A method and system for power generation loss analysis of a photovoltaic power plant

Abstract

The application provides a power generation loss analysis method and system for a photovoltaic power station, and relates to the technical field of photovoltaic technology. The method comprises the following steps: collecting current time information through an environment analysis station, and identifying an obtained shading loss coefficient; collecting a meteorological data sequence in a future preset time range through a meteorological analysis station, performing power generation loss identification and dust influence parameter analysis, obtaining a meteorological loss coefficient and a dust influence parameter; collecting a plurality of dust parameters of all photovoltaic panels or part of the photovoltaic panels in the target power station through a dust analysis station, combining the dust influence parameter, calculating and obtaining a dust parameter array, constructing a dust distribution fitting image, performing cascade image generation analysis, obtaining a dust loss coefficient of the target power station, and combining the shading loss coefficient and the meteorological loss coefficient to calculate and obtain a power generation loss coefficient and a power generation loss amount. The application solves the technical problems of incomplete and inaccurate calculation and analysis of the power generation loss amount of the photovoltaic power station in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of photovoltaic technology, and particularly relates to a power generation loss analysis system for a photovoltaic power station. BACKGROUND

[0002] The power generation operation efficiency of a photovoltaic power station is affected by many factors, such as sunshine hours, radiation, shielding, and dust coverage, among which, dust coverage particularly affects the power generation efficiency, and further affects the power station benefit.

[0003] In the prior art, the loss amount of power generation of a photovoltaic power station needs to be analyzed and calculated to roughly determine the power generation loss of the photovoltaic power station, and further serve as a reference for maintenance analysis and other related work of the power station.

[0004] However, the means for analyzing and calculating the power generation loss of a photovoltaic power station in the prior art generally calculates through rough dust sampling, which is not comprehensive and not accurate, and thus the power generation loss analysis effect in the construction and use of the photovoltaic power station is poor. SUMMARY

[0005] The present application provides a power generation loss analysis method for a photovoltaic power station, which is used to solve the technical problem of the non-comprehensive and non-accurate calculation and analysis of the power generation loss amount of a photovoltaic power station in the prior art.

[0006] In a first aspect, the present application provides a power generation loss analysis method for a photovoltaic power station, which is applied to a power generation loss analysis device for the construction of a photovoltaic power station, the device comprising an environment analysis station, a meteorological analysis station, a dust analysis station, and a comprehensive loss calculation module, and the method comprises:

[0007] Through the environment analysis station, current time information is collected, a shielding identifier trained based on target power station detection data is used to identify the time information, and a shielding loss coefficient is obtained, the target power station being a photovoltaic power station;

[0008] Through the meteorological analysis station, a meteorological data sequence within a future preset time range is collected;

[0009] The meteorological data sequence is subjected to power generation loss identification to obtain a meteorological loss coefficient, and the gas phase data sequence is subjected to dust influence parameter analysis to obtain a dust influence parameter;

[0010] Through the dust analysis station, multiple dust parameters of all photovoltaic panels in the target power station are collected, and the dust parameter array is calculated and obtained in combination with the dust influence parameter;

[0011] According to the dust parameter array, a dust distribution fitting image is constructed, cascade image generation analysis is performed, and a dust loss coefficient of the target power station is obtained;

[0012] According to the shading loss coefficient, the meteorological loss coefficient and the dust loss coefficient, a power generation loss coefficient is calculated and obtained, and a power generation loss amount is calculated and obtained.

[0013] In a second aspect, the application provides a power generation loss analysis system for a photovoltaic power station, which is connected with a power generation loss analysis device for the photovoltaic power station, the device comprising an environment analysis station, a meteorological analysis station, a dust analysis station and a comprehensive loss calculation module, and the system comprising:

[0014] A shading loss analysis module is configured to collect current time information through the environment analysis station, identify the time information based on a shading identifier trained based on detection data of a target power station to obtain a shading loss coefficient, the target power station being a photovoltaic power station.

[0015] A meteorological data collection module is configured to collect a meteorological data sequence within a future preset time range through the meteorological analysis station.

[0016] A meteorological loss analysis module is configured to perform power generation loss identification on the meteorological data sequence to obtain a meteorological loss coefficient, and perform dust influence parameter analysis on the meteorological data sequence to obtain a dust influence parameter.

[0017] A dust parameter processing module is configured to collect a plurality of dust parameters of all or part of photovoltaic panels in a target power station through the dust analysis station, and calculate a dust parameter array in combination with the dust influence parameter.

[0018] A dust loss analysis module is configured to construct a dust distribution fitting image according to the dust parameter array, perform cascade image generation analysis, and obtain a dust loss coefficient of the target power station.

[0019] A power generation loss calculation module is configured to calculate a power generation loss coefficient according to the shading loss coefficient, the meteorological loss coefficient and the dust loss coefficient, and calculate a power generation loss amount.

[0020] In a third aspect, a computer device comprises a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the method in the first aspect when executing the computer program.

[0021] In a fourth aspect, a computer readable storage medium stores a computer program, and the computer program implements the steps of the method in the first aspect when executed by a processor.

[0022] One or more technical solutions provided in the application have at least the following technical effects or advantages:

[0023] The technical scheme provided in the application collects current time information in the photovoltaic power station, analyzes power generation loss caused by surrounding buildings or other objects based on time, collects and analyzes power generation loss caused by current weather changes based on weather data, corrects and calculates dust parameters of part of the photovoltaic panels based on the influence parameters of dust deposition in the photovoltaic power station analyzed based on the weather data, obtains a dust parameter array, constructs a dust distribution fitting image by fitting for dust that has a greater impact on photovoltaic power generation loss, generates and analyzes cascaded images, obtains a power generation loss coefficient caused by dust in the power station, and comprehensively calculates the overall power generation loss in the photovoltaic power station by comprehensively calculating the power generation loss in the three aspects. The technical scheme provided in the application can improve the comprehensiveness of photovoltaic power generation loss analysis by analyzing power generation loss in multiple dimensions, can improve the accuracy of dust impact analysis that has the greatest impact on power generation loss by analyzing the influence parameters of dust deposition based on weather data, and can improve the accuracy and reliability of power generation loss analysis by dust by collecting dust parameters of part of the photovoltaic panels, fitting to form a dust distribution image, and analyzing cascaded images, and the cost of power generation loss monitoring and analysis is reduced without collecting dust data of all photovoltaic panels, thereby achieving the technical effects of improving the comprehensiveness and accuracy of photovoltaic power station power generation loss analysis. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A power generation loss analysis method flowchart for a photovoltaic power station is provided in the application.

[0025] Figure 2 A flowchart for obtaining a dust loss coefficient in a power generation loss analysis method for a photovoltaic power station is provided in the application.

[0026] Figure 3 A power generation loss analysis system structure diagram for a photovoltaic power station is provided in the application.

[0027] Figure 4 An internal structure diagram of a computer device in an embodiment.

[0028] Reference signs: shading loss analysis module 201, weather data collection module 202, weather loss analysis module 203, dust parameter processing module 204, dust loss analysis module 205, and power generation loss calculation module 206. DETAILED DESCRIPTION

[0029] The application provides a power generation loss analysis method and system for a photovoltaic power station to solve the technical problems of incomplete and inaccurate power generation loss calculation and analysis in the prior art.

[0030] Embodiment one

[0031] As Figure 1 shown in the embodiments of the present application, a power generation loss analysis method for a photovoltaic power station is provided, the method is applied to a power generation loss analysis device for photovoltaic power station construction, the device includes an environment analysis station, a meteorological analysis station, a dust analysis station, a comprehensive loss calculation module, and the method includes:

[0032] S101: Collect the current time information through the environment analysis station, identify the time information based on the target power station detection data trained shelter recognizer, and obtain the shelter loss coefficient, the target power station is a photovoltaic power station;

[0033] In the embodiments of the present application, the power generation loss analysis device for photovoltaic power station construction is a device for executing the power generation loss analysis method for photovoltaic power station, which includes memory, processor, etc., which can be used to store the program required for executing the method and perform processing.

[0034] The power generation loss analysis device for photovoltaic power station construction includes an environment analysis station, a meteorological analysis station, a dust analysis station, and a comprehensive loss calculation module, which are used to execute different steps in the method.

[0035] Among them, the target power station is a photovoltaic power station, preferably a larger scale, with more photovoltaic panels to absorb solar radiation for power generation, the target power station will be affected by factors such as shelter, weather and dust deposition, resulting in loss of photovoltaic power generation.

[0036] In the embodiments of the present application, the current time information is collected through the environment analysis station, that is, the real-time time when the method is executed for power generation loss analysis, the time information can be the time accurate to minutes in a day, or the time accurate to days in a year, wherein, due to the movement of the sun, the area of the target power station is changed by other buildings, trees, hills, etc. in the environment, which blocks sunlight, thereby causing the power generation loss to change. And, due to the movement of the sun, the area of the target power station is changed by other buildings, trees, hills, etc. in the environment, which blocks sunlight, thereby causing the power generation loss to change.

[0037] The environment analysis station can be a device with time collection function, which is used to analyze the power generation loss of the target power station caused by the area of the target power station being blocked at different times.

[0038] Among them, based on the detection data of the target power station for power generation detection, combined with the time information collected by the environment analysis station, the shelter recognizer for identifying the power generation loss of the target power station caused by the sunlight being blocked at different times is trained, and the current time information is identified to quickly and accurately obtain the power generation loss coefficient of the target power station caused by the sunlight being blocked at the current time.

[0039] The step S101 in the method provided by the embodiments of the present application comprises:

[0040] According to the monitoring data of the target power station, loss power generation amounts of all photovoltaic panels in the target power station in multiple time periods are obtained, sample time information records and sample shading loss coefficient records are obtained;

[0041] The sample time information records and the sample shading loss coefficient records are used as training data, and a shading identifier meeting a convergence requirement is trained based on machine learning.

[0042] The shading identifier is used to identify the time information, and a shading loss coefficient is obtained.

[0043] In the embodiments of the present application, according to the operation monitoring data of the target power station, specifically, operation time monitoring data and monitoring data of the area of all photovoltaic panels being shaded in the target power station, monitoring data in a period of time can be collected. In this way, sample time information records are obtained.

[0044] Further, according to the monitoring data of the area of all photovoltaic panels being shaded, loss power generation amounts of all photovoltaic panels due to being shaded are calculated, and a ratio of the loss power generation amounts to expected power generation amounts in an expected unshaded condition is further calculated, and sample shading loss coefficient records are obtained.

[0045] The sample time information records and the sample shading loss coefficient records are used as training data, and a shading identifier is trained based on machine learning, for example, based on a feedforward neural network in machine learning in the prior art. The shading identifier is used to analyze a photovoltaic power generation loss coefficient of the target power station due to sunlight being shaded according to time information of the target power station.

[0046] The sample time information records and the sample shading loss coefficient records are used as training data, and are used as input and output of the shading identifier respectively, and the shading identifier is supervised trained. Network parameters in the shading identifier are adjusted according to errors of output values and sample values, and multiple rounds of training are performed until the shading identifier meets a convergence requirement.

[0047] For example, the convergence requirement can be that the accuracy of the shading identifier reaches 90%.

[0048] Based on the shading identifier meeting the convergence requirement, current collected time information can be predicted and identified, and a shading loss coefficient of loss power generation amounts of the target power station due to sunlight being shaded at the time is obtained.

[0049] The embodiment of the present application trains the implicit relationship between the time and the different power generation loss coefficients caused by the solar light being blocked in the target power station based on machine learning, performs identification and prediction of the power generation loss coefficient caused by light blocking, and improves the accuracy and efficiency of the power generation loss analysis caused by solar light blocking.

[0050] S102: Collect a meteorological data sequence in a future preset time range through a meteorological analysis workstation;

[0051] In the embodiment of the present application, the change of weather will cause the change of solar radiation and then affect the power generation of the photovoltaic power station, therefore, the meteorological data is further collected to analyze the proportion of the power generation loss of the target power station caused by the change of weather.

[0052] The meteorological analysis workstation collects a meteorological data sequence in a future preset time range, and the preset time range is a time range of any length of time, that is, a preset time range after the current time, for example, one day or a week.

[0053] The meteorological analysis workstation can be software that can collect meteorological data published by a meteorological publication website, and collect meteorological data in a future preset time range of a region where the target power station is located, for example, the collection can be performed at a certain frequency, for example, once a day or once an hour, to form a meteorological data sequence.

[0054] Exemplarily, the meteorological data can include weather, temperature, wind direction, wind speed and other meteorological data.

[0055] S103: Perform power generation loss identification on the meteorological data sequence to obtain a meteorological loss coefficient, and perform dust influence parameter analysis on the meteorological data sequence to obtain a dust influence parameter;

[0056] In the embodiment of the present application, the meteorological data sequence is subjected to power generation loss identification, that is, the proportion of photovoltaic power generation loss caused by the change of solar radiation caused by the change of weather.

[0057] In addition, the change of weather will cause the change of dust deposition amount in the power station, for example, in low humidity weather, the dust deposition amount will be more, therefore, the influence proportion of the dust deposition amount according to the meteorological data is analyzed as a dust influence parameter, which is used as part of the data for analyzing the power generation loss caused by dust deposition in the subsequent step.

[0058] Through the power generation loss analysis on the meteorological data sequence, the proportion of the power generation loss caused by the meteorological data sequence is obtained as the corresponding meteorological loss coefficient. In addition, the dust influence parameter analysis on the meteorological data sequence that influences the dust deposition amount is performed to obtain the dust influence parameter under the meteorological data sequence.

[0059] The step S103 in the method provided by the embodiment of the present application specifically includes:

[0060] According to the monitoring data of the target power station, a plurality of sample meteorological data sequences and a plurality of sample meteorological loss coefficients are obtained;

[0061] The plurality of sample meteorological data sequences and the plurality of sample meteorological loss coefficients are used as training data, and a meteorological loss identifier that meets a convergence requirement is trained based on machine learning;

[0062] The meteorological loss identifier is used to identify the meteorological data sequence, and the meteorological loss coefficient is obtained.

[0063] In the embodiments of the present application, according to the monitoring data of the target power station, a plurality of sample meteorological data sequences in different preset time ranges within a historical time are obtained, and the ratio of the loss power generation to the expected power generation of the target power station due to meteorological changes is obtained as a plurality of sample meteorological loss coefficients.

[0064] The sample meteorological loss coefficient can be obtained by calculating the change in solar radiation under different meteorological data sequences and the solar radiation under ideal photovoltaic power generation conditions, simulating photovoltaic power generation based on the change, obtaining the simulated loss power generation, and calculating the ratio of the loss power generation to the power generation under ideal power generation conditions.

[0065] The plurality of sample meteorological data sequences and the plurality of sample meteorological loss coefficients are used as training data, and a meteorological loss identifier that meets a convergence requirement is trained based on machine learning.

[0066] The trained meteorological loss identifier is used to identify the currently collected meteorological data sequence, and the meteorological loss coefficient of the target power station under the meteorological data sequence is obtained.

[0067] The embodiments of the present application can enrich the dimension of photovoltaic power generation loss analysis and improve the comprehensiveness and accuracy of photovoltaic power generation loss analysis by using machine learning technology to analyze the photovoltaic power generation loss caused by the influence of meteorological data sequence on solar radiation.

[0068] Further, step S103 in the method provided by the embodiments of the present application further comprises:

[0069] According to the dust monitoring data of the target power station, the change in dust amount under different meteorological data sequences is obtained, and a plurality of sample dust influence parameters are calculated and obtained;

[0070] Based on the decision tree, meteorological data is used as decision input, dust influence parameters are used as decision data, and according to a plurality of sample meteorological data sequences and a plurality of sample dust influence parameters, a dust influence classifier is constructed.

[0071] The dust influence classifier is used to classify the meteorological data sequence, and the dust influence parameter is obtained.

[0072] In the embodiment of the application, according to the dust monitoring data of the target power station, the change amount of the dust amount under different meteorological data sequences is obtained, and a plurality of sample dust influence parameters are calculated and obtained.

[0073] For example, the dust deposition amount on the photovoltaic panel in a predetermined time range under ideal power generation conditions in the target power station can be tested by experiment, for example, the dust deposition area, and the predetermined time period can be 1 day or 1 week. Further, the dust deposition amount on the photovoltaic panel in a predetermined time range under different meteorological data sequences is tested by experiment, and then the change amount of the dust deposition amount under different meteorological data sequences is calculated, and then the deposition rate of the dust under different meteorological data sequences per unit time is calculated as a plurality of sample dust influence parameters.

[0074] Further, based on the decision tree algorithm in the prior art, meteorological data is used as decision input, and dust influence parameters are used as decision data, wherein different multi-layer decision nodes are constructed according to different types of meteorological data in the meteorological data, such as weather data sequences, temperature data series, etc. Each layer of decision node can make classification decision according to the input meteorological data, for example, according to the first data in the weather data sequence, and according to the mapping relationship between the different sample dust influence parameters and the plurality of sample meteorological data sequences, the decision output result of the final decision node is constructed. In this way, the dust influence classifier is constructed.

[0075] Alternatively, other classification algorithms in the prior art can also be used to construct the dust influence classifier according to the plurality of sample meteorological data sequences and the plurality of sample dust influence parameters.

[0076] The dust influence classifier is used to classify the meteorological data sequence in the current meteorological data sequence, and the dust influence parameter in the future predetermined time range in the target power station is obtained by classifying the meteorological data sequence in the current meteorological data sequence according to the different types of meteorological data in the meteorological data sequence.

[0077] The embodiment of the present application obtains the dust influence parameter by analyzing the influence of meteorological change on dust deposition in the target power station according to the meteorological data sequence, improves the accuracy of subsequent analysis of power generation loss in the target power station according to the dust analysis, and obtains the meteorological loss coefficient and the dust influence parameter by simultaneously analyzing a group of meteorological data sequences, improves the data utilization rate, and improves the comprehensiveness of the photovoltaic power generation loss.

[0078] In S104, the dust analysis station collects a plurality of dust parameters of part of the photovoltaic panels in the target power station, and calculates a dust parameter array in combination with the dust influence parameter.

[0079] In the embodiment of the present application, the dust analysis station collects a plurality of dust parameters of part of the photovoltaic panels in the target power station. For example, the area of dust shielding on part of the photovoltaic panels is collected. Optionally, the dust density or the maximum thickness of dust on part of the photovoltaic panels can also be collected.

[0080] The dust analysis station can include an image analysis device arranged on part of the photovoltaic panels. The surface image of the photovoltaic panel can be collected and image recognition can be performed to obtain a plurality of dust parameters of part of the photovoltaic panels.

[0081] Optionally, the dust analysis station can also collect and analyze the dust parameters on part of the photovoltaic panels by manual operation.

[0082] Preferably, part of the photovoltaic panels are a plurality of photovoltaic panels in the central part of the target power station, and the area of part of the photovoltaic panels accounts for 1 / 10 of the area of the target power station.

[0083] By collecting the dust parameters on part of the photovoltaic panels as the data basis for analyzing the dust influence on the photovoltaic power generation loss, the cost of analyzing the photovoltaic power generation loss is reduced on the basis of ensuring the accuracy of the analysis, without collecting the dust parameters on all photovoltaic panels.

[0084] Based on the plurality of dust parameters of part of the photovoltaic panels, the plurality of dust parameters are respectively processed by correction calculation in combination with the dust influence parameter obtained in the above content.

[0085] For example, the dust influence parameter includes the dust deposition rate under the meteorological data sequence. In combination with the current dust parameter, the corrected dust parameter on part of the photovoltaic panels in a future preset time range, such as one week or one day later, is calculated as a plurality of corrected dust parameters of part of the photovoltaic panels.

[0086] Further, based on the photovoltaic panel sequence of part of the photovoltaic panels from which the dust parameters are collected, the plurality of corrected dust parameters are arranged to form a dust parameter array.

[0087] S105: constructing a dust distribution fitting image according to the dust parameter array, performing cascade image generation analysis to obtain a dust loss coefficient of the target power station;

[0088] In the embodiment of the present application, the dust parameter array is converted into a dust distribution fitting image in the target power station, which is used as a dust distribution fitting image. Through the dust distribution fitting image, cascade image generation analysis is performed to analyze the proportion of photovoltaic power generation loss caused by dust shielding in the target power station, and the dust loss coefficient of the target power station is obtained.

[0089] As shown in Figure 2 The step S105 in the method provided by the embodiment of the present application includes:

[0090] Obtaining coordinate parameters of a plurality of photovoltaic panels in the target power station, combining the dust parameter array, and fitting to obtain a dust distribution fitting image in the target power station;

[0091] Based on deep learning, a dust distribution map generator with K-1 layers of dust distribution map generation channels is trained;

[0092] The dust distribution fitting image is processed by using the dust distribution map generator to obtain a generated fine dust distribution image;

[0093] A dust loss identifier is trained;

[0094] The fine dust distribution image is identified by using the dust loss identifier to obtain the dust loss coefficient.

[0095] In the embodiment of the present application, the coordinate parameters of a plurality of photovoltaic panels in the target power station are obtained, the position distribution of part of the photovoltaic panels collecting dust parameters is determined, and based on the position analysis, a plurality of position information of a plurality of pixel points of the fitting image is obtained. For example, according to the position distribution, a mapping relationship between the photovoltaic panels and the plurality of pixel points in the fitting image is constructed.

[0096] The coordinate parameters of the plurality of photovoltaic panels can be obtained according to the construction design data of the photovoltaic power station.

[0097] Combining the dust parameter array, the dust parameter of the photovoltaic panel corresponding to each pixel point is used as the gray value of the pixel point, and thus a preliminary fitted dust distribution image is formed. For example, the gray value corresponding to the average value of the dust parameters in the dust parameter array can be 127, and the corresponding gray value is obtained by calculating and rounding based on the ratio of other dust parameters to the average value of the dust parameters, which is used as the dust distribution image.

[0098] The partial photovoltaic panels are preferably a plurality of photovoltaic panels integrally formed in a rectangular distribution, which can be referred to as a rectangular dust distribution image. Alternatively, if the distribution of the partial photovoltaic panels is irregular, the gray scale values of the blank pixel points in the preliminary fitting dust distribution image without corresponding photovoltaic panels can be obtained by linear difference method, so that the dust distribution fitting image with a regular shape in the target power station is obtained, which is a gray scale image, and the gray scale value of each pixel point in the image can reflect the dust parameter on the photovoltaic panel in the target power station.

[0099] Based on the generative adversarial network in deep learning, a dust distribution image generator with K-1 layer dust distribution image generation channels is trained to perform K-1 layer image adversarial generation on the dust distribution fitting image through the K-1 layer dust distribution image generation channels, so as to obtain a generated fine dust distribution image with dust parameters of all regions in the target power station, and K is an integer greater than 1, preferably 10.

[0100] The method provided in the embodiments of the present application comprises the following steps:

[0101] According to the dust monitoring data in the target power station, a set of K sample dust distribution images with different sizes is obtained.

[0102] Based on the adversarial generative network, the K-1 layer dust distribution image generation channels are constructed, wherein each layer of the dust distribution image generation channel comprises a generator and a discriminator, the sizes of the input image and the output image of the K-1 layer dust distribution image generation channel are in a pyramid distribution, the size of the input image of the first layer dust distribution image generation channel is the same as that of the dust distribution fitting image, and the size of the output image of the K-1 layer dust distribution image generation channel is the same as that of the dust distribution fitting image formed by fitting the dust parameters of all photovoltaic panels in the target power station.

[0103] The first sample dust distribution image set and the second sample dust distribution image set in the set of K sample dust distribution images are used to supervise the training of the first layer dust distribution image generation channel until convergence, wherein random noise is added to the input first sample dust distribution image to generate the generator, and the generated dust distribution image generated by the generator is discriminated by the discriminator.

[0104] After the training of the first layer dust distribution image generation channel is completed, the training of the other K-2 layer dust distribution image generation channels is continued until all the channels converge.

[0105] In the embodiments of the present application, K sample dust distribution image sets are obtained according to dust monitoring data of the target power station in a historical time, and the sample dust distribution images in the K sample dust distribution image sets are of different sizes.

[0106] Each sample dust distribution image in each sample dust distribution image set is fitted based on the method in the foregoing content, and is constructed based on dust parameters of different numbers of partial photovoltaic panels in the target power station.

[0107] For example, the K sample dust distribution image sets are fitted based on dust parameters of photovoltaic panels in the target power station, which are distributed to form a rectangle and have an area of 1 / 10, 2 / 10, 3 / 10, …, 1 times the area of the target power station. The sizes of the sample dust distribution images in the K sample dust distribution image sets can form an image pyramid distribution in an up-sampling manner. Optionally, the sample dust distribution images can be fitted by combining the interpolation method described above.

[0108] Optionally, dust parameters of all photovoltaic panels in the target power station can be collected at multiple time points in the historical time, and K sample dust distribution images can be obtained by fitting, respectively. The K sample dust distribution images form an image pyramid distribution, and a larger sample dust distribution image includes a smaller sample dust distribution image. After the dust distribution map generator is trained, it is not necessary to continue collecting dust parameters of all photovoltaic panels.

[0109] Based on the generative adversarial network, K-1 layers of dust distribution map generation channels are constructed, preferably 9 layers of dust distribution map generation channels, wherein each layer of dust distribution map generation channel includes a generator and a discriminator, the sizes of the input images and the output images of the K-1 layers of dust distribution map generation channels form a pyramid distribution, the size of the output image of the first layer of dust distribution map generation channel is the same as that of the dust distribution fitting image, i.e., the first sample dust distribution image, the size of the output image is the same as that of the second sample dust distribution image, the size of the input image of the K-1 layer of dust distribution map generation channel is the same as that of the ninth sample dust distribution image, and the size of the output image is the same as that of the dust distribution fitting image of all photovoltaic panels in the target power station, i.e., the tenth dust distribution fitting image, for example, 10 times the size of the dust distribution fitting image.

[0110] The generator in each layer of dust distribution map generation channel includes a plurality of convolution blocks composed of a convolution kernel, a normalization layer and a nonlinear change layer, and the discriminator is composed of a plurality of convolution blocks containing convolution, normalization and nonlinear activation function, for example, sigmoid function. Those skilled in the art can construct the generative adversarial network architecture based on the prior art.

[0111] Thus, a K-1 layer dust distribution map generation channel with a cascaded structure is constructed, and the K-1 layer dust distribution map generation channel is trained using K sample dust distribution image sets respectively.

[0112] Specifically, the first and second sample dust distribution image sets within a set of K sample dust distribution images are used to supervise the training of the first dust distribution map generation channel within the K-1 layer dust distribution map generation channel until convergence. This training process is the same as the training process of the generative adversarial network.

[0113] Specifically, the input first sample dust distribution image is added to the generator in the first layer dust distribution image generation channel with random noise, and a generated dust distribution image is randomly generated. The generated dust distribution image generated by the generator is combined with a second sample dust distribution image including the first sample dust distribution image and input into a discriminator. The discriminator performs discrimination. For generated dust distribution images that are close to the second sample dust distribution image, the output of the discriminator is as close to 1 as possible. For generated dust distribution images that are not close to the second sample dust distribution image, the output of the discriminator is as close to 0 as possible.

[0114] This process is repeated multiple times until the accuracy of the first-layer dust distribution map generation channel reaches the required level, for example, the average discrimination result of the generated dust distribution image in the discriminator reaches 0.9, at which point the training is complete.

[0115] Using the same method, the second and third sample dust distribution image sets can be used to train the second layer dust distribution map generation channel, as well as the other K-3 layer dust distribution map generation channels, until all convergence.

[0116] The dust distribution map generator is obtained based on the converged K-1 layer dust distribution map generation channel.

[0117] Based on the completed dust distribution map generator, the fitted dust distribution image can be input into the first-layer dust distribution map generation channel to generate an image, obtaining a generated dust distribution image with a size of 2 / 10 of the fitted dust parameter image of all photovoltaic panels in the target power plant. This image is then input into the second-layer dust distribution map generation channel to obtain a generated dust distribution image with a size of 3 / 10 of the fitted dust parameter image of all photovoltaic panels in the target power plant. Finally, based on the (k-1)th layer dust distribution map generation channel, the final generated dust distribution image, i.e., the fine dust distribution image, is a dust distribution image generated by the dust distribution map generator through multi-level cascading based on the fitted dust distribution image, which can reflect the dust distribution level on all photovoltaic panels in the current target power plant.

[0118] Optionally, the K value can be set to 2, a dust loss identifier including only one layer of dust distribution map generation channel is constructed, and the image is directly generated according to the input dust distribution fitting image, so as to obtain the fine dust distribution image. However, since the fine dust distribution image is generated according to the dust distribution fitting image, more data noise needs to be added, which may cause the accuracy of the dust loss identifier to be low, or the convergence to be slow.

[0119] Based on the fine dust distribution image, the photovoltaic power generation loss caused by dust shielding in the current target power station is performed.

[0120] In the embodiment of the application, the fine dust distribution image is subjected to image feature extraction processing and identification by training the dust loss identifier, the features reflected by the gray values of each pixel point in the fine dust distribution image are identified, and the power generation loss of the target power station caused by dust shielding of the photovoltaic panel under the current dust parameter is obtained.

[0121] The method provided in the embodiment of the application further includes:

[0122] According to the dust monitoring data in the target power station, a sample fine dust distribution image set is obtained;

[0123] Based on the power generation data of the target power station under each sample fine dust distribution image, a sample dust loss coefficient set is calculated and obtained;

[0124] Based on the deep convolutional network, a dust loss identifier is constructed;

[0125] The sample fine dust distribution image set and the sample dust loss coefficient set are used to train the dust loss identifier until the convergence condition is met.

[0126] In the embodiment of the application, based on the method disclosed in the foregoing content, the dust parameters of the target power station at multiple time points can be obtained, the dust distribution fitting image can be fitted and obtained, and the generated fine dust distribution image can be obtained as the sample fine dust distribution image set.

[0127] In addition, based on the dust parameters at multiple time points, photovoltaic power generation test simulation can be performed, the deviation between the power generation under shielding and the power generation under ideal state is obtained, and then the proportion of the photovoltaic power generation loss of the target power station under different dust parameters is calculated and obtained as the sample dust loss coefficient set.

[0128] The embodiment of the application identifies the fine dust distribution image through the deep convolutional network, and outputs the proportion of the power generation loss of the target power station under the dust deposition state of the target power station reflected in the fine dust distribution image.

[0129] Based on the deep convolutional neural network, the dust loss identifier is constructed, which includes multiple convolutional layers, pooling layers and fully connected layers, and can perform convolution feature extraction on the input fine dust distribution image, and then process to obtain the corresponding dust loss coefficient.

[0130] The sample fine dust distribution image set and the sample dust loss coefficient set are used as training data to train the dust loss identifier until the convergence condition is met. For example, the convergence condition can be that the error between the dust loss coefficient output by the dust loss identifier and the sample dust loss coefficient is not more than 5%.

[0131] Based on the trained dust loss identifier, the fine dust distribution image obtained by the current processing is identified by image processing to obtain the dust loss coefficient of the power generation loss caused by dust shielding in the current target power station.

[0132] The embodiments of the present application can improve the efficiency and accuracy of analyzing photovoltaic power generation loss based on dust parameters by collecting dust parameters on part of the photovoltaic panels, using the dust parameters as the gray value in the pixel points, constructing a dust distribution fitting image, and generating a fine dust distribution fitting image including the gray value converted from the dust parameters of all photovoltaic panels through cascading generative adversarial networks, and identifying the dust loss coefficient. The efficiency and accuracy of analyzing photovoltaic power generation loss based on dust parameters can be improved, and only the dust parameters of part of the photovoltaic panels need to be collected, which can reduce the cost of power generation loss analysis for large photovoltaic power stations.

[0133] S106: According to the shielding loss coefficient, the meteorological loss coefficient and the dust loss coefficient, the power generation loss coefficient is calculated and obtained, and the power generation loss amount is calculated and obtained.

[0134] Based on the shielding loss coefficient, the meteorological loss coefficient and the dust loss coefficient of the target power station caused by external object shielding, weather changes and dust deposition analyzed in the foregoing, the total power generation loss coefficient of the target power station is calculated.

[0135] For example, the sum of the shielding loss coefficient, the meteorological loss coefficient and the dust loss coefficient can be calculated as the power generation loss coefficient. Alternatively, the power generation losses caused by the shielding loss coefficient, the meteorological loss coefficient and the dust loss coefficient may have some overlap, and the calculated power generation loss coefficient can be modified based on the experience of photovoltaic power generation loss analysis in the prior art.

[0136] Based on the power generation loss coefficient, the ideal photovoltaic power generation amount of the target power station is combined to calculate and obtain the predicted and analyzed power generation loss amount of the target power station as the analysis result of the photovoltaic power generation loss, which can be used as reference data for the operation and maintenance of the photovoltaic power station.

[0137] The embodiments of the present application achieve at least the following technical effects through the above technical solutions:

[0138] The embodiment of the present application collects current time information in the photovoltaic power station, analyzes the power generation loss caused by the shielding of surrounding buildings or other objects based on time, collects and analyzes the power generation loss caused by current meteorological changes based on meteorological data, corrects and calculates the dust parameters of part of the photovoltaic panels according to the influence parameters of dust deposition in the photovoltaic power station analyzed based on meteorological data, obtains a dust parameter array, constructs a dust distribution fitting image, generates a fine dust distribution fitting image including the gray value converted from the dust parameters of all photovoltaic panels through a cascade generative adversarial network, identifies the dust loss coefficient, obtains the power generation loss coefficient caused by dust in the power station, and comprehensively calculates the power generation loss in three aspects. The technical scheme provided by the present application can improve the comprehensiveness of the photovoltaic power generation loss analysis through multiple dimensions, and can improve the accuracy of the dust impact analysis which has the greatest impact on the power generation loss by analyzing the influence parameters of dust deposition based on meteorological data. By collecting the dust parameters of part of the photovoltaic panels, a dust distribution image is fitted, and through cascade image analysis, the accuracy and reliability of the power generation loss analysis through dust analysis are improved. Moreover, the cost of power generation loss monitoring and analysis is reduced without collecting dust data of all photovoltaic panels, thereby achieving the technical effects of improving the comprehensiveness and accuracy of the photovoltaic power station power generation loss analysis.

[0139] Embodiment two

[0140] Based on the same inventive concept as the power generation loss analysis method for a photovoltaic power station in the foregoing embodiment, as shown in the accompanying drawings, the present application provides a power generation loss analysis system for a photovoltaic power station, which is connected with a power generation loss analysis device for a photovoltaic power station. The device includes an environment analysis station, a meteorological analysis station, a dust analysis station, and a comprehensive loss calculation module. The system includes: Figure 3 The shielding loss analysis module 201 is used to collect current time information through the environment analysis station, identify the time information through the shielding identifier trained based on the target power station detection data, and obtain the shielding loss coefficient. The target power station is a photovoltaic power station.

[0141] The meteorological data acquisition module 202 is used to collect the meteorological data sequence in the future preset time range through the meteorological analysis station.

[0142] The meteorological loss analysis module 203 is used to identify the power generation loss of the meteorological data sequence, obtain the meteorological loss coefficient, and analyze the dust impact parameter of the meteorological data sequence, and obtain the dust impact parameter.

[0143]

[0144] ​The dust parameter processing module 204 is used to collect multiple dust parameters of some photovoltaic panels within all photovoltaic panels in the target power station through a dust analysis station, and calculate and obtain a dust parameter array by combining the dust influence parameters.

[0145] The dust loss analysis module 205 is used to construct a dust distribution fitting image based on the dust parameter array, perform cascaded image generation analysis, and obtain the dust loss coefficient of the target power plant.

[0146] The power generation loss calculation module 206 is used to calculate the power generation loss coefficient and the power generation loss amount based on the shading loss coefficient, meteorological loss coefficient and dust loss coefficient.

[0147] Furthermore, the occlusion loss analysis module 201 is also used to perform the following steps:

[0148] Based on the monitoring data of the target power station, the power generation loss due to the shading of all photovoltaic panels in the target power station within multiple time periods is obtained, and sample time information records and sample shading loss coefficient records are obtained.

[0149] Using the sample time information records and sample occlusion loss coefficient records as training data, an occlusion recognizer that meets the convergence requirements is trained based on machine learning.

[0150] The occlusion recognizer is used to identify the time information and obtain the occlusion loss coefficient.

[0151] Furthermore, the meteorological loss analysis module 203 is also used to perform the following steps:

[0152] Based on the monitoring data of the target power station, multiple sample meteorological data sequences and multiple sample meteorological loss coefficients are obtained;

[0153] Multiple sample meteorological data sequences and multiple sample meteorological loss coefficients were used as training data. Based on machine learning, a meteorological loss recognizer that meets the convergence requirements was trained.

[0154] A meteorological loss identifier is used to identify the meteorological data sequence and obtain the meteorological loss coefficient.

[0155] This also includes:

[0156] Based on the dust monitoring data of the target power plant, the change in dust amount under different meteorological data sequences is obtained, and multiple sample dust impact parameters are calculated.

[0157] Based on decision trees, meteorological data is used as the decision input, and dust impact parameters are used as the decision data. A dust impact classifier is constructed based on multiple sample meteorological data sequences and multiple sample dust impact parameters.

[0158] The weather data sequence is classified by using the dust influence classifier, and the dust influence parameter is obtained.

[0159] Further, the dust loss analysis module 205 is further used to execute the following steps:

[0160] The coordinate parameters of a plurality of photovoltaic panels in the target power station are obtained, and the dust distribution fitting image in the target power station is fitted by combining the dust parameter array;

[0161] Based on deep learning, a dust distribution map generator with K-1 layer dust distribution map generation channels is trained, K being an integer greater than 1;

[0162] The dust distribution fitting image is processed by using the dust distribution map generator, and a generated fine dust distribution image is obtained;

[0163] A dust loss identifier is trained;

[0164] The fine dust distribution image is identified by using the dust loss identifier, and the dust loss coefficient is obtained.

[0165] Wherein, based on deep learning, a dust distribution map generator with K-1 layer dust distribution map generation channels is trained, including:

[0166] According to the dust monitoring data in the target power station, K sample dust distribution image sets are obtained, and the sample dust distribution images in the K sample dust distribution image sets are different in size;

[0167] Based on the generative adversarial network, the K-1 layer dust distribution map generation channel is constructed, wherein each layer of the dust distribution map generation channel includes a generator and a discriminator, the size of the input image and the output image of the K-1 layer dust distribution map generation channel is in a pyramid distribution, the size of the input image of the first layer dust distribution map generation channel is the same as the dust distribution fitting image, and the size of the output image of the K-1 layer dust distribution map generation channel is the same as the dust distribution fitting image formed by fitting the dust parameters of all photovoltaic panels in the target power station;

[0168] The first sample dust distribution image set and the second sample dust distribution image set in the K sample dust distribution image set are used to supervise the training of the first layer dust distribution map generation channel until convergence, wherein random noise is added to the input first sample dust distribution image to generate the input of the generator, and the generated dust distribution image generated by the generator is discriminated by the discriminator;

[0169] After the training of the first layer dust distribution map generation channel is completed, the training of the other K-2 layer dust distribution map generation channels is continued until all the channels converge.

[0170] The training obtains the dust loss identifier, and includes:

[0171] According to the dust monitoring data in the target power station, a sample fine dust distribution image set is obtained;

[0172] Based on the power generation data of the target power station under each sample fine dust distribution image, a sample dust loss coefficient set is calculated and obtained;

[0173] Based on the deep convolutional network, the dust loss identifier is constructed;

[0174] The sample fine dust distribution image set and the sample dust loss coefficient set are used to train the dust loss identifier until the convergence condition is met.

[0175] The foregoing detailed description of the method for analyzing power generation loss of a photovoltaic power station enables those skilled in the art to clearly understand the method for analyzing power generation loss of a photovoltaic power station in the embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant part is described in the method part.

[0176] The foregoing description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0177] The specific embodiments of the system for analyzing power generation loss of a photovoltaic power station can be seen in the embodiments of the method for analyzing power generation loss of a photovoltaic power station described above, and will not be described again here. Each module in the above-described device for analyzing power generation loss of a photovoltaic power station can be realized wholly or partially by software, hardware, and a combination thereof. Each module described above can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module.

[0178] Embodiment Three

[0179] As shown in Figure 4 A computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method in embodiment one when executing the computer program.

[0180] In one embodiment, a computer device, which can be a server, has an internal structure diagram as shown in FIG. 1. Figure 4 The computer device includes a processor, a memory and a network interface connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store news data and time decay factors. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements a power generation loss analysis method for a photovoltaic power station.

[0181] Those skilled in the art can understand that the structure shown in the above embodiments is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. Figure 4

[0182] Embodiment Four

[0183] The embodiment provides a computer readable storage medium, which stores a computer program. The computer program, when executed by a processor, implements the steps of the method in the first embodiment.

[0184] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but it should be considered that any combination of the technical features is within the scope of the present disclosure, as long as the combination does not cause contradiction.

[0185] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be considered as a limitation on the scope of the patent. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, some modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.​

Claims

1. A method for analyzing power generation losses in photovoltaic power plants, characterized in that, The method is applied to a power generation loss analysis device for photovoltaic power plant construction. The device includes an environmental analysis station, a meteorological analysis station, a dust analysis station, and a comprehensive loss calculation module. The method includes: The environmental analysis station collects current time information, and the shading recognition device trained based on the target power station detection data identifies the time information to obtain the shading loss coefficient. The target power station is a photovoltaic power station. Meteorological data sequences are collected within a predetermined time range through meteorological analysis stations. Power generation loss is identified from meteorological data sequences to obtain meteorological loss coefficients; and dust impact parameters are analyzed from gas phase data sequences to obtain dust impact parameters. By using a dust analysis station, multiple dust parameters of some photovoltaic panels within all photovoltaic panels in the target power plant are collected. Combined with the dust impact parameters, a dust parameter array is calculated and obtained. Based on the dust parameter array, a dust distribution fitting image is constructed, and cascaded image generation analysis is performed to obtain the dust loss coefficient of the target power station. Specifically, the coordinate parameters of multiple photovoltaic panels within the target power station are obtained, and combined with the dust parameter array, a dust distribution fitting image within the target power station is fitted. Based on deep learning, a dust distribution map generator with K-1 layers of dust distribution map generation channels is trained and obtained, where K is an integer greater than 1. The dust distribution map generator is used to process the dust distribution fitting image to obtain a generated fine dust distribution image. A dust loss recognizer is trained and obtained. The dust loss recognizer is used to recognize the fine dust distribution image to obtain the dust loss coefficient. Based on the shading loss coefficient, meteorological loss coefficient, and dust loss coefficient, the power generation loss coefficient is calculated, and the power generation loss is calculated.

2. The method according to claim 1, characterized in that, The method includes: Based on the monitoring data of the target power station, the power generation loss due to the shading of all photovoltaic panels in the target power station within multiple time periods is obtained, and sample time information records and sample shading loss coefficient records are obtained. Using the sample time information records and sample occlusion loss coefficient records as training data, an occlusion recognizer that meets the convergence requirements is trained based on machine learning. The occlusion recognizer is used to identify the time information and obtain the occlusion loss coefficient.

3. The method according to claim 1, characterized in that, The method includes: Based on the monitoring data of the target power station, multiple sample meteorological data sequences and multiple sample meteorological loss coefficients are obtained; Multiple sample meteorological data sequences and multiple sample meteorological loss coefficients were used as training data. Based on machine learning, a meteorological loss recognizer that meets the convergence requirements was trained. A meteorological loss identifier is used to identify the meteorological data sequence and obtain the meteorological loss coefficient.

4. The method according to claim 3, characterized in that, The method includes: Based on the dust monitoring data of the target power plant, the change in dust amount under different meteorological data sequences is obtained, and multiple sample dust impact parameters are calculated. Based on decision trees, meteorological data is used as the decision input, and dust impact parameters are used as the decision data. A dust impact classifier is constructed based on multiple sample meteorological data sequences and multiple sample dust impact parameters. A dust impact classifier is used to perform decision classification on the meteorological data sequence to obtain the dust impact parameters.

5. The method according to claim 1, characterized in that, The method includes: Based on the dust monitoring data within the target power plant, K sample dust distribution image sets are obtained, and the sample dust distribution images within the K sample dust distribution image sets are of different sizes; Based on generative adversarial networks, a K-1 layer dust distribution map generation channel is constructed. Each layer of the dust distribution map generation channel includes a generator and a discriminator. The input and output images of the K-1 layer dust distribution map generation channel are distributed in a pyramid shape. The input image of the first layer dust distribution map generation channel is the same size as the dust distribution fitting image. The output image of the K-1 layer dust distribution map generation channel is the same size as the dust distribution fitting image formed by fitting the dust parameters of all photovoltaic panels in the target power station. The first sample dust distribution image set and the second sample dust distribution image set within the K sample dust distribution image set are used to supervise the training of the first layer dust distribution map generation channel until convergence. Random noise is added to the input first sample dust distribution image and the generated dust distribution image is judged by a discriminator. After the first layer of dust distribution map generation channel is trained, the other K-2 layer dust distribution map generation channels are trained until all convergence is achieved.

6. The method according to claim 5, characterized in that, The method includes: Based on the dust monitoring data within the target power plant, a set of detailed dust distribution images was obtained. Based on the power generation data of the target power plant under the fine dust distribution image of each sample, the sample dust loss coefficient set is calculated and obtained; A dust loss detector is constructed based on a deep convolutional network. The dust loss recognizer is trained using the sample fine dust distribution image set and the sample dust loss coefficient set until the convergence condition is met.

7. A power generation loss analysis system for photovoltaic power plants, characterized in that, The system is used to perform the method according to any one of claims 1-6, the system is connected to a power generation loss analysis device for a photovoltaic power station, the device including an environmental analysis station, a meteorological analysis station, a dust analysis station, and a comprehensive loss calculation module, the system comprising: The shading loss analysis module is used to collect current time information through an environmental analysis station, and to identify the time information based on a shading recognizer trained on the target power station detection data to obtain the shading loss coefficient. The target power station is a photovoltaic power station. The meteorological data acquisition module is used to collect meteorological data sequences within a preset time range through a meteorological analysis station; The meteorological loss analysis module is used to identify power generation losses from meteorological data sequences and obtain meteorological loss coefficients, as well as to analyze dust impact parameters from gas phase data sequences and obtain dust impact parameters. The dust parameter processing module is used to collect multiple dust parameters of some photovoltaic panels within all photovoltaic panels in the target power plant through a dust analysis station, and calculate and obtain a dust parameter array by combining the dust influence parameters. The dust loss analysis module is used to construct a dust distribution fitting image based on the dust parameter array, perform cascaded image generation analysis, and obtain the dust loss coefficient of the target power plant. The power generation loss calculation module is used to calculate the power generation loss coefficient and the amount of power generation loss based on the shading loss coefficient, meteorological loss coefficient and dust loss coefficient.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images