Distributed roof photovoltaic processing method and device based on Bayesian optimization, electronic equipment and storage medium

Abstract

The invention discloses a distributed roof photovoltaic processing method and device based on Bayesian optimization, electronic equipment and a storage medium. Advanced data acquisition and preprocessing technologies, accurate grid division and feature extraction, comprehensive multi-dimensional evaluation index calculation and efficient Bayesian optimization modeling are combined. Through a Gaussian process model and a Bayesian optimization algorithm, accurate capacity prediction and resource allocation can be realized in a complex multi-target optimization space. The method has the advantages that multiple factors such as power generation potential, economic benefit and power grid load are comprehensively considered, and the method has dynamic adjustment capability to adapt to changes in actual operation. The invention discloses a distributed roof photovoltaic processing method and device based on Bayesian optimization, electronic equipment and a storage medium, and aims at remarkably improving the overall benefit and resource utilization efficiency of a photovoltaic system and promoting wider application of renewable energy sources.

Description

Technical Field

[0001] The invention relates to a distributed rooftop photovoltaic processing method, device, electronic equipment and storage medium based on Bayesian optimization, and belongs to the technical field of photovoltaic resource planning. Background Art

[0002] As the global demand for renewable energy continues to grow, distributed photovoltaic systems have been widely used on urban rooftops. This green energy solution not only effectively utilizes idle rooftop space, but also reduces power transmission losses and environmental impacts. However, the installed capacity forecasting and resource allocation of photovoltaic systems face challenges under existing methods. Traditional forecasting tools often fail to accurately capture the complex changes in photovoltaic power generation and fail to comprehensively consider multiple influencing factors such as weather changes and policy adjustments.

[0003] There are many difficulties in the existing methods for predicting and allocating rooftop photovoltaic capacity. First, the power generation potential of photovoltaic systems is complexly affected by multiple factors such as roof area, solar radiation intensity and weather conditions. There is a nonlinear relationship between these factors, which is difficult to fully reflect by traditional prediction methods. Secondly, the building distribution density in cities is uneven, and traditional optimization methods often cannot effectively deal with this uneven spatial distribution, resulting in unreasonable resource allocation. In addition, photovoltaic system optimization needs to consider multiple dimensions such as economic benefits, grid load and policy impact. Existing optimization methods mostly focus on a single factor and lack comprehensive evaluation capabilities. Finally, traditional linear programming and heuristic algorithms are inefficient in dealing with complex multi-objective optimization problems, and it is difficult to guarantee the global optimal result.

[0004] Therefore, a more precise and comprehensive optimization method is needed to improve forecasting accuracy and ensure reasonable allocation of resources. Summary of the invention

[0005] Purpose: In order to overcome the deficiencies in the prior art, the present invention provides a distributed rooftop photovoltaic processing method, device, electronic device and storage medium based on Bayesian optimization. To meet the above challenges, precise Bayesian optimization modeling and multi-dimensional evaluation index calculation are used to overcome the limitations of traditional methods and significantly improve the prediction accuracy and resource utilization efficiency of rooftop photovoltaic systems.

[0006] Technical solution: To solve the above technical problems, the technical solution adopted by the present invention is:

[0007] In a first aspect, a distributed rooftop photovoltaic processing method based on Bayesian optimization comprises the following steps:

[0008] Step S1: Acquire data related to the target area and the distributed rooftop photovoltaic system, and preprocess the related data to obtain preprocessed data.

[0009] Step S2: Divide the target area into several grids, input the satellite image of each grid into the neural network, obtain the mask area of ​​the building in each grid, and calculate the effective roof area of ​​each grid based on the mask area of ​​the building in each grid.

[0010] Step S3: Construct a multi-dimensional evaluation index for each grid based on the preprocessed data and the effective roof area of ​​each grid.

[0011] Step S4: construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, optimize the comprehensive benefit optimization objective function using the Gaussian process model and the Bayesian optimization algorithm, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

[0012] As a preferred solution, it also includes: Step S5: correcting the optimal photovoltaic installation capacity allocation plan for each grid in the target area according to the constraint conditions to obtain a corrected photovoltaic installation capacity allocation plan for each grid in the target area.

[0013] As a preferred solution, it also includes: Step S6: adjusting the revised photovoltaic installation capacity allocation plan of each grid in the target area according to the actual implementation situation to obtain an adjusted photovoltaic installation capacity allocation plan of each grid in the target area.

[0014] As a preferred solution, the data related to the target area and the distributed rooftop photovoltaic system include: satellite data, power grid data, economic data and sunshine data.

[0015] The satellite data adopts high-resolution satellite images. The power grid data includes but is not limited to power grid topology, substation and distribution network capacity, historical load and photovoltaic grid connection status. The economic data includes but is not limited to electricity price information, installation and operation and maintenance costs of photovoltaic systems and related subsidy policies. The sunshine data includes but is not limited to sunshine hours and solar radiation intensity.

[0016] As a preferred solution, the step S2 specifically includes:

[0017] Step s2.1: Divide the target area into several uniform or adaptive grids.

[0018] Step s2.2: Input the satellite image of each grid into the Mask R-CNN network to obtain the mask area of ​​each building in each grid. According to the mask area of ​​each building, the area of ​​each building is calculated.

[0019] The area calculation formula for each building is as follows:

[0020]

[0021] in, It is In the grid The area of ​​the building, It is In the grid The mask area of ​​the building, is each type of pixel in the mask area.

[0022] Step s2.3: According to the area of ​​each building in each grid, the effective roof area of ​​each building in each grid is calculated.

[0023] The calculation formula for the effective roof area of ​​each building in each grid is as follows:

[0024]

[0025] in, It is In the grid The effective roof area of ​​a building, It is for The geometric adjustment factor of each building.

[0026] Step s2.4: Calculate the effective roof area of ​​each grid based on the effective roof area of ​​each building in each grid.

[0027] The calculation formula for the effective roof area of ​​each grid is as follows:

[0028]

[0029] in, is the number of buildings.

[0030] As a preferred solution, the multi-dimensional evaluation index of each grid includes: economic benefit index, power grid benefit index, solar radiation condition index and photovoltaic simultaneity index.

[0031] Among them, the economic benefit index is constructed by the electricity cost value index, installed cost index and photovoltaic installation capacity.

[0032] The grid efficiency index is constructed from stability indicators and photovoltaic installation capacity.

[0033] The solar radiation condition index is constructed by the effective sunshine time index, solar radiation index and photovoltaic installation capacity.

[0034] The photovoltaic simultaneous rate index adopts the The photovoltaic synchronization rate index of each grid.

[0035] Among them, the electricity cost value index The calculation formula is as follows:

[0036]

[0037] in, For the The value of electricity bill for each grid, For the The possible power generation of a grid, The local electricity price.

[0038] In the formula, The calculation formula is as follows:

[0039]

[0040] in, For the The effective roof area of ​​the grid, For the The photovoltaic power generation potential per unit area of ​​a grid.

[0041] Installation cost index , the calculation formula is as follows:

[0042]

[0043] in, For the The installation cost of a grid, Installation cost per unit area.

[0044] Stability index The calculation formula is as follows:

[0045]

[0046] in, For the The stability of the grid, is the historical photovoltaic growth, is the residual carrying capacity of the transformer.

[0047] The effective sunshine time index , the calculation formula is as follows:

[0048]

[0049] in, For the The effective sunshine time of each grid is is the total number of hours of sunlight, The number of days in the statistical period.

[0050] The solar radiation index , the calculation formula is as follows:

[0051]

[0052] in, For the The solar radiation of the grid, is the total radiation.

[0053] The said PV synchronicity index for each grid The calculation formula is as follows:

[0054]

[0055] in, For the The actual power generation of the grid, For the The photovoltaic installed capacity of the grid.

[0056] As a preferred solution, step S4 specifically includes:

[0057] Step S4.1: Construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, where the expression of the comprehensive benefit optimization objective function is as follows:

[0058]

[0059] in, It is an indicator of economic benefits. is the grid efficiency indicator, is an indicator of solar radiation conditions. is the photovoltaic simultaneity index, is the weight coefficient of the economic benefit index, is the weight coefficient of the power grid benefit index, is the weight coefficient of the solar radiation condition index and is the weight coefficient of the photovoltaic simultaneous rate indicator, It is the photovoltaic installation capacity allocation plan for each grid in the target area.

[0060] Step S4.2: Use the Gaussian process model to construct a probabilistic proxy model for optimizing the objective function.

[0061] Step S4.3: Use the acquisition function in the Bayesian optimization algorithm to predict the next photovoltaic installation capacity allocation plan to be evaluated based on the covariance function of the probabilistic agent model and the currently known optimal value of the optimization objective function.

[0062] Step s4.4: According to the next photovoltaic installation capacity allocation plan to be evaluated, the sampling result of the optimization objective function is calculated, and the mean function and covariance function of the probability proxy model are updated according to the sampling results to obtain the updated probability proxy model.

[0063] Step s4.5: Replace the currently known optimal value of the optimization objective function in step S4.3 with the sampling result, replace the probability proxy model in step S4.3 with the updated probability proxy model, repeat steps s4.3 and s4.4 until the predetermined termination conditions are met, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

[0064] As a preferred solution, step S5 specifically includes:

[0065] Step S5.1: Obtain the total installed capacity of the target area and the maximum installable capacity of each grid.

[0066] Step S5.2: When the photovoltaic installation capacity allocation plan of the grid is greater than the maximum installable capacity of the grid, the maximum installable capacity of the grid is used as the revised photovoltaic installation capacity allocation plan of the grid.

[0067] Step S5.3: When the sum of the photovoltaic installation capacity allocation plans of each grid in the target area is greater than the total installed capacity of the target area, a revised photovoltaic installation capacity allocation plan of each grid is calculated according to the total installed capacity of the target area.

[0068] The calculation formula of the revised photovoltaic installation capacity allocation plan for each grid is as follows:

[0069]

[0070] Where: is the revised PV installation capacity allocation plan for each grid. It is The optimization objective function value of the grid, It is The optimization objective function value of the grid, is the number of grids, is the total installed capacity of the target area.

[0071] As a preferred solution, the calculation formula for the photovoltaic installation capacity allocation plan of each grid in the adjusted target area is as follows:

[0072]

[0073] In the formula, is the adjusted PV installation capacity allocation plan for each grid in the target area, It is The grid adjustment factor.

[0074] in, The calculation formula is as follows:

[0075]

[0076] In the formula, It is The actual power generation during the implementation of each grid, No. Theoretical power generation of a grid.

[0077] In the second aspect, a distributed rooftop photovoltaic processing device based on Bayesian optimization includes the following modules:

[0078] Data processing module: used for data related to the target area and the distributed rooftop photovoltaic system, and preprocesses the related data to obtain preprocessed data.

[0079] Roof area calculation module: used to divide the target area into several grids, input the satellite image of each grid into the neural network, obtain the mask area of ​​the building in each grid, and calculate the effective roof area of ​​each grid based on the mask area of ​​the building in each grid.

[0080] Evaluation index construction module: used to construct multi-dimensional evaluation indicators for each grid based on the preprocessed data and the effective roof area of ​​each grid.

[0081] Optimal solution acquisition module: used to construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, optimize the comprehensive benefit optimization objective function using the Gaussian process model and Bayesian optimization algorithm, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

[0082] In a third aspect, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements a distributed rooftop photovoltaic processing method based on Bayesian optimization as described in any one of the first aspects.

[0083] According to a fourth aspect, a computer device comprises:

[0084] Memory, used to store instructions.

[0085] A processor is used to execute the instructions so that the computer device performs the operations of a distributed rooftop photovoltaic processing method based on Bayesian optimization as described in any one of the first aspects.

[0086] Beneficial effects: The present invention provides a distributed rooftop photovoltaic processing method, device, electronic device and storage medium based on Bayesian optimization. The method combines advanced data acquisition and preprocessing technology, precise grid division and feature extraction, comprehensive multi-dimensional evaluation index calculation and efficient Bayesian optimization modeling. Through the Gaussian process model and Bayesian optimization algorithm, accurate capacity prediction and resource allocation can be achieved in a complex multi-objective optimization space. Its advantage is that it comprehensively considers multiple factors such as power generation potential, economic benefits, grid load, etc., and has dynamic adjustment capabilities to adapt to changes in actual operation. This comprehensive optimization scheme will significantly improve the overall benefits and resource utilization efficiency of the photovoltaic system, and promote the wider application of renewable energy.

[0087] Compared with the existing technology, the advantages of the present invention are as follows:

[0088] First, the present invention significantly improves the accuracy of photovoltaic installed capacity prediction by introducing the Bayesian optimization algorithm and Gaussian process model. This method can accurately capture the dynamic changes of photovoltaic installed capacity, thereby providing more reliable prediction results and overcoming the error problem in traditional prediction methods.

[0089] Secondly, the present invention is more adaptable. By comprehensively applying the Bayesian optimization algorithm, the capacity allocation strategy can be flexibly adjusted and optimized to cope with changes in different regions and demands. This flexibility ensures that the system can provide the most appropriate optimization solution when facing various policies and regional differences.

[0090] In addition, the present invention has excellent optimization capabilities. The Bayesian optimization algorithm can effectively find the optimal capacity allocation solution under complex constraints, thereby improving the efficiency and effectiveness of resource allocation. Through global search capabilities, the algorithm can optimize in multiple dimensions and multiple objectives, avoiding the limitations of traditional methods.

[0091] Finally, the method of the present invention also performs well in terms of stability. The system ensures the consistency and reliability of the prediction results through systematic data preprocessing and model training. Such stability not only enhances the practical application effect of the model, but also provides a reliable basis to support the planning and allocation decision of photovoltaic installed capacity. These advantages fully verify the feasibility and effectiveness of the method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0093] The following is a clear and complete description of the technical solutions in the examples of the present invention in conjunction with the accompanying drawings in the examples of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present invention.

[0094] The present invention will be further described below in conjunction with specific embodiments.

[0095] Embodiment 1:

[0096] This embodiment introduces a distributed rooftop photovoltaic processing method based on Bayesian optimization. Figure 1 As shown, the following steps are included:

[0097] Step S1: Collect data related to the distributed rooftop photovoltaic system in the target area, including high-resolution satellite images, power grid data, economic data and sunshine data, and preprocess these data to obtain preprocessed data. The preprocessing includes: cleaning, standardization and data fusion to ensure its accuracy and consistency.

[0098] Furthermore, the step 1 specifically includes:

[0099] Step s1.1: Obtain high-resolution satellite images of the target area as satellite data, collect historical sunshine data for the past five years as sunshine data, including sunshine hours and solar radiation intensity, and record the geographical location information of the target area. At the same time, it is also necessary to obtain power grid data, such as power grid topology, substation and distribution network capacity, historical load and photovoltaic grid connection status. In terms of economic data, it is necessary to collect data such as electricity price information, installation and operation and maintenance costs of photovoltaic systems, and related subsidy policies.

[0100] Step s1.2: Preprocess the data collected in step s1.1, specifically including:

[0101] Step s1.2.1: Clean the data, delete duplicate data records, identify and correct outliers and missing values, and ensure the accuracy and completeness of the data.

[0102] Cleaning the data collected in step s1 is a key step to ensure data quality.

[0103] Step s1.2.2: Standardize the data to ensure that all data are within a uniform scale for further analysis and processing.

[0104] The main operations include: normalizing numerical data and encoding categorical data.

[0105] Normalization: Convert numerical data to the range [0,1], the formula is:

[0106] (1)

[0107] in, represents the normalized data, represents the data to be normalized, Indicates the minimum value of the data. Indicates the maximum value of the data.

[0108] Category data encoding: Convert category data into a form suitable for machine learning algorithm processing. This invention adopts one-hot encoding. If the number of categories is , the encoding form of each category k is:

[0109] d encoded = { [ 0 , 0 , … , 1 , … , 0 ] , if it is k category [ 0 , 0 , … , 0 , … , 0 ] , if it is other categories (2)

[0110] in, Represents the encoding of category k.

[0111] Step s1.2.3: Data from different sources are fused to obtain a data set with a unified data framework for comprehensive analysis.

[0112] The main operations of data fusion include:

[0113] Spatial alignment: Align satellite imagery data with other spatial data to ensure all data uses the same coordinate system.

[0114] Data merging: Integrate data from different sources (such as satellite imagery, sunshine data, power grid data, and economic data) into a unified dataset for comprehensive analysis.

[0115] Step S2: Divide the target area into several grids, input the satellite image of each grid into the Mask RCNN network, obtain the mask area of ​​the building in each grid, and calculate the effective roof area of ​​each grid based on the mask area of ​​the building in each grid.

[0116] Furthermore, the step S2 specifically includes:

[0117] Step s2.1: Divide the target area into several uniform or adaptive grids for detailed photovoltaic system evaluation. The purpose of grid division is to divide the area into small units for subsequent analysis and calculation. Since the distribution of buildings in different locations cannot have a uniform distribution relationship, the present invention adopts an adaptive grid division method, and its division rule is to determine the size of the grid according to the density and distribution of the buildings. The grids in densely populated areas are smaller in order to capture more details; the grids in sparsely populated areas are larger to reduce the complexity of calculation. As shown in the following formula:

[0118] (3)

[0119] in, is the area of ​​each grid, is the total area of ​​the target region, is the number of grids. The number of grids in dense areas is large, and the number of grids in sparse areas is small.

[0120] Step s2.2: Input the high-resolution satellite image of each grid into the Mask R-CNN network to obtain the mask area of ​​each building in each grid. According to the mask area of ​​each building, the area of ​​each building is calculated.

[0121] Among them, Mask R-CNN is a powerful instance segmentation and semantic segmentation method that ensures the quality of input data through image preprocessing such as scaling and denoising. Then, the Mask R-CNN network performs instance segmentation on the image and outputs the mask of each building, that is, its pixel-level region identification. Subsequently, by counting the number of pixels in the mask, the actual area of ​​each building is calculated, and the area and category information are integrated into a data set for subsequent analysis and processing.

[0122] (4)

[0123] in, It is In the grid The area of ​​the building, It is In the grid The mask area of ​​the building, is each type of pixel in the mask area.

[0124] Step s2.3: According to the area of ​​each building in each grid, the effective roof area of ​​each building in each grid is calculated.

[0125] The effective roof area refers to the actual roof area that can be used to install a photovoltaic system. Considering the influence of the building's inclination and the surrounding environment on the actual available area of ​​the roof, the extracted roof area is geometrically adjusted to correct the area calculation error caused by inclination or shading.

[0126] (5)

[0127] in, It is In the grid The effective roof area of ​​a building, It is for The geometric adjustment factor of each building.

[0128] Step s2.4: Calculate the effective roof area of ​​each grid according to the effective roof area of ​​each building in each grid.

[0129] (6)

[0130] in, is the number of buildings.

[0131] Step S3: Construct a multi-dimensional evaluation index for each grid based on the preprocessed data and the effective roof area of ​​each grid.

[0132] Among them, the multi-dimensional evaluation indicators include: economic benefit indicators, grid benefit indicators, solar radiation condition indicators and photovoltaic simultaneity indicators. Multiple evaluation indicators are calculated for each grid to comprehensively evaluate the potential and benefits of the photovoltaic system.

[0133] The economic benefit indicators include: possible power generation indicators , Electricity cost value index and installation cost indicators , according to the electricity value index and installation cost indicators And the economic benefit index of photovoltaic installation capacity construction .

[0134] Among them, the possible power generation index It is calculated based on the effective roof area and photovoltaic power generation potential of each grid. The calculation formula is as follows:

[0135] (7)

[0136] in, For the The possible power generation of a grid, For the The effective roof area of ​​the grid, For the The photovoltaic power generation potential per unit area of ​​a grid.

[0137] Electricity cost value index It is calculated based on the possible power generation and local electricity price. The calculation formula is as follows:

[0138] (8)

[0139] in, For the The value of electricity bill for each grid, The local electricity price.

[0140] Installation cost index , the calculation formula is as follows:

[0141] (9)

[0142] in, For the The installation cost of a grid, Installation cost per unit area.

[0143] The grid efficiency index , including stability indicators In order to consider the historical photovoltaic growth and the remaining transformer capacity, the impact of the photovoltaic system on the grid load is evaluated according to the stability index. Calculate the grid efficiency index based on the photovoltaic installation capacity .

[0144] Stability index The calculation formula is as follows:

[0145] (10)

[0146] in, For the The stability of the grid, is the historical photovoltaic growth, is the residual carrying capacity of the transformer.

[0147] The solar radiation condition index includes: effective sunshine time index, solar radiation index, and effective sunshine time index. , Solar radiation index And the photovoltaic installation capacity is constructed to obtain the solar radiation condition index , to assess the availability of solar resources.

[0148] Among them, the effective sunshine time index , the calculation formula is as follows:

[0149] (11)

[0150] in, For the The effective sunshine time of each grid is is the total number of hours of sunlight, The number of days in the statistical period.

[0151] The solar radiation index , the calculation formula is as follows:

[0152] (12)

[0153] in, For the The solar radiation of the grid, is the total radiation.

[0154] The photovoltaic simultaneity index , including actual power generation indicators , No. PV installed capacity of grids , according to PV synchronicity index for each grid As a photovoltaic simultaneous rate indicator .

[0155] No. PV synchronicity index for each grid The calculation formula is as follows:

[0156]

[0157] in, For the The actual power generation of the grid, For the The photovoltaic installed capacity of each grid, For the The photovoltaic simultaneity index of each grid takes into account the difference between actual operating data and theoretical power generation capacity to judge the performance of the photovoltaic system under actual conditions.

[0158] Step S4: construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, optimize the comprehensive benefit optimization objective function using the Gaussian process model and Bayesian optimization algorithm, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

[0159] Furthermore, the step S4 specifically includes:

[0160] Step S4.1: Construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, where the expression of the comprehensive benefit optimization objective function is as follows:

[0161] (14)

[0162] in, It is an indicator of economic benefits. is the grid efficiency indicator, is an indicator of solar radiation conditions. is the photovoltaic simultaneity index, is the weight coefficient of the economic benefit index, is the weight coefficient of the power grid benefit index, is the weight coefficient of the solar radiation condition index and is the weight coefficient of the photovoltaic simultaneous rate indicator, It is the photovoltaic installation capacity allocation plan for each grid in the target area.

[0163] Step S4.2: Use the Gaussian process model to construct a probabilistic proxy model for optimizing the objective function.

[0164] Step S4.3: Use the acquisition function in the Bayesian optimization algorithm to predict the next photovoltaic installation capacity allocation plan to be evaluated based on the covariance function of the probabilistic agent model and the currently known optimal value of the optimization objective function.

[0165] Step s4.4: According to the next photovoltaic installation capacity allocation plan to be evaluated, the sampling result of the optimization objective function is calculated, and the mean function and covariance function of the probability proxy model are updated according to the sampling results to obtain the updated probability proxy model.

[0166] Step s4.5: Replace the currently known optimal value of the optimization objective function in step S4.3 with the sampling result, replace the probability proxy model in step S4.3 with the updated probability proxy model, repeat steps s4.3 and s4.4 until the predetermined termination conditions are met, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

[0167] Furthermore, the probabilistic proxy model for optimizing the objective function is ,in, To predict the mean function of different PV installation capacity allocation schemes, To predict the covariance function of different PV installation capacity allocation schemes.

[0168] Furthermore, the acquisition function includes: an expected improvement function, a probability improvement function and an information gain function.

[0169] Furthermore, the predetermined termination conditions include: reaching a maximum number of iterations, the optimization effect converging, and the benefit improvement being lower than a set threshold.

[0170] Step S5: modifying the optimal photovoltaic installation capacity allocation plan for each grid in the target area according to the constraint conditions to obtain a modified photovoltaic installation capacity allocation plan for each grid in the target area.

[0171] Furthermore, the step S5 specifically includes:

[0172] Obtain the total installed capacity of the target area and the maximum installable capacity of each grid.

[0173] When the photovoltaic installation capacity allocation plan of the grid is greater than the maximum installable capacity of the grid, the maximum installable capacity of the grid is used as the revised photovoltaic installation capacity allocation plan of the grid.

[0174] When the sum of the photovoltaic installation capacity allocation plans of each grid in the target area is greater than the total installed capacity of the target area, the revised photovoltaic installation capacity allocation plan of each grid is calculated based on the total installed capacity of the target area.

[0175] The calculation formula of the revised photovoltaic installation capacity allocation plan for each grid is as follows:

[0176] (15)

[0177] Where: is the revised PV installation capacity allocation plan for each grid. It is The optimization objective function value of the grid, It is The optimization objective function value of the grid, is the number of grids, is the total installed capacity of the target area.

[0178] Among them, the total installed capacity of the target area , the calculation formula is as follows:

[0179] (16)

[0180] In the formula, is the power density of the PV module (i.e. the power generation capacity per unit area).

[0181] According to the power density of photovoltaic modules (i.e. the power generation capacity of photovoltaic modules per unit area), the effective roof area of ​​each grid is converted into potential photovoltaic installation capacity. Then, considering the actual roof structure, inclination and possible shading of the building, the layout of the photovoltaic system of each grid is optimized to ensure maximum space utilization and power generation benefits.

[0182] Step S6: According to the actual implementation situation, the revised photovoltaic installation capacity allocation plan of each grid in the target area is adjusted to obtain an adjusted photovoltaic installation capacity allocation plan of each grid in the target area.

[0183] The calculation formula for the photovoltaic installation capacity allocation plan for each grid in the adjusted target area is as follows:

[0184] (17)

[0185] In the formula, is the adjusted PV installation capacity allocation plan for each grid in the target area, It is The grid adjustment factor.

[0186] in, The calculation formula is as follows:

[0187] (18)

[0188] In the formula, It is The actual power generation during the implementation of each grid, No. Theoretical power generation of a grid.

[0189] Embodiment 2:

[0190] This embodiment introduces a distributed rooftop photovoltaic processing device based on Bayesian optimization, including the following modules:

[0191] Data processing module: used to collect data related to the distributed rooftop photovoltaic system in the target area, including high-resolution satellite images, power grid data, economic data and sunshine data, and pre-process these data to obtain pre-processed data. Pre-processing includes: cleaning, standardization and data fusion to ensure its accuracy and consistency.

[0192] Roof area calculation module: It is used to divide the target area into several grids, input the satellite image of each grid into the Mask RCNN network, obtain the mask area of ​​the building in each grid, and calculate the effective roof area of ​​each grid based on the mask area of ​​the building in each grid.

[0193] Evaluation index construction module: used to construct multi-dimensional evaluation indicators for each grid based on the preprocessed data and the effective roof area of ​​each grid.

[0194] Optimal solution acquisition module: used to construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, optimize the comprehensive benefit optimization objective function using the Gaussian process model and Bayesian optimization algorithm, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

[0195] Correction module: used to correct the optimal photovoltaic installation capacity allocation plan of each grid in the target area according to the constraint conditions, and obtain the corrected photovoltaic installation capacity allocation plan of each grid in the target area.

[0196] Adjustment module: used to adjust the revised photovoltaic installation capacity allocation plan of each grid in the target area according to the actual implementation situation, and obtain the adjusted photovoltaic installation capacity allocation plan of each grid in the target area.

[0197] Embodiment 3:

[0198] This embodiment introduces a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a distributed rooftop photovoltaic processing method based on Bayesian optimization as described in any one of Embodiments 1 is implemented.

[0199] Embodiment 4:

[0200] This embodiment introduces a computer device, including:

[0201] Memory, used to store instructions.

[0202] A processor is used to execute the instructions so that the computer device performs the operations of a distributed rooftop photovoltaic processing method based on Bayesian optimization as described in any one of Embodiment 1.

[0203] Embodiment 5:

[0204] This example introduces the actual application process of the present invention. A certain region plans to install distributed photovoltaic power generation systems on three grids (A, B, and C), and capacity allocation optimization is required. The goal is to comprehensively consider economic benefits, grid benefits, solar radiation conditions, and photovoltaic simultaneity, while satisfying the following constraints:

[0205] Policy restrictions: The installed capacity of each grid must not exceed a certain upper limit.

[0206] Upper limit of total system capacity: The total installed capacity shall not exceed 30MW.

[0207] Maximizing economic benefits: Maximizing comprehensive benefits while satisfying constraints. See Table 1 for details.

[0208] Table 1 Multi-dimensional evaluation indicators of the three grids and the distribution of photovoltaic installation capacity

[0209] Grid Economic benefit (yuan / MW) Grid benefit (yuan / MW) Solar radiation conditions (hours / day) Photovoltaic simultaneous rate (%) Maximum installed capacity (MW) A 3000 2000 6.0 80 10 B 4000 1500 5.5 70 15 C 3500 2500 5.8 75 12

[0210] Weight setting: economic benefit weight is 0.4, grid benefit weight is 0.3, solar radiation weight is 0.2, and photovoltaic synchronization rate weight is 0.1.

[0211] Bayesian Optimization Process

[0212] Step 1: Initialize the model and objective function:

[0213]

[0214] in, , represents the allocation capacity of grids A, B, and C, . Taking the initial sampling point (evenly distributed) as the starting point, calculate the comprehensive benefits.

[0215] Step 2: Select the acquisition function and update the model:

[0216] The expected improvement (EI) is used to find the next set of allocations for the acquisition function, which attempts to balance exploration and exploitation by predicting the uncertainty of the Gaussian process model.

[0217] Generate a new plan:

[0218] C A =8, C B =12, C C =10

[0219] Recalculate the overall benefits and update the Gaussian process model.

[0220] Step 3: Iterative optimization and strategy adjustment:

[0221] After multiple iterations (10 times in total), the initial optimal solution is found:

[0222] C A =8, C B =15, C C =7

[0223] At this time, the comprehensive benefit is maximized and the target value F([8,15,7]) reaches the highest value.

[0224] Step 4: Capacity limit adjustment:

[0225] Check whether the initial plan meets the upper limit of installation:

[0226] For individual caps of grids A, B, C: CA ≤10, C B ≤15, C C ≤12.

[0227] For total system capacity: C A +C B +C C ≤30.

[0228] The preliminary allocation plan complies with the restrictions and no further adjustments are required.

[0229] Step 5: Implementation and dynamic monitoring:

[0230] Based on the optimization results, implement the capacity allocation plan:

[0231] Grid A: 8 MW photovoltaic system installed.

[0232] Grid B: Install a 15 MW photovoltaic system.

[0233] Grid C: Install a 7 MW photovoltaic system.

[0234] After implementation, actual monitoring revealed that the actual power generation deviated from the theoretical value:

[0235] Grid A Actual power generation P actual,A =48MWh, theoretical value P theoretical,A =50 MWh, adjustment factor f A =48 / 50=0.96. Grids B and C are adjusted similarly.

[0236] Ultimately, through dynamic monitoring, the actual allocation strategy is adjusted to further optimize the operating efficiency of the photovoltaic system.

[0237] The present invention is directed to a method for predicting and optimizing the capacity of distributed rooftop photovoltaic power generation, and its characteristic is that it adopts a Bayesian optimization algorithm, which is based on a Gaussian process model and is optimized in combination with multi-dimensional evaluation indicators. The Bayesian optimization algorithm has demonstrated outstanding performance in dealing with optimization problems with high dimensions and complex constraints. Through intelligent sampling and iterative updates, it can effectively explore and approach the optimal solution in a broad solution space. This algorithm not only has global search capabilities, but also can balance exploration and utilization to avoid local optimal solutions. Its flexible model adjustment mechanism enables it to perform well when considering multiple factors such as economic benefits, grid load, and solar radiation conditions, and significantly improves the accuracy and comprehensive benefits of photovoltaic capacity allocation.

[0238] In addition, the present invention provides a solid foundation for Bayesian optimization through precise data collection and processing, advanced feature extraction technology, and comprehensive evaluation index calculation. The multi-dimensional evaluation method and flexible strategy adjustment mechanism used in the optimization process further improve the accuracy and efficiency of capacity prediction and resource allocation. These innovations enable this method to effectively optimize the capacity allocation of photovoltaic systems in practical applications, achieve optimal resource configuration, and maximize overall benefits.

[0239] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A distributed rooftop photovoltaic processing method based on Bayesian optimization, characterized in that: The following steps are involved: Step S1: acquiring data related to the target area and the distributed rooftop photovoltaic system, and preprocessing the related data to obtain preprocessed data; Step S2: Divide the target area into several grids, input the satellite image of each grid into the neural network, obtain the mask area of ​​the building in each grid, and calculate the effective roof area of ​​each grid according to the mask area of ​​the building in each grid; Step S3: construct a multi-dimensional evaluation index for each grid based on the preprocessed data and the effective roof area of ​​each grid; Step S4: construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, optimize the comprehensive benefit optimization objective function using the Gaussian process model and the Bayesian optimization algorithm, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

2. A distributed rooftop photovoltaic processing method based on Bayesian optimization according to claim 1, characterized in that: The method further includes: Step S5: correcting the optimal photovoltaic installation capacity allocation plan for each grid in the target area according to the constraint conditions to obtain a corrected photovoltaic installation capacity allocation plan for each grid in the target area.

3. A distributed rooftop photovoltaic processing method based on Bayesian optimization according to claim 2, characterized in that: Also includes: Step S6: According to the actual implementation situation, the revised photovoltaic installation capacity allocation plan of each grid in the target area is adjusted to obtain an adjusted photovoltaic installation capacity allocation plan of each grid in the target area.

4. The distributed rooftop photovoltaic processing method based on Bayesian optimization according to claim 1 is characterized in that: The step S2 specifically includes: Step s2.1: Divide the target area into several uniform or adaptive grids; Step s2.2: Input the satellite image of each grid into the Mask R-CNN network to obtain the mask area of ​​each building in each grid. According to the mask area of ​​each building, the area of ​​each building is calculated; The area calculation formula for each building is as follows: ； in, It is In the grid The area of ​​the building, It is In the grid The mask area of ​​the building, is each type of pixel in the mask area; Step s2.3: Calculate the effective roof area of ​​each building in each grid according to the area of ​​each building in each grid; The calculation formula for the effective roof area of ​​each building in each grid is as follows: ； in, It is In the grid The effective roof area of ​​a building, It is for The geometric adjustment factor of each building; Step s2.4: Calculate the effective roof area of ​​each grid according to the effective roof area of ​​each building in each grid; The calculation formula for the effective roof area of ​​each grid is as follows: ； in, is the number of buildings.

5. The distributed rooftop photovoltaic processing method based on Bayesian optimization according to claim 1 is characterized in that: The multi-dimensional evaluation indicators of each grid include: economic benefit indicators, power grid benefit indicators, solar radiation condition indicators and photovoltaic simultaneity rate indicators; Among them, the economic benefit index is constructed by the electricity value index, the installed cost index and the photovoltaic installed capacity; The grid efficiency index is constructed from the stability index and the photovoltaic installation capacity; The solar radiation condition index is constructed by the effective sunshine time index, solar radiation index and photovoltaic installation capacity; The photovoltaic simultaneous rate index adopts the The photovoltaic simultaneity index of each grid; Among them, the electricity cost value index The calculation formula is as follows: ； in, For the The value of electricity bill for each grid, For the The possible power generation of a grid, is the local electricity price; In the formula, The calculation formula is as follows: ； in, For the The effective roof area of ​​the grid, For the The photovoltaic power generation potential per unit area of ​​the grid; Installation cost index , the calculation formula is as follows: ； in, For the The installation cost of a grid, Installation cost per unit area; Stability index The calculation formula is as follows: ； in, For the The stability of the grid, is the historical photovoltaic growth, is the residual carrying capacity of the transformer; The effective sunshine time index , the calculation formula is as follows: ； in, For the The effective sunshine time of each grid is is the total number of hours of sunlight, is the number of days in the statistical period; The solar radiation index , the calculation formula is as follows: ； in, For the The solar radiation of the grid, is the total radiation; The said PV synchronicity index for each grid The calculation formula is as follows: ； in, For the The actual power generation of the grid, For the The photovoltaic installed capacity of the grid.

6. A distributed rooftop photovoltaic processing method based on Bayesian optimization according to claim 1, characterized in that: The step S4 specifically includes: Step S4.1: Construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, where the expression of the comprehensive benefit optimization objective function is as follows: ； in, It is an indicator of economic benefits. is the grid efficiency indicator, is an indicator of solar radiation conditions. is the photovoltaic simultaneity index, is the weight coefficient of the economic benefit index, is the weight coefficient of the power grid benefit index, is the weight coefficient of the solar radiation condition index and is the weight coefficient of the photovoltaic simultaneous rate indicator, It is the allocation plan of photovoltaic installation capacity for each grid in the target area; Step S4.2: construct a probabilistic proxy model for optimizing the objective function using the Gaussian process model; Step S4.3: using the acquisition function in the Bayesian optimization algorithm to predict the next photovoltaic installation capacity allocation plan to be evaluated according to the covariance function of the probabilistic agent model and the currently known optimal value of the optimization objective function; Step s4.4: According to the next photovoltaic installation capacity allocation scheme to be evaluated, the sampling result of the optimization objective function is calculated, and the mean function and covariance function of the probability proxy model are updated according to the sampling result to obtain the updated probability proxy model; Step s4.5: Replace the currently known optimal value of the optimization objective function in step S4.3 with the sampling result, replace the probability proxy model in step S4.3 with the updated probability proxy model, repeat steps s4.3 and s4.4 until the predetermined termination conditions are met, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

7. A distributed rooftop photovoltaic processing method based on Bayesian optimization according to claim 2, characterized in that: The step S5 specifically includes: Step S5.1: Obtain the total installed capacity of the target area and the maximum installable capacity of each grid; Step S5.2: when the photovoltaic installation capacity allocation plan of the grid is greater than the maximum installable capacity of the grid, the maximum installable capacity of the grid is used as the revised photovoltaic installation capacity allocation plan of the grid; Step S5.3: When the sum of the photovoltaic installation capacity allocation plans of each grid in the target area is greater than the total installed capacity of the target area, a revised photovoltaic installation capacity allocation plan of each grid is calculated according to the total installed capacity of the target area; The calculation formula of the revised photovoltaic installation capacity allocation plan for each grid is as follows: ； Where: is the revised PV installation capacity allocation plan for each grid. It is The optimization objective function value of the grid, It is The optimization objective function value of the grid, is the number of grids, is the total installed capacity of the target area.

8. A distributed rooftop photovoltaic processing device based on Bayesian optimization, characterized in that: Includes the following modules: Data processing module: used for data related to the target area and the distributed rooftop photovoltaic system, and preprocesses the related data to obtain preprocessed data; Roof area calculation module: used to divide the target area into several grids, input the satellite image of each grid into the neural network, obtain the mask area of ​​the building in each grid, and calculate the effective roof area of ​​each grid based on the mask area of ​​the building in each grid; Evaluation index construction module: used to construct multi-dimensional evaluation indicators for each grid based on the preprocessed data and the effective roof area of ​​each grid; Optimal solution acquisition module: used to construct a comprehensive benefit optimization objective function based on multi-dimensional evaluation indicators, optimize the comprehensive benefit optimization objective function using the Gaussian process model and Bayesian optimization algorithm, and obtain the optimal photovoltaic installation capacity allocation plan for each grid in the target area.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, a distributed rooftop photovoltaic processing method based on Bayesian optimization as described in any one of claims 1 to 7 is implemented.

10. A computer device, characterized in that: include: A memory for storing instructions; A processor is used to execute the instructions so that the computer device performs the operations of a distributed rooftop photovoltaic processing method based on Bayesian optimization as described in any one of claims 1 to 7.

Images