Abnormal fault diagnosis method and system for photovoltaic power generation system based on distributed sensing

Abstract

The present invention relates to the technical field of power generation fault diagnosis, and in particular to a method and system for abnormal fault diagnosis of a photovoltaic power generation system based on distributed sensing. The method comprises the following steps: collecting multimodal sensing data of photovoltaic power generation nodes in a photovoltaic power generation system and correcting the real-time current environmental impact under different environmental conditions to obtain real-time current correction data of power generation nodes; analyzing the current waveforms of different power generation nodes on the real-time current correction data of power generation nodes and identifying the node load distribution to obtain load node distribution data; calculating the probability of false protection triggering between different load nodes on the load node distribution data to obtain node false protection probability data; performing abnormal fault diagnosis learning on the node false protection probability data and formulating abnormal diagnosis strategies for power generation nodes to obtain abnormal diagnosis strategies for power generation nodes. The present invention makes the diagnosis technology more accurate by optimizing the power generation fault diagnosis technology.

Description

Technical Field

[0001] The present invention relates to the technical field of power generation fault diagnosis, and in particular to a distributed sensing-based abnormal fault diagnosis method and system for a photovoltaic power generation system. Background Art

[0002] In modern photovoltaic power generation systems, ensuring efficient operation of the system and timely handling of faults are crucial to improving energy utilization and reducing operating costs. The increasing complexity and scale of photovoltaic systems have challenged traditional fault diagnosis methods. Distributed sensing technology combined with the application of intelligent algorithms has become an important means to solve this problem. Distributed sensing technology collects various operating data such as current, voltage, temperature and radiation intensity in real time by arranging multiple sensors in the photovoltaic array. These sensors are widely distributed in various components and nodes of the photovoltaic system and can reflect the operating status and health of the system in detail. In order to effectively process and analyze these massive data, fault diagnosis systems usually rely on advanced data processing algorithms and machine learning technologies, which can extract key features from the data and identify anomalies and fault types. For example, through time series analysis, pattern recognition and statistical learning, the system can detect problems such as partial shading, component aging, loose connections or faulty modules. The combination of distributed sensing and data processing technology can not only improve the accuracy and response speed of fault diagnosis, but also realize real-time monitoring and preventive maintenance of the system. However, the traditional abnormal fault diagnosis method of photovoltaic power generation system based on distributed sensing has the problems of inaccurate abnormal fault diagnosis analysis and accurate identification of complex interference. Summary of the invention

[0003] Based on this, it is necessary to provide a method and system for abnormal fault diagnosis of a photovoltaic power generation system based on distributed sensing to solve at least one of the above technical problems.

[0004] To achieve the above object, a photovoltaic power generation system abnormal fault diagnosis method based on distributed sensing is provided, the method comprising the following steps:

[0005] Step S1: Based on distributed sensors, multimodal sensing data of photovoltaic power generation nodes of the photovoltaic power generation system are collected to obtain multimodal sensing data of power generation nodes; real-time current data of different power generation nodes are extracted from the multimodal sensing data of power generation nodes to obtain real-time current data of power generation nodes; real-time current environmental impact correction is performed on the real-time current data of power generation nodes under different environmental conditions to obtain real-time current correction data of power generation nodes;

[0006] Step S2: Perform current waveform analysis of different power generation nodes on the real-time current correction data of the power generation node to obtain current waveform data of the power generation node; perform path conduction harmonic superposition effect analysis based on the current waveform data of the power generation node to obtain path conduction harmonic superposition effect data; perform node load distribution identification based on the path conduction harmonic superposition effect data to obtain load node distribution data;

[0007] Step S3: calculating the transient interference order gain between different load nodes on the load node distribution data to obtain transient interference order gain data; simulating the performance failure condition state according to the transient interference order gain data to obtain performance failure condition state data; calculating the misprotection triggering probability between different load nodes on the load node distribution data according to the transient interference order gain data and the performance failure condition state data to obtain node misprotection probability data;

[0008] Step S4: Perform abnormal fault diagnosis learning on the performance failure condition status data and the node misprotection probability data to obtain node abnormal fault diagnosis data; formulate a power generation node abnormal fault diagnosis strategy based on the node abnormal fault diagnosis data to obtain a power generation node abnormal fault diagnosis strategy.

[0009] The present invention uses distributed sensors to collect data from multiple nodes of a photovoltaic power generation system, including data under various environmental conditions, such as temperature, light intensity, and various environmental conditions during photovoltaic power generation. By extracting real-time current data and correcting for environmental impacts, accurate real-time current correction data of power generation nodes is ensured. This helps to improve the operating efficiency and stability of the power generation system, reduce the interference of environmental impacts on abnormal fault diagnosis, perform detailed waveform analysis on the corrected real-time current data, understand the changing trends and characteristics of the current at the power generation nodes in different time periods, and reveal the harmonic superposition effect of the current in the system during conduction through path conduction harmonic superposition effect data analysis, which helps to identify potential problems and optimization opportunities in the system. Through path conduction harmonic superposition effect data analysis, the load distribution of different nodes in the power generation system can be identified and understood. These data are crucial for optimizing system operation, improving energy efficiency, and reducing energy loss. By calculating the transient interference order gain between different load nodes, the system can better understand and quantify the mutual influence and interference degree between load nodes. According to the simulation of performance failure condition status data, the system can predict potential performance failures and take maintenance and repair measures in advance, thereby reducing system downtime and repair costs. By calculating the false protection trigger probability between different load nodes, the possibility of false alarms under abnormal conditions can be evaluated, and then the setting and adjustment of protection equipment can be optimized to reduce the impact of false alarms on system operation. By calculating the false protection trigger probability between different load nodes, the possibility of false alarms under abnormal conditions can be evaluated, and then the setting and adjustment of protection equipment can be optimized to reduce the impact of false alarms on system operation. Based on the performance failure condition status data and false protection trigger probability data, abnormal fault diagnosis learning is carried out. The system can learn and identify abnormal modes and fault characteristics of different load nodes, thereby improving the accuracy and timeliness of fault diagnosis. According to the abnormal fault diagnosis data, targeted abnormal diagnosis strategies for power generation nodes are formulated, including fault detection, location, diagnosis and repair steps, to help the operation and maintenance team respond to and solve problems quickly and minimize system downtime and losses. Therefore, the present invention optimizes the traditional abnormal fault diagnosis method of photovoltaic power generation system based on distributed sensing, solves the problems of inaccurate abnormal fault diagnosis and analysis and accurate identification of complex interference in this method, improves the accuracy of abnormal fault diagnosis and analysis, and improves the ability to accurately identify complex interference.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: collecting multimodal sensing data of photovoltaic power generation nodes of the photovoltaic power generation system based on distributed sensors to obtain multimodal sensing data of power generation nodes; wherein the multimodal sensing data of power generation nodes includes current sensing data of power generation nodes and environmental status data of power generation nodes;

[0012] Step S12: extracting real-time current data of different power generation nodes from the current sensing data of the power generation nodes to obtain real-time current data of the power generation nodes;

[0013] Step S13: Calculating the weights of environmental impact factors of different power generation nodes on the real-time current data of the power generation nodes according to the environmental status data of the power generation nodes, and obtaining environmental impact factor weight data;

[0014] Step S14: Correcting the real-time current environmental impact of different environmental conditions on the real-time current data of the power generation node according to the environmental impact factor weight data to obtain the real-time current correction data of the power generation node.

[0015] The present invention obtains multimodal sensing data of each node in a photovoltaic power generation system through distributed sensors, including current sensing data and environmental status data. The obtained multimodal data can comprehensively monitor the operation status and surrounding environmental conditions of the power generation nodes, and provide basic data for subsequent analysis and optimization. According to the collected data, the real-time current data of each power generation node is extracted to ensure that the system can timely reflect the current situation of the current power generation node, support real-time adjustment and feedback control strategies, thereby improving the response speed and stability of the system. According to the environmental status data of the power generation node, the weights of different environmental impact factors are calculated. These weights reflect the actual degree of influence of environmental factors on current data. Through weight correction, the real real-time current data of the power generation node can be more accurately reflected. According to the calculated environmental impact factor weights, the real-time current data is corrected. The corrected data is more in line with the power generation situation under actual environmental conditions, which helps to reduce the negative impact of environmental changes on power generation efficiency and improve the stability and energy efficiency of the system.

[0016] Preferably, step S2 comprises the following steps:

[0017] Step S21: performing node current conduction topology path analysis on the real-time current correction data of the power generation node to obtain node current conduction topology path data;

[0018] Step S22: performing current waveform analysis of different power generation nodes on the real-time current correction data of the power generation nodes according to the node current conduction topology path data to obtain current waveform data of the power generation nodes;

[0019] Step S23: analyzing the path conduction harmonic superposition effect of the node current conduction topology path data according to the power generation node current waveform data to obtain the path conduction harmonic superposition effect data;

[0020] Step S24: identifying the node load distribution of the node current conduction topology path data according to the path conduction harmonic superposition effect data to obtain load node distribution data.

[0021] The present invention establishes a topological path for current conduction between nodes by analyzing the real-time current correction data of power generation nodes. The obtained topological path data can reveal the current transmission path between different nodes in the power generation system, providing a basis for optimizing the current transmission route and reducing energy loss. The real-time current correction data of the power generation nodes is analyzed to obtain detailed current waveform data. The analysis of the current waveform data is helpful to identify and understand the operating conditions and current change characteristics of the power generation nodes, providing an important basis for subsequent fault diagnosis and prediction. According to the current waveform data, the harmonic superposition effect on the current conduction path between nodes is analyzed. Understanding the harmonic superposition effect is helpful to evaluate the harmonic problems in the system, and measures are taken to reduce the adverse effects of harmonics on the system, thereby improving the power quality and stability of the system. According to the path conduction harmonic superposition effect data, the load distribution of different nodes in the power generation system is identified and understood. These data are helpful to optimize the node load management strategy and improve the energy utilization efficiency and responsiveness of the system.

[0022] Preferably, step S23 includes the following steps:

[0023] Step S231: extracting the asymmetric waveform region of the current waveform data of the power generation node to obtain the current asymmetric waveform region data; performing harmonic identification on the current asymmetric waveform region data using a preset current harmonic identification model to obtain the current asymmetric harmonic data;

[0024] Step S232: performing asymmetric harmonic regional path mapping on the node current conduction topology path data according to the current asymmetric harmonic data to obtain the conduction path asymmetric harmonic data;

[0025] Step S233: calculating the crest factor increment coefficient between different node conduction paths for the asymmetric harmonic data of the conduction path to obtain the harmonic crest factor increment coefficient;

[0026] Step S234: calculating the crest factor progressive magnitude between different node conduction paths for the asymmetric harmonic data of the conduction path according to the harmonic crest factor progressive coefficient, and obtaining crest factor progressive magnitude data;

[0027] Step S235: calculating the magnitude mean difference series of the crest factor progressive magnitude data to obtain the magnitude mean difference series; performing magnitude truncation dynamic modal decomposition on the asymmetric harmonic data of the conduction path according to the magnitude mean difference series to obtain asymmetric harmonic magnitude modal decomposition data;

[0028] Step S236: analyzing the path conduction harmonic superposition effect of the node current conduction topology path data according to the asymmetric harmonic magnitude modal decomposition data and the harmonic crest factor increasing coefficient to obtain the path conduction harmonic superposition effect data.

[0029] The present invention extracts asymmetric waveform area data from the current waveform data of the power generation node, uses a preset current harmonic identification model to analyze and identify the asymmetric waveform area data to determine the existing asymmetric harmonic components, and maps the node current conduction topology path based on the asymmetric harmonic data to obtain the asymmetric harmonic data existing on the conduction path. This step helps to identify and analyze the asymmetric harmonic conduction path existing in the system, providing a basis for subsequent harmonic control and optimization, and calculating the harmonic crest factor increment coefficient between conduction paths at different nodes. The calculation of the increment coefficient can help evaluate the amplification of different conduction paths during the harmonic transmission process, and provide data support for the analysis of the harmonic diffusion effect. Based on the crest factor increment coefficient, the crest factor progressive magnitude data between conduction paths at different nodes is calculated. The calculation of progressive magnitude helps to understand the magnitude difference of harmonic conduction between different paths, and is helpful to determine the strategy of harmonic control and reduction of conduction loss. According to the progressive magnitude data of crest factor, the magnitude mean difference series is calculated, and the magnitude mean difference series is used to perform dynamic modal decomposition on the asymmetric harmonic data of the conduction path to obtain asymmetric harmonic magnitude modal decomposition data. The data analysis of this step helps to formulate targeted harmonic control strategies, reduce the harmonic impact in the system, and improve the power quality and system stability. Based on the asymmetric harmonic magnitude modal decomposition data and the harmonic crest factor increasing coefficient, the node current conduction topology path data is deeply analyzed for the path conduction harmonic superposition effect. The results of this step can reveal the specific impact of the harmonic superposition effect on different conduction paths, and provide an important reference for system operation optimization and harmonic management.

[0030] Preferably, step S3 comprises the following steps:

[0031] Step S31: Calculate the harmonic interference intensity of different load nodes on the load node distribution data according to the path conduction harmonic superposition effect data to obtain the load node harmonic interference intensity data;

[0032] Step S32: calculating transient interference order gain between different load nodes on the load node harmonic interference intensity data to obtain transient interference order gain data;

[0033] Step S33: performing thermal effect trend analysis between different load nodes according to transient interference order gain data and load node harmonic interference intensity data to obtain load node thermal effect trend data;

[0034] Step S34: performing a performance failure condition state simulation according to the transient interference order gain data and the load node thermal effect trend data to obtain performance failure condition state data;

[0035] Step S35: Calculate the false protection triggering probability between different load nodes based on the transient interference order gain data and the performance failure condition status data on the load node distribution data to obtain the node false protection probability data.

[0036] The present invention calculates the harmonic interference intensity of different load nodes according to the path conduction harmonic superposition effect data. By evaluating the harmonic interference intensity of each node, the load nodes in the system that are greatly affected by the harmonics can be identified, providing goals and key points for further problem solving. Based on the load node harmonic interference intensity data, the transient interference order gain (i.e., the increasing order of the intensity of the transient interference) between different load nodes is calculated. The transient interference order gain data reflects the dynamic response characteristics of the system affected by harmonics at different load nodes, which is helpful to evaluate the stability and response capability of the system. According to the transient interference order gain data, the load nodes of different load nodes are analyzed. Thermal effect trend between load nodes. Thermal effect trend data helps predict thermal effect problems of load nodes in the system under the influence of long-term harmonics, providing a basis for preventive maintenance and improved system design. According to transient interference order gain data and load node thermal effect trend data, performance failure condition state simulation is performed. The simulation of this step helps to evaluate the performance failure probability of load nodes under different harmonic interference conditions, providing a reliability analysis basis for system design and operation management. Based on transient interference order gain data and performance failure condition state data, the probability of false protection triggering between different load nodes is calculated. False protection triggering refers to the failure of the load node under non-fault conditions. In some cases, the protection device may operate incorrectly, resulting in unexpected power outage or equipment shutdown. Each load node has different transient interference order gain states and performance failure condition states. At the same time, each power generation node has different sensitivities to different condition states. Therefore, the power generation node may be misprotected, which will cause other nodes to suddenly increase the load of the current node due to the shutdown of the previous node or adjacent node. The load becomes larger, which makes it easier to cause the probability of failure of other equipment nodes, resulting in the overall spread of faults in the power generation system (if in a series configuration, the triggering of a node protection device will cause the entire series circuit to be interrupted, which will affect all downstream nodes in the same series circuit; if in a parallel configuration, it will affect the part connected to the common bus, causing the transmission line to fail). At the same time, it is difficult for technicians to confirm whether the sudden increase in the current node load is caused by a problem with the node itself or by misprotection of other nodes, which will lead to misjudgment for technicians. Therefore, it is crucial to calculate the probability of misprotection. The misprotection trigger probability data helps identify potential misprotection problems and provides data support for the protection configuration and adjustment of the system to improve the reliability and stability of the system.

[0037] Preferably, step S32 includes the following steps:

[0038] Step S321: performing transient harmonic intensity calculation between different load nodes on the load node harmonic interference intensity data to obtain transient harmonic intensity data of the load node; performing transient intensity continuity analysis between different load nodes on the load node transient harmonic intensity data to obtain transient intensity continuity data;

[0039] Step S322: extracting the increasing intervals between different load nodes of the transient harmonic intensity data of the load nodes according to the transient intensity persistence data, and obtaining the transient harmonic intensity increasing intervals;

[0040] Step S323: calculating the overlapping intervals of increasing speed rates between different load nodes for the transient harmonic intensity increasing intervals to obtain the overlapping intervals of increasing speed rates;

[0041] Step S324: sorting the harmonic intensity difference values ​​of the non-overlapping intervals of the transient harmonic intensity increasing intervals according to the overlapping intervals of the increasing rate, to obtain harmonic intensity difference sorting data;

[0042] Step S325: performing piecewise Simpson integration processing on the transient harmonic intensity increasing interval according to the harmonic intensity difference sorting data to obtain transient harmonic intensity approximate data;

[0043] Step S326: Calculate the transient interference order gain between different load nodes according to the transient harmonic intensity approximation data and the harmonic intensity difference sorting data to obtain transient interference order gain data.

[0044] The present invention obtains transient harmonic intensity data between different load nodes by analyzing the harmonic interference intensity data of load nodes. The data reflects the degree of transient harmonic influence of the system at different nodes, performs continuity analysis on the transient harmonic intensity data, evaluates the changing trend and duration of the transient harmonic intensity between different load nodes, thereby identifying the continuity and stability of the harmonic influence, and extracts the increasing interval in the transient harmonic intensity data of the load node based on the transient intensity continuity data. The interval identifies the nodes where the harmonic intensity increases significantly within a specific time period, analyzes the transient harmonic intensity increasing interval, and calculates the increasing rate overlapping interval between different load nodes. The interval indicates the overlapping part of the harmonic intensity growth rate in adjacent nodes, and the harmonic intensity difference of the transient harmonic intensity increasing interval is sorted according to the increasing rate overlapping interval. It is helpful to determine at which nodes the difference in harmonic intensity growth is large, and the transient harmonic intensity increasing interval is processed by piecewise Simpson integration based on the harmonic intensity difference sorting data. This processing method can effectively approximate the overall intensity change of harmonic influence, calculate the transient interference order gain between different load nodes according to the transient harmonic intensity approximate data and the harmonic intensity difference sorting data. This data helps evaluate the frequency response characteristics and interference propagation of the system under the influence of harmonics.

[0045] Preferably, step S34 includes the following steps:

[0046] Step S341: Calculate the node heating rate for the load node thermal effect trend data to obtain the node heating rate data;

[0047] Step S342: Calculate the thermal cycle influence weight according to the node heating rate data and the load node thermal effect trend data to obtain the thermal cycle influence weight data;

[0048] Step S343: performing a thermal accumulation effect simulation on the load node thermal effect trend data according to the thermal cycle influence weight data to obtain thermal accumulation effect data;

[0049] Step S344: performing multi-node heat loss range difference calculation on the load node heat effect trend data according to the heat accumulation effect data and the heat cycle influence weight data to obtain multi-node heat loss range difference data;

[0050] Step S345: performing node load critical state analysis according to transient interference order gain data and multi-node heat loss range difference data to obtain node load critical state data;

[0051] Step S346: Perform performance failure condition state simulation according to the node load critical state data and the multi-node heat loss range difference data to obtain performance failure condition state data.

[0052] The present invention obtains the rate data of node heating by analyzing and calculating the load node thermal effect trend data. The data reflects how the node temperature responds and changes when the load changes or the workload changes. Based on the node heating rate data and the load node thermal effect trend data, the influence weight of the thermal cycle is calculated. The weight reflects the intensity and frequency of the thermal cycle caused by the load change in different nodes during the operation of the system, which helps to identify the potential thermal cycle problems in the system. The load node thermal effect trend data is simulated using the thermal cycle impact weight data to obtain the thermal accumulation effect data. These data reflect the cumulative changes of the node temperature over time during the operation of the system, which helps to predict the thermal effect accumulation under long-term operation. According to the thermal accumulation effect data and the thermal cycle impact weight data, the heat loss range difference between multiple load nodes is calculated. The difference data helps to determine the difference in heat loss between different nodes in the system, indicating the potential uneven temperature distribution or node overload; combined with the transient interference order gain data and the multi-node thermal loss range difference data, the critical state analysis of the node load is performed. This data helps determine under what conditions the load node will reach its load limit, that is, the critical state of the node under the combined action of load and thermal effect. Based on the node load critical state data and multi-node heat loss range difference data, the performance failure condition state is simulated and predicted. This data provides the performance failure of the node when facing extreme load and thermal effect conditions, helping system designers or operation and maintenance personnel prevent potential equipment failures and system failures.

[0053] Preferably, step S35 includes the following steps:

[0054] Step S351: screening the load nodes with high fault risk for the load node distribution data according to the transient interference order gain data and the performance failure condition status data to obtain the load node data with high fault risk;

[0055] Step S352: Calculate the distribution dispersion of the fault high risk load node data to obtain the fault high risk node distribution discrete data;

[0056] Step S353: performing a node distribution geometric correlation analysis based on the discrete data of the high-risk node distribution of failures to obtain high-risk node distribution geometric correlation data;

[0057] Step S354: Calculate the fault chain effect trigger probability of the node distribution geometric association data according to the transient interference order gain data and the performance failure condition state data to obtain the fault chain effect trigger probability data;

[0058] Step S355: Calculate the misprotection triggering probability between different load nodes based on the transient interference order gain data, the performance failure condition status data and the fault chain effect triggering probability data for the load node distribution data to obtain the node misprotection probability data.

[0059] The present invention analyzes and screens the load node distribution data by combining transient interference order gain data and performance failure condition status data, identifies the load nodes with high risks in the system, which are more likely to fail or fail under certain conditions and are the key parts that need special attention in the system. For the identified high-risk load nodes, the discreteness of their distribution is calculated. This data helps to evaluate the unevenness of the nodes in spatial distribution, which in turn affects the fault propagation path and risk diffusion mode between the nodes. Based on the distributed discrete data of the high-risk nodes of the fault, geometric association analysis is performed. This step aims to understand and quantify the geometric arrangement of nodes in space, explore the association between nodes, such as whether there is a concentrated distribution or a specific spatial layout pattern, combine the transient interference order gain data and the performance failure condition status data, further analyze the node distribution geometric association data, and calculate the trigger probability of the fault chain effect. This data indicates the potential path and probability of fault propagation in the system, which helps to identify the complex fault propagation mechanism between multiple nodes. Finally, combined with the transient interference order gain data, the performance failure condition status data and the fault chain effect trigger probability data, the load node distribution data is calculated for the probability of false protection triggering. This data quantifies the possibility of protective action triggering due to misjudgment or false alarm in the system, which is critical to system stability and safety.

[0060] Preferably, step S4 comprises the following steps:

[0061] Step S41: performing abnormal fault diagnosis learning on the performance failure condition status data and the node misprotection probability data based on the policy gradient algorithm to obtain node abnormal fault diagnosis data;

[0062] Step S42: normalizing the node abnormal fault diagnosis data to obtain node abnormal fault diagnosis normalized data;

[0063] Step S43: Formulate a power generation node abnormality diagnosis strategy based on the node abnormality fault diagnosis normalized data to obtain a power generation node abnormality diagnosis strategy.

[0064] The present invention utilizes a policy gradient algorithm, combined with performance failure condition status data and node misprotection probability data, to perform abnormal fault diagnosis learning. The main purpose of this step is to learn, identify and understand abnormal patterns and fault behaviors. Through algorithm learning, the system can grasp the abnormal performance of nodes under different conditions, including its association with performance failure conditions and misprotection events, and normalize the abnormal fault diagnosis data obtained by learning. This process ensures the consistency and comparability of data during quantification and analysis, so that the abnormal conditions of different nodes can be effectively compared and analyzed. The normalized data helps to establish a standardized abnormal diagnosis benchmark for subsequent strategy formulation and decision support, and formulates an abnormal diagnosis strategy for power generation nodes based on the normalized node abnormal fault diagnosis data. The strategy may include setting an early warning threshold, selecting an abnormal pattern recognition algorithm, and responding to abnormal events. By formulating targeted strategies, the system can respond quickly when an abnormality occurs in a node, reduce the impact of the fault, and improve the reliability and stability of the system.

[0065] Preferably, the present invention further provides a photovoltaic power generation system abnormal fault diagnosis system based on distributed sensing, which is used to execute the photovoltaic power generation system abnormal fault diagnosis method based on distributed sensing as described above, and the photovoltaic power generation system abnormal fault diagnosis system based on distributed sensing includes:

[0066] The real-time current correction module is used to collect multi-modal sensor data of photovoltaic power generation nodes in the photovoltaic power generation system based on distributed sensors to obtain multi-modal sensor data of power generation nodes; extract real-time current data of different power generation nodes from the multi-modal sensor data of power generation nodes to obtain real-time current data of power generation nodes; correct the real-time current environmental impact under different environmental conditions for the real-time current data of power generation nodes to obtain real-time current correction data of power generation nodes;

[0067] The harmonic conduction analysis module is used to analyze the current waveforms of different power generation nodes based on the real-time current correction data of the power generation nodes to obtain the current waveform data of the power generation nodes; to analyze the path conduction harmonic superposition effect based on the current waveform data of the power generation nodes to obtain the path conduction harmonic superposition effect data; to identify the node load distribution based on the path conduction harmonic superposition effect data to obtain the load node distribution data;

[0068] The node misprotection analysis module is used to calculate the transient interference order gain between different load nodes on the load node distribution data to obtain the transient interference order gain data; simulate the performance failure condition state according to the transient interference order gain data to obtain the performance failure condition state data; calculate the misprotection trigger probability between different load nodes on the load node distribution data according to the transient interference order gain data and the performance failure condition state data to obtain the node misprotection probability data;

[0069] The fault diagnosis learning module is used to perform abnormal fault diagnosis learning on performance failure condition status data and node misprotection probability data to obtain node abnormal fault diagnosis data; and formulate a power generation node abnormal diagnosis strategy based on the node abnormal fault diagnosis data to obtain a power generation node abnormal diagnosis strategy.

[0070] The beneficial effects of the present invention are:

[0071] The present invention uses distributed sensors to collect data from multiple nodes of a photovoltaic power generation system, including data under various environmental conditions, such as temperature, light intensity, and various environmental conditions during photovoltaic power generation. By extracting real-time current data and correcting for environmental impacts, accurate real-time current correction data of power generation nodes is ensured. This helps to improve the operating efficiency and stability of the power generation system, reduce the interference of environmental impacts on abnormal fault diagnosis, perform detailed waveform analysis on the corrected real-time current data, understand the changing trends and characteristics of the current of the power generation node in different time periods, and reveal the harmonic superposition effect of the current in the system through the path conduction harmonic superposition effect data analysis, which helps to identify potential problems and optimization opportunities in the system. Through the path conduction harmonic superposition effect data analysis, the load distribution of different nodes in the power generation system can be identified and understood. This data is crucial for optimizing system operation, improving energy efficiency, and reducing energy loss. By calculating the transient interference order gain between different load nodes, the system can better understand and quantify the mutual influence and interference degree between load nodes. According to the simulation of performance failure condition status data, the system can predict potential performance failures and take maintenance and repair measures in advance, thereby reducing system downtime and repair costs. By calculating the false protection trigger probability between different load nodes, the possibility of false alarms under abnormal conditions can be evaluated, and then the setting and adjustment of protection equipment can be optimized to reduce the impact of false alarms on system operation. By calculating the false protection trigger probability between different load nodes, the possibility of false alarms under abnormal conditions can be evaluated, and then the setting and adjustment of protection equipment can be optimized to reduce the impact of false alarms on system operation. Based on the performance failure condition status data and false protection trigger probability data, abnormal fault diagnosis learning is carried out. The system can learn and identify abnormal modes and fault characteristics of different load nodes, thereby improving the accuracy and timeliness of fault diagnosis. According to the abnormal fault diagnosis data, targeted abnormal diagnosis strategies for power generation nodes are formulated, including fault detection, location, diagnosis and repair steps, to help the operation and maintenance team respond to and solve problems quickly and minimize system downtime and losses. Therefore, the present invention optimizes the traditional abnormal fault diagnosis method of photovoltaic power generation system based on distributed sensing, solves the problems of inaccurate abnormal fault diagnosis and analysis and accurate identification of complex interference in this method, improves the accuracy of abnormal fault diagnosis and analysis, and improves the ability to accurately identify complex interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic diagram of the steps of a method for diagnosing abnormal faults in a photovoltaic power generation system based on distributed sensing;

[0073] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0074] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0075] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0076] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0077] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0078] It should be understood that although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, and the term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0079] To achieve this, please refer to Figures 1 to 3 , a photovoltaic power generation system abnormal fault diagnosis method based on distributed sensing, the method comprising the following steps:

[0080] Step S1: Based on distributed sensors, multimodal sensing data of photovoltaic power generation nodes of the photovoltaic power generation system are collected to obtain multimodal sensing data of power generation nodes; real-time current data of different power generation nodes are extracted from the multimodal sensing data of power generation nodes to obtain real-time current data of power generation nodes; real-time current environmental impact correction is performed on the real-time current data of power generation nodes under different environmental conditions to obtain real-time current correction data of power generation nodes;

[0081] Step S2: Perform current waveform analysis of different power generation nodes on the real-time current correction data of the power generation node to obtain current waveform data of the power generation node; perform path conduction harmonic superposition effect analysis based on the current waveform data of the power generation node to obtain path conduction harmonic superposition effect data; perform node load distribution identification based on the path conduction harmonic superposition effect data to obtain load node distribution data;

[0082] Step S3: calculating the transient interference order gain between different load nodes on the load node distribution data to obtain transient interference order gain data; simulating the performance failure condition state according to the transient interference order gain data to obtain performance failure condition state data; calculating the misprotection triggering probability between different load nodes on the load node distribution data according to the transient interference order gain data and the performance failure condition state data to obtain node misprotection probability data;

[0083] Step S4: Perform abnormal fault diagnosis learning on the performance failure condition status data and the node misprotection probability data to obtain node abnormal fault diagnosis data; formulate a power generation node abnormal fault diagnosis strategy based on the node abnormal fault diagnosis data to obtain a power generation node abnormal fault diagnosis strategy.

[0084] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for diagnosing abnormal faults of a photovoltaic power generation system based on distributed sensing according to the present invention. In this example, the method for diagnosing abnormal faults of a photovoltaic power generation system based on distributed sensing includes the following steps:

[0085] Step S1: Based on distributed sensors, multimodal sensing data of photovoltaic power generation nodes of the photovoltaic power generation system are collected to obtain multimodal sensing data of power generation nodes; real-time current data of different power generation nodes are extracted from the multimodal sensing data of power generation nodes to obtain real-time current data of power generation nodes; real-time current environmental impact correction is performed on the real-time current data of power generation nodes under different environmental conditions to obtain real-time current correction data of power generation nodes;

[0086] In an embodiment of the present invention, a plurality of types of sensor data of each node of a photovoltaic power generation system are collected through a distributed sensor network, wherein the sensor types include: a current sensor: monitoring the output current of a photovoltaic module; a voltage sensor: monitoring the output voltage of a photovoltaic module; a temperature sensor: measuring the ambient temperature and the temperature of a photovoltaic module; an irradiance sensor: measuring the intensity of solar radiation; a humidity sensor: measuring the air humidity; a wind speed and direction sensor: measuring the wind speed and the wind direction; the above sensors are installed on each photovoltaic power generation module or node, and the data collection frequency of each sensor is determined (for example, collecting data once every 30 minutes); current data is extracted from the sensor data stream, and other non-current data (such as temperature, irradiance, etc.) is classified and stored for subsequent analysis; a relationship model between environmental variables (temperature, irradiance, etc.) and current is established based on historical data and physical models; commonly used methods include linear regression, multivariate regression, and machine learning models; the real-time collected environmental data (temperature, irradiance, etc.) and the established model are used to calculate the corrected current value, and the corrected current data is stored in a database for subsequent analysis and use.

[0087] Step S2: Perform current waveform analysis of different power generation nodes on the real-time current correction data of the power generation node to obtain current waveform data of the power generation node; perform path conduction harmonic superposition effect analysis based on the current waveform data of the power generation node to obtain path conduction harmonic superposition effect data; perform node load distribution identification based on the path conduction harmonic superposition effect data to obtain load node distribution data;

[0088] In the embodiment of the present invention, the time series current data is extracted from the corrected current data obtained in step S1 for waveform analysis, ensuring that the data has a high sampling rate to capture the detailed features in the current waveform, using a low-pass, high-pass or band-pass filter to remove noise in the current signal, using wavelet transform, Fourier transform or fast Fourier transform (FFT) to decompose the current signal into frequency components to analyze harmonic characteristics, extract characteristic parameters of the current waveform, such as harmonic content (THD), peak factor, root mean square value (RMS), obtain the harmonic components of the current waveform through FFT analysis, calculate the amplitude and phase of each harmonic, identify the main harmonic source, determine which device or node these harmonics are caused by, and construct an electrical connection model of each node in the photovoltaic system, including components such as wires, inductors, capacitors, etc., using Circuit simulation tools (such as PSCAD and MATLAB Simulink) simulate the conduction path of harmonics in the system, analyze how the harmonics of different nodes are superimposed in the path, resulting in an increase or decrease in the harmonic current in the system, and calculate the current distribution after the superposition of harmonics on different paths through the superposition effect model. The impact of harmonic superposition on the system, such as the reduction of power quality, overheating of equipment and reduced efficiency, is evaluated, and problem nodes and conduction paths caused by high harmonic levels are identified. According to the current waveform and harmonic characteristics of the node, the load type (such as nonlinear load, linear load, etc.) is identified. The load characteristics are automatically identified and classified through machine learning methods (such as cluster analysis and classification algorithms), and a load distribution model of the system is established to display the load conditions of each node and the total load distribution, and obtain the load node distribution data.

[0089] Step S3: calculating the transient interference order gain between different load nodes on the load node distribution data to obtain transient interference order gain data; simulating the performance failure condition state according to the transient interference order gain data to obtain performance failure condition state data; calculating the misprotection triggering probability between different load nodes on the load node distribution data according to the transient interference order gain data and the performance failure condition state data to obtain node misprotection probability data;

[0090] In the embodiment of the present invention, the order gain of transient interference between load nodes is calculated to quantify the interference transmission effect between different nodes (i.e., the increasing order of the intensity of transient interference). Transient interference refers to short-term current or voltage fluctuations caused by faults or external electromagnetic interference in the system. The load node distribution data and current waveform data are obtained from step S2. A circuit simulation tool (such as PSCAD, MATLAB Simulink) is used to simulate the transmission process of transient interference in the system. The response amplitude of each load node to transient interference is calculated and defined as the order gain. For example, for the nth order interference gain G n , which can be expressed as the transient response gain between node i and node j: It is used to quantify the transmission effect of transient interference between different nodes and represents the transient voltage or current response gain from node i to node j, V i (t), which is the voltage (or current) signal measured at node i at time t, indicating the starting signal of the transient interference source, that is, the input signal, which is the voltage or current change of a specific node when a transient interference event occurs; V j (t) This is the voltage (or current) signal measured at node j at time t, which represents the response signal of transient interference at node j, that is, the output signal, which is the voltage or current change of the node affected by transient interference; for different interference events, extract the transient response amplitude of each node, and calculate the gain between adjacent nodes; simulate the system performance failure state under transient interference conditions, identify key nodes and vulnerable links, define the system performance failure conditions, such as voltage drop exceeding a certain threshold, frequency deviation exceeding the tolerance, set different types of failure conditions (such as short-term voltage interruption, frequency fluctuation, etc.) to simulate various failure scenarios, and use simulation tools (such as MATLAB) Simulink, PSCAD) to establish a detailed simulation model of the photovoltaic system, introduce transient interference gain data into the simulation model, simulate the response of different nodes under transient interference conditions, execute multiple simulation scenarios, simulate the impact of different types of transient interference (such as current mutation, voltage fluctuation, etc.) on the system, record the system response and failure conditions in each simulation scenario, identify the nodes or paths that are prone to system failure, and calculate the probability of false protection triggering between load nodes based on transient interference gain and performance failure simulation data. False protection triggering means that in non-fault conditions, the protection device erroneously operates, resulting in unexpected power outages or equipment shutdowns in the system, which will cause other nodes to shut down due to the previous node or adjacent nodes, causing the current node load to suddenly increase, and the load becomes larger , which makes it easier to cause the probability of failure of other equipment nodes, thereby causing the overall fault spread of the power generation system to become larger and larger; establish a mathematical model of false protection triggering, consider the transient interference response of each node and the triggering characteristics of the protection device, define the probability function of false protection triggering, consider the intensity of transient interference and the protection threshold of the system, obtain the frequency of transient interference triggering the protection device action at different nodes from simulation and historical data, establish a false protection triggering probability model based on the transient interference order gain and performance failure conditions, use probability density function (PDF) or cumulative distribution function (CDF) to represent the false protection triggering probability, calculate the probability of false protection triggering of each node under different interference conditions, and use statistical methods such as Monte Carlo simulation to estimate the false protection triggering probability of the entire system.

[0091] Step S4: Perform abnormal fault diagnosis learning on the performance failure condition status data and the node misprotection probability data to obtain node abnormal fault diagnosis data; formulate a power generation node abnormal fault diagnosis strategy based on the node abnormal fault diagnosis data to obtain a power generation node abnormal fault diagnosis strategy.

[0092] In the embodiment of the present invention, performance failure condition status data is collected and sorted, including the response and failure of the system under different conditions, node misprotection probability data is obtained, reflecting the misprotection triggering probability of different nodes under different interference conditions, and useful features are extracted from the original data, such as voltage waveform features, transient response features, and system status indicators. Anomaly detection models based on machine learning or deep learning, including support vector machines (SVMs), decision trees, random forests, neural networks, and convolutional neural networks (CNNs), consider using supervised learning (fault data with known labels) or unsupervised learning (such as cluster analysis and isolation forest) methods, The data set is divided into a training set and a test set. The model is trained to identify normal and abnormal states, and the accuracy and robustness of the model are evaluated. The parameters of the model are optimized through cross-validation, grid search and other methods. The performance of the model is evaluated using the test data set, including accuracy, recall, and F1 score indicators. Model validation is performed to ensure the effectiveness and reliability of the model under different environmental conditions. According to the diagnostic data, the types of abnormalities that occur in the system are identified and classified, such as transient faults, continuous faults, and equipment aging. Response strategies for different types of abnormalities are formulated, including early warning, isolation, and repair. According to the diagnostic results of the model, automated diagnostic rules and thresholds are defined. For example, when the voltage fluctuation of a node is detected to exceed a certain threshold, an early warning or protection action is triggered. Based on historical data and diagnostic results, the response strategy of the system is optimized to improve the speed and accuracy of fault response. Combined with fault history and environmental data, targeted response measures are formulated to reduce the impact of faults on system operation. The diagnostic strategy is integrated into the system's monitoring and management platform and applied to system operation in real time. Through monitoring and logging, the effectiveness of the diagnostic strategy is continuously evaluated, and necessary adjustments and improvements are made. The effect of the diagnostic strategy is verified in the actual operating environment to ensure its stability and reliability under different conditions.

[0093] Preferably, step S1 comprises the following steps:

[0094] Step S11: collecting multimodal sensing data of photovoltaic power generation nodes of the photovoltaic power generation system based on distributed sensors to obtain multimodal sensing data of power generation nodes; wherein the multimodal sensing data of power generation nodes includes current sensing data of power generation nodes and environmental status data of power generation nodes;

[0095] Step S12: extracting real-time current data of different power generation nodes from the current sensing data of the power generation nodes to obtain real-time current data of the power generation nodes;

[0096] Step S13: Calculating the weights of environmental impact factors of different power generation nodes on the real-time current data of the power generation nodes according to the environmental status data of the power generation nodes, and obtaining environmental impact factor weight data;

[0097] Step S14: Correcting the real-time current environmental impact of different environmental conditions on the real-time current data of the power generation node according to the environmental impact factor weight data to obtain the real-time current correction data of the power generation node.

[0098] In an embodiment of the present invention, a distributed sensor system is used to collect data from each power generation node in a photovoltaic power generation system. The collected data includes current sensing data and environmental status data of the power generation node, which reflects the operating status of the power generation node under different environmental conditions. Real-time current data of each power generation node is extracted from the collected data, and the real-time current data reflects the current power generation performance and operating status of each power generation node. According to the environmental status data of the power generation node, the influence weight of each environmental factor on the real-time current of the power generation node is calculated. The environmental factors include temperature, light intensity, wind speed, etc., which will affect the power generation efficiency of the photovoltaic power generation system. According to the environmental influence factor weight data calculated in step S13, the real-time current data of the power generation node extracted in step S12 is corrected. The corrected data reflects the actual current output of the power generation node under different environmental conditions, and takes into account the influence of environmental factors on the performance of the photovoltaic power generation system.

[0099] Preferably, step S2 comprises the following steps:

[0100] Step S21: performing node current conduction topology path analysis on the real-time current correction data of the power generation node to obtain node current conduction topology path data;

[0101] Step S22: performing current waveform analysis of different power generation nodes on the real-time current correction data of the power generation nodes according to the node current conduction topology path data to obtain current waveform data of the power generation nodes;

[0102] Step S23: analyzing the path conduction harmonic superposition effect of the node current conduction topology path data according to the power generation node current waveform data to obtain the path conduction harmonic superposition effect data;

[0103] Step S24: identifying the node load distribution of the node current conduction topology path data according to the path conduction harmonic superposition effect data to obtain load node distribution data.

[0104] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0105] Step S21: performing node current conduction topology path analysis on the real-time current correction data of the power generation node to obtain node current conduction topology path data;

[0106] In an embodiment of the present invention, the real-time current correction data of the power generation node obtained in step S14 is used as the basic data for analysis to ensure that the data contains sufficient time series and node information for subsequent topological analysis. Based on the current data, the conduction path and connection relationship between each power generation node are determined. This can be achieved by calculating the correlation between the current data or the conduction path between the current values, converting the conduction relationship between the nodes into a topological structure, and forming node current conduction topological path data. The topological path data should include the direct and indirect conduction relationships between the nodes, as well as the strength or weight of the conduction path. The topological path data is analyzed to evaluate the transmission efficiency and path selection of the current in the system.

[0107] Step S22: performing current waveform analysis of different power generation nodes on the real-time current correction data of the power generation nodes according to the node current conduction topology path data to obtain current waveform data of the power generation nodes;

[0108] In the embodiment of the present invention, the current data of the selected node is analyzed in the time domain, including amplitude, frequency, and phase, and the mean, peak, and waveform shape of the current waveform are calculated to understand the basic characteristics of the current, and the frequency distribution of the current waveform is analyzed using tools such as Fourier transform or wavelet transform. The main harmonic components and their impact on system stability and efficiency are determined, and the analyzed current waveform data is visualized, such as drawing a waveform graph or a spectrum graph.

[0109] Step S23: analyzing the path conduction harmonic superposition effect of the node current conduction topology path data according to the power generation node current waveform data to obtain the path conduction harmonic superposition effect data;

[0110] In an embodiment of the present invention, the current waveform data in the time domain is converted to the frequency domain, the harmonic components at each frequency are analyzed, and the main harmonic frequencies and their amplitudes and phases relative to the fundamental wave are determined. These harmonic components are a reflection of the nonlinearity and frequency distortion in the current waveform. According to the topological path data between the power generation nodes, how the harmonics on different paths in the current waveform are superimposed is analyzed, and the harmonic components on each path are considered, and their mutual influence in the conduction process between the nodes is calculated. This helps to understand the diffusion and propagation process of harmonics in the system, and how they affect the stability and power quality of the system. According to the results of the analysis, the path conduction harmonic superposition effect data is generated. The data should include the main harmonic components on different paths and the intensity and direction of their superposition effects. This data is critical for identifying harmonic diffusion paths and optimizing harmonic filtering and compensation strategies in the system.

[0111] Step S24: identifying the node load distribution of the node current conduction topology path data according to the path conduction harmonic superposition effect data to obtain load node distribution data.

[0112] In an embodiment of the present invention, the path conduction harmonic superposition effect data obtained in step S23 is used. Ensure that the data contains sufficient detailed information to distinguish the current load distribution between different nodes. According to the path conduction harmonic superposition effect data, analyze the current load carried by each node. Consider the impact of the harmonic superposition effect on different nodes, and evaluate the current load intensity and distribution of each node. Including the existence of some nodes carrying too much current load or harmonic load, it is necessary to optimize the current distribution or add compensation equipment to improve system performance. Based on the analysis results, generate load node distribution data. The data should provide information on the current load level, harmonic load level and load distribution unevenness of each node. This data is very important for setting the system's load balancing strategy and optimizing the system's operating status.

[0113] Preferably, step S23 includes the following steps:

[0114] Step S231: extracting the asymmetric waveform region of the current waveform data of the power generation node to obtain the current asymmetric waveform region data; performing harmonic identification on the current asymmetric waveform region data using a preset current harmonic identification model to obtain the current asymmetric harmonic data;

[0115] Step S232: performing asymmetric harmonic regional path mapping on the node current conduction topology path data according to the current asymmetric harmonic data to obtain the conduction path asymmetric harmonic data;

[0116] Step S233: calculating the crest factor increment coefficient between different node conduction paths for the asymmetric harmonic data of the conduction path to obtain the harmonic crest factor increment coefficient;

[0117] Step S234: calculating the crest factor progressive magnitude between different node conduction paths for the asymmetric harmonic data of the conduction path according to the harmonic crest factor progressive coefficient, and obtaining crest factor progressive magnitude data;

[0118] Step S235: calculating the magnitude mean difference series of the crest factor progressive magnitude data to obtain the magnitude mean difference series; performing magnitude truncation dynamic modal decomposition on the asymmetric harmonic data of the conduction path according to the magnitude mean difference series to obtain asymmetric harmonic magnitude modal decomposition data;

[0119] Step S236: analyzing the path conduction harmonic superposition effect of the node current conduction topology path data according to the asymmetric harmonic magnitude modal decomposition data and the harmonic crest factor increasing coefficient to obtain the path conduction harmonic superposition effect data.

[0120] In an embodiment of the present invention, the current waveform data of the power generation node obtained in step S22 is used to extract the asymmetric region of each current waveform, that is, to detect and mark the asymmetric part in the waveform. The asymmetric waveform region usually indicates the distortion, harmonics or other nonlinear effects in the current waveform, and a mathematical method is used to detect the asymmetric characteristics in the current waveform. Asymmetric waveforms are usually caused by harmonic distortion or system imbalance. The skewness and kurtosis of the waveform are calculated to quantify the asymmetry of the waveform, and the harmonic components of the asymmetric waveform region data are automatically identified using a preset current harmonic identification model. The model is usually based on frequency domain analysis methods such as Fourier transform or wavelet transform, which is used to detect and identify the harmonic components in the current waveform, determine the main harmonic frequency, amplitude and phase information in each asymmetric waveform region, and generate current asymmetric harmonic data according to the output results of the harmonic identification model. The data should include the harmonic component information in the asymmetric waveform area of ​​each power generation node, such as the main harmonic frequency, amplitude and phase; using the current asymmetric harmonic data obtained in step S231, according to the node current conduction topology path data, the asymmetric harmonic data is mapped to each conduction path, and the main asymmetric harmonic components and their strengths on each path are determined, which helps to understand the distribution and influence of asymmetric harmonics in the system conduction path; using the conduction path asymmetric harmonic data obtained in step S232. Ensure that the data contains the asymmetric harmonic components and their strengths on each conduction path, understand the concept of crest factor, which is usually used to describe the ratio of the peak value to the root mean square value of the current waveform, and is an important indicator for measuring the peak and asymmetry of the waveform. For each conduction path, calculate the crest factor of the asymmetric harmonic. The calculation of the crest factor involves the comparison of the peak value and the root mean square value of the harmonic waveform, or a specific peak calculation of the asymmetric waveform. This calculation needs to consider the main harmonic components and their distribution in the path, and generate the crest factor incremental coefficient data between different node conduction paths based on the calculated crest factor. The incremental coefficient reflects the change and transmission efficiency of the harmonics on the conduction path, and can be used to evaluate the harmonic diffusion and influence between paths, and determine the definition of the progressive magnitude, which usually refers to the comparison and magnitude difference of the crest factors between different conduction paths. The progressive magnitude can help evaluate the relative peakiness or asymmetry of the harmonic waveforms on different paths, thereby affecting the power quality and stability of the system. According to the crest factor incremental coefficient data obtained in step S233, the crest factor progressive magnitude between different node conduction paths is calculated. It involves sorting or normalizing the incremental coefficients to more clearly compare and evaluate the differences in waveform peakiness between paths, and generating crest factor progressive magnitude data based on the calculation results.The data should include the progressive magnitudes and their values ​​between different conduction paths, as well as the path sorting or priority information. The progressive magnitude data should be analyzed to evaluate the waveform changes and impact of harmonics in the system conduction paths. The magnitude mean difference series is used to describe the difference between the series in the crest factor progressive magnitude data. The calculation method of the mean difference series is determined, usually by statistical analysis of the progressive magnitude, including calculating the mean value or standard deviation. According to the crest factor progressive magnitude data, the magnitude mean difference series between different conduction paths is calculated, which involves mathematical statistical analysis of the progressive magnitude data to obtain indicators that describe the difference in series. According to the calculation results, the magnitude mean difference series data is generated, including the mean difference series and their values ​​between different conduction paths, which is used for further harmonic magnitude modal decomposition. The magnitude mean difference series data is used for non- Symmetrical harmonic magnitude modal decomposition, grouping or clustering the harmonic magnitude data to identify the dynamic modes and characteristics of different harmonic magnitudes. The decomposition results should reflect the distribution and influence of different harmonic magnitudes on the conduction path. The asymmetric harmonic magnitude modal decomposition data obtained in step S235 is used to ensure that the data contains the harmonic magnitude distribution on the conduction path and the dynamic modal information of each harmonic magnitude. According to the node current conduction topology path data, the superposition effect of harmonics on the conduction path is analyzed, the propagation and diffusion characteristics of different harmonic magnitude modes are considered, the transmission efficiency and influence of harmonics in the system are evaluated, and the path conduction harmonic superposition effect data is generated according to the analysis results, including the main harmonic superposition effect on different conduction paths, the superposition ratio of harmonic magnitudes and the complexity information of harmonic waveforms.

[0121] Preferably, step S3 comprises the following steps:

[0122] Step S31: Calculate the harmonic interference intensity of different load nodes on the load node distribution data according to the path conduction harmonic superposition effect data to obtain the load node harmonic interference intensity data;

[0123] Step S32: calculating transient interference order gain between different load nodes on the load node harmonic interference intensity data to obtain transient interference order gain data;

[0124] Step S33: performing thermal effect trend analysis between different load nodes according to transient interference order gain data and load node harmonic interference intensity data to obtain load node thermal effect trend data;

[0125] Step S34: performing a performance failure condition state simulation according to the transient interference order gain data and the load node thermal effect trend data to obtain performance failure condition state data;

[0126] Step S35: Calculate the false protection triggering probability between different load nodes based on the transient interference order gain data and the performance failure condition status data on the load node distribution data to obtain the node false protection probability data.

[0127] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0128] Step S31: Calculate the harmonic interference intensity of different load nodes on the load node distribution data according to the path conduction harmonic superposition effect data to obtain the load node harmonic interference intensity data;

[0129] In an embodiment of the present invention, the path conduction harmonic superposition effect data obtained in the above steps is used to ensure that the data includes the harmonic superposition effect on each conduction path and the degree of harmonic influence of each conduction path on each load node, and the distribution data of the load nodes is obtained, including the current characteristics, voltage characteristics and power load characteristics of each node, and the load node data is ensured to include detailed information of each key node in the system. Based on the path conduction harmonic superposition effect data, the harmonic interference intensity at different load nodes is calculated, which involves weighted calculation of the harmonic distribution on the conduction path, considering the current distribution and voltage response of each node, and generating the harmonic interference intensity data of the load node according to the calculation results, including the interference intensity level of each load node at different harmonic frequencies, and the impact assessment on the power load and system operation.

[0130] Step S32: calculating transient interference order gain between different load nodes on the load node harmonic interference intensity data to obtain transient interference order gain data;

[0131] In the embodiment of the present invention, the load node harmonic interference intensity data obtained in step S31 is used to ensure that the data includes the harmonic interference intensity information of each load node, and the transient interference order gain index is used to evaluate the transmission effect of the harmonic interference between different load nodes. The calculation method of the transient interference order gain is defined, which is usually calculated by comparing the harmonic interference intensity between each node, considering the transmission characteristics of the conduction path, and calculating the transient interference order gain between different load nodes according to the load node harmonic interference intensity data. Considering the current transmission path and voltage response characteristics of each node, the propagation speed and transmission effect of the harmonics in the system are evaluated, and the transient interference order gain data is generated according to the calculation results, including the gain coefficient between different load nodes, and the contribution of each node to the diffusion and propagation of harmonics.

[0132] Step S33: performing thermal effect trend analysis between different load nodes according to transient interference order gain data and load node harmonic interference intensity data to obtain load node thermal effect trend data;

[0133] In an embodiment of the present invention, the gain of transient interferences of different orders in the system is obtained, and the harmonic interference intensity at each load node is collected, which is usually expressed as the harmonic component of current or voltage. Based on the laws of physics and system characteristics, a thermal effect model of each load node is established. A heat balance equation is usually used to consider the loss and temperature rise caused by harmonic current. For each load node, its thermal effect under different orders of transient interference gain and harmonic interference intensity is analyzed. A simulation or numerical calculation method is used to predict the temperature change trend of the node under different conditions, determine the thermal effect characteristics of the key load nodes, identify the nodes and working conditions that cause overheating or failure, and organize the analysis results, including the temperature change trend of each load node under different transient interference and harmonic intensity.

[0134] Step S34: performing a performance failure condition state simulation according to the transient interference order gain data and the load node thermal effect trend data to obtain performance failure condition state data;

[0135] In the embodiment of the present invention, the transient interference order gain data and thermal effect trend data obtained in step S33 are used to determine the failure conditions of each load node according to system requirements and design specifications. Usually including: overheating threshold: when the node temperature exceeds this value, it is considered to be a failure, voltage or current exceeding threshold: exceeding the design range is considered to be a failure, and other conditions related to system stability and safety; based on the thermal effect trend data, a probability model of node performance failure is established, and the failure probability of the system under different interference gain and harmonic conditions is simulated at the same time. For each load node, under different transient interference conditions, its state changes within a given time range are simulated to determine the failure probability and state changes of the node under various combinations of conditions, and the simulation results are sorted out, including the failure probability and state data of each load node under different transient interference and harmonic intensities.

[0136] Step S35: Calculate the false protection triggering probability between different load nodes based on the transient interference order gain data and the performance failure condition status data on the load node distribution data to obtain the node false protection probability data.

[0137] In the embodiment of the present invention, the transient interference order gain data and performance failure condition status data obtained in steps S33 and S34 are used. The trigger characteristic data of the protection device are collected or estimated, for example, the action threshold and response time of the protection device, and the false protection triggering situation in the system is identified, such as: the transient current or voltage fluctuation caused by the interference exceeds the threshold set by the protection device, and the false protection action caused by the short-term harmonic interference. According to the transient interference and harmonic data of the load node, the probability of causing false protection triggering is calculated, and the false protection triggering probability is quantified through statistical analysis or Monte Carlo simulation method, and the false protection triggering probability of different load nodes under the same conditions is analyzed. The high-risk nodes and working conditions are identified to provide a basis for system optimization and form the false protection triggering probability data of different load nodes under various conditions.

[0138] Preferably, step S32 includes the following steps:

[0139] Step S321: performing transient harmonic intensity calculation between different load nodes on the load node harmonic interference intensity data to obtain transient harmonic intensity data of the load node; performing transient intensity continuity analysis between different load nodes on the load node transient harmonic intensity data to obtain transient intensity continuity data;

[0140] Step S322: extracting the increasing intervals between different load nodes of the transient harmonic intensity data of the load nodes according to the transient intensity persistence data, and obtaining the transient harmonic intensity increasing intervals;

[0141] Step S323: calculating the overlapping intervals of increasing speed rates between different load nodes for the transient harmonic intensity increasing intervals to obtain the overlapping intervals of increasing speed rates;

[0142] Step S324: sorting the harmonic intensity difference values ​​of the non-overlapping intervals of the transient harmonic intensity increasing intervals according to the overlapping intervals of the increasing rate, to obtain harmonic intensity difference sorting data;

[0143] Step S325: performing piecewise Simpson integration processing on the transient harmonic intensity increasing interval according to the harmonic intensity difference sorting data to obtain transient harmonic intensity approximate data;

[0144] Step S326: Calculate the transient interference order gain between different load nodes according to the transient harmonic intensity approximation data and the harmonic intensity difference sorting data to obtain transient interference order gain data.

[0145] In an embodiment of the present invention, it is ensured that the harmonic interference intensity data of the load nodes include the interference intensity of different load nodes at each harmonic frequency, and the transient harmonic intensity is calculated using the harmonic interference intensity data of the load nodes. The calculation result should reflect the change of the harmonic interference intensity of each load node under different transient conditions. Based on the transient harmonic intensity data, the transient intensity continuity analysis between the load nodes is performed, including the duration of the harmonic interference intensity under transient conditions, the volatility of the intensity, and the periodic change. According to the analysis results, the transient intensity continuity data between the load nodes is generated, including the transient harmonic interference intensity of each node in different time periods during the operation of the system, which is used for subsequent incremental interval extraction and analysis. The transient intensity continuity data obtained in step S321 is used, and based on the transient intensity continuity data, the incremental interval of the transient harmonic intensity of the load node is extracted, and the incremental interval of the harmonic interference intensity of each load node in different time periods is determined, that is, the continuous intensity increase trend area, and the transient harmonic intensity incremental interval data of the load node is generated according to the extraction result. Including the start and end time of different increment intervals of each node during the operation of the system, and the change range information of the harmonic interference intensity; using the load node transient harmonic intensity increment interval data obtained in step S322, for each load node transient harmonic intensity increment interval, calculate the increment rate of its harmonic intensity. The increment rate can be measured by calculating the slope or growth rate of the harmonic intensity, reflecting the change trend of the harmonic interference intensity, and determining the increment rate overlap interval between different load nodes according to the calculated increment rate. The overlapping interval refers to the area where the transient harmonic intensity increment rates of multiple load nodes overlap or intersect on the time axis, and the increment rate overlap interval data between load nodes is generated according to the calculation results. Including the start and end time of each overlapping interval, the information of the participating nodes, and the change of the harmonic intensity increment rate; the transient harmonic intensity of the load node in each increment rate overlap interval is calculated. The difference can be the absolute difference or relative difference of the harmonic intensity, reflecting the change amount of the harmonic intensity of different nodes in the overlapping interval, and the harmonic intensity difference of the non-overlapping interval is sorted according to the calculated harmonic intensity difference. During the sorting process, the harmonic intensity differences are arranged in increasing or decreasing order to further analyze the harmonic intensity change trend of different nodes in the overlapping interval of increasing rate, and the harmonic intensity difference sorting data is generated according to the sorting results. Including the specific values ​​and sorting information of the harmonic intensity differences between nodes in each overlapping interval, providing a reference for interference management and optimization of the system.Use the harmonic intensity difference sorting data recorded by the system, which describes the harmonic characteristics of the system under different working conditions. According to the harmonic intensity difference sorting data, the transient harmonic intensity is segmented according to the increasing interval. This step can use the numerical analysis method to decompose the continuous harmonic intensity data into multiple interval segments, and perform Simpson integral processing on each harmonic intensity increasing interval to obtain the approximate data of the transient harmonic intensity. Simpson integral is a numerical integration method, which is suitable for handling the numerical integration problem of continuous functions and can effectively approximate the value of the transient harmonic intensity in each interval. According to the result of Simpson integral, the approximate data of transient harmonic intensity is generated to describe the harmonic response characteristics of the system under different working conditions. Based on the approximate data of transient harmonic intensity and the sorted data of harmonic intensity difference, the transient interference order gain between different load nodes is calculated. The calculation process can be carried out through numerical simulation or analytical solution method to evaluate the response degree of each node in the system under the action of harmonics. According to the calculation results, the transient interference order gain data is generated to describe the increasing order or increasing law of transient harmonic interference intensity of different load nodes in the system under the action of transient harmonics.

[0146] Preferably, step S34 includes the following steps:

[0147] Step S341: Calculate the node heating rate for the load node thermal effect trend data to obtain the node heating rate data;

[0148] Step S342: Calculate the thermal cycle influence weight according to the node heating rate data and the load node thermal effect trend data to obtain the thermal cycle influence weight data;

[0149] Step S343: performing a thermal accumulation effect simulation on the load node thermal effect trend data according to the thermal cycle influence weight data to obtain thermal accumulation effect data;

[0150] Step S344: performing multi-node heat loss range difference calculation on the load node heat effect trend data according to the heat accumulation effect data and the heat cycle influence weight data to obtain multi-node heat loss range difference data;

[0151] Step S345: performing node load critical state analysis according to transient interference order gain data and multi-node heat loss range difference data to obtain node load critical state data;

[0152] Step S346: Perform performance failure condition state simulation according to the node load critical state data and the multi-node heat loss range difference data to obtain performance failure condition state data.

[0153] In an embodiment of the present invention, the load node thermal effect trend data obtained in step S33 is used, and based on the load node thermal effect trend data, the node heating rate of each load node is calculated, and a numerical differential method, such as a central difference formula, is used. If the data is discrete, a finite difference method can be used for calculation. The node heating rate reflects the temperature change rate of the load node due to the thermal effect during system operation. The node heating rate data of the load node is generated according to the calculation results. Including the heating rate value of each node, which is used for subsequent thermal cycle influence weight calculation and thermal accumulation effect simulation. Using the node heating rate data obtained in step S341, using the load node thermal effect trend data obtained in step S33, according to the node heating rate data and the load node thermal effect trend data, the thermal cycle influence weight is calculated, and a mathematical model between the node temperature change and various influencing factors can be established by regression analysis or machine learning methods. The calculation can take into account the size of the node heating rate, the change trend, and the cumulative influence degree of the thermal effect trend, and generate the thermal cycle influence weight data between the load nodes according to the calculation results. Including the thermal cycle impact weight value of each node, it reflects the potential impact of the temperature change caused by the thermal effect of the node during the operation of the system on the stability of the system. Using the thermal cycle impact weight data obtained in step S342, using the load node thermal effect trend data obtained in step S33, based on the thermal cycle impact weight data and the load node thermal effect trend data, the thermal accumulation effect simulation of the load node is performed. During the simulation process, the cumulative effect of the temperature change of the node is considered, the thermal stability and thermal effect accumulation of the node in long-term operation are analyzed, and the thermal accumulation effect data of the load node is generated according to the simulation results. Including the temperature change trend of each node in the long-term operation of the system, the degree of cumulative effect and the long-term impact assessment on the operation of the system, the node heating rate data can be integrated over time to simulate the cumulative thermal load impact on the node over a long period of time, and the engineering durability analysis methods, such as fatigue analysis, aging models, etc., are used to evaluate the thermal stability of the node under long-term working conditions. According to the thermal accumulation effect data and thermal cycle influence weight data, the thermal effect trend of the load node is analyzed. For each pair of load nodes, the thermal accumulation effect data and thermal cycle influence weight data are combined to calculate the difference in the heat loss range. The temperature variation range of the node under different operating conditions can be evaluated based on the thermal stability model and probability distribution. Through numerical methods such as Monte Carlo simulation or probability density function fitting, the temperature variation range of the node under various conditions is determined. The difference in thermal effect trends between nodes and the difference in temperature variation range of different nodes caused by thermal cycle effects are considered in the calculation process; for each node, the transient interference order gain data and the heat loss range difference data are combined to calculate the node's carrying capacity. The carrying capacity can be expressed as the maximum load that the node can stably operate under given interference and heat loss conditions.The calculation needs to consider the electrical pressure of transient interference on the node and the thermal management pressure of heat loss on the node, compare the node's carrying capacity with the actual load, evaluate whether the node is in a critical state (that is, the load is close to or exceeds its carrying capacity), identify overloaded nodes or nodes with excessive heat loss, and adopt reliability engineering methods to combine the thermal stability and electrical performance of the node to evaluate the reliability of the node under various working conditions, consider the node's failure mode and failure mechanism, and analyze the node's load limit under different operating conditions, so as to further optimize the system or take protective measures. The calculated node load critical state data is stored in the database for subsequent analysis, analyze the load capacity of different nodes under various operating conditions, and identify Identify potential risk points in the system; describe the carrying capacity of each node under the current operating conditions and describe the heat loss characteristics of each node and its range difference according to step S345, set simulation conditions, including different load levels, ambient temperature, transient interference and other variables, use these conditions to simulate the operating status of nodes under different conditions, evaluate whether they will reach or exceed their carrying capacity, build a failure model, take the transient interference gain, heat loss range difference and actual load of the node as input, simulate the failure probability under various conditions, the failure model can be constructed using Monte Carlo simulation, fault tree analysis (FTA) or a model based on machine learning, and analyze which nodes have performance failures under different conditions based on the simulation results.

[0154] Preferably, step S35 includes the following steps:

[0155] Step S351: screening the load nodes with high fault risk for the load node distribution data according to the transient interference order gain data and the performance failure condition status data to obtain the load node data with high fault risk;

[0156] Step S352: Calculate the distribution dispersion of the fault high risk load node data to obtain the fault high risk node distribution discrete data;

[0157] Step S353: performing a node distribution geometric correlation analysis based on the discrete data of the high-risk node distribution of failures to obtain high-risk node distribution geometric correlation data;

[0158] Step S354: Calculate the fault chain effect trigger probability of the node distribution geometric association data according to the transient interference order gain data and the performance failure condition state data to obtain the fault chain effect trigger probability data;

[0159] Step S355: Calculate the misprotection triggering probability between different load nodes based on the transient interference order gain data, the performance failure condition status data and the fault chain effect triggering probability data for the load node distribution data to obtain the node misprotection probability data.

[0160] In the embodiment of the present invention, the high-risk index threshold of the power generation node is obtained from the photovoltaic power generation manufacturer, and the transient interference order gain and performance failure condition status data of each node are checked. The nodes whose gain and failure probability exceed the preset threshold are screened out, and these nodes are marked as high-risk nodes. The distribution dispersion is used to quantify the distribution of high-risk nodes in the system. The discreteness formula or clustering method can be used to calculate and analyze the distribution characteristics of high-risk nodes to identify whether there are concentrated or dispersed patterns; the discrete data of the high-risk node distribution of faults is obtained from step S352, and the spatial analysis tool is used to perform geometric association analysis on the distribution of high-risk nodes, and the geometric characteristics of the node distribution, such as density, aggregation, and path connectivity, are analyzed. According to the distribution characteristics, geometric association parameters are calculated, such as the shortest path between nodes and the maximum connected component; the geometric association data of the high-risk node distribution is obtained from step S353, and the chain effect model is used to calculate the triggering probability of the chain effect of faults between nodes based on the geometric association data; the probability of the fault propagating from one node to an adjacent node is determined based on the transient interference gain and the performance failure condition status; the triggering probability of the chain effect of faults is calculated by combining the transient interference order gain data, the performance failure condition status data and the node distribution geometric association data. This step helps evaluate whether adjacent nodes are easily affected when a node in the system fails, thus forming a chain reaction. The fault chain reaction trigger probability data is generated based on the calculation results to describe the chain reaction between nodes in the system. Combined with transient interference order gain data, performance failure condition status data and fault chain reaction trigger probability data, the probability of false protection triggering between load nodes is calculated. This step evaluates the probability of triggering protection actions between nodes due to misjudgment when a system failure occurs. The false protection trigger probability data of the node is generated based on the calculation results to describe the false protection risk between different load nodes in the system.

[0161] Preferably, step S4 comprises the following steps:

[0162] Step S41: performing abnormal fault diagnosis learning on the performance failure condition status data and the node misprotection probability data based on the policy gradient algorithm to obtain node abnormal fault diagnosis data;

[0163] Step S42: normalizing the node abnormal fault diagnosis data to obtain node abnormal fault diagnosis normalized data;

[0164] Step S43: Formulate a power generation node abnormality diagnosis strategy based on the node abnormality fault diagnosis normalized data to obtain a power generation node abnormality diagnosis strategy.

[0165] In an embodiment of the present invention, the performance failure condition status data of the above steps is used, which describes the performance failure conditions and their probabilities encountered by the node under different working conditions. The node misprotection probability data collected in the system is used to reflect the probability that the node does not actually fail due to abnormal diagnosis. The policy gradient algorithm is used to perform abnormal fault diagnosis learning based on the performance failure condition status data and the node misprotection probability data. The policy gradient algorithm can learn how to accurately judge the abnormal fault conditions of the node under different working conditions by optimizing the policy parameters, and generate the abnormal fault diagnosis data of the node according to the learning results. It includes the abnormal diagnosis strategy of each node under different working conditions, that is, judging whether the node is abnormal according to the input status data, and normalizing the abnormal fault diagnosis data of the node. The normalization process can use statistical methods or standardization techniques to map the abnormal diagnosis results to a unified numerical range, which is convenient for subsequent analysis and strategy formulation, and the abnormal fault diagnosis normalized data of the node is generated according to the results of the normalization process. Including the abnormal diagnosis results of each node under different working conditions, which have been normalized to a unified numerical range. According to the normalized data of abnormal fault diagnosis of the node, the abnormal diagnosis strategy of the power generation node is formulated. The strategy formulation can be based on the numerical range or threshold setting of the normalized data to determine when to perform what type of abnormal diagnosis operation or alarm on the node. According to the formulated strategy, the abnormal diagnosis strategy of the power generation node is generated. The strategy should include the specific method of node abnormal diagnosis, the basis for condition judgment and its implementation steps, and provide guidance for abnormal diagnosis in system operation.

[0166] Preferably, the present invention further provides a photovoltaic power generation system abnormal fault diagnosis system based on distributed sensing, which is used to execute the photovoltaic power generation system abnormal fault diagnosis method based on distributed sensing as described above, and the photovoltaic power generation system abnormal fault diagnosis system based on distributed sensing includes:

[0167] The real-time current correction module is used to collect multi-modal sensor data of photovoltaic power generation nodes in the photovoltaic power generation system based on distributed sensors to obtain multi-modal sensor data of power generation nodes; extract real-time current data of different power generation nodes from the multi-modal sensor data of power generation nodes to obtain real-time current data of power generation nodes; correct the real-time current environmental impact under different environmental conditions for the real-time current data of power generation nodes to obtain real-time current correction data of power generation nodes;

[0168] The harmonic conduction analysis module is used to analyze the current waveforms of different power generation nodes based on the real-time current correction data of the power generation nodes to obtain the current waveform data of the power generation nodes; to analyze the path conduction harmonic superposition effect based on the current waveform data of the power generation nodes to obtain the path conduction harmonic superposition effect data; to identify the node load distribution based on the path conduction harmonic superposition effect data to obtain the load node distribution data;

[0169] The node misprotection analysis module is used to calculate the transient interference order gain between different load nodes on the load node distribution data to obtain the transient interference order gain data; simulate the performance failure condition state according to the transient interference order gain data to obtain the performance failure condition state data; calculate the misprotection trigger probability between different load nodes on the load node distribution data according to the transient interference order gain data and the performance failure condition state data to obtain the node misprotection probability data;

[0170] The fault diagnosis learning module is used to perform abnormal fault diagnosis learning on performance failure condition status data and node misprotection probability data to obtain node abnormal fault diagnosis data; and formulate a power generation node abnormal diagnosis strategy based on the node abnormal fault diagnosis data to obtain a power generation node abnormal diagnosis strategy.

[0171] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0172] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for abnormal fault diagnosis of photovoltaic power generation system based on distributed sensing, characterized in that: The following steps are involved: Step S1: collecting multimodal sensing data of photovoltaic power generation nodes of the photovoltaic power generation system based on distributed sensors to obtain multimodal sensing data of power generation nodes; Extract the real-time current data of different power generation nodes from the multimodal sensor data of the power generation nodes to obtain the real-time current data of the power generation nodes; Correct the real-time current environmental impact of different environmental conditions on the real-time current data of the power generation node to obtain the real-time current correction data of the power generation node; Step S2: Perform current waveform analysis of different power generation nodes on the real-time current correction data of the power generation node to obtain current waveform data of the power generation node; perform path conduction harmonic superposition effect analysis based on the current waveform data of the power generation node to obtain path conduction harmonic superposition effect data; perform node load distribution identification based on the path conduction harmonic superposition effect data to obtain load node distribution data; Step S3: calculating the transient interference order gain between different load nodes on the load node distribution data to obtain transient interference order gain data; performing a performance failure condition state simulation based on the transient interference order gain data to obtain performance failure condition state data; According to the transient interference order gain data and the performance failure condition status data, the load node distribution data is used to calculate the probability of false protection triggering between different load nodes to obtain node false protection probability data; wherein step S3 includes: Step S31: Calculate the harmonic interference intensity of different load nodes on the load node distribution data according to the path conduction harmonic superposition effect data to obtain the load node harmonic interference intensity data; Step S32: Calculate the transient interference order gain between different load nodes for the load node harmonic interference intensity data to obtain transient interference order gain data; wherein step S32 includes: Step S321: performing transient harmonic intensity calculation between different load nodes on the load node harmonic interference intensity data to obtain transient harmonic intensity data of the load node; performing transient intensity continuity analysis between different load nodes on the load node transient harmonic intensity data to obtain transient intensity continuity data; Step S322: extracting the increasing intervals between different load nodes of the transient harmonic intensity data of the load nodes according to the transient intensity persistence data, and obtaining the transient harmonic intensity increasing intervals; Step S323: calculating the overlapping intervals of increasing speed rates between different load nodes for the transient harmonic intensity increasing intervals to obtain the overlapping intervals of increasing speed rates; Step S324: sorting the harmonic intensity difference values ​​of the non-overlapping intervals of the transient harmonic intensity increasing intervals according to the overlapping intervals of the increasing rate, to obtain harmonic intensity difference sorting data; Step S325: performing piecewise Simpson integration processing on the transient harmonic intensity increasing interval according to the harmonic intensity difference sorting data to obtain transient harmonic intensity approximate data; Step S326: Calculate the transient interference order gain between different load nodes according to the transient harmonic intensity approximate data and the harmonic intensity difference sorting data to obtain transient interference order gain data; Step S33: performing thermal effect trend analysis between different load nodes according to transient interference order gain data and load node harmonic interference intensity data to obtain load node thermal effect trend data; Step S34: Performing a performance failure condition state simulation based on the transient interference order gain data and the load node thermal effect trend data to obtain performance failure condition state data; wherein step S34 includes: Step S341: Calculate the node heating rate for the load node thermal effect trend data to obtain the node heating rate data; Step S342: Calculate the thermal cycle influence weight according to the node heating rate data and the load node thermal effect trend data to obtain the thermal cycle influence weight data; Step S343: performing a thermal accumulation effect simulation on the load node thermal effect trend data according to the thermal cycle influence weight data to obtain thermal accumulation effect data; Step S344: performing multi-node heat loss range difference calculation on the load node heat effect trend data according to the heat accumulation effect data and the heat cycle influence weight data to obtain multi-node heat loss range difference data; Step S345: performing node load critical state analysis according to transient interference order gain data and multi-node heat loss range difference data to obtain node load critical state data; Step S346: Perform performance failure condition state simulation according to the node load critical state data and the multi-node heat loss range difference data to obtain performance failure condition state data; Step S35: Calculate the probability of false protection triggering between different load nodes based on the transient interference order gain data and the performance failure condition status data on the load node distribution data to obtain node false protection probability data; Step S4: Perform abnormal fault diagnosis learning on the performance failure condition status data and the node misprotection probability data to obtain node abnormal fault diagnosis data; formulate a power generation node abnormal fault diagnosis strategy based on the node abnormal fault diagnosis data to obtain a power generation node abnormal fault diagnosis strategy.

2. The abnormal fault diagnosis method of photovoltaic power generation system based on distributed sensing according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: collecting multimodal sensing data of photovoltaic power generation nodes of the photovoltaic power generation system based on distributed sensors to obtain multimodal sensing data of power generation nodes; wherein the multimodal sensing data of power generation nodes includes current sensing data of power generation nodes and environmental status data of power generation nodes; Step S12: extracting real-time current data of different power generation nodes from the current sensing data of the power generation nodes to obtain real-time current data of the power generation nodes; Step S13: Calculating the weights of environmental impact factors of different power generation nodes on the real-time current data of the power generation nodes according to the environmental status data of the power generation nodes, and obtaining environmental impact factor weight data; Step S14: Correcting the real-time current environmental impact of different environmental conditions on the real-time current data of the power generation node according to the environmental impact factor weight data to obtain the real-time current correction data of the power generation node.

3. The photovoltaic power generation system abnormal fault diagnosis method based on distributed sensing according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: performing node current conduction topology path analysis on the real-time current correction data of the power generation node to obtain node current conduction topology path data; Step S22: performing current waveform analysis of different power generation nodes on the real-time current correction data of the power generation nodes according to the node current conduction topology path data to obtain current waveform data of the power generation nodes; Step S23: analyzing the path conduction harmonic superposition effect of the node current conduction topology path data according to the power generation node current waveform data to obtain the path conduction harmonic superposition effect data; Step S24: identifying the node load distribution of the node current conduction topology path data according to the path conduction harmonic superposition effect data to obtain load node distribution data.

4. The abnormal fault diagnosis method of photovoltaic power generation system based on distributed sensing according to claim 3 is characterized in that: Step S23 includes the following steps: Step S231: extracting the asymmetric waveform region of the current waveform data of the power generation node to obtain the current asymmetric waveform region data; performing harmonic identification on the current asymmetric waveform region data using a preset current harmonic identification model to obtain the current asymmetric harmonic data; Step S232: performing asymmetric harmonic regional path mapping on the node current conduction topology path data according to the current asymmetric harmonic data to obtain the conduction path asymmetric harmonic data; Step S233: calculating the crest factor increment coefficient between different node conduction paths for the asymmetric harmonic data of the conduction path to obtain the harmonic crest factor increment coefficient; Step S234: calculating the crest factor progressive magnitude between different node conduction paths for the asymmetric harmonic data of the conduction path according to the harmonic crest factor progressive coefficient, and obtaining crest factor progressive magnitude data; Step S235: calculating the magnitude mean difference series of the crest factor progressive magnitude data to obtain the magnitude mean difference series; performing magnitude truncation dynamic modal decomposition on the asymmetric harmonic data of the conduction path according to the magnitude mean difference series to obtain asymmetric harmonic magnitude modal decomposition data; Step S236: analyzing the path conduction harmonic superposition effect of the node current conduction topology path data according to the asymmetric harmonic magnitude modal decomposition data and the harmonic crest factor increasing coefficient to obtain the path conduction harmonic superposition effect data.

5. The abnormal fault diagnosis method of photovoltaic power generation system based on distributed sensing according to claim 1 is characterized in that: Step S35 includes the following steps: Step S351: screening the load nodes with high fault risk for the load node distribution data according to the transient interference order gain data and the performance failure condition status data to obtain the load node data with high fault risk; Step S352: Calculate the distribution dispersion of the fault high risk load node data to obtain the fault high risk node distribution discrete data; Step S353: performing a node distribution geometric correlation analysis based on the discrete data of the high-risk node distribution of failures to obtain high-risk node distribution geometric correlation data; Step S354: Calculate the fault chain effect trigger probability of the node distribution geometric association data according to the transient interference order gain data and the performance failure condition state data to obtain the fault chain effect trigger probability data; Step S355: Calculate the misprotection triggering probability between different load nodes based on the transient interference order gain data, the performance failure condition status data and the fault chain effect triggering probability data for the load node distribution data to obtain the node misprotection probability data.

6. The abnormal fault diagnosis method of photovoltaic power generation system based on distributed sensing according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: performing abnormal fault diagnosis learning on the performance failure condition status data and the node misprotection probability data based on the policy gradient algorithm to obtain node abnormal fault diagnosis data; Step S42: normalizing the node abnormal fault diagnosis data to obtain node abnormal fault diagnosis normalized data; Step S43: Formulate a power generation node abnormality diagnosis strategy based on the node abnormality fault diagnosis normalized data to obtain a power generation node abnormality diagnosis strategy.

7. A photovoltaic power generation system abnormal fault diagnosis system based on distributed sensing, characterized in that: The photovoltaic power generation system abnormal fault diagnosis method based on distributed sensing as claimed in claim 1, the photovoltaic power generation system abnormal fault diagnosis system based on distributed sensing comprises: The real-time current correction module is used to collect multi-modal sensor data of photovoltaic power generation nodes in the photovoltaic power generation system based on distributed sensors to obtain multi-modal sensor data of power generation nodes; extract real-time current data of different power generation nodes from the multi-modal sensor data of power generation nodes to obtain real-time current data of power generation nodes; correct the real-time current environmental impact under different environmental conditions for the real-time current data of power generation nodes to obtain real-time current correction data of power generation nodes; The harmonic conduction analysis module is used to analyze the current waveforms of different power generation nodes based on the real-time current correction data of the power generation nodes to obtain the current waveform data of the power generation nodes; to analyze the path conduction harmonic superposition effect based on the current waveform data of the power generation nodes to obtain the path conduction harmonic superposition effect data; to identify the node load distribution based on the path conduction harmonic superposition effect data to obtain the load node distribution data; The node misprotection analysis module is used to calculate the transient interference order gain between different load nodes on the load node distribution data to obtain the transient interference order gain data; simulate the performance failure condition state according to the transient interference order gain data to obtain the performance failure condition state data; calculate the misprotection trigger probability between different load nodes on the load node distribution data according to the transient interference order gain data and the performance failure condition state data to obtain the node misprotection probability data; The fault diagnosis learning module is used to perform abnormal fault diagnosis learning on performance failure condition status data and node misprotection probability data to obtain node abnormal fault diagnosis data; and formulate a power generation node abnormal diagnosis strategy based on the node abnormal fault diagnosis data to obtain a power generation node abnormal diagnosis strategy.

Images