A method for determining the distribution of sunlight in a region.

Abstract

It provides a tool that allows determining the distribution of sunlight in an area anywhere in the world in the simplest way. [Solution] A phase for collecting data to form a training database (B), and a phase for training a model based on the training database (B) to obtain a trained model, wherein the input of the trained model is an image of an area seen from the sky and the output is a global map of irradiance projected onto each surface of the area captured in the input image, and a phase for operating the trained model, which phase includes a step of receiving an image of an area seen from the sky and a step of determining the global map of irradiance projected onto each surface of the area captured in the received image using the trained model, thereby determining the solar distribution.

Description

Technical Field

[0001] The present invention relates to a method for determining the sunlight distribution in a region. The present invention also relates to a related computer program product.

Background Art

[0002] Generating electricity from renewable energy is an important ongoing issue for our society. This has necessarily been accompanied by the development of specific equipment such as solar panels that enable the production of solar electricity. Solar panels are typically installed on a specific area of a region, such as on the roof, wall, veranda, or ground of a building.

[0003] To support the development of solar facilities, tools have been developed to estimate the sunlight distribution in a region while taking into account the shape and orientation of the specific areas for these solar facilities. These tools aim to improve the distribution of solar panels in a region in order to increase the production of solar electricity. These tools are also used to encourage investment in solar panels by showing the solar potential of the region to individuals and local communities.

[0004] Such tools typically use the 3D data of the local area to evaluate the sunlight distribution in a region.

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, 3D data is not available everywhere in the world. Moreover, 3D reconstruction requires costly resources. Therefore, current tools are not readily available everywhere in the world.

[0006] Therefore, we need a tool that allows us to determine the distribution of sunlight in a specific region anywhere in the world, using the simplest possible method. [Means for solving the problem]

[0007] Therefore, the present invention relates to a method for determining the distribution of sunlight in a region, the method being implemented by a computer, and the method having the following: a. A phase for collecting data to form a training database, wherein the collected data has at least the following: i. Multiple images of different regions viewed from the air, of which at least some have specific features suitable for the installation of solar panels, and ii. For each image, a map showing the overall irradiance projected onto each surface of the region captured in the image. b. A phase for training a model based on a training database to obtain a trained model, wherein the single input to the trained model is just one 2D image of a region viewed from the air, and the output is a whole map of irradiance projected onto each surface of the region captured in the input image. c. A phase for running the trained model, the phase comprising: i. A step of receiving a single 2D image of a region to be analyzed from an aerial view, and, ii. A step in which a trained model determines an overall map of the irradiance projected onto each surface of the region captured in the image to be analyzed.

[0008] The methods according to the present invention may have one or more of the following features, which are considered individually or in any technically possible combination: - The collected and analyzed images are obtained by sensors such as cameras or by satellites. - The collected data also includes, for each image, a mask that divides the image into one or more specific elements captured within the image. - The method has a phase for determining solar data related to one or more specific elements captured in the analysis image, based on the determined overall map and supplementary data, where the supplementary data includes, for example, local weather data and / or solar panel data and / or local location data. - The decision phase involves the following steps for each analyzed image: a. A step of determining a mask that divides one or more specific elements captured in the image, and b. The step of determining a specific map of irradiance projected onto one or more specific elements captured in the image, based on the overall map and the determined mask, Solar data for a given region is determined based on specific maps. - The solar data includes at least one of the following: the solar potential of one or more specific elements captured in the image to be analyzed, and the yield of a solar installation assumed to be installed on one or more specific elements. - The acquisition phase involves filtering the acquired images to obtain images with reduced resolution, and the images that form the training database are images with reduced resolution. - The method has a phase of designing and / or setting up solar panels within a region of the area captured in the image to be analyzed, based on the obtained overall map. -Specific elements are selected from among the building's roof, the building's balcony, and specific parts of the ground such as fields or gardens.

[0009] The present invention also relates to a computer program product having a computer-readable medium containing a computer program having program instructions, wherein the computer program is loadable into a data processing device, and when the computer program is invoked by the data processing device, it triggers the execution of the method described above.

[0010] The present invention will be more easily understood by considering the following description, which is provided merely as an example with reference to the accompanying drawings below.

Brief Description of the Drawings

[0011] [Figure 1] It is a schematic diagram of an example of a computer that permits implementation of a method for determining sunlight distribution in a region. [Figure 2] It is a flowchart of an implementation example of a method for determining sunlight distribution in a region. [Figure 3] It is a schematic diagram of the collected data for forming a training database. [Figure 4] It is a schematic diagram of the input and output of a trained model.

Modes for Carrying Out the Invention

[0012] [[ID=z8]]Computer 10 and computer program product 12 are shown in FIG. 1.

[0013] Computer 10 is preferably a computer.

[0014] More generally, control device 10 is a similar electronic computing device adapted to convert and / or operate on data represented as a physical quantity, such as an electronic quantity in a register and / or memory of a computer or computing system, into other data similarly represented as a physical quantity in a memory, register or other such information storage device, information transmission device or information display device of the computing system.

[0015] Computer 10 exchanges information with computer program product 12.

[0016] As shown in Figure 1, the computer 10 has a processor 14 which includes a data processing device 16, a memory 18, and an information medium reader 20. In the example shown in Figure 1, the computer 10 has a human-machine interface 22 such as a keyboard and a display 24.

[0017] The computer program product 12 has an information medium 26.

[0018] The information medium 26 is a medium that can be read by the computer 10, and usually by the data processing device 16. The readable information medium 26 is a medium suitable for storing electronic instructions and can be connected to a computer system bus.

[0019] Examples of information media 26 include USB keys, floppy disks (registered trademarks) or flexible disks (known as "Floppy discs" in English), optical disks, CD-ROMs, magneto-optical disks, ROM memory, RAM memory, EPROM memory, EEPROM memory, magnetic cards, or optical cards.

[0020] The information medium 26 stores a computer program 12 that includes program instructions.

[0021] The computer program 12 is loadable into the data processing unit 16, and is adapted so that when the computer program 12 is loaded into the processing unit 16 of the computer 10, it inevitably involves the implementation of a method for determining the distribution of sunlight in a region.

[0022] The calculations of computer 10 will then be described with reference to Figure 2, which schematically shows an example implementation of a method for determining the distribution of sunlight in a region, and to Figures 3 and 4, which show specific phases of this method in more detail.

[0023] The decision method has a phase 100 for collecting data to form a training database B. The collection phase 100 is implemented by the computer 10 in its interaction with the computer program product 12, that is, it is implemented by a computer.

[0024] The collected data consists of multiple images of different regions viewed from the air: IM1, ..., IMn. The term "viewed from the sky" is understood to mean that images IM1, ..., IMn were taken from a high vantage point that allows for, for example, imaging of the rooftop of a building.

[0025] At least some of the regions captured in images IM1, ..., IMn have one or more specific elements R suitable for the installation of solar panels. The specific element R is selected from, for example, a specific part of the ground such as the roof of a building, a balcony of a building, or a garden.

[0026] Images IM1, ..., IMn are preferably only two-dimensional images. Images IM1, ..., IMn are preferably color images, such as RGB images (an abbreviation for "Red Green Blue"). Images IM1, ..., IMn are obtained by a sensor, such as a camera. To acquire images IM1, ..., IMn, the sensor is mounted, for example, on an aircraft. In the modified example, images IM1, ..., IMn are satellite images.

[0027] In one particular embodiment, in order to obtain images with reduced resolution, the collected images IM1, ..., IMn are filtered, and the images IM1, ..., IMn that form the training database B are images with reduced resolution. The reduced resolution is preferably such that each pixel of the image corresponds to an actual size of less than 50 centimeters. This makes it possible to obtain images IM1, ..., IMn with a resolution comparable to images that can be easily found in multiple databases.

[0028] The collected data also includes, for each image IM1, ..., IMn, overall maps C1, ..., Cn of the irradiance projected onto each surface of the region captured in images IM1, ..., IMn. Watts per square meter (W / m 2 Irradiance, expressed as ), is the incident radiant flux (power) received by a surface per unit area. Watts per square meter (W / m 2 The irradiance represented and projected as ) is a perspective shadowing-corrected irradiance that takes into account the geometric arrangement and / or shape of the projection surface (roof).

[0029] For example, each map C1, ..., Cn is obtained based on specific measurements taken by sensors in the region corresponding to the images IM1, ..., IMn. The sensor is, for example, a pyranometer. A solar radiation meter is a heat flux sensor. It measures the total solar radiation power in watts per square meter.

[0030] In another example, each map C1, ..., Cn was obtained based on an existing tool that outputs maps C1, ..., Cn as a function of 3D images of the region corresponding to images IM1, ..., IMn. Mapdwell and Google's Project Sunroof are examples of such tools.

[0031] The collected data also optionally includes masks M1, ..., Mn for each image IM1, ..., IMn, which divide one or more specific elements R captured in images IM1, ..., IMn. The partition masks M1, ..., Mn are obtained, for example, from the partitioning algorithm applied to the initial images IM1, ..., IMn. A segmentation algorithm is, for example, an edge detection algorithm.

[0032] Each image IM1, ..., IMn and its associated map C1, ..., Cn, and, if appropriate, its mask M1, ..., Mn, form the training elements E1, ..., En. Each training element E1, ..., En is formed by at most these three components. An example of a training element E1 having image IM1, map C1, and mask M1 is shown in Figure 3.

[0033] The resulting training database B is stored, for example, in the memory 18 of computer 10.

[0034] The decision method is the trained model M T To obtain this, the system has a phase 110 for training a model based on training database B. Preferably, the model is trained end-to-end using only training database B. Training phase 110 is implemented by the computer 10 in interaction with the computer program product 12, that is, it is implemented by a computer.

[0035] As shown in Figure 4, the trained model M T The input is an aerial image (IMi) of a region, and the output is an overall map (Ci) of the irradiance projected onto each surface of the region captured in the input image (IMi). Preferably, a trained model M T The single input is just one 2D image, preferably a color image, of a single region, so there is no need for a 3D image of the region or multiple 2D images of the same part of the region.

[0036] The model is typically a deep learning model. Such models are neural networks, such as convolutional neural networks.

[0037] Models are typically trained according to training techniques. Training techniques include, for example, practicing supervised learning. The training technique allows the neural network to be configured so that it is trained based on the training database B. It is emphasized that the model is trained solely on training database B. The training technique is based, for example, on the Adam optimization algorithm.

[0038] In determining the projected irradiance, training a model based on training database B allows for the incorporation of the surrounding environment, including shadows and vegetation originating from buildings, and any kind of obstacles that obstruct parts of a specific element R. The training also allows for the taking into account specific elements, such as the geometric features of a roof.

[0039] Therefore, in one embodiment, the model is a deep learning model trained end-to-end to map a 2D image to projected irradiance, and does not require the estimation of 3D attributes even in intermediate steps. The training, therefore, allows for the direct estimation of irradiance projected from only two-dimensional images, without intermediate steps, while taking into account three-dimensional aspects (shape, obstacles, and shadowing).

[0040] In one example, part of training database B is used to configure the neural network, while other parts are used to enable the configuration.

[0041] The decision method is the trained model M T It has a phase 120 for operating it. Operation phase 120 is implemented by the computer 10 in its interaction with the computer program product 12, that is, it is implemented by the computer.

[0042] Operation phase 120 includes the step of receiving an image IMi of a region as seen from the air. Advantageously, the image IMi of the region to be evaluated is preferably a color two-dimensional image obtained, for example, by measurement using a sensor such as a camera or by measurement using a satellite. Image IMi is, for example, an image of a building as seen from the air.

[0043] Operation phase 120 then projects the overall map Ci of irradiance projected onto each surface of the captured region onto the received image IMi, using the trained model M. T It has a step that is determined by

[0044] Therefore, a person skilled in the art will understand that the model is first trained on an existing irradiance map during training phase 110. The ongoing operation phase 120 then proceeds to train model M. T This makes it possible to determine an unknown irradiance map.

[0045] The determination method optionally includes a phase 130 for determining solar data associated with one or more specific elements R captured in the analysis image IMi, based on the determined overall map Ci and supplementary data. The decision phase 130 is implemented by the computer 10 in its interaction with the computer program product 12, that is, it is implemented by the computer.

[0046] In one embodiment, the solar data includes at least one data point from among the solar potential of one or more specific elements R captured in the image IMi to be analyzed, and the yield of a solar installation assumed to be installed on one or more specific elements R. Solar potential is the solar power received by a given area, expressed as kilowatts per hour converted to an annual value (kW / hour / year). Yield is the ratio between the energy produced (kWh) and the theoretical power of the equipment (kWp).

[0047] Preferably, the supplementary data includes local weather data and / or solar panel data and / or local location data.

[0048] Weather data refers to weather data for a specific period, such as one year. Solar panel data includes, for example, the specific technology of the solar panel and the dimensions of these solar panels. Location data includes, for example, the latitude and longitude of a region.

[0049] In one example, the determination phase 130 includes a step of determining a mask Mi that divides one or more specific elements R captured in the image IMi. The mask Mi division is obtained, for example, from the division algorithm applied to the initial image IMi. A segmentation algorithm is, for example, an edge detection algorithm.

[0050] Next, the determination phase 130, as shown in Figure 4, includes a step of determining a specific map Csi of irradiance projected onto only one or more specific elements R captured in the image IMi, based on the overall map Ci and the determined mask Mi. For example, if a specific element R is the roof of a building, then the specific map Csi corresponds only to the irradiance projected onto the roof.

[0051] The specific map Csi is displayed on a display, such as the display 24 of a computer 10, which allows the user to optionally control the determined irradiance.

[0052] Decision Phase 130 also includes a step of determining solar data based on a specific map Csi.

[0053] The determination method optionally includes a phase 140 for designing and / or setting up solar panel equipment based on the obtained overall map within the area of ​​a region for which an overall map has been obtained.

[0054] The described method allows for the optimization of a model that learns to estimate irradiance projected onto specific elements (such as roofs) directly from a 2D image. End-to-end training prevents the accumulation of errors that can occur when the task is divided into 3D attribute estimation, segment detection, and shadow heuristics.

[0055] This makes it possible to obtain solar information even in areas where only limited data is available, and especially in areas where 3D data is unavailable. This also enables a rapid first estimate of the solar potential. The obtained irradiance and solar data can be useful in determining appropriate solar installations for a given area.

[0056] Those skilled in the art will understand that the embodiments and modifications described above can be combined to form novel embodiments, provided that there is no technical inconsistency. [Explanation of symbols]

[0057] 10 Calculator 12. Computer Program Products 14 processors 16 Data Processing Devices 18 memory 20 Readers for information media 22 Human-Machine Interface 24 displays 26 Information media 100 Collection Phase 110 Training Phase 120 Operation Phases 130 Decision Phase 140 Design Setup Phase

Claims

1. A method for determining the distribution of sunlight in a region, wherein the method is implemented by a computer, and the method is a. A phase for collecting data to form a training database (B), wherein the collected data has at least the following: i. Multiple images (IM1, ..., IMn) of different regions viewed from the air, wherein at least some of the regions have specific elements (R) suitable for the installation of solar panels, and ii. For each image (IM1, ..., IMn), the overall map of irradiance projected onto each surface of the region captured in the image (IM1, ..., IMn) (C1, ..., Cn), b. Trained model (M Ｔ A phase for training a model based on a training database (B) to obtain the trained model (M Ｔ The single input to this system is a single two-dimensional image (IMi) of a region viewed from the air, and the output is a global map (Ci) of the irradiance projected onto each surface of the region captured in the input image (IMi). c. Trained model (M Ｔ A phase for operating ) and the phase having the following: i. A step of receiving a single two-dimensional image (IMi) of a region to be analyzed from the air, and, ii. The overall map of irradiance (Ci) projected onto each surface of the area captured in the image to be analyzed (IMi) is used with a trained model (M Ｔ The step determined by ) A method of having.

2. The method according to claim 1, wherein the collected images (IM1, ..., IMn) and the analyzed image (IMi) are obtained by a sensor such as a camera or by a satellite.

3. The method according to claim 1 or 2, wherein the collected data also includes, for each image (IM1, ..., IMn), a mask (M1, ..., Mn) that divides one or more specific elements (R) captured in the image (IM1, ..., IMn).

4. The method according to any one of claims 1 to 3, wherein the method comprises a phase for determining solar data related to one or more specific elements (R) captured in an analysis image (IMi) based on a determined overall map (Ci) and supplementary data.

5. The method according to claim 4, wherein the supplementary data includes local weather data and / or solar panel data and / or local location data.

6. The decision phase involves the following steps for each analysis image (IMi): a. A step of determining a mask (Mi) that divides one or more specific elements (R) captured in an image (IMi), and b. The process includes the step of determining a specific map (Csi) of irradiance projected onto only one or more specific elements (R) captured in the image (IMi), based on the overall map (Ci) and the determined mask (Mi), The method according to claim 4, wherein solar data for a region is determined based on a specific map (Csi).

7. The method according to claim 4, wherein the solar data comprises at least one data set from among the solar potential of one or more specific elements (R) captured in the image (IMi) to be analyzed and the yield of a solar facility assumed to be installed on one or more specific elements (R).

8. The method according to any one of claims 1 to 3, wherein the acquisition phase includes a step of filtering the acquired images (IM1, ..., IMn) to obtain images with reduced resolution, and the images (IM1, ..., IMn) forming the training database (B) are images with reduced resolution.

9. The method according to any one of claims 1 to 3, wherein the method comprises a phase for designing and / or setting up solar panels within a region captured in an image to be analyzed (IMi) based on the obtained overall map (Ci).

10. The method according to any one of claims 1 to 3, wherein the specified element (R) is selected from among the roof of a building, the balcony of a building, and a specific part of the ground such as a field or garden.

11. A computer program product having a computer-readable medium containing a computer program having program instructions, wherein the computer program is loadable into a data processing device, and when the computer program is activated by the data processing device, it causes the execution of the method according to any one of claims 1 to 3.

Images