Composite photovoltaic power generation prediction method based on visual identification

Abstract

The invention relates to the technical field of photovoltaic energy, and discloses a composite photovoltaic power generation prediction method based on visual identification, and the method comprises the steps: obtaining the equipment data, weather forecast data and historical power generation data of a photovoltaic station; constructing a photovoltaic power generation prediction model to obtain first prediction data of photovoltaic power generation; the method comprises the following steps: collecting environmental data of a photovoltaic panel before a photovoltaic station prediction day, extracting key features, and calculating to obtain influence correction factor data; and constructing a composite prediction model to obtain second prediction data of photovoltaic power generation. The prediction accuracy is enhanced by acquiring multi-source data, preliminary prediction is performed in combination with weather forecast and equipment data, the environmental data of the photovoltaic panel are acquired through a visual identification technology, key features are extracted, influence correction factors are calculated, and the influence of local environmental change on the generating capacity is captured. And the preliminary prediction result is combined with the influence correction factor to construct a composite prediction model, so that more accurate final power generation capacity prediction is obtained, and the prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic energy, and in particular to a composite photovoltaic power generation prediction method based on visual recognition. Background Art

[0002] Photovoltaic power generation is increasingly being used in the world because it can directly convert solar energy into electricity. However, its energy source is usually unstable, and its actual power generation is also unstable in reality. The volatility of photovoltaic power generation has brought great challenges to the dispatching of power grid systems, power balance, load forecasting, and distribution network operations. In order to ensure the stable operation of the power grid, it is of great significance to conduct photovoltaic power forecasting.

[0003] Currently, the power generation of photovoltaic stations is mainly predicted based on time series analysis and machine learning. First, data is difficult to obtain. Second, weather data is not accurate data with fine granularity. In this case, especially when the sky is clear but there are occasional local clouds, it will affect the accuracy of photovoltaic power generation data prediction. Third, there is no way to obtain the cleanliness of the photovoltaic panels themselves, except in scenarios with robot cleaning. It is believed that the cleanliness of the photovoltaic panels is the highest after robot cleaning, but there are also cases where the robot cleaning is not thorough. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a composite photovoltaic power generation prediction method based on visual recognition to solve the above problems.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a composite photovoltaic power generation prediction method based on visual recognition, comprising: acquiring equipment data, weather forecast data, and historical power generation data of a photovoltaic station; Constructing a photovoltaic power generation prediction model, training the photovoltaic power generation prediction model with the historical power generation data, taking the weather forecast data and equipment data of the prediction day as input, and obtaining first prediction data of photovoltaic power generation; Collect environmental data of photovoltaic panels before the forecast date of the photovoltaic station, extract key features, and calculate the impact correction factor data; Based on the first prediction data and the influencing factor data, a composite prediction model is constructed to obtain second prediction data of photovoltaic power generation as the final photovoltaic power generation prediction power.

[0007] As a preferred solution of the composite photovoltaic power generation prediction method based on visual recognition described in the present invention, obtaining weather forecast data and historical power generation data of the photovoltaic station includes: Acquiring equipment data, wherein the equipment data includes photovoltaic panel location, rated power, conversion efficiency, tilt angle, and azimuth; Acquire weather forecast data for the predicted time, wherein the weather forecast data for the predicted time includes time, weather type, total radiation, direct radiation, scattered radiation, reflected radiation, temperature, humidity, and cloud coverage; The historical power generation data for the past one to three years is obtained, wherein the historical power generation data includes actual weather data, equipment data, and actual power generation data of the site.

[0008] As a preferred solution of the composite photovoltaic power generation prediction method based on visual recognition described in the present invention, the first photovoltaic power generation prediction data is obtained including: Preprocessing the equipment data, weather forecast data and historical power generation data, sorting the historical power generation data in time series, and classifying them according to time periodicity based on data date; Dividing the historical power generation data into different training sets, validation sets and test sets according to time period types and weather types; A photovoltaic power generation prediction model is constructed using a long short-term memory network, and an attention mechanism is introduced. The photovoltaic power generation prediction model is trained using the historical power generation data, and the equipment data and weather forecast data are input into the trained photovoltaic power generation prediction model to obtain the first prediction data of the prediction day.

[0009] As a preferred solution of the composite photovoltaic power generation prediction method based on visual recognition described in the present invention, the environmental data collected before the prediction date of the photovoltaic station includes: A plurality of camera points are set around the photovoltaic panels to regularly collect image data around the photovoltaic panels of the photovoltaic station, and environmental data are extracted from the image data through visual recognition, wherein the environmental data includes sky sunshine data, photovoltaic panel sunshine data, photovoltaic panel heavy snow data, and photovoltaic panel dirt data; According to the weather type of the forecast day, a day closest to the forecast day and with the same weather type is determined as the first reference day, and key features of the first reference day are extracted, including photovoltaic panel sunshine light data and sky sunshine light data; The day before the predicted day is taken as the second reference day, and key features of the second reference day are extracted, including photovoltaic panel heavy snow data and photovoltaic panel dirt data.

[0010] As a preferred solution of the composite photovoltaic power generation prediction method based on visual recognition described in the present invention, the key features are extracted including: The sky sunshine light data represents sky illuminance, collects sky images above the photovoltaic panel during the first reference day in each time period, divides the sky images into multiple sub-areas, identifies the sky illuminance of the sub-areas in each time period, and calculates the sky illuminance in the area in each time period; The photovoltaic panel sunshine light data represents the light intensity of the photovoltaic panel, collecting the first photovoltaic panel image data of the first reference day in each time period, converting the first photovoltaic panel image data into a photovoltaic panel grayscale image, segmenting the photovoltaic panel grayscale image, marking each complete photovoltaic panel area, identifying the grayscale value of each photovoltaic panel area, calculating the average grayscale value of the photovoltaic panel, and obtaining the light intensity of the photovoltaic panel in the area in each time period; The photovoltaic panel heavy snow data represents the heavy snow coverage of the photovoltaic panel, collects the latest second photovoltaic panel image data of the second reference day, performs image preprocessing and image segmentation on the second photovoltaic panel image data, converts the segmented image into a binary image, identifies the snow-covered area, counts the number of white pixels in the binary image, and calculates the heavy snow coverage of the photovoltaic panel; The photovoltaic panel contamination data represents the degree of contamination of the photovoltaic panel. The latest third photovoltaic panel image data of the second reference day is collected, and the third photovoltaic panel image data is subjected to image preprocessing and image segmentation. The pixel quantity of different types of contamination in the third photovoltaic panel image data is identified, and the contamination types include dust, stains and damage. The degree of contamination of the photovoltaic panel is obtained based on the weighted sum of the pixel quantity of dust, stains and damage and the total pixel quantity of the photovoltaic panel.

[0011] As a preferred solution of the composite photovoltaic power generation prediction method based on visual recognition described in the present invention, the calculated influencing correction factor data includes: Compare the sky illuminance in each time period of the first reference day with the illuminance corresponding to the total radiation in the same time period in the weather forecast of the first reference day. If the average value of the difference between the two is greater than the change threshold, calculate the sky illuminance error by the difference between the two to obtain a first correction factor. Compare the light intensity of the photovoltaic panel in each time period of the first reference day with the sky light intensity in each time period, and calculate the light intensity error of the photovoltaic panel by the difference between the two to obtain a second correction factor; Based on the snow coverage of the photovoltaic panel, a first influencing factor is calculated; Based on the degree of contamination of the photovoltaic panel, a second influencing factor is calculated; The first correction factor, the second correction factor, the first impact factor and the second impact factor are combined to generate impact correction factor data.

[0012] As a preferred solution of the composite photovoltaic power generation prediction method based on visual recognition described in the present invention, the second prediction data of photovoltaic power generation is obtained including: Based on the first prediction data and the influencing factor data, a composite prediction model is constructed to obtain the revised second prediction data. The composite prediction model is expressed as: ; in, represents the first forecast data of photovoltaic power generation, represents the first correction factor, represents the second correction factor, represents the first impact factor, represents the second impact factor, Represents the dynamic adjustment factor.

[0013] In a second aspect, the present invention provides a composite photovoltaic power generation prediction system based on visual recognition, comprising: The acquisition module is used to obtain equipment data, weather forecast data and historical power generation data of photovoltaic stations; A construction module is used to construct a photovoltaic power generation prediction model, train the photovoltaic power generation prediction model through the historical power generation data, take the weather forecast data and the equipment data as input, and obtain the first prediction data of photovoltaic power generation; The data processing module is used to collect the environmental data of photovoltaic panels before the forecast date of the photovoltaic station, extract key features, and calculate the impact correction factor data; The output module is used to construct a composite prediction model based on the first prediction data and the influencing factor data to obtain the second prediction data of photovoltaic power generation as the final photovoltaic power generation prediction power.

[0014] In a third aspect, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the composite photovoltaic power generation prediction method based on visual recognition are implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the composite photovoltaic power generation prediction method based on visual recognition.

[0016] Compared with the prior art, the present invention has the following beneficial effects: the present invention provides a comprehensive data basis for the model by acquiring multi-source data, thereby enhancing the accuracy of the prediction, training the model through historical power generation data, and making preliminary predictions in combination with weather forecasts and equipment data, collecting environmental data of photovoltaic panels through visual recognition technology, extracting key features and calculating impact correction factors, capturing the impact of local environmental changes on power generation, combining the preliminary prediction results with the impact correction factors, and constructing a composite prediction model to obtain a more accurate final power generation prediction, which not only improves the accuracy of the prediction, but also enhances the robustness and adaptability of the model, and can better cope with complex and changeable environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0018] Figure 1 The figure is a schematic diagram of the overall process of a composite photovoltaic power generation prediction method based on visual recognition according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0022] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0023] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0024] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0025] Reference Figure 1 , as an embodiment of the present invention, provides a composite photovoltaic power generation prediction method based on visual recognition, comprising: S101, obtaining equipment data, weather forecast data and historical power generation data of the photovoltaic station; S102, constructing a photovoltaic power generation prediction model, training the photovoltaic power generation prediction model through historical power generation data, taking the weather forecast data and equipment data of the prediction day as input, and obtaining first prediction data of photovoltaic power generation; S103, collecting environmental data of photovoltaic panels before the forecast date of the photovoltaic station, extracting key features, and calculating impact correction factor data; S104, constructing a composite prediction model based on the first prediction data and the influencing factor data to obtain second prediction data of photovoltaic power generation as the final photovoltaic power generation prediction power.

[0026] In a preferred implementation of step S101, obtaining weather forecast data and historical power generation data of a photovoltaic station includes: Obtain the latest equipment data, including photovoltaic panel location, rated power, conversion efficiency, tilt angle, and azimuth; Obtain weather forecast data for the forecast time, including time, weather type, total radiation, direct radiation, scattered radiation, reflected radiation, temperature, humidity, and cloud coverage; Obtain historical power generation data for the past one to three years. The historical power generation data includes actual weather data, equipment data, and actual power generation data of the site.

[0027] Specifically, specific parameters in the equipment data, weather forecast data and historical power generation data of the photovoltaic station can be set and obtained according to the actual application scenario, including but not limited to the various parameter types in the present embodiment, among which the weather forecast data can also include wind speed, rainfall, air pressure, etc., the equipment data can also include photovoltaic panel type, inverter characteristics, etc., the weather types can include clear, cloudy, overcast, rainy, etc., and the historical power generation data ensures that the data covers a variety of weather types and seasonal changes.

[0028] It should be noted that by obtaining equipment data, weather forecast data and historical power generation data of photovoltaic sites, the model can obtain rich input information, providing a solid foundation for subsequent model training and prediction, and improving the accuracy of prediction. Equipment data reflects the real-time operating status of the photovoltaic system, weather forecast data provides future environmental conditions, and historical power generation data contains the actual performance of the system under different conditions. The combination of the three can more accurately capture the dynamic changes of the photovoltaic system and reduce prediction errors. Historical power generation data can help the model learn the impact of different weather conditions, equipment status and seasonal changes on power generation, thereby improving the generalization ability of the model and enabling it to make reliable predictions in various environments.

[0029] In a preferred implementation of step S102, building a photovoltaic power generation prediction model to obtain first photovoltaic power generation prediction data includes: Perform data preprocessing on equipment data, weather forecast data and historical power generation data, sort the historical power generation data by time series, and classify them by time periodicity based on the data date; The historical power generation data is divided into different training sets, validation sets and test sets according to the time cycle type and weather type, where the cycle type can be a seasonal cycle or a monthly cycle, etc.; A photovoltaic power generation prediction model is constructed using a long short-term memory network, and an attention mechanism is introduced. The photovoltaic power generation prediction model is trained using historical power generation data. The equipment data and weather forecast data are input into the trained photovoltaic power generation prediction model to obtain the first prediction data for the prediction day.

[0030] Specifically, data preprocessing includes data cleaning and normalization, removing or filling missing values, outliers and noise data, ensuring the integrity and accuracy of the data, normalizing or standardizing features of different dimensions, ensuring the stability of model training, identifying and processing outliers, ensuring data accuracy, and constructing the data in the form of sliding windows. Each window contains historical data for a period of time, which is used to predict the power generation at the next time point.

[0031] The model framework of the photovoltaic power generation prediction model includes: LSTM layers can be stacked in multiple layers, and a Dropout layer can be added after each layer to prevent overfitting. Finally, a Dense layer is used to output the predicted value. At the same time, an attention mechanism is introduced to make the model pay more attention to important time steps or features, further improving the prediction accuracy. The historical power generation data is divided into training set, validation set and test set in a ratio of 7:1.5:1.5 to ensure the generalization of the model on unseen data. The model is trained using the training set data and the loss function is monitored on the validation set. When the loss on the validation set no longer decreases, the training is stopped in advance to avoid overfitting. The model performance is evaluated using methods such as K-fold cross validation, and the hyperparameters of the model are adjusted through grid search, random search or Bayesian optimization to find the optimal configuration. The mean square error (MSE), mean absolute error (MAE), coefficient of determination (R²), etc. are used to select appropriate indicators to measure the prediction effect of the model. Finally, the model is updated in real time. As new data is continuously updated, the model is retrained regularly to ensure that it is always in the best state.

[0032] It should be noted that by constructing a photovoltaic power generation prediction model through LSTM, the characteristics of time series data are fully considered. LSTM can effectively capture the complex dynamic relationship between weather conditions and photovoltaic system power generation, thereby providing accurate prediction results. Through continuous optimization and improvement, the prediction accuracy of the model can be gradually improved, providing strong support for the efficient operation of photovoltaic power generation.

[0033] In a preferred implementation of step S103, collecting environmental data of the photovoltaic station before the forecast date includes: Multiple camera points are set up around the photovoltaic panels to regularly collect image data around the photovoltaic panels of the photovoltaic station, and environmental data is extracted from the image data through visual recognition. The environmental data includes sky sunshine data, photovoltaic panel sunshine data, photovoltaic panel heavy snow data, and photovoltaic panel dirt data; According to the weather type of the forecast day, the day closest to the forecast day and with the same weather type is determined as the first reference day, and the key features of the first reference day are extracted, including the photovoltaic panel sunshine light data and the sky sunshine light data; The day before the forecast day is taken as the second reference day, and the key features of the second reference day are extracted, including the heavy snow data of photovoltaic panels and the dirty data of photovoltaic panels.

[0034] Specifically, when collecting different data, different camera equipment can be selected according to needs to obtain accurate data. Near the photovoltaic panel, ensure that the camera can clearly capture the surface of the photovoltaic panel and its surrounding environment. Multiple cameras can be installed to cover a large area of ​​the photovoltaic array. According to the type of data collected, the camera should maintain a certain angle with the photovoltaic panel to avoid direct light causing overexposure or reflection of the image. The camera can shoot the photovoltaic panel and the surrounding environment from the side to capture the required data. At the same time, the timed collection of various data can be set according to actual conditions. In this embodiment, the data is collected every 15 minutes; wherein, when making photovoltaic power generation predictions, in addition to referring to historical actual power generation data, in order to reduce errors and correct the predicted values, it is also necessary to consider the actual environment of the recent photovoltaic station and the photovoltaic panel conditions. Selecting the photovoltaic panel sunshine light data and the sky sunshine light data of the most recent date under the same weather type can effectively judge the accuracy of the photovoltaic power generation prediction. In addition, the covering of the photovoltaic panel can significantly affect the power generation efficiency of the photovoltaic panel. Collecting the latest real photovoltaic panel heavy snow data and photovoltaic panel dirt data from the prediction day can make key corrections to the power generation prediction value.

[0035] It should be noted that in photovoltaic power generation prediction, the use of visual recognition technology to obtain data such as the sunlight of photovoltaic panels, sunlight in the sky, snow cover, and dirtiness can significantly improve the accuracy of predictions and the maintenance efficiency of the system.

[0036] In a preferred implementation of step S103, extracting key features includes: The sky sunshine light data represents the sky illuminance. The sky image above the photovoltaic panel in the first reference day of each time period is collected, the sky image is divided into multiple sub-areas, the sky illuminance of the sub-area in each time period is identified, and the sky illuminance in the area in each time period is calculated. Photovoltaic panel sunshine light data represents the light intensity of photovoltaic panels. The first photovoltaic panel image data of the first reference day in each time period is collected, and the first photovoltaic panel image is converted into a photovoltaic panel grayscale image. The photovoltaic panel grayscale image is segmented, and each complete photovoltaic panel area is marked. The grayscale value of each photovoltaic panel area is identified, and the average grayscale value of the photovoltaic panel is calculated to obtain the light intensity of the photovoltaic panel in the area in each time period; Photovoltaic panel heavy snow data indicates the heavy snow coverage of photovoltaic panels. The latest second photovoltaic panel image data on the second reference day is collected. The second photovoltaic panel image data is image preprocessed and segmented. The segmented image is converted into a binary image, the snow-covered area is identified, the number of white pixels in the binary image is counted, and the heavy snow coverage of photovoltaic panels is calculated. The photovoltaic panel contamination data represents the degree of contamination of the photovoltaic panel. The latest third photovoltaic panel image data of the second reference day is collected, and image preprocessing and image segmentation are performed on the third photovoltaic panel image data. The pixel quantity of different types of contamination in the third photovoltaic panel image data is identified, and the types of contamination include dust, stains and damage. The degree of contamination of the photovoltaic panel is obtained by taking the weighted sum of the pixel quantity of dust, stains and damage and the total pixel quantity of the photovoltaic panel.

[0037] Specifically, when collecting sky sunlight data, you can use a fisheye lens or a full-sky camera to capture images of the entire sky and capture elements such as the sun, clouds, and blue sky. The full-sky camera can provide a 360-degree field of view to ensure that no important meteorological information is missed. The camera is installed in the center or open area of ​​the photovoltaic station to ensure that there are no obstructions to interfere with the shooting. Use a convolutional neural network (CNN) to identify the cloud types in the image (such as cirrus, cumulus, stratus, etc.). Different cloud types have different effects on solar radiation. The amount of scattered radiation can be estimated based on the thickness and distribution of the clouds. The total area above the photovoltaic station is divided into multiple sub-areas. By analyzing the color and brightness distribution in the image and combining the atmospheric transparency model, the light intensity in all directions of the sky in each sub-area is obtained, and then a weighted average is performed to calculate the sky illumination in each area during each period.

[0038] When collecting photovoltaic panel sunlight data, use image denoising algorithms (such as median filtering and bilateral filtering) to remove noise in the image to ensure image quality. Convert the first photovoltaic panel image into a photovoltaic panel grayscale image. The grayscale value reflects the brightness of each pixel in the image. Use semantic segmentation models (such as U-Net, DeepLab, Mask R-CNN, etc.) to segment the image and distinguish the photovoltaic panel from other backgrounds (such as brackets, ground, sky, etc.). Collect annotated photovoltaic panel images. Accurately outline the outline of the photovoltaic panel during annotation. Annotate each complete photovoltaic panel area. Incomplete photovoltaic panels are ignored. The higher the grayscale value, the greater the light intensity in the area. Calculate the average grayscale value or brightness value of all pixels in the photovoltaic panel area as the average illuminance of the photovoltaic panel. Construct a linear regression model to represent the linear relationship between the grayscale value and the actual illuminance. Establish a mapping relationship between the grayscale value and the illuminance through calibration experiments. The linear regression model is expressed as: ; in, and represents the adjustment parameter, Represents the gray value, Represents the illuminance of photovoltaic panels.

[0039] When collecting heavy snow data for photovoltaic panels, collect the latest second photovoltaic panel image data on the second reference day. The camera regularly captures images of the photovoltaic panels. When the weather changes (such as snow, hail, etc.), increase the shooting frequency to capture real-time conditions. Image preprocessing mainly uses image denoising algorithms (such as median filtering and bilateral filtering) to remove noise in the image to ensure image quality. Image enhancement techniques (such as contrast stretching and histogram equalization) can also be used to improve image clarity. The image is segmented to identify the snow-covered boundaries. Morphological operations (such as dilation, corrosion, and opening and closing operations) are performed on the segmentation results to remove small noise areas and smooth the boundaries to ensure that the segmentation results are more accurate. Edge detection algorithms (such as Canny edge detection) are used to further optimize the segmentation results to ensure that the boundaries of the snow-covered areas are clearly visible. The segmented image is converted into a binary image, in which the snow-covered area is white (value is 1) and the other areas are black (value is 0). The number of white pixels in the binary image is counted, and each pixel represents a unit area. At the same time, the area of ​​the photovoltaic panel is obtained. The snow coverage of the photovoltaic panel can be calculated by dividing the number of pixels in the snow-covered area by the number of pixels in the total area of ​​the photovoltaic panel.

[0040] The photovoltaic panel contamination data is obtained by collecting the latest third photovoltaic panel image data on the second reference day, wherein the contamination type is classified by a multi-category classification model (such as ResNet, VGG, EfficientNet, etc.) to classify the contamination type in the image, and the contamination types include: dust, which is manifested as evenly distributed fine particles with lighter color and stronger reflected light; stains, which are manifested as local dark spots or stripes, which may be caused by bird droppings, leaves, oil stains, etc. Damage: including cracks, scratches, breakages, etc., which are manifested as irregular shapes and obvious edges. A confidence threshold is set for each classification result, and only classification results with high confidence are adopted. Results below the threshold can be marked as "uncertain" and require further manual confirmation. The third photovoltaic panel image data is segmented and marked, and according to the segmentation result, the proportion of the area of ​​the dirty area to the total area of ​​the photovoltaic panel is calculated, that is, the weighted sum of the weights of the number of pixels of dust, stains and damage to the total number of pixels of the photovoltaic panel is used to obtain the degree of contamination of the photovoltaic panel, wherein the weight coefficient of each type of contamination can be set according to the actual application scenario or through model training.

[0041] It should be noted that by obtaining data such as sunlight, sky sunlight, snow coverage and dirtiness of photovoltaic panels through visual recognition technology, richer and more accurate information can be provided for photovoltaic power generation prediction. At the same time, if the snow coverage and dirtiness of the photovoltaic panels exceed the preset thresholds, the automatic cleaning device can be started to clean the surface of the photovoltaic panels. After cleaning, the photovoltaic panel image data can be collected again to obtain the latest snow coverage and dirtiness of the photovoltaic panels. This embodiment can not only improve the power generation efficiency, but also optimize the maintenance and management of the system.

[0042] In a preferred embodiment, calculating the impact correction factor data includes: The sky illuminance in each time period of the first reference day is compared with the illuminance corresponding to the total radiation in the same time period in the weather forecast of the first reference day. If the average value of the difference between the two is greater than the change threshold, the sky illuminance error is calculated by the difference between the two to obtain the first correction factor. If the average value of the difference between the two is less than the change threshold, the first correction factor is 1. The change threshold can be set according to the actual application scenario. For example, sky light error It is expressed as: ; in, represents the number of samples, Indicates the sky illumination in each period of the weather forecast. Indicates the sky illuminance in the area at each time period on the first reference day; Then, calculate the first correction factor , represents the impact of sky illumination error on power generation, the formula is as follows: ; in, Indicates the maximum allowed sky light illumination error.

[0043] The light intensity of the photovoltaic panel in each time period of the first reference day is compared with the sky light intensity in each time period, and the light intensity error of the photovoltaic panel is calculated by the difference between the two to obtain the second correction factor; Exemplarily, the photovoltaic panel light intensity error is expressed as: ; in, represents the number of samples, It represents the sky illuminance in each period of the first reference day. Indicates the light intensity of photovoltaic panels in the area at each time period on the first reference day; Then, calculate the second correction factor , which represents the impact of the photovoltaic panel light intensity error on the power generation, the formula is as follows: ; in, Indicates the maximum permissible photovoltaic panel light intensity error.

[0044] Based on the snow coverage of photovoltaic panels, the first impact factor is calculated; Exemplary, snow coverage of photovoltaic panels obtained by visual recognition technology , takes values ​​[0,1], represents the coverage ratio, and defines the first impact factor , represents the impact of snow cover on power generation efficiency. Snow cover will significantly reduce power generation efficiency. Therefore is a decreasing function, as decreases with the increase of , which can be expressed as: ; in, is a constant that indicates the degree of influence of snow cover on power generation efficiency.

[0045] Based on the degree of contamination of the photovoltaic panels, the second influencing factor is calculated; For example, the degree of contamination of photovoltaic panels obtained by visual recognition technology , takes values ​​of [0,1], represents the dirtiness ratio, and defines the second influencing factor , represents the impact of dirt on power generation efficiency. Dirt will reduce the transmittance of photovoltaic panels, thereby reducing power generation, which is expressed as: ; in, and is a constant that indicates the degree of impact of dirt on power generation efficiency.

[0046] The first correction factor, the second correction factor, the first impact factor, and the second impact factor are combined to generate impact correction factor data.

[0047] In a preferred embodiment, constructing a composite prediction model to obtain second prediction data of photovoltaic power generation includes: Based on the first prediction data and the influencing factor data, a composite prediction model is constructed to obtain the revised second prediction data. The composite prediction model is expressed as: ; in, represents the first forecast data of photovoltaic power generation, represents the first correction factor, represents the second correction factor, represents the first impact factor, represents the second impact factor, Represents the dynamic adjustment factor.

[0048] In order to further improve the accuracy of the prediction, a dynamic adjustment factor is introduced , this factor can be adjusted based on historical data trends, seasonal changes, weather patterns and other factors. The dynamic adjustment factor calculation formula is as follows: ; in, Indicates the actual power generation in the recent period. It represents the predicted power generation in the recent period. The dynamic adjustment factor reflects the deviation of the model and is used to correct the future forecast value.

[0049] It should be noted that this embodiment combines the first prediction data (preliminary prediction based on weather forecast and equipment data) with the impact correction factor data (local environmental information based on visual recognition) through visual recognition and meteorological forecast data to construct a composite prediction model, which can more comprehensively capture the various factors affecting photovoltaic power generation, especially the combination of local information (such as occlusion, dirtiness) and macro weather conditions, which can significantly improve the accuracy of the prediction and is suitable for photovoltaic power stations in complex environments. The introduction of the impact correction factor data enables the model to better cope with extreme weather or emergencies (such as storms, heavy snow, equipment failures, etc.). Even if there is uncertainty in the weather forecast data, the model can still be dynamically adjusted through local environmental information to ensure the stability of the prediction results, which can effectively reduce the prediction deviation caused by local environmental changes and be closer to the actual power generation, thereby improving the accuracy of the prediction.

[0050] The present invention performs time series analysis on the historical power generation data of the photovoltaic power station, and uses a machine learning algorithm (such as LSTM) to predict future power generation trends. Time series analysis and machine learning techniques are used in the present invention to extract patterns from historical data and make predictions. A recurrent neural network is used to predict the power generation of the photovoltaic station through the historical power generation data and weather data of the photovoltaic station. A camera or other image acquisition device is used to obtain environmental images around the photovoltaic power station, and these images are analyzed. Image processing and computer vision techniques are used in the present invention to identify factors that affect power generation, such as cloud distribution and changes in light intensity. A composite prediction model is constructed by combining visual recognition results with historical power data. Data fusion and integrated learning techniques are used in the present invention to integrate multi-source data, improve the accuracy and robustness of the prediction model, and help power system operators to perform power dispatch and optimization. Power system dispatch and optimization techniques are used in the present invention to adjust power generation plans according to the prediction results, and improve the stability and efficiency of the power grid.

[0051] The present invention provides a comprehensive data foundation for the model by acquiring multi-source data, thereby enhancing the accuracy of the prediction. The model is trained through historical power generation data, and preliminary predictions are made in combination with weather forecasts and equipment data. Environmental data of photovoltaic panels are collected through visual recognition technology, key features are extracted and impact correction factors are calculated, and the impact of local environmental changes on power generation is captured. The preliminary prediction results are combined with the impact correction factors to construct a composite prediction model to obtain a more accurate final power generation prediction, which not only improves the accuracy of the prediction, but also enhances the robustness and adaptability of the model, and can better cope with complex and changeable environmental conditions.

[0052] The above is a schematic scheme of a composite photovoltaic power generation prediction method based on visual recognition in this embodiment. It should be noted that the technical scheme of the composite photovoltaic power generation prediction system based on visual recognition and the technical scheme of the composite photovoltaic power generation prediction method based on visual recognition above belong to the same concept. For the details not described in detail in the technical scheme of the composite photovoltaic power generation prediction system based on visual recognition in this embodiment, please refer to the description of the technical scheme of the composite photovoltaic power generation prediction method based on visual recognition above.

[0053] The composite photovoltaic power generation prediction system based on visual recognition in this embodiment includes: The acquisition module is used to obtain equipment data, weather forecast data and historical power generation data of photovoltaic stations; A construction module is used to construct a photovoltaic power generation prediction model, train the photovoltaic power generation prediction model through historical power generation data, take weather forecast data and equipment data as input, and obtain first prediction data of photovoltaic power generation; The data processing module is used to collect the environmental data of photovoltaic panels before the forecast date of the photovoltaic station, extract key features, and calculate the impact correction factor data; The output module is used to construct a composite prediction model based on the first prediction data and the influencing factor data to obtain the second prediction data of photovoltaic power generation as the final photovoltaic power generation prediction power.

[0054] This embodiment also provides an electronic device, which is applicable to the case of composite photovoltaic power generation prediction based on visual recognition, including: Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the composite photovoltaic power generation prediction method based on visual recognition as proposed in the above embodiment.

[0055] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the composite photovoltaic power generation prediction method based on visual recognition as proposed in the above embodiment is implemented.

[0056] The storage medium proposed in this embodiment and the composite photovoltaic power generation prediction method based on visual recognition proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0057] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0058] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A composite photovoltaic power generation prediction method based on visual recognition, characterized in that: include: Obtain equipment data, weather forecast data, and historical power generation data of photovoltaic stations; Constructing a photovoltaic power generation prediction model, training the photovoltaic power generation prediction model with the historical power generation data, taking the weather forecast data and equipment data of the prediction day as input, and obtaining first prediction data of photovoltaic power generation; Collect environmental data of photovoltaic panels before the forecast date of the photovoltaic station, extract key features, and calculate the impact correction factor data; Based on the first prediction data and the influencing factor data, a composite prediction model is constructed to obtain second prediction data of photovoltaic power generation as the final photovoltaic power generation prediction power.

2. The composite photovoltaic power generation prediction method based on visual recognition according to claim 1, characterized in that: Obtaining weather forecast data and historical power generation data for photovoltaic stations includes: Acquiring equipment data, wherein the equipment data includes photovoltaic panel location, rated power, conversion efficiency, tilt angle, and azimuth; Obtaining weather forecast data for the forecast day, the weather forecast data for the forecast day including time, weather type, total radiation, direct radiation, scattered radiation, reflected radiation, temperature, humidity, and cloud coverage; The historical power generation data for the past one to three years is obtained, wherein the historical power generation data includes actual weather data, equipment data, and actual power generation data of the site.

3. The composite photovoltaic power generation prediction method based on visual recognition according to claim 2, characterized in that: The first prediction data of photovoltaic power generation includes: Preprocessing the equipment data, weather forecast data and historical power generation data, sorting the historical power generation data in time series, and classifying them according to time periodicity based on data date; Dividing the historical power generation data into different training sets, validation sets and test sets according to time period types and weather types; A photovoltaic power generation prediction model is constructed using a long short-term memory network, and an attention mechanism is introduced. The photovoltaic power generation prediction model is trained using the historical power generation data, and the equipment data and weather forecast data are input into the trained photovoltaic power generation prediction model to obtain the first prediction data of the prediction day.

4. The composite photovoltaic power generation prediction method based on visual recognition according to claim 1 or 3, characterized in that: Environmental data collected before the forecast date for photovoltaic sites include: A plurality of camera points are set around the photovoltaic panels to regularly collect image data around the photovoltaic panels of the photovoltaic station, and environmental data are extracted from the image data through visual recognition, wherein the environmental data includes sky sunshine data, photovoltaic panel sunshine data, photovoltaic panel heavy snow data, and photovoltaic panel dirt data; According to the weather type of the forecast day, a day closest to the forecast day and with the same weather type is determined as the first reference day, and key features of the first reference day are extracted, including photovoltaic panel sunshine light data and sky sunshine light data; The day before the predicted day is taken as the second reference day, and key features of the second reference day are extracted, including photovoltaic panel heavy snow data and photovoltaic panel dirt data.

5. The composite photovoltaic power generation prediction method based on visual recognition according to claim 4, characterized in that: The key features extracted include: The sky sunshine light data represents sky illuminance, collects sky images above the photovoltaic panel during the first reference day in each time period, divides the sky images into multiple sub-areas, identifies the sky illuminance of the sub-areas in each time period, and calculates the sky illuminance in the area in each time period; The photovoltaic panel sunshine light data represents the light intensity of the photovoltaic panel, collecting the first photovoltaic panel image data of the first reference day in each time period, converting the first photovoltaic panel image data into a photovoltaic panel grayscale image, segmenting the photovoltaic panel grayscale image, marking each complete photovoltaic panel area, identifying the grayscale value of each photovoltaic panel area, calculating the average grayscale value of the photovoltaic panel, and obtaining the light intensity of the photovoltaic panel in the area in each time period; The photovoltaic panel heavy snow data represents the heavy snow coverage of the photovoltaic panel, collects the latest second photovoltaic panel image data of the second reference day, performs image preprocessing and image segmentation on the second photovoltaic panel image data, converts the segmented image into a binary image, identifies the snow-covered area, counts the number of white pixels in the binary image, and calculates the heavy snow coverage of the photovoltaic panel; The photovoltaic panel contamination data represents the degree of contamination of the photovoltaic panel. The latest third photovoltaic panel image data of the second reference day is collected, and the third photovoltaic panel image data is subjected to image preprocessing and image segmentation. The pixel quantity of different types of contamination in the third photovoltaic panel image data is identified, and the contamination types include dust, stains and damage. The degree of contamination of the photovoltaic panel is obtained based on the weighted sum of the pixel quantity of dust, stains and damage and the total pixel quantity of the photovoltaic panel.

6. The composite photovoltaic power generation prediction method based on visual recognition according to claim 5, characterized in that: The calculated impact correction factor data include: Compare the sky illuminance in each time period of the first reference day with the illuminance corresponding to the total radiation in the same time period in the weather forecast of the first reference day. If the average value of the difference between the two is greater than the change threshold, calculate the sky illuminance error by the difference between the two to obtain a first correction factor. Compare the light intensity of the photovoltaic panel in each time period of the first reference day with the sky light intensity in each time period, and calculate the light intensity error of the photovoltaic panel by the difference between the two to obtain a second correction factor; Based on the snow coverage of the photovoltaic panel, a first influencing factor is calculated; Based on the degree of contamination of the photovoltaic panel, a second influencing factor is calculated; The first correction factor, the second correction factor, the first impact factor and the second impact factor are combined to generate impact correction factor data.

7. The composite photovoltaic power generation prediction method based on visual recognition according to claim 6, characterized in that: The second prediction data of photovoltaic power generation includes: Based on the first prediction data and the influencing factor data, a composite prediction model is constructed to obtain the revised second prediction data. The composite prediction model is expressed as: ； in, represents the first forecast data of photovoltaic power generation, represents the first correction factor, represents the second correction factor, represents the first impact factor, represents the second impact factor, Represents the dynamic adjustment factor.

8. A composite photovoltaic power generation prediction system based on visual recognition, used in the composite photovoltaic power generation prediction method based on visual recognition as claimed in any of claims 1 to 7, characterized in that: include, The acquisition module is used to obtain equipment data, weather forecast data and historical power generation data of photovoltaic stations; A construction module is used to construct a photovoltaic power generation prediction model, train the photovoltaic power generation prediction model through the historical power generation data, take the weather forecast data and the equipment data as input, and obtain the first prediction data of photovoltaic power generation; The data processing module is used to collect the environmental data of photovoltaic panels before the forecast date of the photovoltaic station, extract key features, and calculate the impact correction factor data; The output module is used to construct a composite prediction model based on the first prediction data and the influencing factor data to obtain the second prediction data of photovoltaic power generation as the final photovoltaic power generation prediction power.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the composite photovoltaic power generation prediction method based on visual recognition according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the composite photovoltaic power generation prediction method based on visual recognition as described in any one of claims 1 to 7.

Images