Photovoltaic power station management method and system based on artificial intelligence and big data analysis

Abstract

The invention discloses a photovoltaic power station management method and system based on artificial intelligence and big data analysis, and relates to the technical field of photovoltaic power generation. The photovoltaic power generation group comprises a plurality of photovoltaic group strings used for connecting a plurality of photovoltaic panels, at least one combiner box, a direct current cable, a central inverter and an alternating current cable; according to the method, the running state of a photovoltaic power generation set is monitored in real time, abnormal conditions are rapidly recognized based on comparison of predicted power generation capacity and first photovoltaic power generation capacity, faults are subjected to hierarchical management through a comprehensive fault evaluation index, a proper rectification scheme is matched, in addition, the state of a photovoltaic panel is subjected to nondestructive testing through visible light and infrared imaging technologies, and the reliability of the photovoltaic power generation set is improved. According to the method, meteorological data and historical operation parameters are combined, fault reasons and damage degrees are comprehensively evaluated, fault data are updated and uploaded to a photovoltaic data platform after maintenance is finished, model performance is optimized, and the management method improves response speed, reduces fault processing time and reduces maintenance and management cost.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and in particular to a photovoltaic power station management method and system based on artificial intelligence and big data analysis. Background Art

[0002] With the continuous development of technology and the continuous reduction of costs, photovoltaic power generation technology has gradually become a clean energy technology with broad application prospects. At present, photovoltaic power generation systems have been widely used in building roofs, industrial parks, rural areas and other places. It can provide a large amount of clean energy for society and reduce dependence on traditional petrochemical energy and pollution to the environment.

[0003] A photovoltaic power station is a power generation facility that uses solar photovoltaic power generation technology to directly convert sunlight energy into electrical energy. A photovoltaic power station is usually composed of multiple photovoltaic power generation groups, which are composed of multiple photovoltaic strings (photovoltaic panels), at least one junction box, DC cables, central inverters, AC cables, monitoring systems and other auxiliary equipment. Under the traditional operation and maintenance mode of photovoltaic power stations, the real-time operating data of the photovoltaic power generation group is generally monitored to determine whether each component has a fault, but there is a lack of accurate prediction and positioning methods for the type of fault. The operation and maintenance personnel need to investigate the cause of the component failure, which is time-consuming and labor-intensive, and to a certain extent leads to low management and fault handling efficiency of photovoltaic power stations.

[0004] In order to solve this problem, we propose a photovoltaic power station management method and system based on artificial intelligence and big data analysis. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a photovoltaic power station management method and system based on artificial intelligence and big data analysis, which can effectively solve the problems in the above-mentioned background technology.

[0006] To achieve the above object, the technical solution adopted by the present invention is: A photovoltaic power station management method and system based on artificial intelligence and big data analysis, comprising the following steps: Acquiring information on the structure of a photovoltaic power station, wherein the photovoltaic power station structure includes a plurality of photovoltaic power generation groups, the photovoltaic power generation groups include a plurality of photovoltaic strings for connecting a plurality of photovoltaic panels, at least one combiner box, a DC cable, a central inverter and an AC cable, and the photovoltaic power generation groups are used to transmit the converted AC power to a power grid; Constructing a photovoltaic data platform, which is used to store basic information, location information, photovoltaic fault data and historical photovoltaic data of each photovoltaic power generation group in the photovoltaic power station, wherein the historical photovoltaic data includes status information of multiple historical working cycles, the first photovoltaic power generation, meteorological data and historical operating parameters; Build a photovoltaic power generation prediction model based on historical photovoltaic data, obtain current photovoltaic data of the current working cycle, and calculate the predicted power generation of the photovoltaic power generation group in the current working cycle; Obtaining a first photovoltaic power generation in the current photovoltaic data, judging the operating state of the photovoltaic power generation group based on the predicted power generation and the first photovoltaic power generation, and performing a fault analysis on the photovoltaic power generation group if it is in an abnormal operating state; Calculate the photovoltaic power generation interval received by the central inverter based on the predicted power generation, obtain the second photovoltaic power generation received by each combiner box in the current working cycle, and detect the fault based on the relationship between the second photovoltaic power generation and the photovoltaic power generation interval; If the second photovoltaic power generation is within the photovoltaic power generation range, the real-time operating parameters of the current working cycle are obtained, a fault type assessment model is constructed based on the photovoltaic fault data, and the suspected fault types and hazard indexes of the relevant components are obtained based on the real-time operating parameters and packaged into fault assessment data; If the second photovoltaic power generation is not within the photovoltaic power generation range, the photovoltaic string power generation data of each photovoltaic string received by the central inverter in the current working cycle is obtained, and based on the site image of the photovoltaic power generation group, the data is analyzed and processed through the image recognition algorithm to obtain the suspected fault type and hazard index of the photovoltaic string, and packaged into fault assessment data; Use a DC power simulator to simulate the PV string inputting DC power to the combiner box, collect relevant operating parameters, and obtain the suspected fault types and hazard indexes of relevant components based on the fault type assessment model, and package them into fault assessment data; Based on the fault assessment data obtained from the fault assessment platform, a comprehensive fault assessment index is calculated; based on the comprehensive fault assessment index, a suitable rectification plan is matched to the photovoltaic power generation group with failure risk in the photovoltaic power station.

[0007] Preferably, the calculation of the predicted power generation of the photovoltaic power generation group in the current working cycle based on the photovoltaic power generation prediction model specifically includes: Based on the photovoltaic data platform, the status information of multiple historical working cycles in the historical photovoltaic data, the first photovoltaic power generation and meteorological data are obtained; the status information includes the total operation time of the equipment, the operation time of the equipment cycle, performance indicators and electrical parameters; the meteorological data includes ambient temperature, relative humidity, wind speed, irradiation intensity and total radiation; Clean the acquired data, remove outliers and noise data, and use interpolation to process missing values; The data in the same historical working cycle are packaged into a set of data sets, and the data sets are divided into training sets and test sets. The convolutional neural network model is trained using the training set, and the model parameters are adjusted through the back propagation algorithm to make the model output as close as possible to the corresponding first photovoltaic power generation. The trained photovoltaic power generation prediction model is used to calculate the photovoltaic predicted power generation in the current working cycle; during the photovoltaic power generation prediction model training process, cross-validation and grid search are used to perform hyperparameter tuning to optimize the performance of the photovoltaic power generation prediction model; Acquiring current photovoltaic data of a current working cycle, wherein the current photovoltaic data includes state information, first photovoltaic power generation and meteorological data; The status information and meteorological data of the current working cycle are input into the photovoltaic power generation prediction model to calculate the photovoltaic predicted power generation.

[0008] Preferably, judging the operating state of the photovoltaic power generation group based on the predicted power generation and the first photovoltaic power generation specifically includes: Acquire a first photovoltaic power generation amount of current photovoltaic data, where the first photovoltaic power generation amount is the amount of electricity received by the power grid; Set the photovoltaic power generation fluctuation range based on the predicted power generation; If the first photovoltaic power generation is within the photovoltaic power generation fluctuation range, the photovoltaic power generation group in the current working cycle is in a normal operating state; otherwise, the photovoltaic power generation group is in an abnormal operating state, and a fault analysis is performed on the photovoltaic power generation group.

[0009] Preferably, the detecting of the fault based on the relationship between the second photovoltaic power generation and the photovoltaic power generation interval specifically includes: Setting a minimum conversion efficiency, which is the product of the minimum conversion efficiency of the photovoltaic strings, combiner boxes, DC cable losses, central inverters, and AC cables in the photovoltaic power generation group and an adjustment factor, wherein the adjustment factor is a coefficient that quantifies various losses and uncertainties that may be encountered by each component in actual operation, and is a constant between 0.95 and 1; The product of the inverse of the minimum conversion efficiency and the predicted power generation is taken as the maximum acceptable power generation, and the predicted power generation is taken as the minimum acceptable power generation; The photovoltaic power generation range is set, and the minimum acceptable power generation and the maximum acceptable power generation are respectively used as the lower limit and the upper limit of the photovoltaic power generation range; Obtain the second photovoltaic power generation received by each combiner box in the current working cycle; The second photovoltaic power generation of the combiner box is aggregated to obtain the total second photovoltaic power generation of the photovoltaic power generation group. If the total second photovoltaic power generation is within the photovoltaic power generation range, it is determined that there is no fault in each photovoltaic string of the photovoltaic power generation group, and other components in the photovoltaic power generation group are tested, including the combiner box, DC cable, central inverter and AC cable; If the total second photovoltaic power generation is not within the photovoltaic power generation range, it is determined that each photovoltaic string of the photovoltaic power generation group has a fault, and the photovoltaic strings in the photovoltaic power generation group are first tested, and then other components are tested.

[0010] Preferably, the fault type assessment model is constructed based on photovoltaic fault data, and the suspected fault causes and hazard indexes of relevant components are obtained based on real-time operating parameters, including: Obtain photovoltaic fault data and historical operating parameters of historical working cycles; The photovoltaic fault data includes fault type, fault operating parameters, fault number, fault hazard degree and hazard index; According to the fault operation parameters and historical operation parameters, the outlier data of the fault operation parameters are obtained based on the box plot detection; The types of fault operation parameters and historical operation parameters include: input power, output power, DC voltage, DC current, combiner box temperature, combiner box efficiency, input power, output power, DC voltage, AC voltage, DC current, AC current, inverter temperature and inverter efficiency of the central inverter; Based on the type of fault operation parameters, the fault operation parameters of the same type are divided into a group to obtain an outlier data group of the fault operation parameters; Analyze the outlier data of fault operation parameters and obtain the outlier data difference index; Based on data analysis requirements, obtain the outlier data difference index threshold; According to the outlier data difference index and the outlier data difference index threshold, it is determined whether the outlier data of the fault operation parameter is abnormal operation fault data; If the outlier data difference index exceeds the outlier data difference index threshold, the outlier data of the fault operation parameter is erroneous data, and the data is removed; If the outlier data difference index does not exceed the outlier data difference index threshold, the outlier data of the fault operation parameter is available; Matching the above fault operating parameter outlier data with the corresponding fault type, obtaining the fault waveform of the fault operating parameter outlier data, and merging the data to obtain a data set; Based on the fault waveform and fault type of the outlier data of fault operation parameters, a convolutional neural network model is established for training to obtain a fault type assessment model. Acquire real-time operating parameters of the current working cycle, wherein the operating parameters include each parameter in the historical operating parameters; Inputting the real-time operating parameters into the fault type assessment model, detecting and comparing the fault waveform through the fault type assessment model, and determining at least one suspected fault type; Match the hazard index of the suspected fault type, package the suspected fault type and the corresponding hazard index into fault assessment data, and upload it to the fault assessment platform; The outlier data difference index is: In the formula, For the The outlier difference index of the outlier data of the fault operation parameters, For the Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs The outlier data of the fault operating parameters and The interval time between outlier data of operating parameters of each historical fault, For the The total number of fault operating parameter outlier data in the fault operating parameter outlier data group to which the fault operating parameter outlier data belongs.

[0011] Preferably, the obtaining of the fault cause and hazard index of the photovoltaic string specifically includes: Based on the visible light imager and infrared imager installed near the photovoltaic power generation group, the visible light image and infrared image of each photovoltaic panel in the photovoltaic power generation group are captured in real time at preset time intervals; Use filters to remove noise from visible light images and infrared images, and convert the color of visible light images into grayscale to reduce data dimensions; Use edge detection algorithm to detect edge contours of each photovoltaic panel in visible light image and infrared image, extract images within edge contours respectively, and record them as visible light feature images and infrared feature images; Based on the image recognition algorithm, the grayscale value and contrast value of each pixel in the visible light feature image are obtained to obtain the shadow occlusion distribution map; Based on the image recognition algorithm, the temperature value of each pixel in the infrared characteristic image is obtained to obtain the temperature distribution map; Based on the temperature distribution map, the temperature abnormal area is detected, the shadow occlusion distribution map is compared with the temperature abnormal area, and the temperature abnormal area is divided into an abnormal area affected by the occlusion and an abnormal area not affected by the occlusion; If an abnormal area not affected by occlusion is detected, the photovoltaic panel and the photovoltaic string where the photovoltaic panel is located corresponding to the infrared characteristic image are obtained, and the photovoltaic string where the photovoltaic panel is located is judged to be faulty. The photovoltaic panel fault is regarded as a suspected fault type, and the hazard index is calculated based on the number of photovoltaic panel faults and the degree of fault. The suspected fault type and the corresponding hazard index are packaged as fault assessment data and uploaded to the fault assessment platform; If no abnormal area not affected by shading is detected, the reference loss power generation of each photovoltaic panel is calculated based on the shadow shading distribution map of each photovoltaic panel in the current cycle and the total radiation in the meteorological data of the current working cycle, and the reference loss power generation of each photovoltaic string is obtained, wherein the reference loss power generation of a photovoltaic string is the sum of the reference loss power generation of the photovoltaic panels it contains; The photovoltaic string power generation of each photovoltaic string collected by the combiner box is obtained, and the reference loss power generation of each photovoltaic string is added to the photovoltaic string power generation of each photovoltaic string to obtain the power generation evaluation value of each photovoltaic string; The obtained power generation evaluation values ​​are compared, and the data with large differences in the power generation evaluation values ​​are regarded as abnormal data through clustering algorithm; If there is no abnormal data, it is judged that the photovoltaic power generation group is blocked by external shadows for a long time or the surface is covered with dust, resulting in a decrease in conversion rate. It is not recorded as a suspected fault type, and the maintenance personnel are notified to perform maintenance; If there is abnormal data, the string cable in the photovoltaic string corresponding to the abnormal data is judged to be faulty, and the cable fault type is judged according to the abnormal changes in the DC current, DC voltage and power received by the corresponding junction box for the photovoltaic string. The cable fault types include poor contact, short circuit, open circuit and insulation damage of the string cable; the string cable fault type is recorded as a suspected fault type, matched with the hazard index, and the suspected fault type and the corresponding hazard index are packaged as fault assessment data and uploaded to the fault assessment platform.

[0012] Preferably, obtaining the suspected fault type and hazard index of the relevant components specifically includes: Use a DC power simulator to simulate the photovoltaic string inputting DC power to the combiner box, and collect relevant operating parameters, including: the combiner box input power, output power, DC voltage, DC current, combiner box temperature, combiner box efficiency, the central inverter input power, output power, DC voltage, AC voltage, DC current, AC current, inverter temperature and inverter efficiency; The above-mentioned relevant operating parameters are input into the fault type assessment model to determine whether there is a suspected fault type; if so, the suspected fault type is output, the hazard index of the suspected fault type is matched, the suspected fault type and the corresponding hazard index are packaged as fault assessment data, and uploaded to the fault assessment platform.

[0013] Preferably, the fault assessment data acquired based on the fault assessment platform and the calculation of the comprehensive fault assessment index specifically include: The calculation formula of the comprehensive fault assessment index is as follows: Where n is the number of suspected fault types, is the weight coefficient of the i-th suspected fault type, is the hazard index of the i-th suspected fault type; Based on the comprehensive fault assessment index, the medium risk threshold is set as Z1 and the high risk threshold is set as Z2; When the comprehensive fault assessment index is less than Z1, it is recorded as medium-low hazard risk, and the photovoltaic power generation group is initially investigated and optimized; When the comprehensive fault assessment index is greater than Z1 and less than Z2, it is recorded as a high safety risk. Based on the suspected fault type, a suitable rectification plan is matched and a maintenance plan is formulated for the photovoltaic power generation group; When the comprehensive fault assessment index is greater than Z2, it is recorded as an extremely high safety risk. Appropriate rectification plans are matched based on the suspected fault type, and emergency measures are taken for the photovoltaic power generation group.

[0014] Preferably, the present invention further proposes a photovoltaic power station management system based on artificial intelligence and big data analysis, which is used to implement any of the above management methods, including: Photovoltaic data platform, used to store basic information, location information, photovoltaic fault data and historical photovoltaic data of each photovoltaic power generation group in the photovoltaic power station; The power generation prediction module is used to obtain the status information and meteorological data of the historical working cycle from the data platform, use the convolutional neural network to train the photovoltaic power generation prediction model, and calculate the photovoltaic predicted power generation in the current working cycle; An operation status monitoring module is used to obtain the first photovoltaic power generation in the current photovoltaic data, set the photovoltaic power generation fluctuation range based on the predicted power generation, and judge whether the photovoltaic power generation group is in a normal operating state based on the relationship between the first photovoltaic power generation and the predicted power generation. If it is not within the fluctuation range, the fault analysis process is triggered; A fault detection module is used to set the photovoltaic power generation interval, obtain the second photovoltaic power generation received by each combiner box in the current working cycle, and summarize the total second photovoltaic power generation, compare the total second photovoltaic power generation with the photovoltaic power generation interval, and determine whether there is a fault; The fault type assessment module is used to obtain the outlier data of the fault operation parameters through box plot detection and calculate the outlier data difference index. Based on the outlier data of the fault operation parameters and their corresponding fault waveforms, a convolutional neural network model is established and trained to obtain a fault type assessment model. Based on the real-time operation parameters, the suspected fault type is determined and matched with the hazard index, and the data is packaged as fault assessment data and uploaded to the fault assessment platform. The photovoltaic string detection module uses visible light imagers and infrared imagers to capture visible light images and infrared images of photovoltaic panels, obtains shading distribution maps and temperature distribution maps, identifies abnormal temperature areas in infrared characteristic images by superimposing shadow shading distribution maps and temperature distribution maps, and determines photovoltaic panel or string faults; based on the shadow shading distribution map and the total radiation in the meteorological data, calculates the reference loss of power generation of each photovoltaic panel, obtains the power generation evaluation value of each photovoltaic string, and identifies abnormal data through clustering algorithms; The simulation test module uses a DC power simulator to simulate the PV string inputting DC power to the combiner box, collects relevant operating parameters, inputs the collected relevant operating parameters into the fault type assessment model, determines whether there is a suspected fault type, outputs the suspected fault type and matches the hazard index, and packages it into fault assessment data and uploads it to the fault assessment platform; The comprehensive fault assessment module uses the comprehensive fault assessment index formula to calculate the comprehensive fault assessment index based on the fault assessment data obtained by the fault assessment platform, assesses the risk level of the photovoltaic power generation group based on the risk threshold, matches the appropriate rectification plan based on the risk level, and notifies the operation and maintenance management personnel to carry out operation and maintenance management of areas with fault risks; The photovoltaic data platform specifically includes: A data acquisition unit is used to collect basic information, location information, fault data, historical photovoltaic data and current photovoltaic data from each photovoltaic power generation group in the photovoltaic power station; A data storage unit is used to store basic information, location information, fault data and historical photovoltaic data collected by each photovoltaic power generation group in the photovoltaic power station; The data processing unit is used to clean, denoise and interpolate the collected data to ensure the quality and integrity of the data and provide a reliable data basis for subsequent analysis.

[0015] The power generation prediction module specifically includes: The power generation prediction model training unit uses a convolutional neural network machine learning algorithm to train the photovoltaic power generation prediction model based on the preprocessed data set, and uses cross-validation and grid search to optimize the model performance; The prediction calculation unit is used to input the status information and meteorological data of the current working cycle into the trained model to calculate the predicted photovoltaic power generation.

[0016] Preferably, the fault detection module specifically includes: A second photovoltaic power generation calculation unit is used to calculate the minimum acceptable power generation and the maximum acceptable power generation, and set the photovoltaic power generation range; A monitoring data acquisition unit is used to acquire the second photovoltaic power generation received by each combiner box in the current working cycle, and summarize to obtain the total second photovoltaic power generation; The fault detection unit is used to compare the total second photovoltaic power generation with the photovoltaic power generation interval to determine whether there is a fault, and then further detect the photovoltaic strings and other components.

[0017] Compared with the prior art, the present invention provides a photovoltaic power station management method and system based on artificial intelligence and big data analysis, which has the following beneficial effects: This method monitors the operating status of the photovoltaic power generation group in real time, quickly identifies abnormal situations based on the comparison between the predicted power generation and the first photovoltaic power generation, and manages the faults in a graded manner through a comprehensive fault assessment index to match appropriate rectification plans. In addition, the status of the photovoltaic panels is non-destructively detected through visible light and infrared imaging technology, and the cause of the fault and the degree of harm are comprehensively evaluated in combination with meteorological data and historical operating parameters. After maintenance, the fault data is updated and uploaded to the photovoltaic data platform to optimize model performance and form a closed-loop management mechanism. This management method improves the operation and maintenance response speed, reduces fault handling time, and reduces maintenance and management costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of a photovoltaic power station management method based on artificial intelligence and big data analysis of the present invention; Figure 2 This is a framework diagram of a photovoltaic power station management system based on artificial intelligence and big data analysis in the present invention; Figure 3 A flow chart for constructing a fault type assessment model for the present invention. DETAILED DESCRIPTION

[0019] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the present invention will be further described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] In order to address the shortcomings of the existing technology, such as Figure 1 As shown, the present invention provides a photovoltaic power station management method based on artificial intelligence and big data analysis, including: Acquiring information on the structure of a photovoltaic power station, wherein the photovoltaic power station structure includes a plurality of photovoltaic power generation groups, the photovoltaic power generation groups include a plurality of photovoltaic strings for connecting a plurality of photovoltaic panels, at least one combiner box, a DC cable, a central inverter and an AC cable, and the photovoltaic power generation groups are used to transmit the converted AC power to a power grid; Constructing a photovoltaic data platform, which is used to store basic information, location information, photovoltaic fault data and historical photovoltaic data of each photovoltaic power generation group in the photovoltaic power station, wherein the historical photovoltaic data includes status information of multiple historical working cycles, the first photovoltaic power generation, meteorological data and historical operating parameters; Build a photovoltaic power generation prediction model based on historical photovoltaic data, obtain current photovoltaic data of the current working cycle, and calculate the predicted power generation of the photovoltaic power generation group in the current working cycle; Obtaining a first photovoltaic power generation in the current photovoltaic data, judging the operating state of the photovoltaic power generation group based on the predicted power generation and the first photovoltaic power generation, and performing a fault analysis on the photovoltaic power generation group if it is in an abnormal operating state; Calculate the photovoltaic power generation interval received by the central inverter based on the predicted power generation, obtain the second photovoltaic power generation received by each combiner box in the current working cycle, and detect the fault based on the relationship between the second photovoltaic power generation and the photovoltaic power generation interval; If the second photovoltaic power generation is within the photovoltaic power generation range, the real-time operating parameters of the current working cycle are obtained, a fault type assessment model is constructed based on the photovoltaic fault data, and the suspected fault types and hazard indexes of the relevant components are obtained based on the real-time operating parameters and packaged into fault assessment data; If the second photovoltaic power generation is not within the photovoltaic power generation range, the photovoltaic string power generation data of each photovoltaic string received by the central inverter in the current working cycle is obtained, and based on the site image of the photovoltaic power generation group, the data is analyzed and processed through the image recognition algorithm to obtain the suspected fault type and hazard index of the photovoltaic string, and packaged into fault assessment data; Use a DC power simulator to simulate the PV string inputting DC power to the combiner box, collect relevant operating parameters, and obtain the suspected fault types and hazard indexes of relevant components based on the fault type assessment model, and package them into fault assessment data; Based on the fault assessment data obtained from the fault assessment platform, a comprehensive fault assessment index is calculated; based on the comprehensive fault assessment index, a suitable rectification plan is matched to the photovoltaic power generation group with failure risk in the photovoltaic power station.

[0021] Specifically, the calculation of the predicted power generation of the photovoltaic power generation group in the current working cycle based on the photovoltaic power generation prediction model specifically includes: Based on the photovoltaic data platform, the status information of multiple historical working cycles in the historical photovoltaic data, the first photovoltaic power generation and meteorological data are obtained; the status information includes the total operation time of the equipment, the operation time of the equipment cycle, performance indicators and electrical parameters; the meteorological data includes ambient temperature, relative humidity, wind speed, irradiation intensity and total radiation; Clean the acquired data, remove outliers and noise data, and use interpolation to process missing values; The data in the same historical working cycle are packaged into a set of data sets, which are divided into training sets and test sets. The convolutional neural network model is trained using the training set, and the model parameters are adjusted through the back-propagation algorithm. The trained photovoltaic power generation prediction model is used to calculate the photovoltaic predicted power generation in the current working cycle. During the photovoltaic power generation prediction model training process, cross-validation and grid search are used to perform hyperparameter tuning to optimize the performance of the photovoltaic power generation prediction model. Acquiring current photovoltaic data of a current working cycle, wherein the current photovoltaic data includes state information, first photovoltaic power generation and meteorological data; The status information and meteorological data of the current working cycle are input into the photovoltaic power generation prediction model to calculate the photovoltaic predicted power generation.

[0022] Specifically, judging the operating state of the photovoltaic power generation group based on the predicted power generation and the first photovoltaic power generation includes: Acquire a first photovoltaic power generation amount of current photovoltaic data, where the first photovoltaic power generation amount is the amount of electricity received by the power grid; Set the photovoltaic power generation fluctuation range based on the predicted power generation; If the first photovoltaic power generation is within the photovoltaic power generation fluctuation range, the photovoltaic power generation group of the current working cycle is in normal operating state, and the current photovoltaic data collected in the current working cycle is uploaded to the photovoltaic data platform as historical photovoltaic data; otherwise, the photovoltaic power generation group is in abnormal operating state, and a fault analysis is performed on the photovoltaic power generation group.

[0023] Specifically, the fault detection based on the relationship between the second photovoltaic power generation and the photovoltaic power generation interval includes: Setting a minimum conversion efficiency, which is the product of the minimum conversion efficiency of the photovoltaic strings, combiner boxes, DC cable losses, central inverters, and AC cables in the photovoltaic power generation group and an adjustment factor, wherein the adjustment factor is a coefficient that quantifies various losses and uncertainties that may be encountered by each component in actual operation, and is a constant between 0.95 and 1; The product of the inverse of the minimum conversion efficiency and the predicted power generation is taken as the maximum acceptable power generation, and the predicted power generation is taken as the minimum acceptable power generation; The photovoltaic power generation range is set, and the minimum acceptable power generation and the maximum acceptable power generation are respectively used as the lower limit and the upper limit of the photovoltaic power generation range; The photovoltaic power generation range is ,in To predict the power consumption, For the lowest conversion efficiency; Obtain the second photovoltaic power generation received by each combiner box in the current working cycle; Obtain the second photovoltaic power generation received by each combiner box in the current working cycle; The second photovoltaic power generation of the combiner box is aggregated to obtain the total second photovoltaic power generation of the photovoltaic power generation group. If the total second photovoltaic power generation is within the photovoltaic power generation range, it is determined that there is no fault in each photovoltaic string of the photovoltaic power generation group, and other components in the photovoltaic power generation group are tested, including the combiner box, DC cable, central inverter and AC cable; If the total second photovoltaic power generation is not within the photovoltaic power generation range, it is determined that each photovoltaic string of the photovoltaic power generation group has a fault, and the photovoltaic strings in the photovoltaic power generation group are first tested, and then other components are tested.

[0024] Specifically, Figure 3 As shown, the fault type assessment model is constructed based on photovoltaic fault data, and the suspected fault causes and hazard indexes of relevant components are obtained based on real-time operating parameters, including: Obtain photovoltaic fault data and historical operating parameters of historical working cycles; The photovoltaic fault data includes fault type, fault operating parameters, fault number, fault hazard degree and hazard index; According to the fault operation parameters and historical operation parameters, the outlier data of the fault operation parameters are obtained based on the box plot detection; The types of fault operation parameters and historical operation parameters include: input power, output power, DC voltage, DC current, combiner box temperature, combiner box efficiency, input power, output power, DC voltage, AC voltage, DC current, AC current, inverter temperature and inverter efficiency of the central inverter; Based on the type of fault operation parameters, the fault operation parameters of the same type are divided into a group to obtain an outlier data group of the fault operation parameters; Analyze the outlier data of fault operation parameters and obtain the outlier data difference index; Based on data analysis requirements, obtain the outlier data difference index threshold; According to the outlier data difference index and the outlier data difference index threshold, it is determined whether the outlier data of the fault operation parameter is abnormal operation fault data; If the outlier data difference index exceeds the outlier data difference index threshold, the outlier data of the fault operation parameter is erroneous data, and the data is removed; If the outlier data difference index does not exceed the outlier data difference index threshold, the outlier data of the fault operation parameter is available; Matching the above fault operating parameter outlier data with the corresponding fault type, obtaining the fault waveform of the fault operating parameter outlier data, and merging the data to obtain a data set; Based on the fault waveform and fault type of the outlier data of fault operation parameters, a convolutional neural network model is established for training to obtain a fault type assessment model. Acquire real-time operating parameters of the current working cycle, wherein the operating parameters include each parameter in the historical operating parameters; Inputting the real-time operating parameters into the fault type assessment model, detecting and comparing the fault waveform through the fault type assessment model, and determining at least one suspected fault type; Match the hazard index of the suspected fault type, package the suspected fault type and the corresponding hazard index into fault assessment data, and upload it to the fault assessment platform; The outlier data difference index is: In the formula, For the The outlier difference index of the outlier data of the fault operation parameters, For the Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs The outlier data of the fault operating parameters and The interval time between outlier data of operating parameters of each historical fault, For the The total number of fault operating parameter outlier data in the fault operating parameter outlier data group to which the fault operating parameter outlier data belongs.

[0025] Specifically, the obtaining of the fault cause and hazard index of the photovoltaic string includes: Based on the visible light imager and infrared imager installed near the photovoltaic power generation group, the visible light image and infrared image of each photovoltaic panel in the photovoltaic power generation group are captured in real time at preset time intervals; Use filters to remove noise from visible light images and infrared images, and convert the color of visible light images into grayscale to reduce data dimensions; Use edge detection algorithm to detect edge contours of each photovoltaic panel in visible light image and infrared image, extract images within edge contours respectively, and record them as visible light feature images and infrared feature images; Based on the image recognition algorithm, the grayscale value and contrast value of each pixel in the visible light feature image are obtained to obtain the shadow occlusion distribution map; Based on the image recognition algorithm, the temperature value of each pixel in the infrared characteristic image is obtained to obtain the temperature distribution map; Based on the temperature distribution map, the temperature abnormal area is detected, the shadow occlusion distribution map is compared with the temperature abnormal area, and the temperature abnormal area is divided into an abnormal area affected by the occlusion and an abnormal area not affected by the occlusion; If an abnormal area not affected by occlusion is detected, the photovoltaic panel and the photovoltaic string where the photovoltaic panel is located corresponding to the infrared characteristic image are obtained, and the photovoltaic string where the photovoltaic panel is located is judged to be faulty. The photovoltaic panel fault is regarded as a suspected fault type, and the hazard index is calculated based on the number of photovoltaic panel faults and the degree of fault. The suspected fault type and the corresponding hazard index are packaged as fault assessment data and uploaded to the fault assessment platform; If no abnormal area not affected by shading is detected, the reference loss power generation of each photovoltaic panel is calculated based on the shadow shading distribution map of each photovoltaic panel in the current cycle and the total radiation in the meteorological data of the current working cycle, and the reference loss power generation of each photovoltaic string is obtained, wherein the reference loss power generation of a photovoltaic string is the sum of the reference loss power generation of the photovoltaic panels it contains; The photovoltaic string power generation of each photovoltaic string collected by the combiner box is obtained, and the reference loss power generation of each photovoltaic string is added to the photovoltaic string power generation of each photovoltaic string to obtain the power generation evaluation value of each photovoltaic string; The obtained power generation evaluation values ​​are compared, and the data with large differences in the power generation evaluation values ​​are regarded as abnormal data through clustering algorithm; If there is no abnormal data, it is judged that the photovoltaic power generation group is blocked by external shadows for a long time or the surface is covered with dust, resulting in a decrease in conversion rate. It is not recorded as a suspected fault type, and the maintenance personnel are notified to perform maintenance; If there is abnormal data, the string cable in the photovoltaic string corresponding to the abnormal data is judged to be faulty, and the cable fault type is judged according to the abnormal changes in the DC current, DC voltage and power received by the corresponding junction box for the photovoltaic string. The cable fault types include poor contact, short circuit, open circuit and insulation damage of the string cable; the string cable fault type is recorded as a suspected fault type, matched with the hazard index, and the suspected fault type and the corresponding hazard index are packaged as fault assessment data and uploaded to the fault assessment platform.

[0026] Specifically, obtaining the suspected fault type and hazard index of the relevant components specifically includes: Use a DC power simulator to simulate the photovoltaic string inputting DC power to the combiner box, and collect relevant operating parameters, including: the combiner box input power, output power, DC voltage, DC current, combiner box temperature, combiner box efficiency, the central inverter input power, output power, DC voltage, AC voltage, DC current, AC current, inverter temperature and inverter efficiency; The above-mentioned relevant operating parameters are input into the fault type assessment model to determine whether there is a suspected fault type; if so, the suspected fault type is output, the hazard index of the suspected fault type is matched, and the suspected fault type and the corresponding hazard index are packaged as fault assessment data and uploaded to the fault assessment platform; It should be noted that the hazard index matching the suspected fault type includes: Based on the fault type, fault operating parameters, fault frequency and fault hazard degree in the photovoltaic fault data; Obtain the fault hazard level of the fault type corresponding to the suspected fault type. The fault hazard level is scored based on a 1 to 10-point scale set based on the maintenance difficulty and hazard impact of the fault and is obtained through expert scoring. The higher the score, the greater the fault hazard level; The mean value of the fault hazard degree of the fault type in the photovoltaic fault data is used as the hazard index; Specifically, the fault assessment data obtained based on the fault assessment platform is used to calculate the comprehensive fault assessment index, which specifically includes: The calculation formula of the comprehensive fault assessment index is as follows: Where n is the number of suspected fault types, is the weight coefficient of the i-th suspected fault type, is the hazard index of the i-th suspected fault type; Based on the comprehensive fault assessment index, the medium risk threshold is set as Z1 and the high risk threshold is set as Z2; When the comprehensive fault assessment index is less than Z1, it is recorded as medium-low hazard risk, and the photovoltaic power generation group is initially investigated and optimized; When the comprehensive fault assessment index is greater than Z1 and less than Z2, it is recorded as a high safety risk. Based on the suspected fault type, a suitable rectification plan is matched and a maintenance plan is formulated for the photovoltaic power generation group; When the comprehensive fault assessment index is greater than Z2, it is recorded as an extremely high safety risk, and appropriate rectification plans are matched based on the suspected fault type, and emergency measures are taken for the photovoltaic power generation group; It should be noted that after the maintenance is completed, the fault type and fault operating parameters of the photovoltaic power generation group are obtained, and the fault hazard level of this fault type is set through expert scoring based on the maintenance difficulty and hazard impact. The above data is uploaded to the photovoltaic fault data. The new fault data can be used to further train and optimize the fault assessment model; by continuously updating the model, its ability to recognize new fault types and the accuracy of assessing the degree of fault hazard can be improved.

[0027] Specifically, Figure 2 As shown, the present invention also proposes a photovoltaic power station management system based on artificial intelligence and big data analysis, which is used to implement any of the above management methods, including: Photovoltaic data platform, used to store basic information, location information, photovoltaic fault data and historical photovoltaic data of each photovoltaic power generation group in the photovoltaic power station; The power generation prediction module is used to obtain the status information and meteorological data of the historical working cycle from the data platform, use the convolutional neural network to train the photovoltaic power generation prediction model, and calculate the photovoltaic predicted power generation in the current working cycle; An operation status monitoring module is used to obtain the first photovoltaic power generation in the current photovoltaic data, set the photovoltaic power generation fluctuation range based on the predicted power generation, and judge whether the photovoltaic power generation group is in a normal operating state based on the relationship between the first photovoltaic power generation and the predicted power generation. If it is not within the fluctuation range, the fault analysis process is triggered; A fault detection module is used to set the photovoltaic power generation interval, obtain the second photovoltaic power generation received by each combiner box in the current working cycle, and summarize the total second photovoltaic power generation, compare the total second photovoltaic power generation with the photovoltaic power generation interval, and determine whether there is a fault; The fault type assessment module is used to obtain the outlier data of the fault operation parameters through box plot detection and calculate the outlier data difference index. Based on the outlier data of the fault operation parameters and their corresponding fault waveforms, a convolutional neural network model is established and trained to obtain a fault type assessment model. Based on the real-time operation parameters, the suspected fault type is determined and matched with the hazard index, and the data is packaged as fault assessment data and uploaded to the fault assessment platform. The photovoltaic string detection module uses visible light imagers and infrared imagers to capture visible light images and infrared images of photovoltaic panels, obtains shading distribution maps and temperature distribution maps, identifies abnormal temperature areas in infrared characteristic images by superimposing shadow shading distribution maps and temperature distribution maps, and determines photovoltaic panel or string faults; based on the shadow shading distribution map and the total radiation in the meteorological data, calculates the reference loss of power generation of each photovoltaic panel, obtains the power generation evaluation value of each photovoltaic string, and identifies abnormal data through clustering algorithms; The simulation test module uses a DC power simulator to simulate the PV string inputting DC power to the combiner box, collects relevant operating parameters, inputs the collected relevant operating parameters into the fault type assessment model, determines whether there is a suspected fault type, outputs the suspected fault type and matches the hazard index, and packages it into fault assessment data and uploads it to the fault assessment platform; The comprehensive fault assessment module uses the comprehensive fault assessment index formula to calculate the comprehensive fault assessment index based on the fault assessment data obtained by the fault assessment platform, assesses the risk level of the photovoltaic power generation group based on the risk threshold, matches the appropriate rectification plan based on the risk level, and notifies the operation and maintenance management personnel to carry out operation and maintenance management of areas with fault risks; The photovoltaic data platform specifically includes: A data acquisition unit is used to collect basic information, location information, fault data, historical photovoltaic data and current photovoltaic data from each photovoltaic power generation group in the photovoltaic power station; A data storage unit is used to store basic information, location information, fault data and historical photovoltaic data collected by each photovoltaic power generation group in the photovoltaic power station; The data processing unit is used to clean, denoise and interpolate the collected data to ensure the quality and integrity of the data and provide a reliable data basis for subsequent analysis.

[0028] The power generation prediction module specifically includes: The power generation prediction model training unit uses a convolutional neural network machine learning algorithm to train the photovoltaic power generation prediction model based on the preprocessed data set, and uses cross-validation and grid search to optimize the model performance; The prediction calculation unit is used to input the status information and meteorological data of the current working cycle into the trained model to calculate the predicted photovoltaic power generation.

[0029] Specifically, the fault detection module specifically includes: A second photovoltaic power generation calculation unit is used to calculate the minimum acceptable power generation and the maximum acceptable power generation, and set the photovoltaic power generation range; A monitoring data acquisition unit is used to acquire the second photovoltaic power generation received by each combiner box in the current working cycle, and summarize to obtain the total second photovoltaic power generation; The fault detection unit is used to compare the total second photovoltaic power generation with the photovoltaic power generation interval to determine whether there is a fault, and then further detect the photovoltaic strings and other components.

[0030] The fault type assessment module specifically includes: The outlier data analysis unit obtains the outlier data of the fault operation parameters through box plot detection and calculates the outlier data difference index.

[0031] The fault assessment model training unit establishes a convolutional neural network model based on the outlier data of the fault operation parameters and their corresponding fault waveforms, and trains to obtain a fault type assessment model.

[0032] The fault identification unit is used to input the real-time operating parameters into the fault type assessment model, determine the suspected fault type, match the hazard index, and package it into fault assessment data and upload it to the fault assessment platform.

[0033] The photovoltaic string detection module specifically includes The image acquisition unit takes visible light images and infrared images of the photovoltaic panel through a visible light imager and an infrared imager.

[0034] The image processing unit removes image noise through filters, converts visible light images into grayscale images to reduce data dimensions, and extracts image features within edge contours.

[0035] The temperature anomaly detection unit identifies abnormal temperature areas in the infrared characteristic image by superimposing the shadow occlusion distribution map and the temperature distribution map, and determines the fault of the photovoltaic panel or string.

[0036] The power generation loss calculation unit calculates the reference loss power generation of each photovoltaic panel based on the shadow shielding distribution map and the total radiation in the meteorological data, obtains the power generation evaluation value of each photovoltaic string, and identifies abnormal data through a clustering algorithm.

[0037] To sum up, the advantages of the present invention are as follows: the present invention obtains fault assessment data through the fault assessment platform, calculates the comprehensive fault assessment index, and the index comprehensively considers multiple suspected fault types and their hazard indexes. By setting different risk thresholds, the system can perform hierarchical management of photovoltaic power generation groups with failure risks in photovoltaic power stations, match appropriate rectification plans for different levels of risks, and notify operation and maintenance management personnel to take corresponding measures to ensure the safe and stable operation of photovoltaic power stations.

[0038] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A photovoltaic power station management method based on artificial intelligence and big data analysis, characterized in that: The steps include: Acquiring information on the structure of a photovoltaic power station, wherein the photovoltaic power station structure includes a plurality of photovoltaic power generation groups, the photovoltaic power generation groups include a plurality of photovoltaic strings for connecting a plurality of photovoltaic panels, at least one combiner box, a DC cable, a central inverter and an AC cable, and the photovoltaic power generation groups are used to transmit the converted AC power to a power grid; Constructing a photovoltaic data platform, which is used to store basic information, location information, photovoltaic fault data and historical photovoltaic data of each photovoltaic power generation group in the photovoltaic power station, wherein the historical photovoltaic data includes status information of multiple historical working cycles, the first photovoltaic power generation, meteorological data and historical operating parameters; Build a photovoltaic power generation prediction model based on historical photovoltaic data, obtain current photovoltaic data of the current working cycle, and calculate the predicted power generation of the photovoltaic power generation group in the current working cycle; Obtaining a first photovoltaic power generation in the current photovoltaic data, judging the operating state of the photovoltaic power generation group based on the predicted power generation and the first photovoltaic power generation, and performing a fault analysis on the photovoltaic power generation group if it is in an abnormal operating state; Calculate the photovoltaic power generation interval received by the central inverter based on the predicted power generation, obtain the second photovoltaic power generation received by each combiner box in the current working cycle, and detect the fault based on the relationship between the second photovoltaic power generation and the photovoltaic power generation interval; If the second photovoltaic power generation is within the photovoltaic power generation range, the real-time operating parameters of the current working cycle are obtained, a fault type assessment model is constructed based on the photovoltaic fault data, and the suspected fault types and hazard indexes of the relevant components are obtained based on the real-time operating parameters and packaged into fault assessment data; If the second photovoltaic power generation is not within the photovoltaic power generation range, the photovoltaic string power generation data of each photovoltaic string received by the central inverter in the current working cycle is obtained, and based on the site image of the photovoltaic power generation group, the data is analyzed and processed through the image recognition algorithm to obtain the suspected fault type and hazard index of the photovoltaic string, and packaged into fault assessment data; Use a DC power simulator to simulate the PV string inputting DC power to the combiner box, collect relevant operating parameters, and obtain the suspected fault types and hazard indexes of relevant components based on the fault type assessment model, and package them into fault assessment data; Based on the fault assessment data obtained from the fault assessment platform, a comprehensive fault assessment index is calculated; based on the comprehensive fault assessment index, a suitable rectification plan is matched to the photovoltaic power generation group with failure risk in the photovoltaic power station.

2. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 1, characterized in that: The calculation of the predicted power generation of the photovoltaic power generation group in the current working cycle based on the photovoltaic power generation prediction model specifically includes: Based on the photovoltaic data platform, the status information of multiple historical working cycles in the historical photovoltaic data, the first photovoltaic power generation and meteorological data are obtained; the status information includes the total operation time of the equipment, the operation time of the equipment cycle, performance indicators and electrical parameters; the meteorological data includes ambient temperature, relative humidity, wind speed, irradiation intensity and total radiation; Clean the acquired data, remove outliers and noise data, and use interpolation to process missing values; The data in the same historical working cycle are packaged into a set of data sets, which are divided into training sets and test sets. The convolutional neural network model is trained using the training set, and the model parameters are adjusted through the back-propagation algorithm. The trained photovoltaic power generation prediction model is used to calculate the photovoltaic predicted power generation in the current working cycle. During the photovoltaic power generation prediction model training process, cross-validation and grid search are used to perform hyperparameter tuning to optimize the performance of the photovoltaic power generation prediction model. Acquiring current photovoltaic data of a current working cycle, wherein the current photovoltaic data includes state information, first photovoltaic power generation and meteorological data; The status information and meteorological data of the current working cycle are input into the photovoltaic power generation prediction model to calculate the photovoltaic predicted power generation.

3. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 2, characterized in that: The step of judging the operating state of the photovoltaic power generation group based on the predicted power generation and the first photovoltaic power generation specifically includes: Acquire a first photovoltaic power generation amount of current photovoltaic data, where the first photovoltaic power generation amount is the amount of electricity received by the power grid; Set the photovoltaic power generation fluctuation range based on the predicted power generation; If the first photovoltaic power generation is within the photovoltaic power generation fluctuation range, the photovoltaic power generation group in the current working cycle is in a normal operating state; otherwise, the photovoltaic power generation group is in an abnormal operating state, and a fault analysis is performed on the photovoltaic power generation group.

4. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 3, characterized in that: The detecting of the fault based on the relationship between the second photovoltaic power generation and the photovoltaic power generation interval specifically includes: Setting a minimum conversion efficiency, which is the product of the minimum conversion efficiency of the photovoltaic strings, combiner boxes, DC cable losses, central inverters, and AC cables in the photovoltaic power generation group and an adjustment factor, wherein the adjustment factor is a coefficient that quantifies various losses and uncertainties that may be encountered by each component in actual operation, and is a constant between 0.95 and 1; The product of the inverse of the minimum conversion efficiency and the predicted power generation is taken as the maximum acceptable power generation, and the predicted power generation is taken as the minimum acceptable power generation; The photovoltaic power generation range is set, and the minimum acceptable power generation and the maximum acceptable power generation are respectively used as the lower limit and the upper limit of the photovoltaic power generation range; Obtain the second photovoltaic power generation received by each combiner box in the current working cycle; The second photovoltaic power generation of the combiner box is aggregated to obtain the total second photovoltaic power generation of the photovoltaic power generation group. If the total second photovoltaic power generation is within the photovoltaic power generation range, it is determined that there is no fault in each photovoltaic string of the photovoltaic power generation group, and other components in the photovoltaic power generation group are tested, including the combiner box, DC cable, central inverter and AC cable; If the total second photovoltaic power generation is not within the photovoltaic power generation range, it is determined that each photovoltaic string of the photovoltaic power generation group has a fault, and the photovoltaic strings in the photovoltaic power generation group are first tested, and then other components are tested.

5. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 4, characterized in that: The fault type assessment model is constructed based on photovoltaic fault data, and the suspected fault causes and hazard indexes of relevant components are obtained based on real-time operating parameters, including: Obtain photovoltaic fault data and historical operating parameters of historical working cycles; The photovoltaic fault data includes fault type, fault operating parameters, fault number, fault hazard degree and hazard index; According to the fault operation parameters and historical operation parameters, the outlier data of the fault operation parameters are obtained based on the box plot detection; The types of fault operation parameters and historical operation parameters include: input power, output power, DC voltage, DC current, combiner box temperature, combiner box efficiency, input power, output power, DC voltage, AC voltage, DC current, AC current, inverter temperature and inverter efficiency of the central inverter; Based on the type of fault operation parameters, the fault operation parameters of the same type are divided into a group to obtain an outlier data group of the fault operation parameters; Analyze the outlier data of fault operation parameters and obtain the outlier data difference index; Based on data analysis requirements, obtain the outlier data difference index threshold; According to the outlier data difference index and the outlier data difference index threshold, it is determined whether the outlier data of the fault operation parameter is abnormal operation fault data; If the outlier data difference index exceeds the outlier data difference index threshold, the outlier data of the fault operation parameter is erroneous data, and the data is removed; If the outlier data difference index does not exceed the outlier data difference index threshold, the outlier data of the fault operation parameter is available; Matching the above fault operating parameter outlier data with the corresponding fault type, obtaining the fault waveform of the fault operating parameter outlier data, and merging the data to obtain a data set; Based on the fault waveform and fault type of the outlier data of fault operation parameters, a convolutional neural network model is established for training to obtain a fault type assessment model. Acquire real-time operating parameters of the current working cycle, wherein the operating parameters include each parameter in the historical operating parameters; Inputting the real-time operating parameters into the fault type assessment model, detecting and comparing the fault waveform through the fault type assessment model, and determining at least one suspected fault type; Match the hazard index of the suspected fault type, package the suspected fault type and the corresponding hazard index into fault assessment data, and upload it to the fault assessment platform; The outlier data difference index is: In the formula, For the The outlier difference index of the outlier data of the fault operation parameters, For the Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs Outlier data of fault operation parameters, For the The fault operating parameter outlier data group to which the fault operating parameter outlier data belongs The outlier data of the fault operating parameters and The interval time between outlier data of operating parameters of each historical fault, For the The total number of fault operating parameter outlier data in the fault operating parameter outlier data group to which the fault operating parameter outlier data belongs.

6. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 5, characterized in that: The obtaining of the fault cause and hazard index of the photovoltaic string specifically includes: Based on the visible light imager and infrared imager installed near the photovoltaic power generation group, the visible light image and infrared image of each photovoltaic panel in the photovoltaic power generation group are captured in real time at preset time intervals; Use filters to remove noise from visible light images and infrared images, and convert the color of visible light images into grayscale to reduce data dimensions; Use edge detection algorithm to detect edge contours of each photovoltaic panel in visible light image and infrared image, extract images within edge contours respectively, and record them as visible light feature images and infrared feature images; Based on the image recognition algorithm, the grayscale value and contrast value of each pixel in the visible light feature image are obtained to obtain the shadow occlusion distribution map; Based on the image recognition algorithm, the temperature value of each pixel in the infrared characteristic image is obtained to obtain the temperature distribution map; Based on the temperature distribution map, the temperature abnormal area is detected, the shadow occlusion distribution map is compared with the temperature abnormal area, and the temperature abnormal area is divided into an abnormal area affected by the occlusion and an abnormal area not affected by the occlusion; If an abnormal area not affected by occlusion is detected, the photovoltaic panel and the photovoltaic string where the photovoltaic panel is located corresponding to the infrared characteristic image are obtained, and the photovoltaic string where the photovoltaic panel is located is judged to be faulty. The photovoltaic panel fault is regarded as a suspected fault type, and the hazard index is calculated based on the number of photovoltaic panel faults and the degree of fault. The suspected fault type and the corresponding hazard index are packaged as fault assessment data and uploaded to the fault assessment platform; If no abnormal area not affected by shading is detected, the reference loss power generation of each photovoltaic panel is calculated based on the shadow shading distribution map of each photovoltaic panel in the current cycle and the total radiation in the meteorological data of the current working cycle, and the reference loss power generation of each photovoltaic string is obtained, wherein the reference loss power generation of a photovoltaic string is the sum of the reference loss power generation of the photovoltaic panels it contains; The photovoltaic string power generation of each photovoltaic string collected by the combiner box is obtained, and the reference loss power generation of each photovoltaic string is added to the photovoltaic string power generation of each photovoltaic string to obtain the power generation evaluation value of each photovoltaic string; The obtained power generation evaluation values ​​are compared, and the data with large differences in the power generation evaluation values ​​are regarded as abnormal data through clustering algorithm; If there is no abnormal data, it is judged that the photovoltaic power generation group is blocked by external shadows for a long time or the surface is covered with dust, resulting in a decrease in conversion rate. It is not recorded as a suspected fault type, and the maintenance personnel are notified to perform maintenance; If there is abnormal data, the string cable in the photovoltaic string corresponding to the abnormal data is judged to be faulty, and the cable fault type is judged according to the abnormal changes in the DC current, DC voltage and power received by the corresponding junction box for the photovoltaic string. The cable fault types include poor contact, short circuit, open circuit and insulation damage of the string cable; the string cable fault type is recorded as a suspected fault type, matched with the hazard index, and the suspected fault type and the corresponding hazard index are packaged as fault assessment data and uploaded to the fault assessment platform.

7. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 6, characterized in that: The obtaining of the suspected fault type and hazard index of the relevant components specifically includes: Use a DC power simulator to simulate the photovoltaic string inputting DC power to the combiner box, and collect relevant operating parameters, including: the combiner box input power, output power, DC voltage, DC current, combiner box temperature, combiner box efficiency, the central inverter input power, output power, DC voltage, AC voltage, DC current, AC current, inverter temperature and inverter efficiency; Input the above-mentioned relevant operating parameters into the fault type assessment model to determine whether there is a suspected fault type; If it exists, the suspected fault type is output, the hazard index of the suspected fault type is matched, the suspected fault type and the corresponding hazard index are packaged as fault assessment data, and uploaded to the fault assessment platform.

8. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 7, characterized in that: The fault assessment data obtained based on the fault assessment platform is used to calculate the comprehensive fault assessment index, which specifically includes: The calculation formula of the comprehensive fault assessment index is as follows: Where n is the number of suspected fault types, is the weight coefficient of the i-th suspected fault type, is the hazard index of the i-th suspected fault type; Based on the comprehensive fault assessment index, the medium risk threshold is set as Z1 and the high risk threshold is set as Z2; When the comprehensive fault assessment index is less than Z1, it is recorded as medium-low hazard risk, and the photovoltaic power generation group is initially investigated and optimized; When the comprehensive fault assessment index is greater than Z1 and less than Z2, it is recorded as a high safety risk. Based on the suspected fault type, a suitable rectification plan is matched and a maintenance plan is formulated for the photovoltaic power generation group; When the comprehensive fault assessment index is greater than Z2, it is recorded as an extremely high safety risk. Appropriate rectification plans are matched based on the suspected fault type, and emergency measures are taken for the photovoltaic power generation group.

9. A photovoltaic power station management system based on artificial intelligence and big data analysis, used to implement the management method according to any one of claims 1 to 8, characterized in that: include: Photovoltaic data platform, used to store basic information, location information, photovoltaic fault data and historical photovoltaic data of each photovoltaic power generation group in the photovoltaic power station; The power generation prediction module is used to obtain the status information and meteorological data of the historical working cycle from the data platform, use the convolutional neural network to train the photovoltaic power generation prediction model, and calculate the photovoltaic predicted power generation in the current working cycle; An operation status monitoring module is used to obtain the first photovoltaic power generation in the current photovoltaic data, set the photovoltaic power generation fluctuation range based on the predicted power generation, and judge whether the photovoltaic power generation group is in a normal operating state based on the relationship between the first photovoltaic power generation and the predicted power generation. If it is not within the fluctuation range, the fault analysis process is triggered; A fault detection module is used to set the photovoltaic power generation interval, obtain the second photovoltaic power generation received by each combiner box in the current working cycle, and summarize the total second photovoltaic power generation, compare the total second photovoltaic power generation with the photovoltaic power generation interval, and determine whether there is a fault; The fault type assessment module is used to obtain the outlier data of the fault operation parameters through box plot detection and calculate the outlier data difference index. Based on the outlier data of the fault operation parameters and their corresponding fault waveforms, a convolutional neural network model is established and trained to obtain a fault type assessment model. Based on the real-time operation parameters, the suspected fault type is determined and matched with the hazard index, and the data is packaged as fault assessment data and uploaded to the fault assessment platform. The photovoltaic string detection module uses visible light imagers and infrared imagers to capture visible light images and infrared images of photovoltaic panels, obtains shading distribution maps and temperature distribution maps, identifies abnormal temperature areas in infrared characteristic images by superimposing shadow shading distribution maps and temperature distribution maps, and determines photovoltaic panel or string faults; based on the shadow shading distribution map and the total radiation in the meteorological data, calculates the reference loss of power generation of each photovoltaic panel, obtains the power generation evaluation value of each photovoltaic string, and identifies abnormal data through clustering algorithms; The simulation test module uses a DC power simulator to simulate the PV string inputting DC power to the combiner box, collects relevant operating parameters, inputs the collected relevant operating parameters into the fault type assessment model, determines whether there is a suspected fault type, outputs the suspected fault type and matches the hazard index, and packages it into fault assessment data and uploads it to the fault assessment platform; The comprehensive fault assessment module uses the comprehensive fault assessment index formula to calculate the comprehensive fault assessment index based on the fault assessment data obtained by the fault assessment platform, assesses the risk level of the photovoltaic power generation group based on the risk threshold, matches the appropriate rectification plan based on the risk level, and notifies the operation and maintenance management personnel to carry out operation and maintenance management of areas with fault risks; The photovoltaic data platform specifically includes: A data acquisition unit is used to collect basic information, location information, fault data, historical photovoltaic data and current photovoltaic data from each photovoltaic power generation group in the photovoltaic power station; A data storage unit is used to store basic information, location information, fault data and historical photovoltaic data collected by each photovoltaic power generation group in the photovoltaic power station; The data processing unit is used to clean, denoise and interpolate the collected data to ensure the quality and integrity of the data and provide a reliable data basis for subsequent analysis; The power generation prediction module specifically includes: The power generation prediction model training unit uses a convolutional neural network machine learning algorithm to train the photovoltaic power generation prediction model based on the preprocessed data set, and uses cross-validation and grid search to optimize the model performance; The prediction calculation unit is used to input the status information and meteorological data of the current working cycle into the trained model to calculate the predicted photovoltaic power generation.

10. A photovoltaic power station management method based on artificial intelligence and big data analysis according to claim 9, characterized in that: The fault detection module specifically includes: A second photovoltaic power generation calculation unit is used to calculate the minimum acceptable power generation and the maximum acceptable power generation, and set the photovoltaic power generation range; A monitoring data acquisition unit is used to acquire the second photovoltaic power generation received by each combiner box in the current working cycle, and summarize to obtain the total second photovoltaic power generation; The fault detection unit is used to compare the total second photovoltaic power generation with the photovoltaic power generation interval to determine whether there is a fault, and then further detect the photovoltaic strings and other components.

Images