A Comprehensive Intelligent Operation and Maintenance Method and System for Photovoltaic Power Stations Based on Cloud Processing

Abstract

The present invention discloses a comprehensive intelligent operation and maintenance method and system for a photovoltaic power station based on cloud processing, which relates to the technical field of operation and maintenance management of photovoltaic power stations. The comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing includes the following steps: obtaining the characteristic data of operation status data and parameter data and storing them in a cloud platform; using a solar energy conversion efficiency model to evaluate the power generation performance of photovoltaic devices in real time; formulating the optimal cleaning cycle of photovoltaic devices; and inputting the detection results into a photovoltaic device operation fault and anomaly database for fault discrimination. By using an optoelectronic conversion model to evaluate the power generation efficiency in real time, the present invention enables operation and maintenance personnel to understand the power generation status of the photovoltaic power station, avoid frequent cleaning or too long cleaning intervals, thereby improving the discovery and identification ability of fault data, further improving the service life and operation efficiency of equipment, and further improving the operation and maintenance efficiency of the photovoltaic power station.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power station operation and maintenance management. Specifically, it relates to a comprehensive intelligent operation and maintenance method and system for photovoltaic power stations based on cloud processing. Background Art

[0002] A photovoltaic power station is a facility that uses photovoltaic power generation technology to convert solar energy into electrical energy. It consists of a large number of solar panels (photovoltaic modules), which convert solar energy into direct current through photovoltaic conversion. The direct current is converted into alternating current by an inverter and then stepped up by a transformer and connected to the power grid or power supply equipment to achieve the transmission and utilization of electrical energy.

[0003] Intelligent operation and maintenance management is a method that uses advanced technical means and data analysis methods to improve the operation and maintenance efficiency, economic benefits, and reliability of photovoltaic power stations. In the intelligent operation and maintenance management of photovoltaic power stations, power generation prediction is a very important link. Accurate power generation prediction can provide a reliable basis for the planning, operation management, energy market trading, and equipment maintenance of photovoltaic power stations.

[0004] However, in the prior art, since the power generation of a photovoltaic power station is affected by weather conditions such as sunlight intensity, cloud cover, and temperature, it is difficult to predict the power generation of a photovoltaic power station. The evaluation of photovoltaic potential pays more attention to data collection and real-time monitoring, and less to data analysis and prediction, which is not conducive to comprehensively evaluating the photovoltaic potential, and thus not conducive to improving the accuracy and reliability of power generation prediction. The operation efficiency and economic benefits of photovoltaic power stations are gradually decreasing.

[0005] In response to the problems in the related art, no effective solution has been proposed yet. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention proposes a comprehensive intelligent operation and maintenance method and system for photovoltaic power stations based on cloud processing, which solves the problems mentioned in the above background art that the power generation of existing photovoltaic power stations is affected by weather conditions such as sunlight intensity, cloud cover, and temperature, so it is difficult to predict the power generation of photovoltaic power stations. The evaluation of photovoltaic potential pays more attention to data collection and real-time monitoring, and less to data analysis and prediction, which is not conducive to comprehensively evaluating the photovoltaic potential, and thus not conducive to improving the accuracy and reliability of power generation prediction. The operation efficiency and economic benefits of photovoltaic power stations are gradually decreasing.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions:

[0008] According to one aspect of the present invention, there is provided a comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing. The comprehensive intelligent operation and maintenance method for the photovoltaic power station based on cloud processing includes the following steps:

[0009] S1. Obtain the operation status data and parameter data of the photovoltaic power station, and preprocess the obtained operation status data and parameter data to obtain the characteristic data of the operation status data and parameter data, and store them in the cloud platform;

[0010] S2. Use cloud processing analysis technology to establish a solar energy conversion efficiency model for the photovoltaic power station, and use the solar energy conversion efficiency model to evaluate the power generation performance of photovoltaic devices in real time;

[0011] S3. Use an analysis algorithm to analyze the evaluation results of the power generation performance, judge the accumulation degree of obstacles on the surface of the photovoltaic device, formulate the best cleaning cycle of the photovoltaic device, and use an automatic cleaning device to clean it;

[0012] S4. Establish a database for abnormal operation failures of photovoltaic devices, perform anomaly detection on the characteristic data in the cloud platform through an anomaly detection algorithm, and input the detection results into the database for abnormal operation failures of photovoltaic devices for fault discrimination;

[0013] S5. Based on the historical data in the database for abnormal operation failures and the characteristic data in the cloud platform, use the beamforming method and time series analysis method to construct a fault prediction model in the cloud platform, and use the fault prediction model to predict the occurrence of faults at the next moment;

[0014] S6. According to the prediction results, formulate corresponding maintenance strategies and emergency plans.

[0015] Further, obtaining the operation status data and parameter data of the photovoltaic power station, and preprocessing the obtained operation status data and parameter data to obtain the characteristic data of the operation status data and parameter data, and storing them in the cloud platform includes the following steps:

[0016] S11. Collect the duplicate data, missing values, and abnormal values of the obtained operation status data and parameter data, and perform denoising, filtering, and smoothing processing on the duplicate data, missing values, and abnormal values;

[0017] S12. Perform data verification on the operation status data and parameter data after denoising, filtering, and smoothing processing, extract the valid data, and perform normalization processing to generate accurate operation status data and parameter data;

[0018] S13. Use the principal component analysis method to fuse the generated accurate operation status data and parameter data into the same dataset;

[0019] S14. Extract relevant features from the fused dataset to obtain the feature data of the operating status data and parameter data, and store them in the cloud platform.

[0020] Further, use cloud processing analysis technology to establish a solar energy conversion efficiency model for a photovoltaic power station, and use the solar energy conversion efficiency model to evaluate the power generation performance of photovoltaic equipment in real time, including the following steps:

[0021] S21. Use cloud processing analysis technology to analyze the operating status data of photovoltaic equipment and extract features related to photovoltaic modules.

[0022] S22. Convert the extracted relevant features into the characteristics of photovoltaic modules and the principle of photovoltaic conversion, and construct a solar energy conversion efficiency model.

[0023] S23. By fitting the actually measured power generation power and irradiance data with the constructed solar energy conversion efficiency model, estimate the parameter values in the solar energy conversion efficiency model.

[0024] S24. Use the established solar energy conversion efficiency model to calculate the actual power generation efficiency of photovoltaic equipment, and compare the actual power generation efficiency with the expected efficiency to evaluate the power generation performance of photovoltaic equipment.

[0025] Further, use an analysis algorithm to analyze the evaluation results of power generation performance, judge the accumulation degree of obstacles on the surface of photovoltaic equipment, formulate the optimal cleaning cycle of photovoltaic equipment, and use an automatic cleaning device to clean it, including the following steps:

[0026] S31. Import the actual power generation efficiency and expected efficiency, initialize, and use the shortest path algorithm to calculate the accumulation degree of obstacles between any two detection points on the surface of photovoltaic equipment.

[0027] S32. Set up several analysis strategies. Each analysis strategy represents a correlation analysis method between the accumulation degree of obstacles and the power generation performance of the equipment. The first N numbers represent the numbers of the selected detection points, and the last N numbers represent the corresponding power generation performance levels of the photovoltaic equipment.

[0028] S33. According to the preset performance maintenance target of photovoltaic equipment, find the global optimal analysis strategy, and add the environmental impact factor to jointly affect the optimization of the analysis strategy with each group of local optimal analysis strategies and the operation trend of photovoltaic equipment.

[0029] S34. Substitute the optimized analysis strategy into the Kent mapping, and compare the analysis strategy after Kent mapping with the analysis strategy before optimization with the power generation performance of photovoltaic equipment and the accumulation degree of obstacles as the evaluation criteria.

[0030] S35. Apply the elitist retention strategy and replace the analysis strategy with the worst effect with the globally sub-optimal analysis strategy according to the photovoltaic device performance maintenance goal;

[0031] S36. Loop through steps S33 to S35. If the number of iterations exceeds the threshold, end the iteration to obtain the best judgment method for evaluating the power generation performance of the photovoltaic device and the degree of accumulation of obstacles;

[0032] S37. According to the best judgment method, formulate the best cleaning cycle of the photovoltaic device and use the automatic cleaning equipment to clean it.

[0033] Furthermore, according to the preset photovoltaic device performance maintenance goal, find the globally optimal analysis strategy. The optimization of adding the environmental impact factor and jointly affecting the analysis strategy with each group of locally optimal analysis strategies and the operation trend of the photovoltaic device includes the following steps:

[0034] S331. Define the objective function of the photovoltaic device performance maintenance goal to evaluate the fitness of the analysis strategy;

[0035] S332. Initialize the analysis strategy population and randomly generate the parameters of each analysis strategy;

[0036] S333. Calculate the fitness of each analysis strategy according to the objective function and find the globally optimal analysis strategy as the best solution;

[0037] S334. Randomly generate environmental impact factors for each group of best analysis strategies and calculate the within-group average parameters of each group of locally best analysis strategies as the group population trend;

[0038] S335. Substitute the original analysis strategy parameters into the update formula, calculate the parameters of the new analysis strategy, and update the parameters of each analysis strategy;

[0039] S336. Repeat steps S333 to S335 until the termination condition is met.

[0040] Furthermore, substitute the optimized analysis strategy into the Kent mapping, and compare the analysis strategy after Kent mapping with the analysis strategy before optimization using the power generation performance of the photovoltaic device and the degree of accumulation of obstacles as the evaluation criteria, including the following steps:

[0041] S341. Train and iterate the analysis strategy population to obtain the optimized analysis strategy population;

[0042] S342. Perform data normalization processing on the optimized analysis strategy population to make it in a preset state;

[0043] S343. Randomly generate the parameters of the Kent mapping and preset the value range;

[0044] S344. Map each optimized analysis strategy using the Kent mapping formula to generate new analysis strategy individuals;

[0045] S345. Compare the fitness of each analysis strategy with its Kent mapping result, and retain the analysis strategy individual with the best fitness as the Kent mapping analysis strategy;

[0046] S346. Compare the fitness of the Kent mapping analysis strategy with the original analysis strategy population, and retain the analysis strategy individual with the best fitness to enter the next generation of analysis strategy population;

[0047] S347. Repeat steps S342 to S346 until the iteration termination condition is reached.

[0048] Furthermore, the update formula is:

[0049] ;

[0050] Wherein, represents the parameters of the optimized new analysis strategy;

[0051] represents the parameters of the original analysis strategy;

[0052] m represents the m th analysis strategy;

[0053] i represents the i th iteration;

[0054] b represents the b th dimension parameter value;

[0055] x represents the parameter vector of the analysis strategy;

[0056] l 1 represents the weight of the environmental impact factor after normalization;

[0057] l 2 represents the growth factor δ 1's random weight impact factor after normalization;

[0058] l 3 represents the growth factor δ 2's random weight impact factor after normalization.

[0059] Further, establish a database for abnormal operation faults of photovoltaic equipment, perform anomaly detection on the feature data in the cloud platform through an anomaly detection algorithm, and input the detection results into the database for abnormal operation faults of photovoltaic equipment for fault discrimination, including the following steps:

[0060] S41. Establish a database for abnormal operation faults of photovoltaic equipment;

[0061] S42. Input the dataset of performance parameters of photovoltaic equipment into the cloud platform, preset the height of the decision tree in the anomaly detection algorithm, and initialize the random forest model;

[0062] S43. Use the performance parameter data of photovoltaic equipment to construct several decision trees to form an initial random forest model;

[0063] S44. Use the performance parameters of normally operating photovoltaic equipment as the training set to train the initial random forest model;

[0064] S45. According to the differences and accuracies of the decision trees, use a probability search algorithm to screen out decision trees with higher fitness from the initial forest model and combine them into a new random forest model;

[0065] S46. Apply the new random forest model to the performance parameters of photovoltaic equipment in the cloud platform, predict whether there are faults in the photovoltaic equipment according to the input performance parameters, and input the prediction results into the database for abnormal operation faults of photovoltaic equipment for fault discrimination and recording.

[0066] Further, based on the historical data in the database for abnormal faults and the feature data in the cloud platform, use the beamforming method and time series analysis method to construct a fault prediction model in the cloud platform, and use the fault prediction model to predict the occurrence of faults at the next moment, including the following steps:

[0067] S51. Use a polynomial regression model to analyze whether there are long-period trend terms in the historical data and feature data. If so, remove the long-period trend terms. If not, continue to analyze the periodic changes in the historical data and feature data;

[0068] S52. According to the analysis results, perform the beamforming method on the feature data in the historical data and feature data after removing the trend terms to obtain the amplitudes and phases of each frequency component;

[0069] S53. Use a significance test to judge whether each frequency component is significant, extract the significant periodic terms to construct a periodic term model;

[0070] S54. Consider the residuals after removing the trend terms and periodic terms as random variations and construct a residual prediction model;

[0071] S55. Superimpose the polynomial regression model, the periodic term model, and the residual prediction model to obtain a fault prediction model;

[0072] S56. Predict the historical data and feature data at the next moment through the fault prediction model;

[0073] S57. Perform corresponding weighted processing on the predicted historical data and feature data with the weight values to obtain the comprehensive prediction output of the occurrence of a fault.

[0074] According to another aspect of the present invention, there is also provided a cloud - processing - based integrated intelligent operation and maintenance system for a photovoltaic power station. The cloud - processing - based integrated intelligent operation and maintenance system for a photovoltaic power station includes:

[0075] A data acquisition and data pre - processing module, configured to obtain the operation status data and parameter data of the photovoltaic power station, and pre - process the obtained operation status data and parameter data to obtain the feature data of the operation status data and parameter data and store them in the cloud platform;

[0076] A performance evaluation module, configured to establish a solar energy conversion efficiency model of the photovoltaic power station by using cloud - processing analysis technology, and use the solar energy conversion efficiency model to evaluate the power generation performance of the photovoltaic equipment in real - time;

[0077] A data analysis and maintenance management module, configured to analyze the evaluation results of the power generation performance by using an analysis algorithm, judge the accumulation degree of the obstacles on the surface of the photovoltaic equipment, formulate the optimal cleaning cycle of the photovoltaic equipment, and use an automatic cleaning device to clean it;

[0078] An abnormal database establishment and fault discrimination module, configured to establish an abnormal database for the operation faults of the photovoltaic equipment, perform abnormal detection on the feature data in the cloud platform through an abnormal detection algorithm, and input the detection results into the abnormal database for the operation faults of the photovoltaic equipment for fault discrimination;

[0079] A fault prediction model construction module, configured to construct a fault prediction model in the cloud platform based on the historical data in the fault abnormal database and the feature data in the cloud platform by using the beamforming method and the time - series analysis method, and use the fault prediction model to predict the occurrence of a fault at the next moment;

[0080] A maintenance strategy formulation and fault prevention management module, configured to formulate corresponding maintenance strategies and emergency plans according to the prediction results;

[0081] Among them, the data acquisition and data pre - processing module is connected to the performance evaluation module and the data analysis and maintenance management module. The data analysis and maintenance management module is connected to the abnormal database establishment and fault discrimination module and the fault prediction model construction module. The fault prediction model construction module is connected to the maintenance strategy formulation and fault prevention management module.

[0082] The beneficial effects of the present invention are as follows:

[0083] 1. The present invention evaluates the power generation efficiency in real time through an optoelectronic conversion model, enabling the operation and maintenance personnel to understand the power generation status of the photovoltaic power station, helping the operation and maintenance personnel adjust the operation strategy according to the change of the power generation efficiency, optimizing the cleaning cycle of the photovoltaic equipment according to the evaluation result, avoiding frequent cleaning or too long cleaning intervals, reducing energy consumption and cleaning costs, increasing the service life of the equipment. At the same time, through the detection by the anomaly detection algorithm, it can be avoided that normal data is wrongly marked as an outlier, thereby improving the discovery and identification ability of fault data, reducing the false alarm rate. By predicting and timely handling potential faults, the decline of equipment performance can be effectively avoided, increasing the service life and operation efficiency of the equipment, and thus improving the operation and maintenance efficiency of the photovoltaic power station.

[0084] 2. The present invention analyzes the evaluation result of the power generation performance through an analysis algorithm, enabling a better judgment of the reasons for the reduction of the power generation efficiency, thereby enhancing the overall power generation efficiency of the photovoltaic power station. And through the correlation analysis of the accumulation degree of the shielding objects and the power generation performance of the equipment, the cleaning cycle of the photovoltaic equipment can be formulated more accurately, ensuring that the equipment is in the best working state, reducing the maintenance cost. At the same time, by repeatedly executing the optimization strategy and combining with the elite retention strategy, the maintenance method can be continuously improved, ensuring that the photovoltaic equipment remains efficient and stable during long-term operation, and thus improving the operation efficiency and reliability of the photovoltaic equipment, while reducing the maintenance cost, bringing long-term economic and technical benefits to the photovoltaic power station.

[0085] 3. The present invention performs anomaly detection on the feature data in the cloud platform through an anomaly detection algorithm, enabling the real-time performance parameter data set to be input into the cloud platform and analyzed through a random forest model, so as to monitor the equipment status in real time, thereby being able to more accurately identify and predict potential faults of the photovoltaic equipment, reducing false alarms and missed alarms, and thus improving the accuracy and timeliness of fault detection, optimizing the resource allocation, enhancing the stability and reliability of the photovoltaic power station, and bringing benefits to the intelligent operation and maintenance of the photovoltaic power station.

[0086] 4. The present invention enables the prediction of possible problems in advance before the occurrence of a fault through a fault prediction model, providing time for the maintenance personnel to take preventive measures or make corresponding preparations, thereby reducing the impact of the fault on the equipment operation. And by predicting and timely handling potential faults, the serious decline of equipment performance can be effectively avoided, thereby increasing the service life and operation efficiency of the equipment, and thus improving the evaluation efficiency of the potential of the photovoltaic equipment. Brief Description of the Drawings

[0087] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0088] Figure 1 is a flowchart of a comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to an embodiment of the present invention;

[0089] Figure 2 is a schematic block diagram of a comprehensive intelligent operation and maintenance system for a photovoltaic power station based on cloud processing according to an embodiment of the present invention.

[0090] In the figure:

[0091] 1. Data acquisition and data preprocessing module; 2. Performance evaluation module; 3. Data analysis and maintenance management module; 4. Abnormal database establishment and fault discrimination module; 5. Fault prediction model construction module; 6. Maintenance strategy formulation and fault prevention management module. Specific embodiments

[0092] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments.

[0093] In the description of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0094] According to an embodiment of the present invention, a comprehensive intelligent operation and maintenance method and system for a photovoltaic power station based on cloud processing are provided.

[0095] Now, the present invention will be further described in conjunction with the drawings and specific embodiments. As Figure 1 shown, the comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to an embodiment of the present invention includes the following steps:

[0096] S1. Obtain the operation status data and parameter data of the photovoltaic power station, and preprocess the obtained operation status data and parameter data to obtain the characteristic data of the operation status data and parameter data and store them in the cloud platform;

[0097] Specifically, the operation status data and parameter data of the PV power station are obtained through sensors and a monitoring system (including hardware and software components. The monitoring system not only collects the data provided by the sensors, but also includes video monitoring devices, communication interfaces, and data processing software. It can display the operation status of the power station in real time, record historical data, and also provide remote control functions).

[0098] Specifically, the operation status data refers to various data during the actual operation of the PV power station, such as voltage, current, power output, temperature, light intensity, wind speed and direction, environmental radiation, and health status indicators, etc.

[0099] Specifically, the parameter data refers to the set parameters or configuration information of the PV power station, such as the type, capacity, layout, inclination angle, geographical information, climate data, building data of the PV panels, etc.

[0100] S2. Use cloud processing analysis technology to establish a solar energy conversion efficiency model for the PV power station, and use the solar energy conversion efficiency model to evaluate the power generation performance of the PV equipment in real time;

[0101] Specifically, cloud processing analysis technology refers to the technology that uses the resources and capabilities of cloud computing for data processing and analysis.

[0102] S3. Use analysis algorithms to analyze the evaluation results of the power generation performance, judge the accumulation degree of the obstacles on the surface of the PV equipment, formulate the optimal cleaning cycle for the PV equipment, and use automatic cleaning equipment to clean it;

[0103] Specifically, the obstacles include fog, rain, snow, dust, bird droppings, fallen leaves, etc.

[0104] Specifically, the automatic cleaning equipment is a machine specially designed for cleaning the PV panels of the PV power station. The automatic cleaning equipment includes:

[0105] Intelligent control system: The automatic cleaning equipment is usually equipped with an advanced control system, which can automatically start the cleaning process according to a preset plan or according to actual needs (such as the accumulation degree of the obstacles judged by the analysis algorithm).

[0106] Cleaning mechanical device: The device includes brushes, spray nozzles, etc., which are used to effectively remove the dust, dirt and other obstacles on the surface of the PV panels. The cleaning methods involve dry cleaning (such as using brushes or blowing) and wet cleaning (such as using water or cleaning agents), etc.

[0107] Moving mechanism: Many automatic cleaning equipment are designed to be able to move along the PV panels, so as to achieve comprehensive and efficient cleaning. The moving mechanism can include tracks, wheels or other moving devices, etc.

[0108] S4. Establish a photovoltaic equipment operation failure and anomaly database, perform anomaly detection on the feature data in the cloud platform through an anomaly detection algorithm, and input the detection results into the photovoltaic equipment operation failure and anomaly database for fault discrimination;

[0109] S5. Based on the historical data in the fault and anomaly database and the feature data in the cloud platform, use the beamforming method and time series analysis method to construct a fault prediction model in the cloud platform, and use the fault prediction model to predict the occurrence of faults at the next moment;

[0110] S6. According to the prediction results, formulate corresponding maintenance strategies and emergency plans.

[0111] Specifically, formulating corresponding maintenance strategies and emergency plans includes:

[0112] Fault prediction analysis: Analyze the results of fault prediction to determine the possible types of problems, severity, and expected occurrence time.

[0113] Priority setting: Set priorities for different maintenance tasks according to the severity and urgency of the prediction results. For example, give higher priority to faults that will have a greater impact on system performance in the prediction results.

[0114] Resource allocation: Consider available maintenance resources such as personnel, equipment, materials, etc., and allocate them reasonably to deal with the predicted faults.

[0115] Preventive maintenance plan: Develop a preventive maintenance plan to reduce the likelihood of faults occurring. Include measures such as regular inspections, replacement of vulnerable parts, and cleaning.

[0116] Emergency plan formulation: For suddenly occurring faults, a clear emergency plan is required. Include response measures during the fault, emergency repair procedures, startup of backup systems, etc.

[0117] Continuous monitoring and evaluation: During the implementation of maintenance strategies and emergency plans, continuously monitor the operating status of the photovoltaic system, and adjust the plans and strategies according to the actual situation.

[0118] Communication mechanism: Establish an effective communication mechanism to ensure that all relevant personnel can quickly receive information and take corresponding actions when needed.

[0119] Record and feedback: Keep detailed records of maintenance activities and fault response processes, and provide feedback and optimization according to the actual situation.

[0120] Preferably, obtaining the operating status data and parameter data of the photovoltaic power station, and preprocessing the obtained operating status data and parameter data to obtain the feature data of the operating status data and parameter data and storing them in the cloud platform includes the following steps:

[0121] S11. Collect duplicate data, missing values, and outliers in the obtained operation status data and parameter data, and perform denoising, filtering, and smoothing on the duplicate data, missing values, and outliers;

[0122] S12. Perform data verification on the operation status data and parameter data after denoising, filtering, and smoothing, extract valid data, and perform normalization to generate accurate operation status data and parameter data;

[0123] S13. Use the principal component analysis method to fuse the generated accurate operation status data and parameter data into the same dataset;

[0124] Specifically, the principal component analysis method (Principal Component Analysis, abbreviated as PCA) is a commonly used data analysis technique, especially in feature extraction and dimensionality reduction. First, it is necessary to standardize the data to ensure that the importance of each feature in the analysis is consistent. Then calculate the covariance matrix of the data, extract the eigenvalues and eigenvectors to determine the principal components. Finally, select a certain number of principal components as needed for data dimensionality reduction and feature representation. In this way, the key information in the photovoltaic power station data can be effectively extracted and utilized.

[0125] S14. Extract relevant features from the fused dataset to obtain the feature data of the operation status data and parameter data, and store them in the cloud platform.

[0126] Specifically, the relevant features include electrical characteristic data (such as voltage, current, power, power generation, etc., which are data directly reflecting the power generation performance of photovoltaic panels), environmental characteristic data (such as temperature, light intensity, wind speed, humidity, etc., which help understand the impact of the external environment on the performance of photovoltaic panels), equipment status data (including equipment operation time, on / off status, fault records, maintenance history, etc.), performance degradation data (the degradation of the performance of photovoltaic panels during long-term operation, such as the trend of efficiency decline, the occurrence of hot spot effects, etc.), and occlusion accumulation data (data on the accumulation of occlusions (such as dust, leaves) collected through image recognition or other sensors), etc.

[0127] Preferably, establishing a solar energy conversion efficiency model of a photovoltaic power station using cloud processing analysis technology and using the solar energy conversion efficiency model to evaluate the power generation performance of photovoltaic equipment in real time includes the following steps:

[0128] S21. Use cloud processing analysis technology to analyze the operation status data of photovoltaic equipment and extract features related to photovoltaic modules;

[0129] S22. Convert the extracted relevant features into photovoltaic module characteristics and the principle of photovoltaic power conversion, and construct a solar conversion efficiency model;

[0130] S23. Estimate the parameter values in the solar conversion efficiency model by fitting the measured power generation and irradiance data with the constructed solar conversion efficiency model;

[0131] Specifically, the measured power generation and irradiance data refer to the numerical data of power generation and irradiance obtained by actual measurement of photovoltaic cells or photovoltaic power plants in reality. By fitting the measured power generation and irradiance data with the constructed photovoltaic power conversion model, the parameter values in the model can be estimated. The fitting process will attempt to adjust the parameters in the model to make the power generation calculated by the model close to the measured power generation.

[0132] S24. Calculate the actual power generation efficiency of the photovoltaic device using the established solar conversion efficiency model, and compare the actual power generation efficiency with the expected efficiency to evaluate the power generation performance of the photovoltaic device.

[0133] Specifically, the solar conversion efficiency model is a device that converts light energy into electrical energy, and its working principle is based on the photovoltaic effect and the photo-generated current effect. The photovoltaic power conversion model includes the single diode model (Single Diode Model), also known as the equivalent circuit model. The photovoltaic power conversion model is based on circuit theory and semiconductor physics principles, and abstracts the photovoltaic cell as a combination of a current source, a voltage source, and two diode equivalent circuits.

[0134] Preferably, use an analysis algorithm to analyze the evaluation results of the power generation performance, judge the accumulation degree of the obstacles on the surface of the photovoltaic device, formulate the optimal cleaning cycle of the photovoltaic device, and clean it using an automatic cleaning device, including the following steps:

[0135] S31. Import the actual power generation efficiency and the expected efficiency, initialize, and use the shortest path algorithm to calculate the accumulation degree of the obstacles between any two detection points on the surface of the photovoltaic device;

[0136] Specifically, the shortest path algorithm is the Floyd-Warshall algorithm, which is a classic algorithm for searching the shortest path between any two points in a weighted graph. It uses the idea of dynamic programming to gradually improve the path to obtain the shortest path length, divides the path through intermediate nodes, and compares and selects a better indirect path.

[0137] Specifically, the surface of the photovoltaic panel is divided into multiple detection points, and each point represents a specific position on the photovoltaic panel. These detection points can be physical measurement points or virtual points defined by image analysis software. The shortest path refers to the shortest distance path on the surface of the photovoltaic panel from one detection point to another, and the degree of accumulation of obstacles on this path is the lowest.

[0138] S32. Set up several analysis strategies. Each analysis strategy (i.e., each coyote in the improved coyote optimization algorithm) represents an association analysis method between the degree of accumulation of obstacles and the power generation performance of the device. The first N numbers represent the numbers of the selected detection points, and the last N numbers represent the corresponding power generation performance levels of the photovoltaic devices;

[0139] Specifically, the numbers of the selected detection points refer to the obstacle detection points set at specific positions on the surface of the photovoltaic panel. These detection points will be used to collect data on the degree of accumulation of obstacles for the purpose of evaluating and predicting the degree of accumulation of obstacles.

[0140] S33. According to the preset maintenance objectives of the photovoltaic device performance, find the global optimal analysis strategy, and add the environmental impact factor to jointly affect the optimization of the analysis strategy with each group of local optimal analysis strategies and the operation trend of the photovoltaic device;

[0141] S34. Substitute the optimized analysis strategy into the Kent mapping, and compare the analysis strategy after Kent mapping with the analysis strategy before optimization using the power generation performance of the photovoltaic device and the degree of accumulation of obstacles as the evaluation criteria;

[0142] Specifically, the Kent mapping is a chaotic mapping. When the value is 0.5, it is a symmetric chaotic mapping.

[0143] S35. Apply the elitist retention strategy, and according to the maintenance objectives of the photovoltaic device performance, replace the analysis strategy with the worst effect with the global sub-optimal analysis strategy;

[0144] Specifically, the elitist retention strategy itself is to replace the worst solution with the optimal solution to avoid the algorithm eliminating excellent solutions during the optimization process and achieve the purpose of retaining the genes of excellent solutions. Therefore, the elitist selection strategy is adjusted to name the sub-optimal wolf before the growth of the entire population as the Beta wolf, and use the Beta wolf to replace the wolf with the worst social adaptability in the wolf pack after growth, which improves the search efficiency while retaining the genes of excellent solutions within the population.

[0145] S36. Loop and execute steps S33 to S35. If the number of iterations exceeds the threshold, end the iteration to obtain the best judgment method for evaluating the power generation performance of the photovoltaic device and the degree of accumulation of obstacles;

[0146] Specifically, the optimal judgment method is an algorithm or decision-making process. In the present invention, the optimal judgment method refers to an analysis result that can accurately and efficiently determine the degree of influence of the accumulation of obstacles (such as dust, dirt, etc.) on the power generation performance.

[0147] Specifically, the setting of the threshold is a process that needs to comprehensively consider multiple factors and requires multiple experiments and iterations for optimization. The threshold is based on algorithm performance evaluation metrics, such as error rate, accuracy, recall rate, or a certain metric specific to the problem. By adjusting these metrics, one or more thresholds are set to obtain the optimal judgment method in practical applications.

[0148] S37. According to the optimal judgment method, formulate the optimal cleaning cycle of the photovoltaic device and clean it using an automatic cleaning device.

[0149] Specifically, formulating the optimal cleaning cycle of the photovoltaic device according to the optimal judgment method includes:

[0150] Analyze the evaluation result: Using the evaluation result obtained by the optimal judgment method, analyze the degree of accumulation of obstacles on the surface of the photovoltaic device and its influence on the power generation performance. This includes identifying the types of obstacles, the accumulation rate, and the specific impact on the performance of the photovoltaic panel.

[0151] Determine the performance threshold: According to the design parameters and operating requirements of the photovoltaic device, set a threshold for the power generation performance. When the performance of the photovoltaic panel drops below this threshold, it indicates that cleaning is required.

[0152] Predict the accumulation trend: Use the optimal judgment method to predict the accumulation trend of obstacles and estimate the time required to reach the performance threshold. This time interval is an important reference for the cleaning cycle.

[0153] Consider environmental factors: The cleaning cycle of the photovoltaic device also needs to consider environmental factors, such as seasonal changes, local climate conditions, air pollution levels, etc., which will all affect the accumulation rate of obstacles.

[0154] Formulate a cleaning plan: Based on the above information, formulate a cleaning plan, which includes the cleaning cycle, cleaning method (such as manual or automatic), cleaning time (such as day or night), etc.

[0155] Specifically, the analysis algorithm is an improved coyote optimization algorithm. In the traditional coyote optimization algorithm, the growth of coyotes is the main way to obtain new solutions. The coyote optimization algorithm groups the entire coyote population for growth. During the growth process, the growth of coyotes within the group is guided by the group alpha wolf and the group cultural trend cult, which greatly limits the population diversity of COA coyotes and the information exchange within the population. The global traversability of coyotes during the growth process is relatively low. At the same time, due to the low birth rate of COA coyotes and the low population diversity, the exploration ability of the algorithm is poor and it is easy to fall into local optimum. The present invention mainly improves the growth mode of coyotes in the coyote optimization algorithm (i.e., the update formula in the present invention).

[0156] Preferably, according to the preset photovoltaic device performance maintenance target, a global optimal analysis strategy is searched, and the optimization of the analysis strategy jointly affected by the environmental impact factor, each group of local optimal analysis strategies and the photovoltaic device operation trend includes the following steps:

[0157] S331. Define the objective function of the photovoltaic device performance maintenance target for evaluating the fitness of the analysis strategy;

[0158] Specifically, the photovoltaic device performance maintenance target is mainly to ensure the best operation state of the device, improve the power generation efficiency, reduce the failure rate, extend the device life, and ensure safe and reliable power output. In order to achieve the photovoltaic device performance maintenance target, an objective function is defined to evaluate and optimize the maintenance strategy of the photovoltaic device. When specifically defining the objective function, the following aspects need to be considered:

[0159] Power generation efficiency: The objective function should consider the difference between the actual power generation efficiency of the photovoltaic device and the theoretical or expected efficiency. The closer the efficiency is to the theoretical maximum, the better.

[0160] Operation stability: It includes the failure frequency and severity of the device. The goal is to reduce the number and impact of failures.

[0161] Operation and maintenance cost: Consider the cost of maintenance activities, such as cleaning, repair, replacement of parts, etc. The goal is to minimize these costs on the premise of ensuring the device performance.

[0162] Device life: Consider the impact of the maintenance strategy on the device life. The ideal strategy should be able to extend the effective service life of the device.

[0163] Full performance: Ensure that the operation and maintenance activities of the photovoltaic device comply with safety standards.

[0164] Environmental factors: Consider the impact of environmental changes on the performance of the photovoltaic device, such as temperature, humidity, dust, etc.

[0165] S332. Initialize the analysis strategy population (i.e., the coyote population), and randomly generate the parameters of each analysis strategy (i.e., the positions of the coyotes in the improved coyote optimization algorithm);

[0166] S333. Calculate the fitness of each analysis strategy according to the objective function, and find the global optimal analysis strategy as the best solution;

[0167] S334. Randomly generate environmental impact factors for each group of the best analysis strategies, and calculate the average parameters within each group of the local best analysis strategies as the group trend;

[0168] Specifically, the environmental impact factor is a parameter used to simulate the influence of randomness in the natural environment on the development of the coyote population. In nature, the growth of organisms and the development of populations are not only determined by the characteristics of the organisms themselves, but also affected by environmental factors such as food supply, climate change, and predator pressure. The introduction of the environmental impact factor in the algorithm is to imitate this natural phenomenon and increase the practical applicability and effect of the algorithm.

[0169] S335. Substitute the original analysis strategy parameters into the update formula to calculate the parameters of the new analysis strategy, and update the parameters of each analysis strategy;

[0170] Specifically, the update formula is the constructed new growth mode of the coyote.

[0171] S336. Repeat steps S333 to S335 until the termination condition is met.

[0172] Preferably, substituting the optimized analysis strategy into the Kent mapping, and comparing the analysis strategy after Kent mapping with the analysis strategy before optimization with the power generation performance of the photovoltaic device and the degree of accumulation of the shielding object as the evaluation criteria includes the following steps:

[0173] S341. Train and iterate the analysis strategy population to obtain the optimized analysis strategy population;

[0174] S342. Perform data normalization processing on the optimized analysis strategy population to make it in a preset state;

[0175] Specifically, normalization processing: Normalization processing is to unify the dimension of the data, make different data comparable, and perform normalization processing on the prediction results, which helps to eliminate the influence of the data dimension and facilitates subsequent weighted processing and summation calculation.

[0176] S343. Randomly generate the parameters of the Kent mapping and preset the value range (0, 1);

[0177] S344. Map each optimized analysis strategy using the Kent mapping formula to generate new analysis strategy individuals;

[0178] Specifically, each optimized analysis strategy is each grown coyote.

[0179] S345. Compare the fitness of each analysis strategy with its Kent mapping result, and retain the analysis strategy individual with the best fitness as the Kent mapping analysis strategy;

[0180] S346. Compare the Kent mapping analysis strategy with the original analysis strategy population in terms of fitness, and retain the analysis strategy individual with the best fitness to enter the next-generation analysis strategy population;

[0181] S347. Repeat steps S342 to S346 until the iteration termination condition is reached.

[0182] Preferably, the update formula is:

[0183] ;

[0184] Where represents the parameters of the new optimized analysis strategy;

[0185] represents the parameters of the original analysis strategy;

[0186] m represents the m th analysis strategy;

[0187] i represents the i th iteration;

[0188] b represents the b th parameter value of the dimension;

[0189] x represents the parameter vector of the analysis strategy;

[0190] l 1 represents the weight of the environmental impact factor after normalization;

[0191] l 2 represents the δ random weight impact factor of growth factor

[0192] l 1 after normalization; δ 3 represents the random weight impact factor of growth factor

[0193] 2 after normalization.Preferably, a database for abnormal operation faults of photovoltaic equipment is established, and the characteristic data in the cloud platform is detected for abnormalities through an anomaly detection algorithm, and the detection results are input into the database for abnormal operation faults of photovoltaic equipment for fault discrimination, including the following steps:

[0194] S41. Establish a database for abnormal operation faults of photovoltaic equipment;

[0195] S42. Input the dataset of performance parameters of the photovoltaic equipment into the cloud platform, preset the height of the decision tree in the anomaly detection algorithm, and initialize the random forest model;

[0196] S43. Use the performance parameter data of the photovoltaic equipment to construct several decision trees to form an initial random forest model;

[0197] S44. Use the performance parameters of the normally operating photovoltaic equipment as the training set to train the initial random forest model;

[0198] S45. According to the difference and accuracy of the decision trees, use a probability search algorithm to screen out the decision trees with higher fitness from the initial forest model and combine them into a new random forest model;

[0199] Specifically, the accuracy value of each decision tree is calculated by the leave-one-out method, and the difference value between the decision trees is calculated by the statistical method.

[0200] Specifically, the statistical method is the Q-statistic method, which is a statistical method for calculating the difference between two classifiers. In the present invention, the Q-statistic method is used to evaluate the difference degree of the classification results between every two decision trees. By calculating the Q value between the decision trees, the difference size between the decision trees can be obtained, and based on this, the decision trees with larger differences are selected, so as to improve the generalization ability of the random forest.

[0201] Specifically, the probability search algorithm is the simulated annealing algorithm. The simulated annealing algorithm is a heuristic optimization algorithm that gradually optimizes the objective function value by simulating the cooling process in the solid annealing process. In the present invention, the simulated annealing algorithm is used to find the optimal combination of decision trees, which come from an initial random forest model. The fitness value of each decision tree combination is calculated according to the difference and accuracy of the decision trees. The combination with large difference and high accuracy has a larger fitness value. The simulated annealing algorithm searches for the decision tree combination with the largest fitness value in the solution space.

[0202] Specifically, according to the difference and accuracy of the decision trees, using a probability search algorithm to screen out the decision trees with higher fitness from the initial forest model and combine them into a new random forest model includes the following steps:

[0203] S451. Initialize the probabilistic search algorithm and set the initial temperature and initial solution;

[0204] S452. According to the set initial temperature, repeat steps S563 to step 566;

[0205] S453. Randomly perturb the set initial solution (i.e., the current decision tree combination) to generate a new solution (a new decision tree combination);

[0206] S454. Calculate the difference between the fitness value of the new solution and the fitness value of the initial solution;

[0207] S455. If the fitness value of the new solution is higher than the initial solution (i.e., the difference is less than 0), then accept the new solution as the initial solution. Otherwise, according to the Metropolis rule, calculate the acceptance probability of the new solution. If the acceptance probability is greater than the random number, then accept the new solution as the initial solution. Otherwise, retain the initial solution;

[0208] S456. If the set termination condition is satisfied, then output the initial solution as the optimal solution. The termination condition is that the new solution has not been accepted in a continuous number of Metropolis chains or the end temperature is reached. Otherwise, after attenuating the temperature through the attenuation function, return to step S452;

[0209] Specifically, the Metropolis rule is an acceptance - rejection criterion used in Monte Carlo simulation. The basic idea of this rule is: If the new state has lower energy (or higher probability) compared to the current state, then accept the new state; otherwise, accept the new state with a probability that is the negative exponent of the energy difference (or probability ratio) between the new state and the current state.

[0210] S457. Repeat steps S453 to S456 to select several decision trees that meet the preset fitness value from the initial forest model and combine them into a new random forest model.

[0211] S46. Apply the new random forest model to the performance parameters of the photovoltaic equipment in the cloud platform, predict whether there is a fault in the photovoltaic equipment according to the input performance parameters, and input the prediction result into the photovoltaic equipment operation fault exception database for fault discrimination and recording.

[0212] Specifically, the anomaly detection algorithm is a data anomaly detection algorithm based on SA - iForest (Subspace Isolation Forest), which is an improved data anomaly detection algorithm based on the Isolation Forest algorithm.

[0213] Specifically, when the Isolation Forest (iForest) algorithm processes high-dimensional data, it may lead to a decline in algorithm performance. Therefore, the Subspace-based Isolation Forest (SA-iForest) algorithm is proposed. When SA-iForest splits data each time, instead of performing on all features, it randomly selects a feature subspace and selects features within the subspace for splitting. Thus, on the one hand, the algorithm complexity can be reduced, and on the other hand, the accuracy of anomaly detection can be improved.

[0214] Preferably, based on the historical data in the fault anomaly database and the feature data in the cloud platform, a fault prediction model is constructed in the cloud platform using the beamforming method and the time series analysis method, and predicting the occurrence of faults at the next moment using the fault prediction model includes the following steps:

[0215] S51. Use a polynomial regression model to analyze whether there is a long-term trend term in the historical data and the feature data. If there is, remove the long-term trend term. If not, continue to analyze the periodic changes in the historical data and the feature data;

[0216] S52. According to the analysis results, perform the beamforming method on the feature data in the historical data and the feature data after removing the trend term to obtain the amplitude and phase of each frequency component;

[0217] Specifically, the amplitude represents the change amount of the historical data or the feature data at a certain frequency, and the phase represents the time-delay characteristic of the change of the historical data or the feature data.

[0218] S53. Use a significance test to determine whether each frequency component is significant, and extract the significant periodic terms to construct a periodic term model;

[0219] S54. The residuals after removing the trend term and the periodic term are regarded as random changes, and a residual prediction model is constructed;

[0220] S55. Superimpose the polynomial regression model, the periodic term model, and the residual prediction model to obtain a fault prediction model;

[0221] S56. Predict the historical data and the feature data at the next moment through the fault prediction model;

[0222] S57. Perform corresponding weighted processing on the predicted historical data and the feature data with the weight values to obtain the comprehensive prediction output of the occurrence of faults.

[0223] Specifically, calculating the weight values of the historical data and the feature data using information gain includes:

[0224] Collect the sample data sets of the historical data and the feature data;

[0225] Calculate the information entropy of the target variable in the sample dataset to measure the uncertainty of the target variable;

[0226] Specifically, the target variable refers to the variable used to determine whether the photovoltaic power station operation and maintenance system is faulty.

[0227] For each feature, calculate the information gain between it and the target variable respectively;

[0228] Take the information gain value as the weight value of the feature and normalize the weight value;

[0229] Analyze the normalized weight value to understand the importance of historical data and feature data.

[0230] Specifically, significance test: The significance test is a statistical method used to test whether there is a significant difference between the observed data and a certain hypothesis. In time series analysis, the significance test is often used to determine whether the periodic components in the data are statistically significant. The result of the significance test is usually represented by the p-value. The smaller the p-value, the more significant the difference between the observed data and the hypothesis.

[0231] Specifically, periodic term model: The periodic term model is mainly used to describe the periodic components in time series data. In photovoltaic power station fault prediction, the periodic term model can help capture the periodic changes of faults at different time scales such as within a day and within a week. The methods for constructing the periodic term model include Fourier analysis, periodic regression, etc.

[0232] Specifically, removing the trend term and the periodic term: In time series analysis, the data can usually be decomposed into a trend term, a periodic term, and random variations (residuals). By removing the trend term and the periodic term, the regular components in the data can be eliminated, so as to better focus on the random variations. The methods for removing the trend term and the periodic term include differencing, filtering, etc.

[0233] Specifically, residual prediction model: The residual prediction model is used to describe the random variations (residuals) in time series data. After removing the trend term and the periodic term, the residuals are regarded as random variations, and the residual prediction model can be constructed to estimate the fluctuations at the next moment. Commonly used residual prediction models include autoregressive moving average model (ARIMA), exponential smoothing model (ETS), etc.

[0234] According to another embodiment of the present invention, as Figure 2 shown, there is also provided a cloud processing-based integrated intelligent operation and maintenance system for photovoltaic power stations, and the cloud processing-based integrated intelligent operation and maintenance system for photovoltaic power stations includes:

[0235] The data acquisition and data preprocessing module 1 is used to obtain the operation status data and parameter data of the photovoltaic power station, preprocess the obtained operation status data and parameter data, obtain the characteristic data of the operation status data and parameter data, and store them in the cloud platform;

[0236] The performance evaluation module 2 is used to establish a solar energy conversion efficiency model of the photovoltaic power station by using cloud processing and analysis technology, and use the solar energy conversion efficiency model to evaluate the power generation performance of the photovoltaic equipment in real time;

[0237] The data analysis and maintenance management module 3 is used to analyze the evaluation results of the power generation performance by using analysis algorithms, judge the accumulation degree of the obstacles on the surface of the photovoltaic equipment, formulate the optimal cleaning cycle of the photovoltaic equipment, and use the automatic cleaning equipment to clean it;

[0238] The abnormal database establishment and fault discrimination module 4 is used to establish an abnormal database of the operation faults of the photovoltaic equipment, perform abnormal detection on the characteristic data in the cloud platform through an abnormal detection algorithm, and input the detection results into the abnormal database of the operation faults of the photovoltaic equipment for fault discrimination;

[0239] The fault prediction model construction module 5 is used to construct a fault prediction model in the cloud platform by using the beamforming method and the time series analysis method based on the historical data in the abnormal database of faults and the characteristic data in the cloud platform, and use the fault prediction model to predict the occurrence of faults at the next moment;

[0240] The maintenance strategy formulation and fault prevention management module 6 is used to formulate corresponding maintenance strategies and emergency plans according to the prediction results;

[0241] Among them, the data acquisition and data preprocessing module 1 is connected to the performance evaluation module 2 and the data analysis and maintenance management module 3. The data analysis and maintenance management module 3 is connected to the abnormal database establishment and fault discrimination module 4 and the fault prediction model construction module 5. The fault prediction model construction module 5 is connected to the maintenance strategy formulation and fault prevention management module 6.

[0242] In summary, by means of the above technical solutions of the present invention, the present invention analyzes the evaluation results of power generation performance through an analysis algorithm, enabling better judgment of the reasons for the reduction of power generation efficiency, thereby enhancing the overall power generation efficiency of the photovoltaic power station. And through the correlation analysis of the accumulation degree of the shielding object and the power generation performance of the equipment, the cleaning cycle of the photovoltaic equipment can be more accurately formulated to ensure that the equipment is in the best working state, reduce the maintenance cost. At the same time, by repeatedly executing the optimization strategy and combining the elite retention strategy, the maintenance method can be continuously improved to ensure that the photovoltaic equipment maintains high efficiency and stability during long-term operation, thereby improving the operation efficiency and reliability of the photovoltaic equipment, reducing the maintenance cost, and bringing long-term economic and technical benefits to the photovoltaic power station. The present invention performs anomaly detection on the feature data in the cloud platform through an anomaly detection algorithm, enabling the real-time performance parameter data set to be input into the cloud platform and analyzed through a random forest model, so as to monitor the equipment status in real time, thereby being able to more accurately identify and predict potential faults of the photovoltaic equipment, reduce false alarms and missed alarms, and then improve the accuracy and timeliness of fault detection, optimize resource allocation, enhance the stability and reliability of the photovoltaic power station, and bring benefits to the intelligent operation and maintenance of the photovoltaic power station. The present invention enables the prediction of possible problems in advance before the occurrence of a fault through a fault prediction model, providing time for maintenance personnel to take preventive measures or make corresponding preparations, thereby reducing the impact of the fault on the operation of the equipment. And by predicting and timely handling potential faults, the serious decline of equipment performance can be effectively avoided, thereby improving the service life and operation efficiency of the equipment, and then improving the evaluation efficiency of the potential of the photovoltaic equipment.

[0243] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing, characterized in that The cloud - processing - based integrated intelligent operation and maintenance method for a photovoltaic power station includes the following steps: S1. Obtain the operation status data and parameter data of the photovoltaic power station, pre - process the obtained operation status data and parameter data to obtain the characteristic data of the operation status data and parameter data, and store them in the cloud platform; S2. Use cloud - processing analysis technology to establish a solar energy conversion efficiency model for the photovoltaic power station, and use the solar energy conversion efficiency model to evaluate the power generation performance of photovoltaic devices in real - time; S3. Use an analysis algorithm to analyze the evaluation results of the power generation performance, judge the accumulation degree of obstacles on the surface of photovoltaic devices, formulate the optimal cleaning cycle of photovoltaic devices, and use an automatic cleaning device to clean them; The step of using an analysis algorithm to analyze the evaluation results of the power generation performance, judge the accumulation degree of obstacles on the surface of photovoltaic devices, formulate the optimal cleaning cycle of photovoltaic devices, and use an automatic cleaning device to clean them includes the following steps: S31. Import the actual power generation efficiency and the expected efficiency, perform initialization, and use the shortest - path algorithm to calculate the accumulation degree of obstacles between any two detection points on the surface of the photovoltaic device; S32. Set up several analysis strategies, each analysis strategy represents a correlation analysis method between the accumulation degree of obstacles and the power generation performance of the device. The first N numbers represent the numbers of the selected detection points, and the last N numbers represent the corresponding power generation performance levels of the photovoltaic devices; S33. According to the preset performance maintenance target of the photovoltaic device, find the global optimal analysis strategy, and add the environmental impact factor to jointly affect the optimization of the analysis strategy with each group of local optimal analysis strategies and the operation trend of the photovoltaic device; S34. Substitute the optimized analysis strategy into the Kent mapping, and compare the analysis strategy after Kent mapping with the analysis strategy before optimization with the power generation performance of the photovoltaic device and the accumulation degree of obstacles as the judgment criteria; S35. Use the elitist retention strategy, and according to the performance maintenance target of the photovoltaic device, replace the analysis strategy with the worst effect with the global sub - optimal analysis strategy; S36. Loop through steps S33 to S35. If the number of iterations exceeds the threshold, end the iteration to obtain the best judgment method for evaluating the power generation performance and the accumulation degree of obstacles of the photovoltaic device; S37. According to the best judgment method, formulate the optimal cleaning cycle of the photovoltaic device, and use an automatic cleaning device to clean it; S4. Establish a database for abnormal operation faults of photovoltaic devices, perform abnormal detection on the characteristic data in the cloud platform through an abnormal detection algorithm, and input the detection results into the database for abnormal operation faults of photovoltaic devices for fault discrimination; S5. Based on the historical data in the database for abnormal operation faults and the characteristic data in the cloud platform, use the beam - forming method and the time - series analysis method to construct a fault prediction model in the cloud platform, and use the fault prediction model to predict the occurrence of faults at the next moment; S6. According to the prediction results, formulate corresponding maintenance strategies and emergency plans.

2. The integrated intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to claim 1, characterized in that, Obtaining the operation status data and parameter data of a photovoltaic power station, preprocessing the obtained operation status data and parameter data, and obtaining the characteristic data of the operation status data and parameter data and storing them in a cloud platform includes the following steps: S11. Collect the duplicate data, missing values, and outliers in the obtained operation status data and parameter data, and perform denoising, filtering, and smoothing processing on the duplicate data, missing values, and outliers; S12. Perform data verification on the operation status data and parameter data after denoising, filtering, and smoothing processing, extract the valid data, and perform normalization processing to generate accurate operation status data and parameter data; S13. Use the principal component analysis method to fuse the generated accurate operation status data and parameter data into the same dataset; S14. Extract relevant features from the fused dataset to obtain the characteristic data of the operation status data and parameter data, and store them in the cloud platform.

3. The integrated intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to claim 1, wherein, Using cloud processing analysis technology to establish a solar energy conversion efficiency model for a photovoltaic power station, and using the solar energy conversion efficiency model to evaluate the power generation performance of photovoltaic equipment in real time includes the following steps: S21. Use cloud processing analysis technology to analyze the operation status data of photovoltaic equipment and extract the features related to photovoltaic modules; S22. Convert the extracted relevant features into the characteristics of photovoltaic modules and the principle of photovoltaic conversion, and construct a solar energy conversion efficiency model; S23. By fitting the actually measured power generation power and irradiance data with the constructed solar energy conversion efficiency model, estimate the parameter values in the solar energy conversion efficiency model; S24. Use the established solar energy conversion efficiency model to calculate the actual power generation efficiency of photovoltaic equipment, and compare the actual power generation efficiency with the expected efficiency to evaluate the power generation performance of photovoltaic equipment.

4. The integrated intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to claim 1, wherein According to the preset performance maintenance target of photovoltaic equipment, finding the global optimal analysis strategy, and optimizing by adding environmental impact factors and each group of local optimal analysis strategies and the operation trend of photovoltaic equipment to jointly affect the analysis strategy includes the following steps: S331. Define the objective function of the performance maintenance target of photovoltaic equipment, which is used to evaluate the fitness of the analysis strategy; S332. Initialize the analysis strategy population and randomly generate the parameters of each analysis strategy; S333. Calculate the fitness of each analysis strategy according to the objective function, and find the global optimal analysis strategy as the best solution; S334. Randomly generate environmental impact factors for each group of best analysis strategies, and calculate the average parameter within each group of local best analysis strategies as the group population trend; S335. Substitute the original analysis strategy parameters into the update formula, calculate the parameters of the new analysis strategy, and update the parameters of each analysis strategy; S336. Repeat steps S333 to S335 until the termination condition is met.

5. The integrated intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to claim 1, characterized in that, Substituting the optimized analysis strategy into the Kent mapping, and comparing the analysis strategy after Kent mapping with the analysis strategy before optimization using the power generation performance of photovoltaic equipment and the degree of accumulation of obstacles as the evaluation criteria includes the following steps: S341. Train and iterate the analysis strategy population to obtain the optimized analysis strategy population; S342. Perform data normalization on the optimized analysis strategy population to make it in a preset state; S343. Randomly generate the parameters of the Kent mapping and preset the value range; S344. Use the Kent mapping formula to map each optimized analysis strategy to generate new analysis strategy individuals; S345. Compare the fitness of each analysis strategy with its Kent mapping result, and retain the analysis strategy individual with the best fitness as the Kent mapping analysis strategy; S346. Compare the Kent mapping analysis strategy with the original analysis strategy population in terms of fitness, and retain the analysis strategy individual with the best fitness to enter the next-generation analysis strategy population; S347. Repeat steps S342 to S346 until the iteration termination condition is reached.

6. The integrated intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to claim 4, wherein, The update formula is: ； Among them, represents the parameters of the optimized new analysis strategy; Represent the parameters of the original analysis strategy; m Indicates the m th analysis strategy; i Indicates the i th iteration; b Indicates the b parameter value of the dimension; x Denote the parameter vector of the analysis strategy; l 1 represents the weight of the environmental impact factor after normalization; l 2 represents the growth factor δ The weight of the random weight influence factor of 1 after normalization; l 3 represents the growth factor δ The weight after normalization of the random weight influence factor of 2 7. A comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to claim 1, characterized in that, The steps of establishing a photovoltaic device operation fault anomaly database, performing anomaly detection on the feature data in the cloud platform through an anomaly detection algorithm, and inputting the detection result into the photovoltaic device operation fault anomaly database for fault discrimination include the following steps: S41. Establish a photovoltaic device operation fault anomaly database; S42. Input the performance parameter dataset of the photovoltaic device into the cloud platform, preset the height of the decision tree in the anomaly detection algorithm, and initialize the random forest model; S43. Use the performance parameter data of the photovoltaic device to construct several decision trees to form an initial random forest model; S44. Use the performance parameters of the normally operating photovoltaic device as the training set to train the initial random forest model; S45. According to the difference and accuracy of the decision trees, use a probability search algorithm to screen out the decision trees with higher fitness from the initial forest model and combine them into a new random forest model; S46. Apply the new random forest model to the performance parameters of the photovoltaic device in the cloud platform, predict whether the photovoltaic device has a fault according to the input performance parameters, and input the prediction result into the photovoltaic device operation fault anomaly database for fault discrimination and recording.

8. A comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to claim 1, characterized in that, The steps of constructing a fault prediction model in the cloud platform based on the historical data in the fault anomaly database and the feature data in the cloud platform using the beamforming method and the time series analysis method, and using the fault prediction model to predict the occurrence of faults at the next moment include the following steps: S51. Use a polynomial regression model to analyze whether there is a long-period trend term in the historical data and feature data. If so, remove the long-period trend term. If not, continue to analyze the periodic changes in the historical data and feature data; S52. According to the analysis result, perform the beamforming method on the feature data in the historical data and feature data after removing the trend term to obtain the amplitude and phase of each frequency component; S53. Use a significance test to judge whether each frequency component is significant, and extract the significant periodic terms to construct a periodic term model; S54. Consider the residual after removing the trend term and periodic term as random variation and construct a residual prediction model; S55. Superimpose the polynomial regression model, the periodic term model, and the residual prediction model to obtain a fault prediction model; S56. Predict the historical data and characteristic data at the next moment through the fault prediction model; S57. Perform corresponding weighted processing on the predicted historical data and characteristic data with the weight values to obtain the comprehensive prediction output of the occurrence of a fault.

9. A comprehensive intelligent operation and maintenance system for a photovoltaic power station based on cloud processing, which is used to implement the comprehensive intelligent operation and maintenance method for a photovoltaic power station based on cloud processing according to any one of claims 1-8, characterized in that, The integrated intelligent operation and maintenance system for a photovoltaic power station based on cloud processing includes: A data acquisition and data preprocessing module, which is used to obtain the operation status data and parameter data of the photovoltaic power station, and preprocess the obtained operation status data and parameter data to obtain the characteristic data of the operation status data and parameter data and store them in the cloud platform; A performance evaluation module, which is used to establish a solar energy conversion efficiency model of the photovoltaic power station by using cloud processing analysis technology, and use the solar energy conversion efficiency model to evaluate the power generation performance of the photovoltaic equipment in real time; A data analysis and maintenance management module, which is used to analyze the evaluation results of the power generation performance by using analysis algorithms, judge the accumulation degree of the obstacles on the surface of the photovoltaic equipment, formulate the best cleaning cycle of the photovoltaic equipment, and use the automatic cleaning equipment to clean it; An abnormal database establishment and fault discrimination module, which is used to establish an abnormal database of the operation faults of the photovoltaic equipment, perform abnormal detection on the characteristic data in the cloud platform through an abnormal detection algorithm, and input the detection results into the abnormal database of the operation faults of the photovoltaic equipment for fault discrimination; A fault prediction model construction module, which is used to construct a fault prediction model in the cloud platform based on the historical data in the fault abnormal database and the characteristic data in the cloud platform by using the beamforming method and the time series analysis method, and use the fault prediction model to predict the occurrence of a fault at the next moment; A maintenance strategy formulation and fault prevention management module, which is used to formulate corresponding maintenance strategies and emergency plans according to the prediction results; Among them, the data acquisition and data preprocessing module is connected through the performance evaluation module and the data analysis and maintenance management module, the data analysis and maintenance management module is connected through the abnormal database establishment and fault discrimination module and the fault prediction model construction module, and the fault prediction model construction module is connected to the maintenance strategy formulation and fault prevention management module.

Images