A spatiotemporal prediction method for rooftop photovoltaics
Through GIS view and machine vision technology, a spatial and temporal information correlation model was established, which solved the problem of roof photovoltaic stock resource distribution evaluation error, achieved accurate prediction of power grid planning, and improved grid operation efficiency.
Patent Information
- Application Number
- CN202111352756.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-16
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2041-11-16
AI Technical Summary
The existing technology cannot accurately evaluate and predict the spatiotemporal distribution of roof photovoltaic stock resources, resulting in grid planning errors and affecting social and economic benefits.
Intelligent technology based on GIS view and machine vision is adopted to identify the stock resources of roof photovoltaics, combine resource development intention and scale survey, establish a photovoltaic spatio-temporal information correlation model to conduct accurate perception and future incremental resource prediction.
By accurately sensing roof photovoltaic resources, providing accurate time and space predictions, it helps grid companies make more scientific planning decisions and improve grid operation efficiency and economic benefits.
Smart Images

Figure CN114332597B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of photovoltaics, and in particular to a spatiotemporal prediction method for rooftop photovoltaics. Background Art
[0002] Conventional energy sources are very limited. Solar energy is an inexhaustible renewable energy source for human beings. It has the advantages of full cleanliness, absolute safety, relative extensiveness, long life and maintenance-free, sufficient resources and potential economy. It plays an important role in the long-term energy strategy. Photovoltaic power generation is a technology that uses the photovoltaic effect of semiconductor interfaces to directly convert light energy into electrical energy. It is mainly composed of three parts: solar panels (assemblies), controllers and inverters, and the main components are composed of electronic components. After the solar cells are connected in series and packaged for protection, they can form a large-area solar cell module, and then combined with power controllers and other components to form a photovoltaic power generation device. With the gradual development of the information age, the intelligent interconnection of distributed rooftop photovoltaic power stock resource data is of great significance for photovoltaic spatiotemporal prediction. By the end of 2020, the total installed capacity of photovoltaic power in Zhizao New City will be 86.9MW, of which there are 24 photovoltaic power stations connected to 10kV, with a total installed capacity of 52.14MW, and the access mode is full access to the Internet, self-generation and self-use, and surplus power to the Internet. There are 538 units connected to low voltage with a total installed capacity of 20.239MW, in two modes: "self-generation for own use, surplus power connected to the grid" and "full power connected to the grid".
[0003] However, as existing technologies are unable to evaluate the spatiotemporal distribution of rooftop photovoltaic stock resources, there are errors in the prediction of photovoltaic power supply conditions, and the inability to accurately associate grid data makes it difficult for grid companies to make correct decisions on grid planning, affecting local social and economic benefits. Summary of the invention
[0004] The present invention aims to overcome the problem of grid planning errors caused by distribution errors of rooftop photovoltaic stock resources in the prior art, and provides a spatiotemporal prediction method for rooftop photovoltaics. Based on the intelligent technology of GIS view and machine vision, the method identifies the rooftop photovoltaic stock resources in the manufacturing new city, further compares them with the rooftop resource development intention and scale survey, accurately perceives the current new energy resources in the manufacturing new city, and predicts future incremental resources based on control planning and developable plots.
[0005] In order to achieve the above object, the present invention adopts the following technical solutions:
[0006] A spatiotemporal prediction method for rooftop photovoltaics, including the following contents:
[0007] S1. Evaluate the rooftop photovoltaic potential in urban areas;
[0008] S2. Establishing a rooftop photovoltaic spatiotemporal information correlation model;
[0009] S3. Predict the photovoltaic installation scale based on the photovoltaic spatiotemporal information correlation model.
[0010] Rooftop photovoltaics uses intelligent technology based on GIS views and machine vision to identify the existing photovoltaic resources on the roofs of manufacturing new cities, further compare them with the willingness and scale of rooftop resource development, accurately perceive the current new energy resources in manufacturing new cities, and predict future incremental resources based on control plans and developable plots.
[0011] Preferably, the S1 includes the following contents:
[0012] S101, Generate a digital elevation model, using the stereo remote sensing data of the associated RPC to generate a digital surface model of the area, and next generate a digital terrain model, looking for a local area with a specific tile size at a specific tile size, using a tile size related to the size of the building can produce a reliable digital terrain model; the tile size should be slightly larger than the feature being removed, such as buildings, trees, etc.;
[0013] S102, selecting one of the images of the digital elevation model for construction detection, the image including all panchromatic spectra and multispectral "MS", and here, selecting an image close to the lowest point image;
[0014] Then, using the DSM layer, the corresponding image coordinates of each ground point are calculated corresponding to the ground X, Y, Z and the 3D coordinates of the associated image RPC;
[0015] S103, performing image processing, before image segmentation, in order to use the Pan and MS bands at the same time, using the UNB pan-sharpening method to retain the color of the object;
[0016] S104, vegetation suppression, all vegetation areas were removed from the study area, and vegetation was detected using a variety of different vegetation indices, such as the standardized difference vegetation index and near infrared infrared ratio;
[0017] Since elevation parameters will play a key role in building detection, other elevated features, such as trees, must be suppressed;
[0018] S105, after the vegetation suppression segmentation, the segments with lift values greater than a specific threshold are assigned to the building class, using shape features, the segments belonging to the building with high rectangles and ellipses, and high altitudes are assigned to the building class, and the installable amount of rooftop photovoltaic resources is summarized, and the building types are divided into residential buildings and industrial buildings, and the respective roof areas are summarized;
[0019] The specific threshold is generally 2 meters or less; considering the noise and errors in NDSM generation, the threshold is considered to be slightly lower than the actual building height to reduce classification omission errors.
[0020] In order to evaluate the potential of rooftop photovoltaics in urban areas, building detection through high-resolution remote sensing images is required. This direction is one of the most demanding applications of high-resolution remote sensing images because there are various functions that require building boundaries, such as map updates and hazard impact assessments. However, due to the complexity of the above tasks, a large number of studies use auxiliary data, such as road maps and old building maps, to perform building detection in high-resolution remote sensing images. Typically, buildings are elevated rectangular objects. Therefore, shape and elevation are of great significance for building detection in object-based analysis.
[0021] Preferably, the building types in S105 are further subdivided into roofs of health system buildings, roofs of education system buildings, roofs of government buildings, roofs of township and village-level buildings, farm roofs, industrial roofs, urban residential roofs, photovoltaic building integration and farmers' roofs.
[0022] Preferably, S2 includes the following contents:
[0023] S201, summarizing and classifying the collected and summarized power grid data of the roof area;
[0024] S202, the power grid data is divided into several cell arrays according to the building type, each cell array corresponds to a building type, and the cell array is represented by c i ={a i ,b i ,d i ,e i}, i represents different building types, a i represents the building volume, b i represents the roof area, d i Indicates the photovoltaic installation area, e i It indicates the estimated installed capacity of photovoltaic power;
[0025] S203, establishing a rooftop photovoltaic spatiotemporal information association model including photovoltaic installable area and photovoltaic estimated installable capacity;
[0026]
[0027] η represents the roof wall utilization coefficient, ρ represents the photovoltaic power generation intensity;
[0028] S204, selecting samples from all cell arrays to perform photovoltaic intensity statistics;
[0029] S205. Obtain different photovoltaic output curves based on a clustering algorithm according to photovoltaic intensity statistics.
[0030] Preferably, S3 includes the following contents:
[0031] The photovoltaic spatiotemporal information correlation model is combined with the evaluation of rooftop photovoltaic potential in urban areas. The bottom-up curve superposition method is used to quantify the typical daily rooftop photovoltaic output curves of the grid and each block, and the boundary conditions are considered to meet the load demand under the load extreme photovoltaic curve.
[0032] Preferably, S202 further includes the following contents: the cell arrays are classified according to the building types, and each cell array is divided into multiple multi-level data nodes according to the land use code, the land use code represents the land use nature, different land use codes represent different land use natures, and there are also multiple user properties under the land use nature; a cell array among several cell arrays is selected as the initial clustering center for clustering, and all data nodes are clustered to the most similar cell array according to the similarity of user properties and geographical locations.
[0033] Preferably, the similarity is calculated in the following manner:
[0034]
[0035] λ xy is the correlation between the power consumption of user x and the power consumption of user y. The larger the value, the more correlation there is, and the clustering is done in the same cell array. xi represents the power consumption of user x at the current time node. represents the average power consumption of user x in a time interval, and yi represents the power consumption of user y at the current time node. Represents the average power consumption of user y in a time interval.
[0036] Therefore, the present invention has the following beneficial effects:
[0037] The present invention collects and surveys the current status of rooftop photovoltaics based on remote sensing images, performs spatiotemporal prediction based on power grid data, and adopts a clustering algorithm to calculate the relationship between photovoltaics and electricity consumption of each building type. The clustering algorithm is used to obtain photovoltaic output curves of different typical days in Qujiang Intelligent Manufacturing New City, accurately perceive the current new energy resources of the manufacturing new city, and predict future incremental resources based on control plans and developable plots. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 This is the digital elevation model generation image of this embodiment.
[0039] Figure 2 This is the UNB pan-sharpened picture obtained by image processing in this embodiment.
[0040] Figure 3 It is the vegetation suppression map of this embodiment.
[0041] Figure 4 This is the image recognition roof resource map of this embodiment.
[0042] Figure 5 This is a statistical diagram of the current photovoltaic rooftop development intensity of this embodiment.
[0043] Figure 6 This is a typical photovoltaic curve cluster diagram of Qujiang Intelligent Manufacturing New City in this embodiment.
[0044] Figure 7 This is the representative photovoltaic output curve on different typical days in the Qujiang Intelligent Manufacturing New City area of this embodiment.
[0045] Figure 8 This is the typical daily photovoltaic curve of the grid in 2025 (summer clear) of this embodiment.
[0046] Fig. 9 This is the typical daily photovoltaic curve (summer rain) of the grid in 2025 of this embodiment. DETAILED DESCRIPTION
[0047] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.
[0048] Example:
[0049] This embodiment provides a spatiotemporal prediction method for rooftop photovoltaics, including the following contents:
[0050] S1. Evaluate the rooftop photovoltaic potential in urban areas;
[0051] By the end of 2020, the total installed photovoltaic capacity of the Smart Manufacturing New City was 86.9MW, of which there were 24 photovoltaic power stations connected to 10kV with a total installed capacity of 52.14MW. The grid-connected mode was full access to the grid, self-generation for self-use, and surplus power to the grid.
[0052] There are 538 low-voltage power stations with a total installed capacity of 20.239MW, in two modes: "self-generation for self-use, surplus power connected to the grid" and "full power connected to the grid". Table 1 shows the statistics of 10kV distributed power supply in the Smart New City.
[0053] Table 1
[0054]
[0055]
[0056] S1 includes the following:
[0057] S101, generate a digital elevation model, such as Figure 1 As shown, using the stereo remote sensing data of the associated RPC, a digital surface model of the area is generated. The next step is to generate a digital terrain model. The local area at a specific tile size is found. Using tile sizes related to the building size can produce a reliable digital terrain model;
[0058] S102, selecting one of the images of the digital elevation model for construction detection, where an image close to the lowest point image is selected; then, using the DSM layer, corresponding to the 3D coordinates of the ground X, Y, Z and the relevant image RPC, the corresponding image coordinates of each ground point are calculated;
[0059] S103, image processing is performed. Before image segmentation, in order to use the Pan and MS bands at the same time, the UNB pan-sharpening method is used to retain the color of the object, such as Figure 2 As shown;
[0060] S104, vegetation suppression, removes all vegetation areas from the study area, and uses a variety of vegetation indices, such as the standardized difference vegetation index and near infrared infrared ratio, to detect vegetation such as Figure 4 As shown;
[0061] S105. After the vegetation suppression segmentation, the segments with lift values greater than a certain threshold are assigned to the building class. Using shape features, the segments belonging to buildings with high rectangles and ellipses, as well as high altitudes are assigned to the building class, and the installable amount of rooftop photovoltaic resources is summarized. The building types are divided into residential buildings and industrial buildings, and their respective roof areas are summarized.
[0062] The building types in S105 are further subdivided into roofs of health system buildings, roofs of education system buildings, roofs of government buildings, roofs of township and village-level buildings, roofs of farms and ranches, industrial roofs, roofs of urban residential buildings, photovoltaic building integration and roofs of farmers' households.
[0063] According to the survey of roof resources in the Qujiang Intelligent Manufacturing New City area, the building types include roofs of health system buildings, roofs of education system buildings, roofs of government buildings, roofs of township and village buildings, farm roofs, industrial roofs, urban residential roofs, photovoltaic building integration, and farmer roofs. The total photovoltaic area that can be developed is 16.324 square kilometers, and the total photovoltaic installed capacity is 66.04 megawatts. The details are shown in Table 2.
[0064] Table 2
[0065]
[0066]
[0067] S2. Establishing a rooftop photovoltaic spatiotemporal information correlation model;
[0068] S2 includes the following:
[0069] S201, summarizing and classifying the collected and summarized power grid data of the roof area;
[0070] Figure 5This is a statistical diagram of the current photovoltaic rooftop development intensity of this embodiment.
[0071] S202, the power grid data is divided into several cell arrays according to the building type, each cell array corresponds to a building type, and the cell array is represented by c i ={a i ,b i ,d i ,e i}, i represents different building types, a i represents the building volume, b i represents the roof area, d i Indicates the photovoltaic installation area, e i It indicates the estimated installed capacity of photovoltaic power;
[0072] S202 also includes the following contents: the cell arrays are classified according to the building types, and each cell array is divided into multiple multi-level data nodes according to the land use code. The land use code represents the land use nature. Different land use codes represent different land use natures, and there are also multiple user properties under the land use nature; a cell array among several cell arrays is selected as the initial clustering center for clustering, and all data nodes are clustered to the most similar cell array according to the similarity of user properties and geographical locations.
[0073] The similarity is calculated as follows:
[0074]
[0075] λ xy is the correlation between the power consumption of user x and the power consumption of user y. The larger the value, the more correlation there is, and the clustering is done in the same cell array. xi represents the power consumption of user x at the current time node. represents the average power consumption of user x in a time interval, and yi represents the power consumption of user y at the current time node. Represents the average power consumption of user y in a time interval.
[0076] S203, establishing a rooftop photovoltaic spatiotemporal information association model including photovoltaic installable area and photovoltaic estimated installable capacity;
[0077]
[0078] η represents the roof wall utilization coefficient, ρ represents the photovoltaic power generation intensity;
[0079] (1) Photovoltaic installation area (m 2 ) = Land area (m 2 )×Ratio of roof area to floor area (%)×Roof wall utilization coefficient (%)
[0080] (2) Photovoltaic installed capacity (kw) = Photovoltaic installation area (m 2 )×Rooftop photovoltaic power generation intensity (kw / m 2 )
[0081] According to the regional positioning and social life characteristics of Qujiang District, the ratio of roof area to construction land area, roof wall utilization coefficient, and photovoltaic panel power generation efficiency parameters of each land use type are selected as shown in Table 3.
[0082] Table 3
[0083]
[0084] Table 4 is a detailed list of installable photovoltaic power plants according to land use code and land use nature.
[0085] Table 4
[0086]
[0087]
[0088] S204, selecting samples from all cell arrays to perform photovoltaic intensity statistics;
[0089] S205. Obtain different photovoltaic output curves based on a clustering algorithm according to photovoltaic intensity statistics.
[0090] The planning area has abundant lighting resources, with an annual equivalent number of hours of approximately 996 hours.
[0091] 100 samples of photovoltaic power were selected for the current planning area to conduct photovoltaic intensity statistics. The statistical results are as follows: Figure 5 As shown in the figure, it can be seen that the development intensity of photovoltaic roofs in the region is roughly similar, with a distribution range of 34W / m 2 ~134W / m 2 , the median is about 69W / m 2 Using SPSS software to perform mean clustering algorithm test, it can be found that the associated probability value is greater than the significance level, which shows that the development intensity of photovoltaic roofs is around 68W / m 2 The assumption of a Gaussian distribution holds.
[0092] When considering the grid load balance of photovoltaic power, simple peak subtraction will cause large errors. The load and photovoltaic output characteristics of different periods must be considered. The photovoltaic daily output curve of Qujiang Intelligent Manufacturing New City collected from OPEN3000 also shows the typical photovoltaic daily curve obtained based on the clustering algorithm. Using the clustering algorithm, the photovoltaic output curves of different typical days in Qujiang Intelligent Manufacturing New City can be obtained as follows: Figure 6 shown.
[0093] Figure 7 This is the representative photovoltaic output curve on different typical days in the Qujiang Intelligent Manufacturing New City area of this embodiment.
[0094] S3. Predict the photovoltaic installation scale based on the photovoltaic spatiotemporal information correlation model.
[0095] S3 includes the following:
[0096] The photovoltaic spatiotemporal information correlation model is combined with the evaluation of rooftop photovoltaic potential in urban areas. The bottom-up curve superposition method is used to quantify the typical daily rooftop photovoltaic output curves of the grid and each block, and the boundary conditions are considered to meet the load demand under the load extreme photovoltaic curve.
[0097] Based on the typical photovoltaic curve, combined with the photovoltaic installed capacity forecast results obtained through rooftop resource surveys, the bottom-up curve superposition method is used to quantify the typical daily rooftop photovoltaic output curves of the grid and each block. Since the load extremes considered in the planning generally occur in summer, and the photovoltaic peak reduction is not obvious on rainy days in summer, the load is relatively large at this time. Therefore, the boundary conditions that should be considered in the planning are to meet the load demand under the typical photovoltaic curve on rainy days in summer. Figure 7 It can be seen that the maximum photovoltaic output at this time is 40% of the theoretical peak value.
[0098] The peak photovoltaic output of Qujiang Intelligent Manufacturing New City in 2025 will be 126.65MW in sunny summer, and 44.3MW in rainy summer. The peak will occur around 1:00 p.m. The specific typical daily photovoltaic curve is shown in Figure 8.
[0099] The peak photovoltaic output in the summer of Qujiang Intelligent Manufacturing New City is 298.88MW, and the peak photovoltaic output in the summer rain is 104.54MW. The peak occurs at around 1:00 pm. The specific typical daily photovoltaic curve is as follows: Fig. 9 shown.
[0100] Combined with the recent land development and housing construction, the recent photovoltaic output forecast for Qujiang Intelligent Manufacturing City is shown in Table 5.
[0101] Table 5 is the forecast of installed capacity of Qujiang Intelligent Manufacturing New City.
[0102] Table 5
[0103]
[0104] The specific description of the present invention in the above embodiments is only used to further illustrate the present invention and cannot be understood as limiting the scope of protection of the present invention. Technical engineers in this field may make some non-essential improvements and adjustments to the present invention based on the contents of the above invention, which fall within the scope of protection of the present invention.
Claims
1. A spatiotemporal prediction method for rooftop photovoltaics, Its characteristics are: Includes the following: S1. Evaluate the potential of rooftop photovoltaics in urban areas: Establish digital elevation models of urban areas based on GIS views and machine vision technology; Select one of the images of the digital elevation model for construction detection, process the selected image, and use the UNB pan-sharpening method to preserve the object color before image segmentation; vegetation suppression, removing vegetated areas from the study area, and using a number of different vegetation indices to detect vegetation; Segments with lift values greater than a certain threshold are assigned to the building class. Using shape features, segments belonging to buildings with high rectangles, ellipses, and high altitudes are assigned to the building class to obtain a summary of the installable rooftop PV resources. Building types are divided into residential buildings and industrial buildings, and their respective roof areas are summarized; S2. Use clustering algorithm to calculate the relationship between photovoltaic power generation and power consumption of each building type, and establish a rooftop photovoltaic spatiotemporal information correlation model; S3. Predict the photovoltaic installation scale based on the photovoltaic spatiotemporal information correlation model.
2. A spatiotemporal prediction method for rooftop photovoltaics according to claim 1, Its characteristics are: The S1 includes the following contents: S101, generating a digital elevation model, using the stereo remote sensing data of the associated RPC, generating a digital surface model of the urban area, and next generating a digital terrain model, finding a local area of a specific tile size at a specific tile size, and using the tile size related to the building size to generate a reliable digital terrain model; S102, select one of the images of the digital elevation model for construction detection, here, select the image close to the lowest point image; then, use the DSM layer, corresponding to the 3D coordinates of the ground X, Y, Z and the relevant image RPC, to calculate the corresponding image coordinates of each ground point.
3. A spatiotemporal prediction method for rooftop photovoltaics according to claim 2, Its characteristics are: The building types in S105 are further subdivided into roofs of health system buildings, roofs of education system buildings, roofs of government buildings, roofs of township and village-level buildings, roofs of farms and pastures, industrial roofs, urban residential roofs, photovoltaic building integration and farmer's roofs.
4. A spatiotemporal prediction method for rooftop photovoltaics according to claim 1, Its characteristics are: The S2 includes the following contents: S201, summarizing and classifying the collected and summarized power grid data of the roof area; S202, the power grid data is divided into several cell arrays according to the building type, each cell array corresponds to a building type, and the cell array is represented by c i ={a i ,b i ,d i ,e i }, i represents different building types, a i represents the building volume, b i represents the roof area, d i Indicates the photovoltaic installation area, e i It indicates the estimated installed capacity of photovoltaic power; S203, establishing a rooftop photovoltaic spatiotemporal information association model including photovoltaic installable area and photovoltaic estimated installable capacity; η represents the roof wall utilization coefficient, ρ represents the photovoltaic power generation intensity; S204, selecting samples from all cell arrays to perform photovoltaic intensity statistics; S205. Obtain different photovoltaic output curves based on a clustering algorithm according to photovoltaic intensity statistics.
5. A spatiotemporal prediction method for rooftop photovoltaics according to claim 1, Its characteristics are: The S3 includes the following: The photovoltaic spatiotemporal information association model is combined with the results of rooftop photovoltaic potential assessment in urban areas. The bottom-up curve superposition method is used to quantify the grid and the typical daily rooftop photovoltaic output curve of each block is quantified. The boundary conditions are considered to meet the load demand under the load extreme photovoltaic curve.
6. A spatiotemporal prediction method for rooftop photovoltaics according to claim 4, Its characteristics are: The S202 further includes the following contents: the cell array is classified according to the building type, and each cell array is divided into multiple multi-level data nodes according to the land use code, the land use code represents the land use nature, different land use codes represent different land use natures, and there are also multiple user properties under the land use nature; A cell array among several cell arrays is selected as the initial clustering center for clustering, and all data nodes are clustered to their most similar cell arrays according to the similarity of user properties and geographical locations.
7. A spatiotemporal prediction method for rooftop photovoltaics according to claim 6, Its characteristics are: The similarity is calculated in the following way: λ xy is the correlation between the power consumption of user x and the power consumption of user y. The larger the value, the more correlation there is, and the clustering is done in the same cell array. xi represents the power consumption of user x at the current time node. represents the average power consumption of user x in a time interval, and yi represents the power consumption of user y at the current time node. Represents the average power consumption of user y in a time interval.
Citation Information
Patent Citations
Quick three-dimensional virtual scene constructing method
CN104835202A
Automatic detection method for overstretched region of optical remote sensing satellite orthoimage
CN108335261A
