Roof photovoltaic potential detection and short-term power prediction method and system

Abstract

The invention discloses a roof photovoltaic potential detection and short-term power prediction method and system, and relates to the technical field of photovoltaic system simulation, and the method comprises the steps: collecting the data of a satellite image, and carrying out the preprocessing of the data; and training a semantic segmentation model, and adjusting hyper-parameters to optimize model performance. And inputting the processed data into a semantic segmentation model to obtain a building roof image, and calculating the total number of pixel points in the contour. And calculating the actual area of the building roof according to the satellite remote sensing image, and correcting the data. And estimating the photovoltaic available area and the photovoltaic generating capacity in combination with multi-source data. According to the method provided by the invention, remarkable effects are achieved in a plurality of aspects of photovoltaic potential evaluation precision, short-term power prediction accuracy, data processing efficiency, model adaptability and the like, and better effects are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of photovoltaic system simulation, and in particular to a method and system for rooftop photovoltaic potential detection and short-term power prediction. Background Art

[0002] As global energy and environmental issues become increasingly severe, photovoltaic power generation, as a clean and efficient form of energy, has received widespread attention. Photovoltaic power generation can not only reduce dependence on fossil fuels, but also significantly reduce greenhouse gas emissions, helping to achieve the goal of "carbon peak and carbon neutrality". However, in the actual promotion process, due to the lack of effective rooftop photovoltaic potential assessment and short-term power prediction methods, many projects face problems such as blind investment and waste of resources during implementation. Therefore, it is particularly important to develop a scientific and accurate method for rooftop photovoltaic potential detection and short-term power prediction.

[0003] The traditional prediction method obtains the power curve from the power curve database of the rooftop distributed photovoltaics in the whole county, and calculates the similarity between each rooftop distributed photovoltaic power curve and each type of cluster center curve according to the comprehensive similarity measurement formula of Euclidean distance and slope distance, and then calculates and adjusts the cluster centers of each type of cluster center curve category group according to the k-mediod algorithm to determine the distributed photovoltaic designated areas and the central power curves of each zone within the whole county. There are the following shortcomings: It is impossible to realize the detection of rooftop photovoltaic potential to assist planning and design. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the existing rooftop photovoltaic potential assessment and short-term power prediction methods have low assessment accuracy, low data processing efficiency, and imperfect prediction models, and how to collect and preprocess satellite image data, train and optimize semantic segmentation models, and combine the actual configuration of photovoltaic components and meteorological data to estimate power generation, effectively solving the shortcomings of the existing technology in photovoltaic potential assessment accuracy, data processing efficiency and short-term power prediction accuracy. The problem of image segmentation and contour detection is carried out by introducing the OpenCV module, and verification is carried out in combination with PVsyst.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a method for detecting rooftop photovoltaic potential and predicting short-term power, comprising collecting satellite image data and preprocessing the data;

[0007] Train the semantic segmentation model and adjust hyperparameters to optimize model performance;

[0008] The processed data is fed into the semantic segmentation model to obtain the building roof image and calculate the total number of pixels in the outline;

[0009] Calculate the actual area of ​​the building's roof based on satellite remote sensing images and correct the data;

[0010] The available photovoltaic area and photovoltaic power generation are estimated by combining multi-source data.

[0011] As a preferred solution of the rooftop photovoltaic potential detection and short-term power prediction method described in the present invention, the acquisition preprocessing process includes data source acquisition, data format conversion, image optimization, data cropping, data labeling, data optimization, data set segmentation, and label normalization.

[0012] As a preferred solution of the rooftop photovoltaic potential detection and short-term power prediction method described in the present invention, the training and optimization of the semantic segmentation model includes selecting a basic model and setting various hyperparameters of the model according to specific needs, pre-training weight migration, freezing training and unfreezing training, model training and tuning according to needs, and evaluating and comparing the models.

[0013] As a preferred solution of the rooftop photovoltaic potential detection and short-term power prediction method described in the present invention, the building roof image extraction process includes inputting the image into a semantic segmentation model, segmenting and binarizing the input image, and performing contour detection and area calculation on the segmented and binarized image.

[0014] As a preferred solution of the rooftop photovoltaic potential detection and short-term power prediction method described in the present invention, the roof area estimation includes writing a python script, using the OpenCV module to perform contour detection on multi-angle binary segmentation images, calculating the number of pixels in the contour, and then converting it into the actual area of ​​the building roof according to the surface resolution of the satellite remote sensing image, estimating the roof area of ​​the building within each contour.

[0015] As a preferred solution of the rooftop photovoltaic potential detection and short-term power prediction method described in the present invention, the estimation of the photovoltaic usable area includes calculating the photovoltaic usable area of ​​the building roof based on the model, layout and inclination factors of the photovoltaic components.

[0016] As a preferred solution of the rooftop photovoltaic potential detection and short-term power prediction method described in the present invention, the estimation of photovoltaic power generation includes calculating the footprint of photovoltaic modules and performing shading analysis based on the specifications and installation angles of different photovoltaic modules in combination with the local longitude and latitude, so as to ensure that the estimation result is more accurate.

[0017] Another object of the present invention is to provide a rooftop photovoltaic potential detection and short-term power prediction system, which can collect and pre-process satellite image data, train semantic segmentation models and perform contour detection, and combine the actual configuration, layout and meteorological data of photovoltaic components to perform power prediction, thereby solving the problems of low assessment accuracy, poor data processing efficiency and inaccurate photovoltaic power generation prediction in current rooftop photovoltaic potential assessment and short-term power prediction technologies.

[0018] As a preferred solution of the rooftop photovoltaic potential detection and short-term power prediction system described in the present invention, it includes: a data acquisition and processing module, a model training and optimization module, an image extraction module, and a model evaluation module. The data acquisition and processing module is used to collect and pre-process satellite image data.

[0019] The model training optimization module is used to train the model and adjust hyperparameters to optimize model performance.

[0020] The image extraction module is used to input the preprocessed satellite image into the trained semantic segmentation model, the model outputs the binary segmentation image of the building roof, performs contour detection on the segmentation image, and calculates the total number of pixels in the contour.

[0021] The model evaluation module is used to take into account the tilt angle of satellite images, correct the roof area, calculate the photovoltaic usable area of ​​the building roof according to the model, layout and tilt factor of the photovoltaic modules, calculate the average received radiation energy of each photovoltaic module in combination with meteorological data, and use the formula to calculate the photovoltaic power generation, and use PVsyst for verification and optimization.

[0022] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a method for rooftop photovoltaic potential detection and short-term power prediction.

[0023] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for detecting rooftop photovoltaic potential and predicting short-term power.

[0024] Beneficial effects of the present invention: The method for detecting the potential of rooftop photovoltaic power and predicting short-term power provided by the present invention, the data acquisition and preprocessing scheme ensures data accuracy and reliability through high-quality data acquisition and preprocessing, provides a solid foundation, and provides accurate data for subsequent image segmentation and analysis. The semantic segmentation model training and optimization scheme improves the accuracy and robustness of the model by optimizing the hyperparameters of the semantic segmentation model, can accurately segment the roof image, and improve the accuracy of photovoltaic potential detection. The roof image extraction and contour detection scheme extracts the roof image and performs contour detection through the semantic segmentation model, accurately calculates the available area of ​​the roof, and avoids the errors in the traditional method. The roof area estimation scheme uses the OpenCV module to accurately calculate the roof area, avoids the errors in area estimation in the traditional method, and provides real roof area data. The photovoltaic available area estimation scheme combines the photovoltaic module configuration and environmental factors to accurately estimate the photovoltaic available area and optimize the design of the photovoltaic system. The photovoltaic power generation estimation scheme accurately predicts the power generation of photovoltaic modules by combining meteorological data, photovoltaic module specifications and installation angles, and improves the accuracy of short-term power prediction. The model evaluation and optimization scheme uses tools such as PVsyst for optimization and verification, improves the actual application effect of the model, and ensures the accuracy of the prediction. The present invention has achieved remarkable results in many aspects, such as the accuracy of photovoltaic potential assessment, the accuracy of short-term power prediction, the data processing efficiency and the adaptability of the model, and has achieved better results. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] 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 work.

[0026] Figure 1 An overall flow chart of a method for rooftop photovoltaic potential detection and short-term power prediction provided in the first embodiment of the present invention.

[0027] Figure 2 A confusion matrix diagram of a method for rooftop photovoltaic potential detection and short-term power prediction provided by the first embodiment of the present invention.

[0028] Figure 3 An extraction flow chart of a rooftop photovoltaic potential detection and short-term power prediction method provided for the first embodiment of the present invention. DETAILED DESCRIPTION

[0029] 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.

[0030] Example 1, reference Figure 1-3 , which is an embodiment of the present invention, provides a method for detecting rooftop photovoltaic potential and predicting short-term power, comprising:

[0031] S1: Collect satellite image data and preprocess the data.

[0032] Download high-resolution multi-spectral, multi-angle, and multi-temporal images of the target area from the authoritative satellite remote sensing image database, convert the original satellite image data into a unified format suitable for deep learning model processing, perform geometric correction and radiation correction on the image, and crop the sub-image containing the roof from the original image according to the position of the building roof. Mark the position of the building roof on the satellite image to generate the corresponding training data set, optimize the image data by rotation, scaling, flipping, and noise addition, divide the labeled data set into training set, validation set, and test set in proportion, convert all labeled images into grayscale value representation, and denoise the labeled images.

[0033] S2: Train the semantic segmentation model and adjust the hyperparameters to optimize the model performance.

[0034] PSPNet and DeepLabV3+ are selected as the basic semantic segmentation model framework. PSPNet uses a deep residual network as the backbone feature extraction network, and DeepLabV3+ uses Xception or MobileNetV2 as the backbone feature extraction network. The hyperparameters of the model are set according to specific needs, including input image size, classification categories and number, number of iterations and batch size, optimizer and learning rate, sampling factor, and pre-trained weight files are imported. The existing knowledge and experience are used to accelerate the model convergence process. The training is first frozen and then gradually unfrozen. The model parameters are gradually adjusted to adapt to the current task. The loss value changes are monitored in real time during the training process. The hyperparameters and optimizer strategies are adjusted to ensure that the model converges quickly and achieves the best performance. The confusion matrix and other evaluation indicators are used to compare and analyze the performance of different models.

[0035] S3: Input the processed data into the semantic segmentation model to obtain the building roof image and calculate the total number of pixels in the outline.

[0036] When evaluating a semantic segmentation model, the accuracy of feature extraction is judged by comparing the classification results of the predicted output with the ground truth at the pixel level using a confusion matrix.

[0037] Further, for the confusion matrix of the binary classification problem, TP is the number of pixels of the correctly extracted positive examples, TN is the number of pixels of the correctly extracted negative examples, FP is the number of pixels of the incorrectly extracted negative examples as positive examples, and FN is the number of pixels of the incorrectly extracted positive examples as negative examples. After obtaining the confusion matrix, the following evaluation parameters are used for evaluation. Pixel accuracy refers to the overlap ratio of the pixels predicted by the standard answer and the pixels predicted by the segmentation model. The higher the accuracy of the model, the better the effect of the model. PixelAccuracy=(TP+TN) / (TP+FN+FP+TN), and the average accuracy is the average value of the pixel accuracy. The calculation method is the sum of the pixel accuracy of all categories / the number of categories. mPA=(PA background+PA building roof) / 2, and the intersection-over-union ratio is calculated for each category, that is, the average value of the ratio of the intersection area of ​​the predicted pixel and the standard pixel to their union area. mIOU is one of the more mainstream evaluation indicators in semantic segmentation evaluation. IOU=TP / (TP+FP+FN), mIOU=(IOU background+IOU building roof) / 2. The F1 score is the harmonic mean of precision and recall, that is, the sum of precision and recall divided by 2. Precision, also known as the call rate, indicates the proportion of samples that are truly positive among the samples identified as positive by the model. Recall, also known as the recall rate, indicates the ratio of the number of samples correctly identified as positive by the model to the total number of positive samples. The F1 score is an indicator used in statistics to measure the accuracy of a binary classification model. The value of F1-Score ranges from 0 to 1, with 1 being the best and 0 being the worst. The pixel-level F1 score is the average of the F1 scores of all pixels. Precision = TP / (TP+FP), Recall = TP / (TP+FN), F1 score = 2×(Precision×Recall) / (Precision+Recall) When evaluating a semantic segmentation model, the corresponding evaluation parameters should be selected according to the different requirements of the task and dataset to accurately evaluate the performance and advantages and disadvantages of the model. The preprocessed satellite image data is input into the trained semantic segmentation model, and the model outputs a binary segmentation image of the building roof, accurately identifying the roof area.

[0038] S4: Calculate the actual area of ​​the building roof based on satellite remote sensing images and correct the data.

[0039] The OpenCV module is used to perform contour detection on binary segmentation images at multiple angles. The total number of pixels in the contour is calculated and converted into the actual area of ​​the building roof according to the surface resolution of the satellite remote sensing image. The roof area is corrected considering the inclination angle of the satellite image to ensure the accuracy of the estimation result.

[0040] Furthermore, the roof image of the building is obtained through satellite images, and semantic segmentation is performed to obtain a binary image. In the image, the roof part is white with a value of 1, and the background is black with a value of 0. The image is converted into a binary image using a threshold algorithm. This facilitates subsequent contour detection. The cv2.findContours() function in OpenCV is used to detect contours in binary images. This function returns the contour point set and hierarchy in the image. The contour is the boundary of the roof area, which can be used to calculate the area after extraction. The pixel area of ​​each contour can be calculated through the cv2.contourArea() function. These area values ​​are the number of pixels of the contour in the image, reflecting the relative size of the roof area. If the roof has multiple contours, such as segments or different structural parts, all contours can be traversed, their pixel areas can be accumulated, and the total roof area can be obtained. Satellite images usually have a resolution, and each pixel corresponds to a certain actual distance on the ground. To convert the pixel area of ​​the image into the actual area, the resolution must be known. Satellite images may be tilted, and the tilt angle will cause a deviation between the measured area and the actual ground area. Correction can be calculated based on the tilt angle θ. The corrected area is the actual PV usable area of ​​the building roof, which can be used for subsequent PV power generation estimation and optimization, and the results can be displayed through logs or images.

[0041] S5: Estimate the available photovoltaic area and photovoltaic power generation based on multi-source data.

[0042] Estimate the roof area based on the area corresponding to the pixel points. Write a python script, use the OpenCV module to perform contour detection on the binary segmentation images from multiple angles, calculate the number of pixels in the contour, convert it into the actual area of ​​the building roof based on the surface resolution of the satellite remote sensing image, and estimate the building roof area within each contour.

[0043] S=N×PPI 2

[0044] S is the roof area of ​​the building extracted by the algorithm, N is the number of pixels in the extracted building roof image, and PPI is the surface resolution of the satellite remote sensing image, which indicates how much space each pixel represents.

[0045] Furthermore, the estimation of the PV usable area is carried out based on factors such as the model, layout and inclination of the PV modules, as well as the shading effect.

[0046] S r =S×f0×f1×f2

[0047] Sr is the available photovoltaic area on the roof of the building in the area to be evaluated, S is the calculated roof area of ​​the building, f0 is the roof inclination coefficient of the building, f1 is the shading coefficient of the building roof, and f2 is the available area coefficient of the building roof.

[0048] Furthermore, the estimation of photovoltaic power generation specifically includes parameter calculation method and PVsyst software analysis method, which calculates the footprint of photovoltaic modules based on the specifications and installation angles of different photovoltaic modules, combined with the local longitude and latitude. The optimal installation angle of photovoltaic modules needs to be determined in combination with the solar energy resource conditions of different cities and the azimuth and altitude angles of the sun.

[0049]

[0050] S pv is the floor area of ​​the photovoltaic module, h is the width of the photovoltaic module, l is the length of the photovoltaic module, β is the optimal installation angle, is the latitude.

[0051] Obtain the light data of the area to be estimated to calculate the average radiation energy received by each photovoltaic module, which is generally measured in kilowatt-hours. It can be obtained from meteorological stations, meteorological departments, local governments, the National Meteorological Center and other channels, and combined with the radiation of photovoltaic modules to predict photovoltaic power generation.

[0052]

[0053] Where E p is the total development potential of photovoltaics, G t is the average radiation received by each photovoltaic module, α is the conversion efficiency of the photovoltaic module, and μ is the operating efficiency of the photovoltaic system.

[0054] It should be noted that, considering the shading problem in the city, 3D modeling is performed on the objects that cause shading of photovoltaic panels. The shading area is determined through the model, and the formula is used to quantify the shading effect and calculate the actual power generation.

[0055]

[0056] Among them, I 实际 It represents the radiation intensity after considering the shielding, I 无遮挡 Indicates the radiation intensity without shielding, F 遮挡 Represents the occlusion factor.

[0057] P 发电 =P 最大 ×F遮挡

[0058] Among them, P 发电 Indicates the actual power generation, P 最大 Indicates the maximum power generation.

[0059] P 最大 =A 组件 ×G STC ×η

[0060] Among them, A 组件 Represents the effective area of ​​the photovoltaic module, G STC It represents the radiation intensity under standard test conditions, and η represents the conversion efficiency of photovoltaic modules.

[0061] Example 2 is an embodiment of the present invention, which provides a method for rooftop photovoltaic potential detection and short-term power prediction. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0062] First, data collection is performed to obtain satellite image data from different regions. These data include high-definition satellite images of building roofs and related remote sensing image data. The collected data come from multiple geographical locations to ensure the representativeness and wide applicability of the experimental results. After data collection, the satellite images are first preprocessed. The preprocessing includes: data format conversion, image optimization, data cropping, data annotation, and data optimization. All these processing steps use automated tools to improve efficiency and reduce human errors. The semantic segmentation model is trained using the collected preprocessed images. U-Net is selected as the base model and its hyperparameters are tuned. During the training process, the performance of the model is optimized by adjusting hyperparameters such as learning rate, batch size, and loss function. The pre-trained weights are used and fine-tuned to better adapt to the characteristics of building roofs in different regions. After training, the model can accurately separate the building roof from the surrounding environment and obtain an accurate image of the roof area. The processed satellite image data is input into the trained semantic segmentation model to obtain a binary image of the building roof. The total number of pixels in the roof part of the image is calculated by the contour detection method. The OpenCV module is used for contour detection, and the pixel area of ​​the contour is calculated by the cv2.contourArea() function. The pixel area of ​​each contour represents the proportion of the roof in the image. According to the resolution of the satellite remote sensing image, the pixel area is converted to the actual roof area of ​​the building. Since satellite images usually have a known resolution, the actual area of ​​the roof can be obtained by multiplying the pixel area and the resolution. Since satellite images have a certain tilt angle, the roof area is corrected for the tilt angle. The correction adjusts the area calculation result through the relationship between the tilt angle and the image acquisition angle to ensure the accuracy of the estimated area. Combined with the photovoltaic module parameters of the building and real-time meteorological data, the photovoltaic available area and short-term photovoltaic power generation of the roof are estimated. Specifically, by combining the performance curve of the photovoltaic module and the local meteorological data, the power generation of different photovoltaic panels is estimated, and the existing PVsyst software is used for further verification and optimization.

[0063] Table 1 Experimental data table

[0064]

[0065] The data in the table show that higher-resolution satellite images can more accurately capture the roof features of buildings, thereby reducing errors in the conversion between pixel area and actual area. For example, building A uses a satellite image with a resolution of 0.5 meters / pixel, and its actual roof area after conversion is 6,000 square meters, while building B uses a resolution of 1 meter / pixel, and its actual roof area after conversion is 15,000 square meters, resulting in a large difference in area. The tilt angle of building A is 10°, and the roof area after correction increases from 6,000 square meters to 6,200 square meters. The tilt angle of building B is larger, and its corrected area increases from 15,000 square meters to 15,150 square meters. The accuracy of roof area estimation is improved by correcting the tilt angle. In building A, the photovoltaic usable area is 70%, while the photovoltaic usable area of ​​building D is higher, reaching 80%. These data show the photovoltaic utilization potential of building roofs under different layouts and designs. The annual power generation of building A is 1,000 kWh, and that of building D is 2,100 kWh. By combining the performance of PV panels and the actual installation angle, power generation estimation is more accurate than traditional methods and can provide users with reliable PV power generation expectations.

[0066] Embodiment 3 is an embodiment of the present invention, which provides a rooftop photovoltaic potential detection and short-term power prediction system, including a data acquisition and processing module, a model training and optimization module, an image extraction module, and a model evaluation module.

[0067] The data acquisition and processing module is used to acquire and pre-process satellite image data.

[0068] The model training optimization module is used to train the model and adjust hyperparameters to optimize model performance.

[0069] The image extraction module is used to input the preprocessed satellite image into the trained semantic segmentation model, the model outputs the binary segmentation image of the building roof, performs contour detection on the segmentation image, and calculates the total number of pixels in the contour.

[0070] The model evaluation module is used to take into account the tilt angle of satellite images, correct the roof area, calculate the photovoltaic usable area of ​​the building roof according to the model, layout and tilt factor of the photovoltaic modules, calculate the average received radiation energy of each photovoltaic module in combination with meteorological data, and use the formula to calculate the photovoltaic power generation, and use PVsyst for verification and optimization.

[0071] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0072] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0073] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0074] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0075] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. 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 method for detecting rooftop photovoltaic potential and predicting short-term power, characterized in that: include: Collect satellite image data and pre-process the data; Train the semantic segmentation model and adjust hyperparameters to optimize model performance; The processed data is fed into the semantic segmentation model to obtain the building roof image and calculate the total number of pixels in the outline; Calculate the actual area of ​​the building's roof based on satellite remote sensing images and correct the data; The available photovoltaic area and photovoltaic power generation are estimated by combining multi-source data.

2. The method for detecting rooftop photovoltaic potential and predicting short-term power according to claim 1, characterized in that: The acquisition preprocessing process includes data source acquisition, data format conversion, image optimization, data cropping, data labeling, data optimization, data set segmentation, and label normalization.

3. The method for detecting rooftop photovoltaic potential and predicting short-term power according to claim 2, characterized in that: The training and optimization of the semantic segmentation model includes selecting a basic model and setting various hyperparameters of the model according to specific needs, pre-training weight migration according to needs, freezing training and unfreezing training, model training and tuning, and evaluating and comparing the models.

4. The method for detecting rooftop photovoltaic potential and predicting short-term power according to claim 3, characterized in that: The building roof image extraction process includes inputting an image into a semantic segmentation model, segmenting and binarizing the input image, and performing contour detection and area calculation on the segmented and binarized image.

5. The method for detecting rooftop photovoltaic potential and predicting short-term power according to claim 4, characterized in that: The roof area estimation includes writing a python script, using the OpenCV module to perform contour detection on multi-angle binary segmentation images, calculating the number of pixels in the contour, and then converting it into the actual area of ​​the building roof based on the surface resolution of the satellite remote sensing image to estimate the roof area of ​​the building within each contour.

6. The method for detecting rooftop photovoltaic potential and predicting short-term power according to claim 5, characterized in that: The estimation of the photovoltaic usable area includes calculating the photovoltaic usable area on the roof of the building based on the model, layout and inclination factors of the photovoltaic modules.

7. The method for detecting rooftop photovoltaic potential and predicting short-term power according to claim 6, characterized in that: The estimation of photovoltaic power generation includes calculating the footprint of photovoltaic modules and performing shading analysis based on the specifications and installation angles of different photovoltaic modules in combination with the local longitude and latitude to ensure a more accurate estimation result.

8. A system using the rooftop photovoltaic potential detection and short-term power prediction method as claimed in any one of claims 1 to 7, characterized in that: It includes data acquisition and processing module, model training and optimization module, image extraction module and model evaluation module; The data acquisition and processing module is used to collect and pre-process satellite image data; The model training optimization module is used to train the model and adjust the hyperparameters to optimize the model performance; The image extraction module is used to input the preprocessed satellite image into the trained semantic segmentation model, the model outputs a binary segmented image of the building roof, performs contour detection on the segmented image, and calculates the total number of pixels in the contour; The model evaluation module is used to take into account the tilt angle of satellite images, correct the roof area, calculate the photovoltaic usable area of ​​the building roof according to the model, layout and tilt factor of the photovoltaic modules, calculate the average received radiation energy of each photovoltaic module in combination with meteorological data, and use the formula to calculate the photovoltaic power generation, and use PVsyst for verification and optimization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the rooftop photovoltaic potential detection and short-term power prediction method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the rooftop photovoltaic potential detection and short-term power prediction method according to any one of claims 1 to 7 are implemented.

Images