Detection Method for Preventing Grounding Faults of Photovoltaic Modules Based on Multi-Source Data Fusion

Abstract

The present invention discloses a detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, including: obtaining data sources of photovoltaic modules in different spatio-temporal dimensions; selecting different data analysis methods to analyze the states of corresponding types of data sources; fusing and verifying the state analysis results of each data source to obtain the grounding fault probability of the photovoltaic module. By performing fusion analysis on multi-dimensional data sources, the present invention realizes accurate prediction of the grounding failure rate of photovoltaic modules, effectively prevents grounding faults of photovoltaic modules, and ensures continuous and stable power generation of photovoltaic modules.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and more specifically, to a detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion. Background Art

[0002] As the core equipment of a photovoltaic power generation system, the working efficiency of photovoltaic modules will directly affect the economic benefits of enterprises. Fault handling and daily maintenance of photovoltaic modules are important measures to ensure the availability of core power generation equipment. More and more on-site problems and accidents indicate that grounding faults of photovoltaic modules are one of the most common faults in photovoltaic power plants.

[0003] Currently, the detection method for checking grounding faults of photovoltaic modules mainly uses a detection circuit. When a grounding fault of a photovoltaic module is detected, a power outage loss has already occurred.

[0004] Therefore, how to provide a detection method that can prevent grounding faults of photovoltaic modules is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion. By performing fusion analysis on multi-dimensional data sources, accurate prediction of the grounding failure rate of photovoltaic modules is achieved, effective prevention of grounding faults of photovoltaic modules is realized, and continuous and stable power generation of photovoltaic modules is ensured.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, comprising:

[0008] Obtaining data sources of photovoltaic modules in different space-time dimensions;

[0009] Selecting different data analysis methods to analyze the status of the corresponding type of data source;

[0010] Fusing and verifying the status analysis results of each data source to obtain the grounding fault probability of the photovoltaic module.

[0011] Preferably, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, it further comprises:

[0012] Querying corresponding countermeasures from a pre-constructed knowledge base based on the grounding fault probability of the photovoltaic module and the current location of the photovoltaic module, and distributing them to the corresponding maintenance terminals.

[0013] Preferably, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, it further includes:

[0014] Writing the data source status, grounding fault probability, actual grounding fault, and countermeasures of the photovoltaic module into the knowledge base, and updating the knowledge base.

[0015] Optionally, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, the types of the data sources at least include: video imaging data of photovoltaic modules, family defect data, online operation parameters, operation and maintenance data, vibration sensor data, weather forecast data, and work ticket data of the photovoltaic power station.

[0016] Optionally, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, the video imaging data of the photovoltaic module is obtained by a drone or a low-altitude balloon-mounted camera; the family defect data is obtained from a pre-constructed knowledge base; the weather forecast data is obtained from a server interface; the online operation parameters, the operation and maintenance data, and the vibration sensor data are obtained from a photovoltaic operation and maintenance management system.

[0017] Optionally, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, an image search and matching algorithm is used to analyze the video imaging data of the photovoltaic module; a fuzzy mathematics method is used to analyze the family defect data; an LSTM algorithm is used to analyze the short-term and short-term weather forecast data in the weather forecast data, a BP neural network algorithm is used to analyze the medium-term weather forecast data in the weather forecast data, and then a weighted average is performed on the analysis results of the short-term, short-term, and medium-term weather forecast data; a statistical method is used to analyze the online operation parameters, the operation and maintenance data, and the vibration sensor data; a query or search method is used to analyze the work ticket data of the photovoltaic power station.

[0018] Optionally, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, the process of analyzing the video imaging data of the photovoltaic module includes:

[0019] Extracting video frames from the video data of the photovoltaic module, and using an image search and matching algorithm to compare the video frames with the original state image of the photovoltaic module area to judge the degree of consistency between the two;

[0020] Marking the video frames with a consistency degree lower than a preset value as warning images;

[0021] Identifying the surface state of the photovoltaic module in the warning image and judging whether there is on-site construction work around.

[0022] Optionally, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, the step of fusing and verifying the status analysis results of each data source to obtain the grounding fault probability of the photovoltaic module includes:

[0023] Using a data fusion algorithm to perform weighted averaging on the status analysis results of different data sources to obtain the calculation result of the fault information of the photovoltaic module; the calculation result includes the fault type, the location where the fault occurs, and the grounding fault probability value;

[0024] Using a BP neural network algorithm to perform fitting verification on the calculation result of the data fusion algorithm, and predicting and outputting the grounding fault probability value of the photovoltaic module;

[0025] Comparing the grounding fault probability value in the calculation result of the data fusion algorithm with the grounding fault probability value output by the BP neural network algorithm. If the deviation value between the two is within the preset value, the current calculation result is valid; if the deviation value between the two is not within the preset value, the current calculation result is invalid.

[0026] Optionally, in the above detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, the expression of the data fusion algorithm is as follows:

[0027] Y1 = ∑y i b i ;

[0028] where i represents the data source type, y i represents the status analysis result of the i-th type of data source, b i represents the weight of the status analysis result of the i-th type of data source, and 1 = ∑b i .

[0029] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion. This method performs data fusion on the video imaging data of photovoltaic modules, short-term, medium-term, and long-term data of weather forecasts, family information of photovoltaic modules, operation and maintenance data, online operation parameters of photovoltaic modules, vibration sensor data, and work tickets of photovoltaic power stations, calculates the probability of grounding faults of photovoltaic modules, and automatically generates countermeasures after grounding faults of photovoltaic modules, realizing effective prevention of grounding faults of photovoltaic modules and ensuring continuous and stable power generation of photovoltaic modules. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on the provided accompanying drawings.

[0031] Figure 1 The accompanying drawing is a flowchart of a detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion provided by the present invention in one embodiment;

[0032] Figure 2 The accompanying drawing is a flowchart of a detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion provided by the present invention in another embodiment. Detailed implementation manners

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0034] As Figure 1 shown, the embodiments of the present invention disclose a detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, including the following steps:

[0035] S1. Obtain data sources of photovoltaic modules in different spatio-temporal dimensions;

[0036] S2. Select different data analysis methods to analyze the states of the corresponding types of the data sources;

[0037] S3. Fuse and verify the state analysis results of each of the data sources to obtain the grounding fault probability of the photovoltaic modules.

[0038] In one embodiment, it further includes:

[0039] S4. Based on the grounding fault probability of the photovoltaic modules and the current positions of the photovoltaic modules, query the corresponding countermeasures from a pre-constructed knowledge base and distribute them to the corresponding maintenance terminals. Among them, the family defect information of the photovoltaic modules is obtained from the knowledge base. When the grounding fault probability exceeds a set threshold, it is determined that a fault exists, and countermeasures are queried from the knowledge base.

[0040] In other embodiments, it further includes:

[0041] As Figure 2As shown in the figure, S5: Write the data source status, grounding fault probability, actual grounding fault, and countermeasures of the photovoltaic module into the knowledge base to update the knowledge base.

[0042] The actual grounding fault is obtained based on actual records. After each fault occurs, relevant information such as the type, time, and handling measures of the actually occurred fault will be recorded in the knowledge base to update the knowledge base. Strictly speaking, each fault is different, with different occurrence times and conditions. Adding the actually occurred faults to the knowledge base will enhance the capabilities of the knowledge base.

[0043] The embodiment of the present invention can also automatically set weights, which are obtained based on the knowledge base of photovoltaic modules.

[0044] The embodiment of the present invention automatically sets weights using different setting methods according to the set goals; for example, when focusing on weather conditions, the weight value of weather conditions will be increased when setting weights.

[0045] The above steps are further described below.

[0046] In S1, the types of the data source at least include: video imaging data of photovoltaic modules, family defect data, online operation parameters, operation and maintenance data, vibration sensor data, weather forecast data, and work ticket data of the photovoltaic power station.

[0047] Among them, the data sources related to time are the family defect data of photovoltaic modules, operation and maintenance data, short-term, medium-term, and long-term weather forecast data, and work tickets of the photovoltaic power station. The data sources related to space are the video imaging data of photovoltaic modules, vibration sensor data, and work tickets of the photovoltaic power station. The basic data source of photovoltaic modules is the operation parameters of photovoltaic modules, such as voltage, current, heat, position information, and self-position information.

[0048] Family defect data refers to the defect data of photovoltaic modules of the same manufacturer; different manufacturers belong to different families. In actual applications, if the photovoltaic modules of a certain manufacturer have defects, their products may generally have defects. Using this characteristic, a higher weight value can be assigned to this indicator.

[0049] The work ticket of the photovoltaic power station records the time, location, area range, work category, and content of construction, maintenance, and other work in the photovoltaic power station. If there are operations such as digging, demolition, and installation near a photovoltaic module, the probability of damage to the photovoltaic module will increase significantly. For example, there are often cases where construction work digs through cables or damages the foundation of photovoltaic modules.

[0050] The acquisition methods of each data source are as follows:

[0051] The video imaging data of the photovoltaic module is collected by an aerial camera. Specifically, a camera is mounted on a drone or a low-altitude balloon and transmitted to a video acquisition service device via wireless 4G and 5G signals. Then, the video acquisition service device transmits the data to a monitoring service platform via optical fiber for video analysis and data processing to identify the surface state of the photovoltaic module and determine whether there is on-site construction work. Among them, the video acquisition service device is responsible for collecting and storing the video image data of the camera and performing some preliminary processing; the monitoring service platform is responsible for monitoring the operation status of the photovoltaic modules in the photovoltaic power station, including video data and acquisition data from sensors such as voltage, current, power, and power generation. Among them, the amount of video image data is very large, while the data such as voltage, current, power, and power generation of the photovoltaic module, although there are many data points, each data point occupies a very small storage space, only 4 bytes per point.

[0052] The family defect data is obtained from a pre-constructed knowledge base of photovoltaic modules.

[0053] The short-term, medium-term, and long-term weather forecast data is obtained from a server interface.

[0054] The operation and maintenance data, operation parameters, and vibration sensor data are obtained from the photovoltaic operation and maintenance management system.

[0055] The work tickets of the photovoltaic power station are issued by the dispatching system and obtained from the photovoltaic operation and maintenance management system.

[0056] In S2, the specific process of selecting different data analysis methods to analyze the status of the corresponding type of data source is as follows:

[0057] An image search and matching algorithm is used to analyze the video imaging data of the photovoltaic module. Specifically:

[0058] 1. Extract the video frames from the video data of the photovoltaic module, and use an image search and matching algorithm to compare the video frames with the original state image of the photovoltaic module area to judge the degree of consistency between the two;

[0059] 2. Mark the video frames with a consistency degree lower than the preset value as warning images; if the consistency degree is above 95%, it is a valid image; if the consistency degree is lower than 95%, the image obtained from the camera mounted on the low-altitude balloon shall be used as the standard, and an alarm flag shall be set as a warning image.

[0060] 3. Identify the surface state of the photovoltaic module in the warning image and determine whether there is on-site construction work around. Use the method of image recognition to identify the abnormal states of the photovoltaic module, such as missing, flipped, damaged, etc. At the same time, analyze the on-site construction area around the photovoltaic module and set key areas.

[0061] Analyze the family defect data using the method of fuzzy mathematics; classify the product quality into three levels: good, average, and poor, with corresponding weights of 2, 1, and 0.

[0062] Analyze the short-term and short-term weather forecast data in the weather forecast data using the LSTM algorithm, analyze the medium-term weather forecast data in the weather forecast data using the BP neural network algorithm, and then perform a weighted average on the analysis results of the short-term, short-term, and medium-term weather forecast data, with each weight being 1 / 3.

[0063] Analyze the online operating parameters, the operation and maintenance data, and the vibration sensor data using statistical methods; calculate the maximum value, minimum value, mean square deviation, etc., and identify whether there are any abnormalities in the above data.

[0064] Obtain the position of the photovoltaic module from the work ticket of the photovoltaic power station. The influence degree of the photovoltaic module at this position is 1, and the influence degree of the photovoltaic modules at other positions is 0.. The current work ticket will also form a record in the system when printed on paper; the work ticket includes work content, time, personnel, location, measures, etc. Therefore, it is easy to obtain the position information of the photovoltaic module from the work ticket using the query or search method.

[0065] In S3, fuse and verify the status analysis results of each data source to obtain the grounding fault probability of the photovoltaic module, specifically including:

[0066] Use the data fusion algorithm to perform a weighted average on the status analysis results of different data sources to obtain the calculation result of the photovoltaic module fault information; the calculation result includes the fault type, the position where the fault occurs, and the grounding fault probability;

[0067] Use the BP neural network algorithm to perform fitting verification on the calculation result of the data fusion algorithm, and predict and output the grounding fault probability value of the photovoltaic module;

[0068] Compare the grounding fault probability value in the calculation result of the data fusion algorithm with the grounding fault probability value output by the BP neural network algorithm. If the deviation value between the two is within the preset value, the current calculation result is valid; if the deviation value between the two is not within the preset value, the current calculation result is invalid.

[0069] Specifically, among them, the expression of the data fusion algorithm is as follows:

[0070] Y1 = ∑y i b i ;

[0071] Among them, i represents the data source type, and y i represents the status analysis result of the i-th type of data source, and b iRepresents the weight of the status analysis result of the i-th type of data source, and 1 = ∑b i .

[0072] In a specific embodiment, y i (i = 1, 2, 3, 4, 5, 6) are respectively the video analysis result, the short-term, short-term, medium-term weather forecast data processing result, the online status parameter processing result of photovoltaic modules, the vibration sensor data processing result of photovoltaic modules, the work ticket processing result of photovoltaic power stations, and the family defect information processing result of photovoltaic modules.

[0073] b i (i = 1, 2, 3, 4, 5, 6) are respectively the weights of the video analysis result, the short-term, short-term, medium-term weather forecast result, the online status parameter processing result of photovoltaic modules, the vibration sensor processing result of photovoltaic modules, the work ticket processing result of photovoltaic power stations, and the family defect information processing result of photovoltaic modules.

[0074] The processing results of the above six data sources Yi are obtained by different algorithms of different data sources. Specifically:

[0075] The video analysis result is obtained by processing the video image data of a camera mounted on a drone or a low-altitude balloon. The content of the video analysis result includes: data source type, photovoltaic module failure type, location of the photovoltaic module, and probability of failure.

[0076] The short-term, short-term, medium-term weather forecast data processing result is to calculate the weight of the future impact based on the obtained weather forecast data, and this weight value will increase the contribution value of other calculation results. When the future weather deteriorates (i.e., the time without sunlight is long), the weight value is large; when the future weather is sunny, the weight value is small and approaches 0. Using the method of fuzzy mathematics, the future weather conditions from good to bad correspond to the weight values {0.01, 0.2, 0.6, 0.8, 1.5, 1.8}, namely {very good, good, general, poor, very poor, extremely bad}. Very good means that the weather is sunny and there is sufficient sunlight within the next two weeks. Good means that the weather is sunny and there is sufficient sunlight within the next week, but there may be clouds, and the weather is uncertain after one week. General means that the weather is good within the next two days, there may be clouds, and there are clouds or overcast days after two days. Poor means that it is cloudy within the next three days. Very poor means that there is no sunlight within the next week, and there are strong winds, precipitation, etc. Extremely bad means that there is no sunlight within the next two weeks, and the weather has rain, snow, hail, strong winds, etc. These weather conditions correspond to the above weight values.

[0077] The processing results of the online status parameters of photovoltaic modules are used to determine the status of photovoltaic modules based on parameters such as the voltage, current, and power generation of the monitored photovoltaic modules, including the position of the photovoltaic modules and the status of the photovoltaic modules. The status of the photovoltaic modules is divided into {very good, good, average, bad, very bad} according to the method of fuzzy mathematics, and the corresponding numerical values are {1.0, 0.8, 0.5, 0.2, 0.1}.

[0078] The processing results of the vibration sensor data of photovoltaic modules are used to distinguish the status based on the online data of the vibration sensor. When the data of the vibration sensor exceeds the specified value, it is in a bad state; when the data of the vibration sensor meets the requirements, it is in a good state. The corresponding data values are {0.2, 1.0}. The data processing results also include the position information of the photovoltaic modules.

[0079] The processing results of the work tickets of photovoltaic power stations and the processing results of the family defect information of photovoltaic modules are similar to the above.

[0080] The weight bi allocation principle for the above six data sources is as follows:

[0081] Sorted from largest to smallest: the processing results of the online status parameters of photovoltaic modules, the processing results of the vibration sensors of photovoltaic modules, the video analysis results, the processing results of the work tickets of photovoltaic power stations, the short-term, medium-term, and long-term weather forecast results, and the processing results of the family defect information of photovoltaic modules. Their weight values add up to 1. For example, they can take values {0.30, 0.25, 0.20, 0.15, 0.08, 0.02}. 0.30 + 0.25 + 0.20 + 0.15 + 0.08 + 0.02 = 1.0.

[0082] The specific process of verifying the first ground fault probability value using the BP neural network algorithm is as follows:

[0083] Each calculation will store the actual results and calculation results of the data fusion algorithm; among them, the actual results are stored as the calculation results after the fault occurs; when the fault does not occur, 0 is stored. In this way, after accumulating a large amount of data, these data will become the training data of the BP neural network algorithm.

[0084] The BP neural network algorithm predicts new data based on the accumulated data; the new data is compared with the calculation results of the data fusion algorithm to observe the difference level. If the difference is not obvious, the result is valid. If the difference is obvious, the result is invalid.

[0085] Perform BP neural network training on the calculation results of the data fusion algorithm, then predict the future results, and compare and verify the predicted results with the results of the first algorithm for multi-source data fusion analysis. Generally, there will be some deviations; if the deviation value is large and the relative value exceeds 30%, there is a problem and the current calculation result is invalid; if the deviation is very small, within 10%, the current calculation result is valid.

[0086] The results output by the data fusion algorithm are the fault information of the photovoltaic modules, including the fault type, the location where the fault occurs, and the probability of the fault occurring; importantly, the probability value. Its essence is to obtain the probability value after classification to determine the maximum possibility of the fault occurring.

[0087] When the BP neural network makes a prediction, it is related to the output results of the data fusion algorithm, which is equivalent to using the calculation results of the data fusion algorithm to make another prediction. After the prediction, it is compared with the results of the data fusion algorithm. That is to say, the data fusion algorithm is similar to a rough screening process, and the BP neural network algorithm is similar to a reprocessing. Since the BP neural network algorithm is a non-linear algorithm, the results of the two times are not exactly the same. When the deviation is small, the predicted results are considered reliable. When the deviation is large, it means that the predicted results are not credible and cannot be judged as a fault, and the reasons need to be further analyzed.

[0088] In S4, the knowledge base records the location information, fault type, fault description, cause of the fault, countermeasures for the fault, etc. of the photovoltaic modules; according to the probability of the grounding fault of the photovoltaic modules, determine the fault type and fault description, and combine the location information of the photovoltaic modules to find the countermeasures. Because for the same kind of fault in different locations, the countermeasures are different. Task automatic distribution is to form a task list and operation guide for the countermeasures of photovoltaic faults and distribute them to the corresponding maintenance terminals for maintenance personnel to handle.

[0089] In S5, compare the grounding fault probability of the photovoltaic modules with the actual faults, and write the data source status, actual grounding fault, and countermeasures of the photovoltaic modules corresponding to the grounding fault probability of the photovoltaic modules into the knowledge base to update the knowledge base. The purpose is to generate valuable knowledge information to provide or retrieve information when the state of new photovoltaic modules and grounding faults of photovoltaic modules occur later.

[0090] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0091] The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily 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. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion, characterized in that, Including: Obtaining data sources of photovoltaic modules in different spatio-temporal dimensions; The types of the data sources at least include: video imaging data of photovoltaic modules, family defect data, online operation parameters, operation and maintenance data, vibration sensor data, weather forecast data, and work ticket data of photovoltaic power stations; Selecting different data analysis methods to analyze the states of the corresponding types of the data sources; Fusing and verifying the state analysis results of each of the data sources to obtain the grounding fault probability of the photovoltaic module, specifically including: Using a data fusion algorithm to perform weighted averaging on the state analysis results of different data sources to obtain the calculation results of the fault information of the photovoltaic module; the calculation results include the fault type, the location where the fault occurs, and the grounding fault probability value; for each calculation, the actual results and the calculation results of the data fusion algorithm will be stored; among them, the actual results are stored as the calculation results after the fault occurs; when the fault does not occur, 0 is stored, and the accumulated multiple groups of data are used as the training data of the BP neural network algorithm; the BP neural network algorithm predicts new data based on the accumulated data; Using the BP neural network algorithm to perform fitting verification on the calculation results of the data fusion algorithm and predict and output the grounding fault probability value of the photovoltaic module; Comparing the grounding fault probability value in the calculation results of the data fusion algorithm with the grounding fault probability value output by the BP neural network algorithm, if the deviation value between the two is within the preset value, the current calculation result is valid, if the deviation value between the two is not within the preset value, the current calculation result is invalid.

2. The detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion according to claim 1, characterized in that, Also including: Based on the grounding fault probability of the photovoltaic module and the current location of the photovoltaic module, querying the corresponding countermeasures from a pre-constructed knowledge base and distributing them to the corresponding maintenance terminals.

3. The detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion according to claim 2, wherein, Also including: Comparing the grounding fault probability of the photovoltaic module with the actually occurred fault, and writing the data source state, actual grounding fault, and countermeasures of the photovoltaic module corresponding to the grounding fault probability of the photovoltaic module into the knowledge base to update the knowledge base.

4. A detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion according to claim 1, characterized in that, The video imaging data of the photovoltaic module is obtained by a drone or a low-altitude balloon-mounted camera; the family defect data is obtained from a pre-constructed knowledge base; the weather forecast data is obtained from a server interface; the online operation parameters, the operation and maintenance data, and the vibration sensor data are obtained from a photovoltaic operation and maintenance management system.

5. The detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion according to claim 1, wherein Using an image search and matching algorithm to analyze the video imaging data of the photovoltaic module; using the method of fuzzy mathematics to analyze the family defect data; using the LSTM algorithm to analyze the short-term and short-term weather forecast data in the weather forecast data, using the BP neural network algorithm to analyze the medium-term weather forecast data in the weather forecast data, and then performing weighted averaging on the analysis results of the short-term, short-term, and medium-term weather forecast data; using statistical methods to analyze the online operation parameters, the operation and maintenance data, and the vibration sensor data; using query or search methods to analyze the work ticket data of the photovoltaic power station.

6. The detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion according to claim 1, characterized in that, The process of analyzing the video imaging data of the photovoltaic module includes: Extract video frames from the video data of the photovoltaic module, and use an image search and matching algorithm to compare the video frames with the original state image of the photovoltaic module area to determine the degree of consistency between the two; Mark the video frames with a consistency degree lower than the preset value as warning images; Identify the surface state of the photovoltaic module in the warning image and determine whether there is on-site construction work around; 7. A detection method for preventing grounding faults of photovoltaic modules based on multi-source data fusion according to claim 1, characterized in that, The expression of the data fusion algorithm is as follows: Y1 = ∑y i b i ; Among them, i represents the data source type, and y i represents the status analysis result of the i-th type of data source, and b i represents the weight of the status analysis result of the i-th type of data source, and 1 = ∑b i .

Images