship between graph nodes, a physical and electrical dual weight association mechanism is constructed, and the topological edge weight of any two graph nodes is assigned. Based on the device locations of the graph nodes, the connection relationships between the graph nodes, and the topological edge weights between any two graph nodes, an initial topological graph is constructed.

[0027] S12. Extract the spatiotemporal features of the graph nodes, construct a spatiotemporal feature pool, filter steady-state spatiotemporal feature data within a preset period, input the initial topological graph, and generate a baseline feature matrix under normal operating conditions.

[0028] The system collects DC voltage, DC current, real-time output power, component backplane temperature, ambient irradiance, and equipment operating status data for each spectrum node at a preset frequency. Noise reduction can be achieved using a Kalman filter algorithm, missing values ​​can be repaired using linear interpolation, and invalid data caused by momentary sensor jitter can be eliminated, ensuring the continuity of time-series data.

[0029] The temporal and spatial features of the preprocessed time series data are extracted, normalized, and then weighted and fused based on feature importance to construct a spatiotemporal feature pool.

[0030] Steady-state spatiotemporal feature data within a preset period are selected from the spatiotemporal feature pool as the training set. A pre-constructed initial topological map is input, and a baseline feature matrix under normal operating conditions is constructed through iterative learning via a graph attention mechanism. S13. Calculate the real-time map correlation degree based on the benchmark feature matrix, and locate the abnormal node region based on the calculation result.

[0031] Using the spatiotemporal features and baseline feature matrix in the spatiotemporal feature pool, the correlation degree of the current map node is obtained, the hidden anomaly and coupled anomaly node are autonomously identified, and the anomaly node region is located.

[0032] S14. Perform topological backtracking on the abnormal region, trace the fault propagation path, and confirm the fault type of the abnormal node region.

[0033] Locate the abnormal node region in step S13, take the abnormal node set as the endpoint, perform a reverse depth-first traversal of the topology edges, sort out all fault propagation paths, construct a multi-branch fault propagation link tree, record the topology edge weight and propagation delay of each propagation path, evaluate the fault propagation path based on the topology edge weight and propagation delay, determine the coupling fault level, and locate the specific photovoltaic string number, branch number, and inverter port.

[0034] In summary, the photovoltaic power generation anomaly diagnosis method provided by this invention overcomes the technical limitations of traditional supervised learning and single time-series analysis, and innovatively proposes an unsupervised spatiotemporal coupling graph autonomous modeling mechanism. This mechanism integrates the inherent physical topology of photovoltaic power plants with the characteristics of time-series operational data, requiring no manual fault labeling samples. It autonomously learns normal operating condition benchmarks based on the normal spatiotemporal correlation patterns between equipment nodes, and accurately uncovers latent anomalies through operating condition deviation. Simultaneously, it utilizes topological weights and reverse tracing to achieve hierarchical decomposition, source point location, and quantitative assessment of multi-coupled faults. This fundamentally solves the industry's technical bottlenecks of difficult identification of latent photovoltaic faults, difficulty in tracing concurrent faults, and difficulty in training due to scarce samples. Compared to existing technologies, this invention represents substantial innovation and significant technological progress. This invention overcomes the dependence on the number of fault samples, is adaptable to both newly built and old photovoltaic power plants, and integrates physical topology and spatiotemporal dual-dimensional features. By uncovering subtle anomalies through inter-node correlation deviations, it can accurately identify various latent, concurrent, and coupled faults that are imperceptible by traditional technologies, exhibiting excellent diagnostic depth and coverage.

[0035] Example 2 Please see Figure 2 The figure shows a photovoltaic power generation anomaly diagnosis method in the second embodiment of the present invention, including steps S21-S24.

[0036] S21. Collect basic topology data and construct an initial topology map based on the basic topology data, wherein the initial topology map contains several map nodes, and any map node is an independent monitoring unit of the photovoltaic power station.

[0037] Import original design data and equipment ledger data of photovoltaic power plants in batches, including the layout coordinates of photovoltaic module arrays, the string-parallel connection relationship, the branch allocation rules of combiner boxes, and the corresponding relationship of inverter access ports, clarifying the physical location and electrical subordination of each photovoltaic module, each branch, and each inverter.

[0038] All independent monitoring units of the power station are defined as map nodes, with the smallest unit refined to a single photovoltaic string. Node combinations are defined as... , To determine the total number of photovoltaic strings and core equipment in the entire station, a unified node code is established, and a binding mapping relationship is created between the unique ID of the node and the actual equipment, so as to achieve a one-to-one correspondence between the map nodes and the physical equipment.

[0039] Construct a physical and electrical dual-weight association mechanism, with the topological edge set as follows: Any two graph nodes , Weights between for:

[0040] in, As a spatial physical weight, the smaller the physical spacing between devices and the higher the adjacency of their arrangement, the larger the parameter value. Electrical coupling weights, devices in the same string or branch. The value is 1, and it decreases across array devices according to the electrical association gradient. and These are weighting coefficients. It should be noted that, in order to conform to the characteristics of electrical coupling operation, this embodiment uses... Set it to 0.3, Set it to 0.7.

[0041] Integrate all graph nodes, topological edges, and dual weight parameters to generate an initial topological graph. , This is a global weight matrix. The initial topology graph supports dynamic adaptation, automatically updating the topology nodes and weight parameters for scenarios such as power plant component expansion and line modification, thus completing the construction of a dynamic topology graph specific to the photovoltaic system.

[0042] S22. Extract the spatiotemporal features of the graph nodes, construct a spatiotemporal feature pool, filter steady-state spatiotemporal feature data within a preset period, input the initial topology graph, and generate a baseline feature matrix under normal operating conditions.

[0043] S221. Data Acquisition and Preprocessing.

[0044] The system collects DC voltage, DC current, real-time output power, component backplane temperature, ambient irradiance, and equipment operating status data at preset frequencies for each spectrum node. Noise reduction is achieved using a Kalman filter algorithm, missing values ​​are repaired using linear interpolation, and invalid data caused by momentary sensor jitter is eliminated to ensure the continuity of time-series data.

[0045] S222, Time dimension feature extraction.

[0046] Set a time-domain window to extract the time dimension features of the graph nodes. Specifically, define short-term and long-term time-domain windows respectively, and use them to extract the short-term fluctuation features and long-term decay features of the graph nodes.

[0047] The short-term time-domain window for short-term fluctuation characteristics can be set to 5 minutes. Short-term fluctuation characteristics can include time series variance, power fluctuation amplitude, time series change slope, etc., to capture short-term small disturbances.

[0048] The long-term time-domain window for long-term decay characteristics can be set to 24 hours. Long-term decay characteristics can include parameters such as irradiance-output matching degree and daily average output decay rate.

[0049] S223, Spatial Dimension Feature Extraction.

[0050] Using adjacent nodes or nodes on the same branch in the initial topology map as comparison objects, the spatial correlation deviation of nodes is calculated, with the core indicator being the processing balance deviation on the same branch. :

[0051] In the formula, For graph nodes contribute, For graph nodes Average power output of surrounding adjacent nodes The larger the value, the more pronounced the abnormal price control characteristics.

[0052] S224. Construction of the spatiotemporal feature pool.

[0053] The extracted temporal and spatial features are normalized to eliminate dimensional differences, and then weighted and fused by feature importance to construct a fixed and refined spatiotemporal feature pool.

[0054] Steady-state spatiotemporal feature data within a preset period are selected from the spatiotemporal feature pool as the training set. The pre-constructed initial topological map is input, and the baseline feature matrix under normal working conditions is constructed through iterative learning via the graph attention mechanism.

[0055] S23. Calculate the real-time map correlation degree based on the benchmark feature matrix, and locate the abnormal node region based on the calculation result.

[0056] S231, Real-time graph correlation calculation.

[0057] The spatiotemporal features and baseline feature matrices within the spatiotemporal feature pool are used to calculate the node feature deviation of the current map node. :

[0058] in, The normalized spatiotemporal feature vector within the spatiotemporal feature pool. The baseline feature matrix; Node coupling consistency of synchronous computation graph nodes :

[0059] In the formula, This represents the number of adjacent nodes. For topological edge weights, For average topological weights, The characteristic Pearson correlation coefficient is given.

[0060] S232, Anomaly Detection: Under steady-state conditions, although the coupling consistency of the graph nodes may fluctuate, their parameter values ​​remain within a certain range. If the feature deviation and node coupling consistency of the current graph node both exceed the threshold set by the system and continue for a certain period of time, the current graph node can be determined to be an abnormal node.

[0061] Furthermore, based on the physical location of each node in the initial topology map constructed in the previous steps, physical spatial coordinates are mapped one by one to all the identified abnormal node sets. These coordinates correspond to the actual arrangement of photovoltaic strings in the power station array, generating an abnormal node spatial location dataset, which provides a location basis for spatial density clustering.

[0062] The neighborhood radius is set based on the maximum physical distance between adjacent strings in the same array to ensure that only physically adjacent and electrically coupled strings can be included in the same neighborhood. A density threshold is set, which is the minimum number of abnormal nodes in the neighborhood. If the number of abnormal nodes in the neighborhood radius of an abnormal node is greater than the density threshold, the abnormal node is determined to be a core node and belongs to the area coupling fault. If there are no other abnormal nodes in the neighborhood, it is determined to be an isolated scattered anomaly. If an abnormal node is located in the neighborhood of a core node and its own neighborhood density is less than the density threshold, it belongs to the fault area of ​​the corresponding core node and is determined to be a boundary abnormal node.

[0063] Traverse all anomalous nodes and retrieve the number of other anomalous nodes within the neighborhood radius of each anomalous node to complete the node type classification. Aggregate all mutually density-reachable core nodes and boundary nodes to form a continuous independent anomalous region cluster; connected domains of different densities are independent of each other, do not intersect or overlap, and finally generate one or more spatially independent anomalous region clusters; all isolated anomalous nodes do not participate in region aggregation and are retained as single-point anomalous nodes.

[0064] Based on the number of nodes within the aggregated regional cluster, two types of fault modes are strictly distinguished, while transient interference pseudo-anomalies are filtered out: Single-point independent latent anomaly: There is only one isolated anomalous node with no adjacent anomalous nodes. It is determined to be a minor latent fault in a single string (early decay, small hot spot), without fault linkage and conduction characteristics; Regional coupling concurrent anomaly: If there are ≥3 adjacent abnormal nodes in the aggregation area, it is determined to be a multi-branch, multi-string coupled linkage fault, with fault propagation and mutual disturbance characteristics, and enters the subsequent coupling fault tracing and dismantling process. False anomaly filtering: Scattered, unconnected, and transient isolated points that do not meet the 10-minute duration anomaly condition are directly removed and not included in the fault area marking range.

[0065] All confirmed anomalies are structurally labeled, and complete anomaly information that can be used for subsequent tracing is output: a unique region ID is assigned to each aggregated anomaly region, and the node numbers, physical locations, and branch affiliations of all strings covered by the region are recorded; single-point anomalies are individually labeled with a single-point ID and precise location; and the feature deviation of each anomaly node is bound. Coupling consistency score Abnormal start time.

[0066] S24. Perform topological backtracking on the abnormal region, trace the fault propagation path, and confirm the fault type of the abnormal node region.

[0067] Locate the abnormal region in step S232, and perform a reverse depth-first traversal of the topological edges, using the set of abnormal nodes as the endpoint. We sort out all fault propagation paths, construct a multi-branch fault propagation link tree, and record the topological weight and propagation delay of each propagation path.

[0068] Distinguish between primary and secondary faults in any fault propagation path:

[0069] in, These are the fault impact factors of abnormal nodes in the fault propagation path. The average deviation of the abnormal area. The global radiation weights of the graph nodes. , , It is a fixed coefficient.

[0070] When in the abnormal area The node with the largest parameter value in the graph is the core primary fault source, characterized by the earliest occurrence time, strongest feature deviation, and widest radiation range; secondary linked faults are triggered by the propagation of the primary fault source. The parameter values ​​are too low, the anomaly occurs after the main fault, and there are no independent abnormal characteristics.

[0071] It filters out pseudo-abnormal nodes caused by linkage interference, locks the core fault source, accurately locates the specific photovoltaic string number, branch number, and inverter port, and outputs the complete fault propagation link and hierarchical disassembly results.

[0072] In summary, the photovoltaic power generation anomaly diagnosis method provided by this invention overcomes the technical limitations of traditional supervised learning and single time-series analysis, and innovatively proposes an unsupervised spatiotemporal coupling graph autonomous modeling mechanism. This mechanism integrates the inherent physical topology of photovoltaic power plants with the characteristics of time-series operational data, requiring no manual fault labeling samples. It autonomously learns normal operating condition benchmarks based on the normal spatiotemporal correlation patterns between equipment nodes, and accurately uncovers latent anomalies through operating condition deviation. Simultaneously, it utilizes topological weights and reverse tracing to achieve hierarchical decomposition, source point location, and quantitative assessment of multi-coupled faults. This fundamentally solves the industry's technical bottlenecks of difficult identification of latent photovoltaic faults, difficulty in tracing concurrent faults, and difficulty in training due to scarce samples. Compared to existing technologies, this invention represents substantial innovation and significant technological progress. This invention overcomes the dependence on the number of fault samples, is adaptable to both newly built and old photovoltaic power plants, and integrates physical topology and spatiotemporal dual-dimensional features. By uncovering subtle anomalies through inter-node correlation deviations, it can accurately identify various latent, concurrent, and coupled faults that are imperceptible by traditional technologies, exhibiting excellent diagnostic depth and coverage.

[0073] Example 3 In another aspect, this invention also provides a photovoltaic power generation anomaly diagnosis system, please refer to [link / reference needed]. Figure 3 The diagram shown is a structural schematic of the photovoltaic power generation anomaly diagnosis system in Embodiment 2 of the present invention. The photovoltaic power generation anomaly diagnosis system includes: The graph construction module 11 is used to collect basic topology data and construct an initial topology graph based on the basic topology data. The initial topology graph contains a number of graph nodes, and any graph node is an independent monitoring unit of the photovoltaic power station. The feature extraction module 12 is used to extract the spatiotemporal features of the graph nodes, construct a spatiotemporal feature pool, filter steady-state spatiotemporal feature data within a preset period, input the initial topological graph, and generate a baseline feature matrix under normal operating conditions. Anomaly localization module 13 is used to calculate the real-time map correlation degree based on the benchmark feature matrix and locate the abnormal node region based on the calculation result; The type confirmation module 14 is used to perform topological backtracking on the abnormal area, sort out the fault propagation path, and confirm the fault type of the abnormal node area.

[0074] Example 4 In another aspect, the present invention also proposes a computer-readable storage medium having stored thereon one or more computer programs that, when executed by a processor, implement the above-described photovoltaic power generation anomaly diagnosis method.

[0075] Those skilled in the art will understand that the logic or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable storage medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable storage medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0076] More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable storage media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0077] Example 5 Figure 4 This is a structural block diagram of an electronic device provided in Embodiment 4. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the photovoltaic power generation anomaly diagnosis method described in the above embodiments. Figure 4 The electronic device 30 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0078] like Figure 4 As shown, the electronic device 30 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).

[0079] Bus 33 includes a data bus, an address bus, and a control bus.

[0080] The memory 32 may include volatile memory, such as RAM 321 (random access memory), and / or cache memory 322, and may further include ROM 323 (read-only memory).

[0081] The memory 32 may also include a program tool 325 having a set (at least one) of program modules 324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0082] The processor 31 executes various functional applications and data processing by running computer programs stored in the memory 32, such as the photovoltaic power generation anomaly diagnosis method described above.

[0083] Electronic device 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed via I / O interface 35 (input / output interface). Furthermore, electronic device 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 36. Figure 3 As shown, network adapter 36 communicates with other modules of the model-generated electronic device 30 via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the model-generated electronic device 30, including but not limited to: microcode, device drivers, redundant processors, disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0084] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0085] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0086] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for diagnosing photovoltaic power generation anomalies, characterized in that, The methods include: Collect basic topology data and construct an initial topology map based on the basic topology data. The initial topology map contains several map nodes, and any map node is an independent monitoring unit of the photovoltaic power station. Extract the spatiotemporal features of the graph nodes, construct a spatiotemporal feature pool, filter steady-state spatiotemporal feature data within a preset period, input the initial topological graph, and generate a baseline feature matrix under normal operating conditions. Calculate the real-time map correlation degree based on the benchmark feature matrix, and locate abnormal node regions based on the calculation results; Perform topological backtracking on the abnormal region to trace the fault propagation path and confirm the fault type of the abnormal node region.

2. The photovoltaic power generation anomaly diagnosis method according to claim 1, characterized in that, The steps of collecting basic topology data and constructing an initial topology map based on the basic topology data include: Batch collection of original design data and equipment ledger data of photovoltaic power plants; definition of all independent monitoring units of photovoltaic power plants as map nodes; unified map node coding; and establishment of a mapping relationship between the unique identifier of map nodes and actual equipment. Based on the physical location and electrical relationship between graph nodes, a physical and electrical dual-weight association mechanism is constructed, and the topological edge weights of any two graph nodes are assigned. Based on the device locations of the graph nodes, the connection relationships between the graph nodes, and the topological edge weights between any two graph nodes, an initial topological graph is constructed.

3. The photovoltaic power generation anomaly diagnosis method according to claim 1, characterized in that, The steps of extracting the spatiotemporal features of the graph nodes, constructing a spatiotemporal feature pool, filtering steady-state spatiotemporal feature data within a preset period, inputting the initial topological graph, and generating a baseline feature matrix under normal operating conditions include: The device operation data of each map node is collected at a preset frequency, and the device operation data is preprocessed. A time-domain window is set to extract the time dimension features of the graph nodes. Specifically, a short-term time-domain window and a long-term time-domain window are defined and used to extract the short-term fluctuation features and long-term decay features of the graph nodes, respectively. Using adjacent or same branch nodes in the initial topological graph as comparison objects, the spatial correlation deviation between different graph nodes is calculated to extract spatial dimension features. The time-dimensional features and the spatial-dimensional features are normalized and then weighted and fused based on feature importance to construct a spatiotemporal feature pool.

4. The photovoltaic power generation anomaly diagnosis method according to claim 3, characterized in that, The steps following the construction of the spatiotemporal feature pool are as follows: Steady-state spatiotemporal feature data within a preset period are selected from the spatiotemporal feature pool as the training set. The pre-constructed initial topological map is input, and the baseline feature matrix under normal working conditions is constructed through iterative learning via the graph attention mechanism.

5. The photovoltaic power generation anomaly diagnosis method according to claim 1, characterized in that, The step of calculating the real-time map correlation degree based on the benchmark feature matrix and locating abnormal node regions based on the calculation results includes: Based on the spatiotemporal features and the baseline feature matrix in the spatiotemporal feature pool, calculate the node feature deviation of the current map node. : in, The normalized spatiotemporal feature vector within the spatiotemporal feature pool. The baseline feature matrix; Node coupling consistency of synchronous computation graph nodes : In the formula, This represents the number of adjacent nodes. For topological edge weights, For average topological weights, The characteristic Pearson correlation coefficient is given.

6. The photovoltaic power generation anomaly diagnosis method according to claim 5, characterized in that, After calculating and obtaining the node feature deviation degree and node coupling consistency, when the node feature deviation degree, the node coupling consistency, and the duration all reach a preset threshold, the current map node is determined to be an abnormal node. Based on the physical location of each node in the initial topology graph, obtain the physical spatial coordinates of all abnormal nodes, retrieve the number of abnormal nodes within the neighborhood radius of any abnormal node, and complete the classification of abnormal node types. Aggregate all mutually density-reachable core nodes and boundary nodes to form a continuous independent anomalous region cluster, and locate the independent anomalous region cluster as an anomalous node region.

7. The photovoltaic power generation anomaly diagnosis method according to claim 1, characterized in that, The steps of performing topological backtracking on the abnormal region, tracing the fault propagation path, and confirming the fault type of the abnormal node region include: Starting from the abnormal node region, traverse the topological edges in reverse to sort out the fault propagation path and record the topological edge weights and propagation delays of the fault propagation path. The fault impact factor of each graph node on the fault propagation path is calculated based on the topological edge weight and the propagation delay. Based on the parameter value of the fault impact factor of each graph node, the level of coupled fault is determined and located to the specific photovoltaic string number, branch number, and inverter port.

8. A photovoltaic power generation anomaly diagnosis system, characterized in that, The photovoltaic power generation anomaly diagnosis system is used to implement the photovoltaic power generation anomaly diagnosis method according to any one of claims 1-7, the system comprising: The graph construction module is used to collect basic topology data and construct an initial topology graph based on the basic topology data. The initial topology graph contains several graph nodes, and any graph node is an independent monitoring unit of the photovoltaic power station. The feature extraction module is used to extract the spatiotemporal features of the graph nodes, construct a spatiotemporal feature pool, filter steady-state spatiotemporal feature data within a preset period, input the initial topological graph, and generate a baseline feature matrix under normal operating conditions. The anomaly localization module is used to calculate the real-time map correlation degree based on the benchmark feature matrix and locate the abnormal node region based on the calculation result. The type confirmation module is used to perform topological backtracking on the abnormal region, trace the fault propagation path, and confirm the fault type of the abnormal node region.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the photovoltaic power generation anomaly diagnosis method as described in any one of claims 1-7.

10. An electronic device 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 a computer program, it implements the photovoltaic power generation anomaly diagnosis method as described in any one of claims 1-7.

Images