Device and method for determining at least one component of hemispherical irradiance of solar radiation in an arbitrary plane

Abstract

The invention relates to a method (S100, S200, S300, S400, S500) for determining at least one component of a hemispherical irradiance (120) of solar radiation in an arbitrary plane, wherein the at least one component has a diffuse irradiance (122) and / or a direct irradiance (124), comprising the following steps: (i) collecting measurement data relating to the hemispherical irradiance (120) using a radiation sensor unit (12) in a field of view (26) over a plane (46) of the radiation sensor unit (12); (ii) capturing an image (110) of the sky (50) using a camera (14) in a field of vision (32) over a plane (48) of the camera (14); (iii) extracting (S130, S230, S330, S430, S530) features from the image (110) of the sky (50) by means of a first machine learning model and producing a result data set; (iv) merging (S140, S240, S340, S440, S540) the measurement data to form a common data set; (v) determining (S150, S250, S350, S450, S550) the at least one component of the hemispherical irradiance (120) in the arbitrary plane from the data set by means of a second machine learning model.

Description

[0001] Description

[0002] Apparatus and method for determining at least one component of hemispheric irradiance of solar radiation in any plane

[0003] State of the art

[0004] The invention relates to a method and a device for determining at least one component of a hemispheric irradiance of solar radiation in any plane. Furthermore, the invention relates to a method for training a first machine learning model for extracting features, in particular structures, from a camera image of the sky, as well as a computer program for determining at least one component of a hemispheric irradiance of solar radiation in any plane, a data processing device, a trained first machine learning model for extracting features, in particular structures, from an image of the sky, and a trained second machine learning model for determining at least one component of a hemispheric irradiance of solar radiation in any plane.

[0005] Large-scale solar energy systems, such as solar power plants and photovoltaic systems, which are installed or planned at different latitudes around the world, are becoming increasingly important for the energy supply. The economic viability of operating the solar energy system is becoming a key factor in the decision for or against its installation. Due to the varying intensities of solar radiation at different latitudes, the location and efficiency of the solar energy system are of great importance for its economical operation. For example, to plan the location for the installation and economical operation of the photovoltaic system, the irradiance of the solar radiation at the respective location must be known or measurable.Solar irradiance, also known as solar irradiance, is the intensity and thus the light output of the solar radiation that hits one square meter of a solar panel in a photovoltaic system to be installed. To maximize solar radiation on the solar panels and maximize efficiency, solar panels are often tracked to change their azimuth and tilt angle. In bifacial photovoltaic systems, solar radiation from the back of the solar panel is utilized in addition to the front of the solar panel.Since solar radiation has individual components, such as diffuse irradiance in the horizontal plane (DHI), ground-reflected radiation, and direct normal irradiance (DNI), all of which contribute differently to the overall performance of a solar system, it is necessary to have access to the individual components of solar irradiance in addition to the hemispheric solar irradiance. It is of interest to measure and know the hemispheric irradiance and its components at any desired plane.

[0006] It is known that a radiation sensor unit, typically a pyranometer, is used to measure solar irradiance. Measurements are always taken in the field of view of a half-space above the installed plane of the radiation sensor unit, typically the pyranometer. The half-space describes a hemisphere, corresponding to a solid angle of at least approximately 2μm steradians. It is also known to record the degree of sky coverage using a surveillance camera or cloud camera. However, determining diffuse horizontal radiation and direct normal radiation requires extensive measurement effort. For example, a sun-tracking pyrheliometer and a sun-tracking pyranometer can be used for this purpose. The measured data are analyzed using transposition and decomposition models, which, however, are subject to significant errors.

[0007] WO 2021 / 219570 A1 discloses a measuring device comprising a radiation sensor unit, for example, a pyranometer, and a camera. Using a method, the measuring device can be used to determine the hemispheric solar irradiance at any plane by analyzing the measured data. Furthermore, the measuring device and method can also be used to determine the radiance distribution of the sky and the resulting fractions of diffuse radiation.

[0008] Disclosure of the invention

[0009] The object of the invention is to provide an improved method and an improved device for determining at least one component of a hemispheric irradiance of solar radiation in any plane.

[0010] Furthermore, it is the object of the invention to provide an improved method for training a first machine learning model for extracting features, in particular structures, from the image of the sky.

[0011] Furthermore, the object of the invention is to provide an improved computer program for determining at least one component of a hemispheric irradiance of solar radiation in any plane. Furthermore, the object of the invention is to provide a data processing device for a device for determining at least one component of a hemispheric irradiance of solar radiation in any plane.

[0012] Furthermore, it is the object of the invention to provide an improved trained first machine learning model for extracting features, in particular structures, from the image of the sky and an improved trained second machine learning model for determining at least one component of a hemispheric irradiance of solar radiation in any plane.

[0013] The objects are achieved by the features of the independent claims. Advantageous embodiments and advantages of the invention emerge from the further claims, the description, and the drawings.

[0014] The features listed individually in the patent claims can be combined with one another in a technologically meaningful manner and can be supplemented by explanatory facts from the description and by details from the figures, whereby further embodiments of the invention are shown.

[0015] According to a first aspect of the invention, a method is proposed for determining at least one component of a hemispheric irradiance of solar radiation in any plane, wherein the at least one component comprises a diffuse irradiance and / or a direct irradiance. The method comprises the following steps:

[0016] (i) determining measurement data of the hemispherical irradiance with a radiation sensor unit in a field of view above a plane of the radiation sensor unit;

[0017] (ii) capturing an image of the sky with a camera in a field of view above a plane of the camera; (iii) extracting features, in particular structures, from the image of the sky using a first machine learning model and generating a result dataset;

[0018] (iv) merging the measurement data from the radiation sensor unit and the result data set from the camera into a common data set;

[0019] (v) determining the at least one component of the hemispheric irradiance in the arbitrary plane, in particular the diffuse irradiance and / or the direct irradiance, from the data set by means of a second machine learning model.

[0020] Any plane shall be understood to be a horizontal plane or an inclined plane.

[0021] The radiation sensor unit and the camera form a measuring unit, which forms the basis of the proposed method. The measuring unit can thus provide a first measurement data set from the radiation sensor unit, which comprises the hemispheric irradiance measured in a field of view of a half-space above the radiation sensor unit. The half-space can be a hemisphere and / or hemisphere corresponding to a solid angle of at least approximately 2μm steradians. The measuring device can provide a second measurement data set, which comprises measurement data measured by the camera. The measurement data set from the camera comprises an image of the sky, which depicts the sky in a field of view of a half-space above the camera. Thus, the method is based on two measurement data sets from two different instruments: the radiation sensor unit and the camera. The radiation sensor unit can be a pyranometer, for example.The camera can be a cloud camera or a surveillance camera. The advantage of this is that improved accuracy can be achieved in determining one of the components of hemispheric irradiance. In particular, measurement inaccuracies and / or errors of the radiation sensor unit or the camera can be compensated. The reason for this is that the camera generally cannot measure irradiance as accurately as the radiation sensor unit. Thus, the accuracy is improved compared to the state of the art, which typically uses only one camera.

[0022] The image of the sky typically contains the sun and structures around the sun, such as cloud cover and arrangement, shading of the sun, blurring by cirrus clouds, and / or image artifacts caused by the influence of direct radiation on the camera image. Thus, the image of the sky contains structures and textures. These structures and textures provide the image features extracted by the first machine learning model. In particular, the image features around the sun can be used. The first machine learning model can extract the image features from the image of the sky. This allows effects such as lens refraction and image saturation to be detected. A type of spectral correction can be applied. These image features can be used in the subsequent analysis of the combined dataset using the second machine learning model.

[0023] The image of the sky can contain the structures and textures, each as angle-resolved information, due to the information stored pixel by pixel, particularly the measured intensities. The structures and textures can be used as image features for analyzing the measurement data by the first machine learning model and form the resulting dataset. The first machine learning model can, for example, be a so-called convolutional neural network (CNN). The CNN algorithm is a widely used algorithm that is used in a wide range of applications and, in particular, delivers reliable iterative results.

[0024] In contrast to the analysis of the sky image with the first machine learning model, the proposed method uses a physical model based on physical assumptions and corrections. The first machine learning model, on the other hand, can directly analyze the image features of the structures and textures present in the sky image. This can have the advantage of preventing any information loss.

[0025] The second machine-learning model can correct the measurement errors of a camera-based radiation measurement, as an independent radiation measurement is performed using the radiation sensor unit and additional image features are used. The second machine-learning model determines the contribution of diffuse radiation to the hemispheric irradiance based on the diffuse irradiance in the horizontal plane (DHI) and / or the direct irradiance (DNI). Thus, in the present method, the second machine-learning model takes on the role of a decomposition model of the prior art method.

[0026] The diffuse irradiance can be determined in the horizontal plane and / or in an inclined plane. The direct irradiance can be determined in the horizontal plane and / or in the inclined plane. The corrections applied in the horizontal plane can preferably be used for the inclined plane. According to a favorable embodiment of the method, features can be extracted in an image analysis using the camera image, applying a physical camera model.

[0027] The camera model can be a mathematical model that describes the imaging properties of the camera, especially the camera used. The camera model can, in particular, serve to precisely mathematically describe the imaging ratios of the cameras used. The camera model can also be referred to as a physical camera model because it includes physical properties. For example, the most important calculation rules for distance calculation, especially for a non-parallel stereo setup, can be derived.

[0028] In its simplest form, the camera model can be a simple pinhole camera model. In an improved version, it can also include a collinearity model. In this case, a reference coordinate system external to the camera can be taken into account. The camera model can typically require a high algorithmic effort for stereo analysis systems. This has the advantage that a squinting camera arrangement can be achieved with an increase in accuracy. For example, the camera model can also contain an image transformation, such as image rotation and / or image distortion correction. In this case, the camera model can simulate a tracking camera. For example, an improved analysis of the images of clouds in the image can be achieved, in particular a more precise differentiation between cloud components and diffuse sky. Furthermore, this allows the sun's disk to always be positioned in the center of the image.

[0029] The camera model may include a nonlinear gamma correction applied independently to each channel (R, G, B). The camera model may include a constant mixing matrix, in particular a full matrix of size 3 x 3. The camera model may consider the spectral sensitivities of the camera per color channel. The camera model may consider the area covered by each pixel. The camera model may consider the wavelength interval in which the camera responds to the spectral irradiance. The camera model may consider the camera's dark signal.

[0030] Applying the physical camera model can provide the advantage of more accurately determining the diffuse irradiance in the horizontal plane, referred to as DHI. The camera model can be a parameter of the algorithm in the first machine learning model and / or the second machine learning model, and / or results obtained with the camera model can be used as input data for the first machine learning model.

[0031] In one embodiment of the method, the image of the camera can be evaluated via the physical camera model and a preliminary measurement of the diffuse irradiance in the horizontal or in an inclined plane can be obtained, wherein this preliminary measurement of the diffuse irradiance or a direct radiation calculated therefrom is included as a feature by the second machine learning model and / or wherein a direct radiation is determined with the aid of the camera model and the measurement of the hemispherical irradiance and wherein this parameter is included as a feature by the second machine learning model.In a further embodiment of the method, the camera image can be evaluated using the physical camera model, and with the aid of the camera model, circumsolar radiation and / or illuminance or radiation information for each color channel and / or a ratio of the illuminances or radiation information for two different color channels and / or a number of saturated image pixels is determined, and these parameters are included as features by the second machine learning model. Circumsolar radiation is understood to mean the radiation emitted from an area immediately around the sun's disk. It can be advantageous to determine the circumsolar radiation from the camera image, since the radiation from the immediate vicinity of the sun's disk is part of the DNI.In particular, circumsolar diffuse radiation, which results from the scattering of radiation in the atmosphere, can contribute to the DNI. It can also be advantageous to know the illuminance per color channel, since the distribution of illuminance across the color channels reflects the spectral distribution of the solar spectrum. It can be advantageous to determine the number of saturated image pixels from the camera image, as this provides information about the intensity of the radiation hitting the camera, particularly its distribution within the camera image.

[0032] If the procedure is designed in a favourable manner, the following procedural steps can be carried out:

[0033] (i) observed areas of the sky can be assigned to image pixels;

[0034] (ii) and / or a situation-adapted broadband correction can be applied, which is a function of radiance calculated across the camera's various color channels, in particular diffuse irradiance; (iii) and / or the radiance can be integrated or summed in different sky regions, in particular to determine circumsolar radiation or diffuse irradiance in the horizontal plane;

[0035] (iv) and / or intensity values ​​of the color channels or parameters derived therefrom can be summed in a weighted manner;

[0036] (i) and / or image metadata may be included to compensate for any influence of the camera's exposure control;

[0037] (ii) and / or a calibration of the camera using the measurements of the hemispherical irradiance (120) of the radiation sensor unit can be carried out.

[0038] After further structuring the procedure, the following procedural steps can be carried out:

[0039] (i) an estimate of the sky radiance from the camera image based on a camera model can be obtained;

[0040] (ii) and / or the camera model may take into account the spectral sensitivity of the camera per color channel or color matching functions adapted to the color space used;

[0041] (iii) and / or an intrinsic and external geometric calibration of the camera can be used.

[0042] According to a favorable embodiment of the method, the second machine learning model can be different from the first machine learning model, wherein the first machine learning model is trained with first training data sets, and wherein the second machine learning model is trained with second training data sets. The first training data sets can preferably contain reference data. The second machine learning model typically comprises a different algorithm than the first machine learning model. The second machine learning model can have a so-called multilayer perception algorithm (MLP: multilayer perceptron). The data set underlying the second machine learning model includes both the measured value of the hemispheric irradiance of the radiation sensor unit and the result data set of the image of the sky analyzed with the first machine learning model. The data set is also referred to as a data vector or vector data set.

[0043] The second machine-learning model estimates the diffuse irradiance in the horizontal plane (DHI) and / or the direct irradiance (DNI) from the shared data set, in particular the data vector, depending on the second training data used. Typically, the second machine-learning model can be trained for the diffuse irradiance in the horizontal plane (DHI). The hemispheric horizontal irradiance (GHI) and the diffuse irradiance in the horizontal plane (DHI) and the direct irradiance (DNI) are related by a known equation: GHI - DHI = DNI x sin(sun elevation angle). Here, the sun elevation angle is known at any time of day. Thus, the DNI, for example, can be determined from the DHI extracted from the sky image using the second machine-learning model and the measured GHI.

[0044] With the second machine learning model, in addition to the measurement of the hemispheric irradiance with the radiation sensor unit and the image features from the structure and texture of the sky, the following features can also be used to estimate the diffuse irradiance in the horizontal plane (DHI) and / or the direct irradiance (DNI): solar elevation, number of saturated pixels in the image captured by the camera, DHI measurement from the classical analysis using a physical model, DNI measurement from the classical analysis using a physical model, irradiance in the horizontal plane from an angular range of, for example, 25° around the sun, intermediate results from the classical analysis using a physical model, in particular a DHI measurement per color channel of the camera and ratios of these DHI measurements per color channel.The advantage here is that measurement errors caused, for example, by lens errors, overexposure and / or lens refraction effects can be detected.

[0045] The first training data sets for training the first machine learning model can advantageously be extended with the additional features listed above.

[0046] According to a further advantageous embodiment of the method, features can be derived from the intensities of the image pixels during image analysis using the camera image, with a number of saturated pixels being included for feature extraction. From the number of saturated pixels, a statement about the irradiance, in particular about the distribution of the irradiance across the image, can be derived. In this case, setting a threshold value is not necessary.

[0047] According to a further advantageous embodiment of the method, features can be extracted in an image analysis using the camera image. In particular, features can optionally also be extracted from the measurement data of the radiation sensor unit. The image analysis can use information contained in the image pixels. For example, features can be calculated from the intensities of the image pixels. These can be, for example, a number of saturated image pixels, weighted sums of the intensities of the image pixels across the entire image and / or across the image area around the sun, and ratios of these sums.

[0048] According to a further advantageous embodiment of the method, features can be derived from the intensities of the image pixels in the image analysis using the camera image. In particular, a physical camera model can be applied. Alternatively or additionally, a number of saturated pixels and / or at least one sum and / or weighted sum of the intensities of the image pixels and / or at least one ratio of sums and / or weighted sums of the intensities of the image pixels can be included for feature extraction.In this way, advantageous features can be obtained, particularly for the second machine learning model: solar elevation, number of saturated pixels in the image captured by the camera, DHI measurement from the classical analysis using a physical model, DNI measurement from the classical analysis using a physical model, irradiance in the horizontal plane from an angular range of, for example, 25° around the sun, intermediate results from the classical analysis using a physical model, in particular a DHI measurement per color channel of the camera and ratios of these DHI measurements per color channel.

[0049] According to a favorable embodiment of the method, the features extracted from the camera image and / or from the measurement data of the radiation sensor unit can be transferred to a third machine learning model and processed using the third machine learning model, and a radiation sensor unit data set can be generated. The radiation sensor unit data set and the camera result data set can be merged before the second machine learning model is applied. The second machine learning model can be applied to a common data set consisting of features extracted from the measurement data of the radiation sensor unit using the third machine learning model and measurement data of the camera analyzed using the first machine learning model. The common data set can also be referred to as a data vector or vector data set.

[0050] In a favorable design of the method, an image transformation of the sky image can be performed before applying the first machine-learning model. This has the advantage that only image regions with high information content regarding direct radiation are transmitted to the first machine-learning model. This is important when resources are limited.

[0051] A raw image of the sky captured by the camera is subjected to an image transformation. The raw image contains the camera's measurement data prior to any analysis. This allows a transformed image of the sky to be generated. This has the advantage that only image regions with high information content, particularly regarding direct radiation, are transmitted to the first machine learning model. This is important when resources are limited. In addition, errors in diffuse radiation measurements can also be passed on to the first machine learning model. One or more of the following steps can be performed as an image transformation: (i) Applying image rectification and image rotation and image distortion according to a camera model such that the solar disk is always located in the image center.This image transformation can simulate camera tracking, (ii) performing an image crop to a square area around the sun, encompassing angular intervals of approximately 25° around the sun, and (iii) downscaling the resolution of the camera image. For example, a size of 224x224 pixels can be used. This is a common size and has the advantage of allowing the use of known pre-trained initial machine learning models from other applications.

[0052] According to a favorable embodiment of the method, the first machine learning model and / or the second machine learning model can be trained with reference data of at least one of the components of the hemispheric irradiance, in particular the diffuse irradiance in the horizontal plane (DHI), in particular from solar trackers and / or the direct irradiance (DNI), in particular from a pyrheliometer tracking the sun, and / or with regular input data, in particular with images of the camera, data of the hemispheric irradiance and / or intermediate results.

[0053] Intermediate results can be analysis results of measurement data recorded with the measuring device consisting of the radiation sensor unit and the camera, whereby the measurement data are then evaluated using a physical model as described, for example, in WO 2021 / 219570 A1.

[0054] According to a favorable embodiment of the method, the first machine learning model can be pre-trained using at least one of the following steps:

[0055] (i) Use of publicly available weights;

[0056] (ii) training using at least one unsupervised or self-supervised training approach, for example using a DeepCluster v2 method according to CARON, Mathilde et al., in particular with a number of k=30 clusters / groups;

[0057] (iii) augmenting data with Gaussian blurs and / or distorting image colors and / or mirroring images;

[0058] (iv) Training with images of the

[0059] Himmplq The number k=30 is intended as an example. Fewer clusters / groups can also be used, for example, k=25. More clusters / groups can also be used, for example, k=35 or k=40. The number of clusters / groups should be understood as indicating the order of magnitude of the clusters / groups.

[0060] The advantage, especially of step (iii), may be the acquisition of additional training data.

[0061] Pre-training the first machine learning model typically means training the first machine learning model with known datasets. The known datasets preferably contain structures and textures that are similar or identical to those in the sky images measured with the camera.

[0062] According to a favorable embodiment of the method, a so-called convolutional neural algorithm, in particular a convolutional neural network (CNN), can be used in the first machine learning model. Preferably, image features that occur around the sun's surface can be extracted. For example, artifacts such as rays and / or circles around the sun's disk in the image of the sky can indicate the occurrence of lens refraction effects.

[0063] According to a favorable embodiment of the method, at least one algorithm from a multilayer perceptron (MLP), random forest algorithm, recurrent neural network (RNN), LSTM long-short-term memory (LSTM) algorithm, transformer model algorithm, k-nearest neighbor (k-NN) algorithm, and / or support vector machine (SVM) can be used in the second machine learning model. According to a favorable embodiment of the method, the first training data sets and / or the second training data sets can be filtered so that different atmospheric conditions are represented with similar intensity. The atmospheric conditions can be, for example, cloud cover, veil clouds, or partial cloud cover.

[0064] Advantageously, one-minute mean values ​​can be used as the time resolution for processing measurement data according to the method.

[0065] According to a favorable embodiment of the method, the first machine learning model, the second machine learning model, and / or the third machine learning model can be trained in a supervised manner, using input data and / or correct reference data from at least one of the components of the hemispheric irradiance. The at least one component can be diffuse irradiance in the horizontal plane (DHI) or direct irradiance (DNI).

[0066] According to a favorable design of the method, the first machine learning model, the second machine learning model, and the third machine learning model can also be trained jointly, particularly under supervision. This is referred to as fine-tuning. Joint training can take place, in particular, after the individual training of the first machine learning model, the second machine learning model, and the third machine learning model.

[0067] It can also be provided to train the first machine learning model unsupervised. Existing images from a cloud camera can typically be used for this purpose. For example, unsupervised training can use only input data and no reference data. According to a favorable embodiment of the method, image features from the sky image can be used as additional input data for the third machine learning model or for merging, using estimated values ​​of a red channel, a blue channel, and / or a green channel from the measurement data of the sky image. Advantageously, intensity values ​​of the red channel, the blue channel, and / or the green channel can be used.

[0068] According to a further aspect of the invention, a device is proposed for carrying out a method for determining at least one component of a hemispheric irradiance of solar radiation in any plane, wherein the component comprises a diffuse irradiance and / or a direct irradiance. The device comprises at least one radiation sensor unit, a camera, and an evaluation unit which is provided for evaluating measurement data from the radiation sensor unit and the camera. The radiation sensor unit determines the irradiance of solar radiation in a field of view of 180° above a plane, wherein the plane is a horizontal plane or an inclined plane. The camera captures measurement data comprising an image of the sky in a field of view of 180° above a plane, wherein the plane is a horizontal plane or an inclined plane.The evaluation unit has at least a first machine learning model and a second machine learning model. The evaluation unit is configured to perform at least the following steps:

[0069] (i) extracting features, in particular structures, from the image of the sky using a first machine learning model and generating a result data set;

[0070] (ii) Combining the measurement data from the radiation sensor unit and the resulting data set from the camera into a common data set; (iii) Determining the at least one component of the hemispheric irradiance in the arbitrary plane, in particular the diffuse irradiance and / or the direct irradiance, from the data set using a second machine learning model.

[0071] The image of the sky includes an image of the sky within a field of view of up to 180° above the camera. The camera can be, in particular, a surveillance camera or a cloud camera. The image of the sky can also be referred to as a sky image.

[0072] Based on the analysis with the first machine learning model, one or more components of the hemispheric irradiance can be determined from the hemispheric irradiance measured by the radiation sensor unit and the sky image from the camera. In particular, a proportion of the respective components can be determined.

[0073] The at least one component can be determined in any plane. The advantage is that the determined components each exhibit a high degree of accuracy. This is particularly true compared to prior art methods in which the hemispheric irradiance used is determined solely with the camera.

[0074] According to a favorable embodiment of the device, the camera's measurement data can contain angle-resolved radiation information. The structures and textures in the sky image can each be present as angle-resolved information. The structures and textures can be used as image features for analyzing the measurement data by the camera's first machine learning model and form the result data set. According to a further aspect of the invention, a method for training a first machine learning model for extracting features, in particular structures, from an image of the sky is proposed, comprising the following steps:

[0075] (i) Determining or using data sets of images of the sky taken with a camera, in particular a cloud camera

[0076] (ii) Training the first machine learning model with the data sets of images of the sky or

[0077] (iii) Training the first machine learning model with known images of the sky taken at other locations.

[0078] In this way, the first machine learning model can be efficiently trained and used to extract features, especially structures, from an image of the sky.

[0079] According to a favorable embodiment of the method, pre-training with at least one of the following steps can be used to train the first machine learning model:

[0080] (i) Use of publicly available weights;

[0081] (ii) training using at least one unsupervised or self-supervised training approach;

[0082] (iii) augmenting data with Gaussian blurs and / or distorting image colors and / or mirroring the images;

[0083] (iv) Training with images of the sky captured by the camera. When training using at least one unsupervised or self-supervised training approach, DeepCIuster-v2, a method according to CARON, Mathilde et al., can be used, particularly with a number of k=30 clusters / groups.

[0084] In this way, the first machine learning model can be efficiently trained and used to extract features, especially structures, from one of the images of the sky.

[0085] The broadband correction described in the above method and / or the application of the camera model, in particular the physical camera model, is described below in detailed steps.

[0086] The proposed physical method for measuring GHI, DHI, DNI, and finally GTI in a combined ASI (All Sky Imager, cloud camera) and pyranometer setup is described. This method can be used to generate parameters or features that serve as input data for machine learning models. The measurement procedure consists of sequential steps, which are explained below. First, the sky radiance (hereafter referred to as radiance) is calculated from the ASI image based on a camera model. Then, the radiance and the GHI measured by the pyranometer are combined to obtain all irradiance components of interest. This last step is supported by a series of corrections based on the GHI and information from the ASI image processing.Finally, these corrections are parameterized, and self-calibration of the ASI-based DHI to the pyranometer-based GHI is introduced, which is based on conditions with negligible direct irradiance. 1 ) Radiance estimation from ASI-RGB images (cloud camera images):

[0087] The following describes the method for estimating radiance based on an RGB (red-green-blue) image captured by an all-sky imager (ASI). First, the camera model provides the relationship between the spectral irradiance incident on a pixel and the intensity in the ASI image. Estimates of the spectral composition of daylight then enable a pixel-by-pixel estimation of the broadband irradiance. Finally, knowledge of the camera optics allows the solid angle of the sky and pixels to be related, thus determining the sky radiance.

[0088] The camera model relates the spectral irradiance incident on the chip's pixels to the intensities in the R, G, and B color channels of each pixel in the camera image. The radiometric camera model corresponds to the model used by Kuhn et al. (2017) (see Kuhn, P., Wilbert, S., Prahl, C., Schüler, D., Haase, T., Hirsch, T., Wittmann, M., Ramirez, L., Zarzalejo, L., Meyer, A., Vuilleumier, L., Blanc, P., Pitz-Paal, R., 2017. Shadow camera system for the generation of solar irradiance maps. Sol. Energy 157, 157-170. http: / / dx.doi.org / 10.1016 / j.solener.2017.05.074). The camera model includes a non-linear gamma correction that is applied to each channel (R, G, B) independently.

[0089] The camera model includes a constant mixing matrix, a full matrix of size 3 x 3. The camera model takes into account the spectral sensitivities of the camera per color channel. The camera model takes into account the area covered by each pixel. The camera model takes into account the wavelength interval in which the camera responds to the spectral irradiance. The camera model takes into account the camera's dark signal. The camera model has been adapted to the ASI type used, specifically the fisheye camera type.

[0090] Gamma correction was determined according to the method of Grossberg and Nayar (2002) (see Grossberg, MD, Nayar, SK, 2002. What can be known about the radiometric response from images? In: Computer Vision — ECCV 2002 Proceedings, Part IV / 7th European Conference on Computer Vision. Copenhagen, Denmark, 28-31 May 2002. pp. 189-205. http: / / dx.doi.org / 10.1007 / 3-540-47979-1_13) by comparing images with different exposure times. On this basis, the gamma correction applied by the camera is reversed. This is used to calculate intensities of the linearized RGB image. The dark signal is expected to cause a positive offset of the received image intensity. Images taken under completely dark conditions are used to characterize the dark signal for the acquisition settings used. The exposure time is kept constant over time by the camera settings used.In a refined version of the process, series of images are taken with different, precisely defined exposure times.

[0091] The camera's spectral sensitivities are assumed to be proportional to the quantum efficiencies of the camera chip over the wavelength. The constant blending matrix is ​​assumed to be applied by the camera firmware. It corrects for deviations between the actual spectral sensitivities of the camera chip per channel and the spectral sensitivity caused by the color space of the delivered image, which is usually sRGB. The spectral sensitivities and the blending matrix are usually not disclosed by the manufacturer. However, in digital photography, a blending matrix is ​​typically used, resulting in an sRGB image. The product of the spectral sensitivities and the blending matrix can then advantageously be replaced with sRGB-compliant spectral sensitivities.

[0092] The camera used is sensitive in the visible light wavelength range. Based on the specifications of the respective camera, the sensitive wavelength range is conveniently determined, e.g., between 390 nm and 700 nm. Radiation outside this wavelength range is expected to be suppressed by an optical filter. Therefore, the camera model used integrates over this wavelength range to obtain an irradiance in the visible wavelength range. Each pixel mn covers an area on the sensor chip. It is assumed that the irradiance is homogeneously distributed within this small area. Overall, these considerations result in an adapted version of the camera model.

[0093] Next, the irradiance in the visible wavelength range Emn, isible is estimated based on the image's three color channels, which describe the visible spectrum. To do this, a grayscale image is first calculated that responds as evenly as possible to all wavelengths of the visible spectrum.

[0094] ASI images are acquired with a white balance setting corresponding to a constant color temperature. Depending on the color temperature set for the white balance, there is a specific shape of the irradiance spectrum to which the pixels in each color channel c respond with identical intensity. The white balance is reversed to obtain an image S" in which the ratio of the intensities of the color channels corresponds to the ratio of the respective received energy. The image S" is obtained by weighting each color channel with a factor ßc. The factor ßc gives the ratio of the response of channel c when irradiated with a standard daylight spectrum E,CT of a specific color temperature CT compared to its response to illumination with white irradiance EX,white, characterized by a constant spectral irradiance across all X.

[0095] Only the ratio of the responses of the c channels is of interest. Accordingly, the constant EX, white can be chosen arbitrarily. EX, CT is approximated by the spectrum of a blackbody radiator with the corresponding (color) temperature. This determines ß for the color temperature used by the camera.

[0096] To accurately measure Emn, visible from the measured value of a pixel S" mn, a large number of color channels c, each with a known and unique spectral sensitivity, would be required. Based on the three available channels, Emn, isible is measured by summing the channel intensities of the respective pixel in the image S" and scaling with a pixel-wise calibration factor kmn.

[0097] The pixel-wise calibration factor is typically determined by radiometric calibration and assumed to be constant for all scenes. In the setup used, it is determined based only on the ASI and the pyranometer in an automatic self-calibration by comparing the ASI-derived DHI and the pyranometer's GHI measurement under vanishing DNI conditions. This will be explained in more detail later. A further approximation is required to obtain the broadband irradiance. According to the specification of thermopile pyranometers, the broadband irradiance is defined here as the integral of the spectral irradiance over wavelengths in the range [0.3 pm, 3 pm]. Consequently, the irradiance received in the visible spectrum, to which the camera is sensitive, is scaled with a broadband correction factor to estimate the broadband irradiance.

[0098] Using the SMARTS model (Gueymard, 2005) (see Gueymard, CA, 2005. SMARTS Code, Version 2.9.5: Users Manual. Technical Report, Solar Consulting Services, URL: https: / / www.solarconsultingservices.com / SMARTS295_Users_Manual_PC.pdf), the diffuse irradiance spectrum is calculated under clear skies and the broadband correction factor is determined. For the model developed by Wilbert et al. (2016) (see Wilbert, S., Kleindiek, S., Nouri, B., Geuder, N., Habte, A., Schwandt, M., Vignola, F., 2016. Uncertainty of rotating shadowband irradiometers and si-pyranometers including the spectral irradiance error. AIP Conf. Proc. 1734 (1 ), 150009. http: / / dx.doi.org / 10.1063 / 1.4949241 ).

[0099] Calibration conditions with an air mass of 1.4 at an altitude of 500 m above sea level, an optical aerosol depth of 0.1 at a wavelength of 500 nm, and a precipitable precipitation water column of 1.45 cm result in a value of 1.56. This parameter is set constant for all sky conditions. In subsequent processing steps, this broadband correction factor is replaced by a situation-dependent broadband correction factor.

[0100] Finally, the radiance is estimated from the ASI's RGB image by assigning each pixel to the area of ​​sky it observes. ASIs are generally geometrically calibrated using an intrinsic and an external calibration. These calibrations allow each pixel to be assigned to a field of view defined by intervals of the zenith and azimuth angles. The radiance integrated over the solid angle describing the pixel's field of view corresponds to the irradiance received by a pixel for the camera type used.

[0101] The radiance is assumed to be constant over the small solid angle covered by a pixel. Therefore, a discrete representation of the sky's radiance is approximated based on the irradiance recorded by the pixel. If not a single image capture but rather a series of exposures per timestamp is acquired, the procedure described above is performed for each individual capture.

[0102] The radiance matrices of all images are then combined to create an HDR (High Dynamic Range) image of the radiance. HDR algorithms from photography are used as the basis for this step. The combination is realized, for example, as a pixel-by-pixel weighted sum of the individual images, which takes into account the individual exposure times and the intensity of the respective pixel in each image. In this way, the combination excludes information from pixels that are under- or overexposed in the individual images.

[0103] 2) Diffuse radiation:

[0104] The diffuse sky irradiance in a horizontal and possibly inclined plane is calculated by projecting the sky radiance onto the plane and integrating this projected radiance over the part of the sky dome that lies within the plane's field of view (see, e.g., Li, DH, Lam, JC, 2004. Predicting solar irradiance on inclined surfaces using sky radiance data. Energy Convers. Manage. 45 (11-12), 1771-1783. http: / / dx.doi.org / 10.1016 / j.enconman.2003.09.020). The evaluated plane is characterized by the inclination angle of the plane, or the azimuth angle about north. The projection is determined by the angle of incidence between the respective point in the sky dome and the plane normal, such as B. by Westbrook (2015) (see Westbrook, OW, 2015. A sky radiance-based approach to diffuse irradiance transposition. In: 2015 IEEE 42nd Photovoltaic Specialist Conference (PVSC). New Orleans, LA, USA 14-19 June 2015. pp. 1-5. http: / / dx.doi.org / 10.1109 / PVSC.2015. 7356210). The part of the sky over which integration is performed excludes the solid angle of the solar disk and the circumsolar region.

[0105] The circumsolar region is defined in this particular case as an area of ​​the sky with a maximum solar distance angle a < 2.5°, which represents the aperture angle of conventional pyrheliometers and the area blocked by a shadow ball above shaded pyranometers.

[0106] The angle of incidence and radiance are approximated as constant over the solid angle represented by the pixel. In accordance with Li and Lam (2004) (see Li, DH, Lam, JC, 2004. Predicting solar irradiance on inclined surfaces using sky radiance data. Energy Convers. Manage. 45 (11-12), 1771-1783. http: / / dx.doi.org / 10.1016 / j.enconman.2003.09.020), the integration over the sky dome is replaced by a weighted sum in the present case.

[0107] The angle of incidence covered by the pixel is known through intrinsic and external calibration. The diffuse sky irradiance can thus be measured in any plane. The diffuse horizontal irradiance (DHI) is determined in the specific case of diffuse sky irradiance in a plane with a zero tilt angle. The measurement of diffuse sky irradiance relies on the pixel-wise calibration factor. In previous studies, it was determined individually for each pixel through radiometric calibration and assumed to be independent of the observed scene. Accordingly, in a very simplified version of the proposed method, the pixel-wise calibration factor can be adjusted uniformly for all pixels to minimize the deviation between the DHI measured by an ASI and a shaded reference pyranometer.In tests conducted, this resulted in deviations that may only be acceptable for some applications. In this simplified version, the correlation between the reference DHI and the proposed DHI measurement is not yet satisfactory. Therefore, additional correction terms are introduced, as described below.

[0108] 3) Correction and calibration of measurements:

[0109] To improve the accuracy compared to the previously described baseline model and to obtain a model that is more transferable between times, locations, and instruments, the model equation is adjusted. In the simple version of the inventive method described so far, a relatively wide scatter of measurements is observed, especially at low DHI, and a nonlinearity of the observed measurement errors is observed.

[0110] First, a situation-dependent broadband correction factor is applied, replacing the aforementioned static factor. The spectral DHI was measured using spectroradiometers. The inventive measurement system and a solar tracker measuring GHI, DHI, and DNI were operated simultaneously. This allows the modeled and measured DHI from the various sources to be compared. In particular, a function can be determined that maps the ratio of DHI from the red channel of the ASI to the DHI from the blue channel of the ASI to the applicable broadband correction factor. This function was defined as a piecewise linear function, and its parameters were determined by regression, using the ratio of DHI received from the red channel of the ASI to the DHI received from the blue channel of the ASI as the independent variable and the ratio of broadband DHI to the fraction of DHI in the visible wavelength range as the dependent variable.The broadband DHI and the fraction of DHI in the visible wavelength range were both calculated from the spectroradiometer measurements.

[0111] The pixel-wise calibration factor is modeled by two multiplicative parameters, kexp and ksens. Where kexp depends on the exposure control and describes the variable effect of the camera firmware's internal exposure control on the camera's sensitivity, ksens specifies a static sensitivity of a single camera. Both factors are scalars that are applied uniformly to all pixels.

[0112] The determination of ksens and kexp is described later. This simplification neglects vignetting—the potentially reduced sensitivity of the camera near the edge of its field of view. Previous work aimed at measuring sky radiance with ASIs typically used a radiometric calibration that also compensated for vignetting. Following the argument of Chauvin et al. (2015) (see Chauvin, R., Nou, J., Thil, S., Grieu, S., 2015. Modelling the clear-sky intensity distribution using a sky imager. Sol. Energy 119, 1-17. http: / / dx.doi.org / 10.1016 / j.solener.2015.06.026), it can be assumed, however, that vignetting is mainly explained by differences in the size of the solid angles observed by the pixels in the case of a fisheye lens.

[0113] Since this influence can be covered by geometric calibrations in the method according to the invention, a possibly remaining vignetting effect is neglected.

[0114] In addition, the measurement can be extended by an additive correction factor kadd to obtain the final measurement. To account for two overlapping mechanisms that are likely to affect image acquisition near the Sun, the additive correction is chosen as follows: kadd - ksat * nsat - kglare * DNI

[0115] The DNI required in the equation is calculated based on the GHI from the pyranometer and the DHI from the camera image according to the basic relationship

[0116] DNI = (GHI - DHI) / cos(sun_elevation_angle). For the correction, DNI is calculated based on the received DHI without applying the additive correction factor.

[0117] The camera's limited dynamic range leads to saturation of some pixels near the sun, even outside the circumsolar region defined for analysis. Especially in the case of cloud enhancement events (i.e., in the case of an increase in global radiation due to clouds in the angular range around the sun), the diffuse irradiance is then underestimated; ksat accounts for this effect. Nsat is the number of saturated pixels. On the other hand, lens refraction distributes a portion of the DNI over larger areas of the ASI image, which may appear bright and exhibit characteristic rays. Consequently, the diffuse irradiance is overestimated by a fraction kglare of the DNI.

[0118] 4) On-the-fly adjustment of camera sensitivity parameters:

[0119] Illuminance data are used to determine kexp.

[0120] The fisheye camera used in this example, like many other commercial cameras that can be used as all-sky imagers, provides a measurement of the illuminance. At the same time, the illuminance can be estimated from the ASI image. Comparing the two measurements allows us to understand the effects of the camera's exposure control on the image intensity. Furthermore, comparing the two measurements allows for a dynamic determination and continuous update of kexp.

[0121] First, the illuminance is calculated from the ASI image. Similar to calculating the irradiance received by a pixel, the illuminance received by a pixel is calculated from the intensity in the ASI image. However, in this case, the color channels are multiplied by a different weighting (nu_c), and no broadband correction factor is applied.

[0122] Of the correction terms kexp, ksens, ksat, and kglare, only kexp is applied in this step. The ASI firmware is expected to correct for influences caused by its internal exposure control and digital image processing from the measured illuminance. These corrections are represented by kexp. In contrast, the remaining correction parameters are related to the camera's optical and electronic hardware and cannot be considered by the ASI firmware. nu_c specifies the average luminous efficacy of the color channel c in human perception. Assuming an sRGB image and based on the definition of sRGB together with the known color temperature of the image, a special set of nu_c values ​​is used.Similar to the irradiance measurement above, the illuminance recorded by the camera is calculated as an integral, or in a simpler implementation, as the sum of the illuminance per pixel across the camera's field of view. A projection into the sensor plane is required to obtain the actual illuminance for the plane in which the camera is mounted.

[0123] However, since this projection may not be taken into account by the camera firmware, it can be omitted if necessary. This is the case with the camera used in the example, which is why the projection is also neglected in the calculation.

[0124] The illuminance calculated from the image and that provided by the camera firmware are compared.

[0125] In situations with high direct irradiance, the ratio of the two illuminance measurements remains almost constant. This indicates that the camera control system intervenes only minimally in the image exposure process.

[0126] In certain situations, however, especially in darker ones, an increased ratio between the calculated illuminance and the illuminance provided by the camera firmware indicates images that are significantly influenced by the camera's exposure control. It can be detected, in particular, when the camera artificially increases image brightness.

[0127] The ratio of the illuminance measurements is used to correct the irradiance measurement on the fly. The correction factor is automatically set for each timestamp: Kexp = calculated illuminance / firmware illuminance. For this correction, the accuracy of firmware illuminance is not of primary importance. Rather, firmware illuminance serves as an indicator of the camera's internal gain, which is increased in darkness. Overall, this situation-dependent correction factor leads to a noticeable increase in the accuracy of the proposed method.

[0128] 5) Online determination of camera sensitivity using a coupled pyranometer:

[0129] The combined setup of ASI and pyranometer allows for continuous calibration of the ASI's sensitivity: Under diffuse conditions (DNI = 0 Wfn2), GHI should be equal to DHI. A diffuse situation is detected from the ASI image when the number of saturated pixels is less than 100. During development, it was found that this threshold reliably filters out situations with DNI > 0 Wfn2. The value will likely need to be slightly adjusted for other camera types.

[0130] After a sufficient number of scenes with vanishing DNI have been observed, ksens is set to minimize the RMSD (root mean square error) between GHI and DHI for this filtered dataset. For these diffuse situations, the additive corrections kadd are very close to zero and are therefore not applied.

[0131] Since diffuse situations are detected automatically and without additional information using this method, ksens can be determined using this method for any other ASI and at any location. 6) Image saturation:

[0132] A weakness of commercially available fisheye surveillance cameras that can be used as ASIs is their limited dynamic range. Therefore, the images often exhibit saturation in the circumsolar region. High turbidity and optically very thin clouds increase the scattering of irradiance, resulting in the impression of a large solar disk in the image. Furthermore, clouds reflect solar radiation during cloud enhancement events. These clouds can also appear saturated in the ASI image. In parts of the sky dome corresponding to such saturated pixels, the radiance is underestimated. Consequently, in situations of image saturation, the diffuse irradiance can also be underestimated by the method. This effect can be analyzed by looking at the relative frequency with which pairs of DHI measurement deviations and a given number of saturated pixels were observed.This investigation revealed an additive error that scales linearly with the number of saturated pixels nsat.

[0133] The ksat parameter is estimated by multivariate regression performed jointly on the ksat and kglare parameters. The RMSD between the ASI-based final DHI and the reference DHI is used as the cost to be minimized. The ksat correction factor is assumed to be constant between locations as long as the camera type remains unchanged.

[0134] 7) Lens refraction in the lens:

[0135] Conventional ASIs sample the entire sky dome without a shadowing device. This setup is susceptible to lens refraction effects under direct irradiance. These effects can lead to a systematic positive bias in the measured radiance near the Sun, as shown by Mejia et al. (2016) (see Mejia, F.A., Kurtz, B., Murray, K., Hinkelman, L.M., Sengupta, M., Xie, Y., Kleissl, J., 2016. Coupling sky images with radiative transfer models: a new method to estimate cloud optical depth. Atmos. Meas. Tech. 9 (8), 4151—4165. http: / / dx.doi.org / 10.5194 / amt-9-4151-2016). Lens refraction by a point light source is expected to spread the radiance from that point over a larger image area. It is conclusive that the scattered irradiance scales linearly with the irradiance of the light source and overlays the actual sky radiation of interest.It is therefore advisable to use the additive lens refractive correction kglare.

[0136] The extent of lens refraction depends on the optical properties of the camera lens. Therefore, an attempt is made to determine kglare for each individual camera model and camera based on clear-sky periods, using only the combination of ASI and pyranometer. The GHI time series are checked for potentially clear periods. In the second step, the ASI images of these periods are examined, and only the periods with very low cloud cover are considered.

[0137] These steps are performed manually but can be easily automated by cloud segmentation techniques capable of reliably detecting such clear skies (e.g., Fabel, Y., Nouri, B., Wilbert, S., Blum, N., Triebel, R., Hasenbalg, M., Kuhn, P., Zarzalejo, LF, Pitz-Paal, R., 2021. Applying self-supervised learning for semantic cloud segmentation of all-sky images. Atmos. Meas. Tech. Discuss. 2021 , 1-20. http: / / dx.doi.Org / 10.5194 / amt-2021 -1 ). For the remaining clear-sky periods, GHI and DNI, and subsequently DHI, are modeled by the Ineichen clear-sky irradiance model (see Ineichen, P., Perez, R., 2002. A new airmass independent formulation for the Linke turbidity coefficient. Sol. Energy 73 (3), 151-157. http: / / dx.doi.org / 10.1016 / S0038-092X(02)00045-2), which relies only on the Linke turbidity, the current solar geometry, and the site elevation.For each clear-sky period, a numerical solver is used to find a pair of Linke turbidity and kglare that minimizes the sum of the MADs (mean absolute value of error) of measured and modeled GHI, DNI, and DHI, respectively. A Linke turbidity of 2 and kglare = 0.030 are used as the starting point for the minimization.

[0138] The DN I, used as input for lens refraction correction, is calculated from GHI and DHI as described above, using the preliminary DHI that has not yet been corrected for lens refraction. When subsequently measuring the final diffuse irradiance in a horizontal or inclined plane, the DNI calculation is repeated using the corrected DHI.

[0139] Both additive corrections together significantly improve the measurement accuracy.

[0140] The described physical method for measuring DNI and DHI yields numerous parameters that serve as important input parameters for the machine learning models. In particular, the following parameters, determined using the method described above, have proven to be very beneficial for achieving high accuracy with the machine learning models:

[0141] - Global radiation GHI

[0142] - Diffuse radiation DHI with or without application of kadd

[0143] - Direct radiation DNI with or without application of kadd

[0144] Circumsolar radiation. Diffuse radiation received from a region of the sky with a solar distance angle of up to 10°–25°, preferably 25°, with or without the use of kadd.

[0145] - Radiation information calculated from the camera image for each color channel, in particular illuminance for each color channel (red, green, blue)

[0146] - Ratio of the illuminance of the green colour channel calculated from the camera image to the illuminance of the red colour channel or ratio of the radiation information of the green colour channel calculated from the camera image to the radiation information of the red colour channel or

[0147] - Ratio of the illuminance of the blue color channel calculated from the camera image to the illuminance of the red color channel or ratio of the radiation information of the blue color channel calculated from the camera image to the radiation information of the red color channel

[0148] - Sun elevation angle

[0149] - Number of saturated pixels nsat.

[0150] According to a further aspect of the invention, a computer program for determining at least one component of a hemispheric irradiance of solar radiation in any plane is proposed. The program comprises instructions which, when executed by a computer, cause the computer to perform the steps of a method for determining at least one component of a hemispheric irradiance of solar radiation in any plane. The computer program can advantageously be used in a computer of a data processing device of a device for determining at least one component of a hemispheric irradiance of solar radiation in any plane.

[0151] According to a further aspect of the invention, a data processing device for a device for determining at least one component of a hemispheric irradiance of solar radiation is proposed, which comprises at least one measurement data acquisition unit, an evaluation unit and a computer.

[0152] The data processing device can advantageously be used in a device for determining at least one component of a hemispherical irradiance of solar radiation in any plane.

[0153] According to a further aspect of the invention, a trained first machine learning model is proposed for a method for determining at least one component of a hemispheric irradiance of solar radiation in any plane. The trained first machine learning model serves to extract features, in particular structures, from an image of the sky and is trained according to a method for training the first machine learning model.

[0154] The trained machine learning model can advantageously have been trained with initial training data sets that contain image features relevant to an image of the sky.

[0155] The trained first machine learning model can advantageously be used in a device for determining at least one component of a hemispheric irradiance of solar radiation in any plane. According to a further aspect of the invention, a trained second machine learning model is proposed for a method for determining at least one component of a hemispheric irradiance of solar radiation in any plane. The trained second machine learning model serves to determine at least one component of a hemispheric irradiance of solar radiation in any plane. The trained second machine learning model can be trained in a supervised manner, wherein, in particular, input data and / or reference data can be used. The trained second machine learning model has, in particular, a multilayer perceptron algorithm.The trained second machine learning model can advantageously be trained using second training data sets. The second training data sets can contain image features.

[0156] The trained second machine learning model can advantageously be used in a device for determining at least one component of a hemispheric irradiance of solar radiation in an arbitrary plane.

[0157] drawing

[0158] Further advantages will become apparent from the following description of the drawings. The figures illustrate exemplary embodiments of the invention. The figures, the description, and the claims contain numerous features in combination. Those skilled in the art will also expediently consider the features individually and combine them into useful further combinations.

[0159] Examples include:

[0160] Fig. 1 shows a schematic structure of a device for determining at least one component of a hemispherical irradiance of solar radiation in any plane in a plan view; Fig. 2 shows the structure of the device according to Fig. 1 according to an embodiment of the invention in a side view;

[0161] Fig. 3 shows the structure of the device according to Fig. 1 in an alternative setup in a side view, wherein the planes of the radiation sensor unit and the camera are inclined relative to the horizontal plane;

[0162] Fig. 4 shows a schematic representation of an image of the sky taken with a camera;

[0163] Fig. 5 is a flowchart of an embodiment of the method for determining at least one component of a hemispheric irradiance of solar radiation;

[0164] Fig. 6 is a flowchart of a further embodiment of the method for determining at least one component of a hemispheric irradiance of solar radiation;

[0165] Fig. 7 is a flowchart of a further embodiment of the method for determining at least one component of a hemispheric irradiance of solar radiation;

[0166] Fig. 8 is a flowchart of a further embodiment of the method for determining at least one component of a hemispheric irradiance of solar radiation;

[0167] Fig. 9 is a flowchart of another embodiment of the method for determining at least one component of a hemispheric irradiance of solar radiation;

[0168] Fig. 10 is a flowchart of the method according to Fig. 9 in a further embodiment;

[0169] Fig. 11 Parameters for training the first machine learning model.

[0170] Embodiments of the invention

[0171] In the figures, similar or similarly functioning components are numbered with the same reference numerals. The figures show only examples and are not to be understood as limiting. Before the invention is described in detail, it should be pointed out that it is not limited to the respective components of the device or the respective method steps, since these components and methods can vary. The terms used here are intended only to describe particular embodiments and are not used in a limiting manner. Furthermore, when the singular or indefinite article is used in the description or in the claims, this also refers to the plural of these elements, unless the overall context clearly indicates otherwise.

[0172] The directional terminology used below, including terms such as “left”, “right”, “top”, “bottom”, “before”, “behind”, “after” and the like, is intended solely to improve understanding of the figures and is in no way intended to limit the scope of the illustrations.

[0173] The components and elements shown, their design and use may vary according to the considerations of a specialist and may be adapted to the respective applications.

[0174] Figure 1 shows a schematic structure of the device 5 for determining a hemispherical irradiance of solar radiation in any plane and its components: direct irradiance in a horizontal plane (GHI) or in an inclined plane (GTI), diffuse irradiance in the horizontal plane (DHI), direct irradiance (DNI), and irradiance reflected from the ground, in a plan view. The device 5 has a measuring unit 10, which has a radiation sensor unit 12 and a camera 14. Furthermore, a data processing device 20 with a measurement data acquisition unit 16 and an evaluation unit 18 is provided. The evaluation unit 18 is provided for evaluating measurement data from the radiation sensor unit 12 and / or the camera 14.The data processing device 20 with a control and regulation unit, a measurement data acquisition unit 16 and an evaluation unit 18 can, for example, realize a synchronous measurement data acquisition of the radiation sensor unit 12 and the camera 14.

[0175] The measurement data acquisition unit 16 is configured to acquire the measurement data of the radiation sensor unit 12 and the camera 14. The

[0176] Evaluation unit 18 is set up to process the measurement data of the

[0177] radiation sensor unit 12 and the camera 14. The

[0178] Evaluation unit 18 can have at least one first machine learning model, preferably a second and third machine learning model. The data processing device 20 or the evaluation unit 18 has a computer 19. The at least one first machine learning model is configured to analyze the measurement data from the camera 14.

[0179] Typically, the first trained machine learning model is implemented in the evaluation unit 18. It may be provided to train the first machine learning model in or with the evaluation unit 18. The particular design depends on the size and structure of the data processing device 20. The data processing device 20 may be a computer with a small storage capacity and low computing power. In this case, the training of the first machine learning model can take place on a server with a large storage capacity and high computing power, and the trained first machine learning model can only be stored on the data processing device 20. If necessary, the evaluation of the raw data during measurement operation can also take place on a server in order to use even fewer resources on the computer.The radiation sensor unit 12 and the camera 14 can advantageously be arranged in close proximity to each other, horizontally leveled and at the same height. In addition, the installation location can be expediently selected so that further obstacles in the fields of view 26, 32 (the fields of view 26, 32 are shown in Figures 2 and 3) of the radiation sensor unit 12 and the camera 14 are avoided.

[0180] According to one embodiment of the invention, the radiation sensor unit 12 and the camera 14 can be arranged in the horizontal plane 36, which is shown in Figure 2. In another embodiment, which is shown in Figure 3, the radiation sensor unit 12 and / or the camera 14 are each arranged in a plane 46 and 48 inclined to the horizontal plane.

[0181] The radiation sensor unit 12 comprises a housing 22 and a sensor 24. The sensor 24 detects radiation in an upper half-space with a field of view 26 of 180°.

[0182] The radiation sensor unit 12 measures the irradiance of solar radiation in an upper half-space in a field of view 26 of 180° above the plane 38 in which it is arranged.

[0183] The half-space refers to a hemisphere, also called a hemisphere, which covers a solid angle of at least approximately 2TT steradians.

[0184] The radiation sensor unit 12 can be configured to detect solar radiation in a wavelength range from 0.3 pm to 3 pm. In particular, the sensor 24 can detect light in the wavelength range from 0.3 pm to 3 pm. The radiation sensor unit 12 is configured, for example, as a pyranometer, in particular a thermopile pyranometer. The radiation sensor unit 12 can alternatively also be configured as a photodiode or photovoltaic reference cell if lower requirements are placed on the accuracy of determining the hemispheric irradiance.

[0185] The camera 14 is designed, for example, as a fisheye surveillance camera, in particular, for example, as a Mobotix Q25 surveillance camera or as a cloud camera. The camera 14 comprises a housing 28, above which a sensor 30 is arranged to capture the upper half-space with a field of view 32 of 180°. The camera 14 is preferably designed to record the sky in the field of view 32. The camera 14 generates an image 110 of the sky 50 (this is shown in Figure 4). The image 110 of the sky 50 contains, for example, information about clouds, the solar disk, and areas around the solar disk.

[0186] The camera 14 is advantageously designed so that it can capture the entire field of view 32 in a single shot without the need for mechanically moving parts.

[0187] The camera 14, in particular the properties of the camera 14, are described in a camera model. The camera model typically includes the imaging properties of the camera 14 and is or includes a mathematical description of the properties, in particular the imaging properties of the camera 14. The camera model is used in the evaluation of an image captured by the camera 14.

[0188] The camera model includes gamma correction, specifically nonlinear gamma correction. The gamma correction is determined according to the method of Grossberg and Nayar (2002) (see Grossberg, MD, Nayar, SK, 2002. What can be known about the radiometric response from images? In: Computer Vision — ECCV 2002 Proceedings, Part IV / 7th European Conference on Computer Vision. Copenhagen, Denmark, 28-31 May 2002. pp. 189-205. http: / / dx.doi.org / 10.1007 / 3-540-47979-1_13) by comparing images with different exposure times.

[0189] The camera model further includes a constant mixing matrix, in particular a full matrix of size 3 x ß. The camera model takes into account the spectral sensitivities of the camera per color channel.

[0190] The camera model considers the area covered by each pixel. The camera model considers the wavelength range in which the camera responds to the spectral irradiance. The camera model considers the camera's dark signal.

[0191] The camera model is adapted to the camera type used, specifically the fisheye camera type. Based on this, the gamma correction applied by the camera is reversed. This is used to calculate the intensities of the linearized RGB image. The dark signal is expected to cause a positive offset in the received image intensity. Images taken under completely dark conditions are used to characterize the dark signal for the recording settings used.

[0192] The exposure time is kept constant over time by the camera settings used. In a refined version of the process, image series are taken with different, precisely defined exposure times.

[0193] The camera's spectral sensitivities are assumed to be proportional to the quantum efficiencies of the camera chip over the wavelength. The constant blending matrix is ​​assumed to be applied by the camera firmware. It corrects for deviations between the actual spectral sensitivities of the camera chip per channel and the spectral sensitivity caused by the color space of the delivered image, which is usually sRGB. The spectral sensitivities and the blending matrix are usually not disclosed by the manufacturer. However, in digital photography, a blending matrix is ​​typically used, resulting in an sRGB image. The product of the spectral sensitivities and the blending matrix can then advantageously be replaced with sRGB-compliant spectral sensitivities.

[0194] The camera model integrates over a fixed wavelength range for the reasons listed below to obtain an irradiance in the visible wavelength range.

[0195] The camera used is sensitive in the visible light wavelength range. Based on the specifications of the respective camera, the sensitive wavelength range is conveniently determined, for example, between 390 nm and 700 nm. Radiation outside this wavelength range is expected to be suppressed by an optical filter. Each pixel mn covers an area on the sensor chip. It is assumed that the irradiance is distributed homogeneously within this small area. Overall, these considerations result in an adapted version of the camera model.

[0196] The camera model also takes into account a white balance, which is reversed to obtain an image S" in which the ratio of the intensities of the color channels corresponds to the ratio of the respective received energy. The image S" is obtained by weighting each color channel with a factor ßc. The factor ßc gives the ratio of the response of channel c when irradiated with a standard daylight spectrum E,CT of a certain color temperature CT compared to its response to illumination with white irradiance EX,white, characterized by a constant spectral irradiance over all X. Only the ratio of the responses of the channels c is of interest. Accordingly, the constant EX,white can be chosen arbitrarily. EX,CT is approximated by the spectrum of a blackbody radiator with the corresponding (color) temperature. This is how ß is determined for the color temperature used by the camera.

[0197] Furthermore, the camera model has a pixel-wise calibration factor, which is typically determined by a radiometric calibration. This is interesting because measuring Emn, visible accurately from the measured value of a pixel S" mn, would require a large number of color channels c, each with a known and unique spectral sensitivity. Based on the three available channels, Emn, visible is measured by summing the channel intensities of the respective pixel in the image S" and scaling by the pixel-wise calibration factor.

[0198] The radiation sensor unit 12 and the camera 14 are arranged at a distance 34, each measured from a center point of the sensors 24 and 30. The distance 34 is preferably less than 3 m. The minimum distance 34 is determined by the diameters of the respective housings 22 and 28.

[0199] Figure 2 shows the device 5 in a schematic representation in a side view. The horizontal plane on which the radiation sensor unit 12 and the camera 14 are arranged is designated by the reference numeral 36. The sensor 24 of the radiation sensor unit 12 is arranged in or above a plane 38. The sensor 30 of the camera 14 is arranged in or above a plane 40. In the exemplary embodiment shown in Figure 2, the planes 38 and 40 run parallel to the horizontal plane 36. Figure 3 shows the structure of the device 5 according to Figure 1 in an alternative arrangement in which the plane of the radiation sensor unit 12 and the plane of the camera 14 are inclined relative to the horizontal plane 36. The camera 14 is arranged in the plane 48. The radiation sensor unit 12 is arranged in the plane 46. The sensor 24 of the radiation sensor unit 12 is arranged in or above the plane 46. The sensor 30 of the camera 14 is arranged in or above the plane 48.

[0200] The device 5 comprises the radiation sensor unit 12, the camera 14, as well as the data processing device 20 with the measurement data acquisition unit 16 and the evaluation unit 18 (shown in Figure 1), which is provided for evaluating measurement data from the radiation sensor unit 12 and / or the camera 14, and the computer 19. The radiation sensor unit 12 is provided for determining the irradiance of solar radiation in a field of view 26 of 180°, i.e. the half-space above the plane 38, 46, and the camera 14 is also provided for detecting a field of view 32 of 180°, i.e. the half-space above the plane 40, 48.

[0201] In the setup of the device 5 shown in Figure 1, the two planes 38, 40 are aligned in the horizontal plane 36 and coincide with the horizontal plane 36.

[0202] In the setup of the device 5 shown in Figure 3, the planes 46, 48 of the radiation sensor unit 12 and the camera 14 are each inclined at an angle 42 and an angle 44 to the horizontal plane 36. The angle 44 can be set between 0° and 90°. In this exemplary embodiment, the radiation sensor unit 12 and the camera 14 are arranged, for example, on a north-south axis, with the radiation sensor unit 12 being arranged north of the camera 14 in the northern hemisphere, as shown in Figure 2. In the southern hemisphere, the radiation sensor unit 12 would be arranged south of the camera 14.In the advantageous embodiment shown here, the sensor 24 of the radiation sensor unit 12 and the sensor 30 of the camera 14 are each arranged in the horizontal plane 36 such that the field of view 26, 32 of the two sensors 24, 30 is each above the horizontal plane 36 and ends with the horizontal plane 36.

[0203] The distance 34 between the radiation sensor unit 12 and the camera 14 can be adjusted such that the sensor 24 of the radiation sensor unit 12 is visible in the field of view 32 of the camera 14 with an elevation 43 of at most 10°.

[0204] The radiation sensor unit 12 and the camera 14 are coupled in such a way that the measurement data are recorded by the radiation sensor unit 12 and the camera 14 in a synchronized manner. The measurement data can be conveniently evaluated in the evaluation unit 18.

[0205] The measurement data recording of the radiation sensor unit 12 can advantageously be carried out with a high temporal resolution, in particular with a temporal resolution in the second range, in particular of less than 10 s, preferably less than 5 s, particularly preferably less than or equal to 1 s.

[0206] The camera 14 can expediently be configured such that a single image is captured at a fixed time interval, in particular every half and full minute. Advantageously, the at least one sensor 30 of the camera 14 can have a constant color temperature. The camera 14 can advantageously have a constant exposure time for each individual image. For the exposure control of the camera 14, a predetermined minimum value of an average image brightness can be set, wherein the exposure time remains unchanged at a higher image brightness. In particular, the predetermined minimum value of an average image brightness can preferably be at most 10%, more preferably at most 8%, most preferably at least 5%.

[0207] The image from camera 14 provides the image 110 of the sky 50 in real time, which is shown in Figure 4. The image 110 of the sky 50 contains raw data about the sky.

[0208] Shown is a typical camera image, specifically a fisheye lens with a field of view 26 of 180° in the half-space above the horizontal. Near the zenith, the bright solar disk 54 can be seen, while clouds 52 can be seen at the edge of the image, i.e., more toward the horizon.

[0209] The image of the sky can advantageously contain the following information:

[0210] - Clouds and cloud cover,

[0211] - Thin cirrus clouds,

[0212] - solar disk, shaded solar disk,

[0213] - lens refraction effects around the solar disk,

[0214] - Shape of the solar disk, cloudiness of the solar disk,

[0215] - Clarity of the edge of the solar disk,

[0216] - Decrease in brightness in the vicinity of the solar disk.

[0217] This information is referred to as image features. If a first machine learning model is applied to image 110 of sky 50, these image features can be recognized and evaluated.

[0218] The image 110 of the sky 50 can be recorded as a digital image. The digital image 110, also referred to as the sky image 110, has a matrix of pixels, referred to as a pixel matrix, wherein the information is contained as the intensity of each pixel, referred to as the pixel intensity. Each pixel has a red channel, a green channel, and a blue channel. Each of the channels thus has its own pixel intensity. Thus, the information of the image is contained in the pixel intensities of the red channel, the green channel, and the blue channel and in the pixel position in the pixel matrix. The pixel position contains the information about the location or position in the sky assigned to the respective pixel. Thus, the pixel intensity of the pixel matrix contains radiation information. The image of the sky contains angle-resolved radiation information about the pixel position and the pixel intensity of the pixel in the pixel matrix.

[0219] Figures 5 to 9 describe methods S100, S200, S300, S400, S500 that use machine learning models to analyze images 110 from camera 14 and measurement data from radiation sensor unit 12. Reference numeral 122 denotes the diffuse irradiance in the horizontal plane extracted using one of methods S100, S200, S300, S400, S500.

[0220] Reference numeral 124 denotes the direct irradiance extracted using one of the methods S100, S200, S300, S400, S500. The diffuse irradiance in the horizontal plane and the direct irradiance, which were determined or measured by other means, are used without reference numerals.

[0221] Figure 5 shows an embodiment of a method S100 for analyzing the information recorded by the device 5. The method steps S130 to S150 are listed below:

[0222] In method step S130, an image 110 captured with the camera 14 is analyzed with a first machine learning model by extracting image features, in particular structures and textures, from the image 110. In method step S140, the image features and a value of the hemispheric irradiance 120 captured with the radiation sensor unit 12 are concatenated, and a common data set, in particular a data vector, also referred to as a vector data set, is formed.

[0223] In process step S150, the joint data set is analyzed with a second machine learning model and values ​​for the diffuse irradiance 122 in the horizontal plane (DHI) and the direct irradiance (DNI) 124 are extracted.

[0224] The first machine learning model is preferably a pre-trained machine learning model. The first machine learning model can, for example, be a so-called convolutional neural network (CNN). The CNN is a so-called deep learning algorithm that is primarily used for analyzing images and videos. It extracts visual features—image features—from images.

[0225] The image features are present in the images as structures and / or textures that form a pattern. The CNN is typically trained with a large amount of data, especially images with structures and textures, and then applied to new images to analyze them, especially automatically.

[0226] The term neural network refers to a logical structure inspired by the human brain and can form the basis for deep learning algorithms. For example, the term deep learning can be used to describe a multi-layer neural network that learns from large data sets and from image features contained in the data sets. The term convolutional neural network means that convolutional layers with filters are applied to the input data to learn the image features. Typically, there are multiple convolutional layers connected by pooling layers. The first convolutional layers extract general or low-level features such as lines and edges, while the later layers learn finer details or high-level features such as the sun's disk.Pooling layers are used to reduce the size of the convolutional features and thus the computing time and thus the computing costs.

[0227] Fully connected layers learn global patterns based on the high-level features output by the convolutional and pooling layers, generating global patterns, such as those for the solar disk or clouds. After the input data passes through the fully connected layer, the final layer activates the machine learning model, allowing the neural network to make its predictions.

[0228] The first machine learning model can be trained differently than the second machine learning model. For example, a multilayer perceptron (MLP) model can be used as the second machine learning model. Different training means that different data sets are used and the algorithm is different.

[0229] Figure 6 shows an extended embodiment of the method S100 described in Figure 5, which is referred to as method S200.

[0230] The method S200 comprises the following steps:

[0231] In method step S230, the first machine learning model is applied to the camera image 110, and image features are extracted from the image 110 of the sky 50. In method step S235, an image analysis is performed using the image 110 from the camera 14 and / or other information. Optionally, measured values ​​of the hemispheric irradiance 120 can also be used to extract features.

[0232] Advantageously, in the image analysis in step S235, using the image 110 from the camera 14, features can be derived from the intensities of the image pixels. In particular, a physical camera model can be applied. Alternatively or additionally, a number of saturated pixels and / or sums and / or weighted sums of the intensities of the image pixels and / or ratios of sums and / or weighted sums of the intensities of the image pixels can be included for feature extraction.

[0233] In this way, advantageous features can be obtained, particularly for the second machine learning model: solar elevation, number of saturated pixels in the image captured by the camera, DHI measurement from the classical analysis using a physical model, DNI measurement from the classical analysis using a physical model, irradiance in the horizontal plane from an angular range of, for example, 25° around the sun, intermediate results from the classical analysis using a physical model, in particular a DHI measurement per color channel of the camera and ratios of these DHI measurements per color channel.

[0234] When analyzing the camera image 110 in the image analysis in method step S235, the following features can advantageously be obtained:

[0235] - solar elevation,

[0236] - Number of saturated pixels in the image 110,

[0237] - Measurement of diffuse irradiance in the horizontal plane using physical models or analyzed values ​​of diffuse irradiance from the analysis of the camera image using the physical model,

[0238] - Determining the direct normal radiation using physical models or analyzed values ​​of the direct radiation from the analysis of the camera image using the physical model,

[0239] - Measurement of irradiance in the horizontal plane from an angular range of 25° around the sun, using only direct radiation

[0240] - Estimated illuminance (radiance) from the red channel, the green channel and the blue channel of image 110,

[0241] - Ratio of illuminance estimates based on each channel.

[0242] The direct normal radiation as well as the diffuse irradiance is determined using physical models or analyzed values ​​from the analysis of the camera image using the physical model.

[0243] The features listed above are particularly suitable for determining at least one component of a hemispheric irradiance of solar radiation using the second machine learning model, as this allows for the determination of errors inherent in evaluations using physical models. This is because there is a close relationship between the features listed above and such a measurement of diffuse irradiance in the horizontal plane. For example, the irradiance in the horizontal plane from an angular range of 25° around the sun shows measurement errors that may have been caused by overexposure and / or lens refraction effects. Irradiance estimates based on the individual color channels (red channel, green channel, blue channel) and their ratios provide information on the spectral composition of the diffuse irradiance.The spectral composition has an influence on the measured values ​​of the diffuse irradiance 122, in particular the accuracy of the measurement of the intensity of the diffuse irradiance 122. If the spectral composition of the diffuse irradiance 122 is known, the measured values ​​of the diffuse irradiance 122 can be corrected accordingly.

[0244] In process step S240, the extracted data from process step S230 and process step S235, as well as the hemispheric irradiance 120, are concatenated. Features of the resulting data set, also referred to as a vector data set, contain data from S230, S235, and the hemispheric irradiance 120.

[0245] In method step S250, the second machine learning model is applied to the vector data set to extract the diffuse irradiance 122 in the horizontal plane and / or the direct irradiance 124.

[0246] Figure 7 shows a further extended process S300 with the following process steps:

[0247] In method step S315, an image transformation is performed using the measured values ​​of the image 110 captured by the camera 14. In method step S315, an image rectification and / or image rotation are / is performed. The image rectification and / or image rotation is applied according to the camera model. As a result, the sun disk 54 is always positioned in the center of the image in the camera image 110. This simulates a camera 14 tracking the sun 54. Furthermore, in method step S315, the image is cropped to a quasi-square area around the sun. This includes angular distances of up to 25° around the sun. Furthermore, the resolution of the camera image is scaled down, for example, to a format of 224x224 pixels. This enables pre-trained machine learning models from other applications to be used to analyze the image data from the camera 14.This saves the time and cost of pre-training the initial machine learning model. Furthermore, larger data sets are available for use as training datasets.

[0248] In method step S330, the first machine learning model is applied to the camera image 110 to extract image features from the image 110 of the sky 50.

[0249] In method step S335, the camera image 110 is analyzed as described in method step S235 and optionally using the hemispheric irradiance 120 and / or other information.

[0250] In method step S340, the extracted data from method step S330 and method step S335, as well as the hemispheric irradiance 120, are concatenated, and a vector data set is generated. Features of the data set include data from S330, S335, and the hemispheric irradiance 120. In method step 350, the second machine learning model is applied to the vector data set to extract the diffuse irradiance 122 in the horizontal plane and / or the direct irradiance 124.

[0251] Figure 8 shows a further embodiment of a method S400 for determining the diffuse irradiance 122 in the horizontal plane and / or the direct irradiance 124. The method S400 is similar to the method S300 shown in Figure 7. However, no image analysis, as described in step S335, is performed. Image features are extracted solely using the first machine learning model in method step S430.

[0252] The S400 process comprises the following process steps:

[0253] In method step S415, an image transformation is carried out which corresponds to the image transformation described in method step S315.

[0254] In method step S430, the first machine learning model is applied to the camera image 110 after method step S315 to extract image features from the image 110 of the sky 50.

[0255] In process step S440, the extracted data from process step S430 and the hemispheric irradiance 120 are concatenated. Features of the data set contain data from S430 as well as values ​​of the hemispheric irradiance 120.

[0256] In method step S450, the second machine learning model is applied to the vector data set to determine the diffuse irradiance 122 in the horizontal plane and / or the direct irradiance 124. Figure 9 shows a further embodiment of a method S500, which has the following steps. In this case, the method S300 is extended by a third machine learning model in step S538.

[0257] In method step S515, an image transformation of the image 110 of the sun 54 is carried out, which corresponds to the image transformation described in method step S315.

[0258] In method step S530, the first machine learning model is applied to the camera image 110 after method step S515 to extract image features from the image 110 of the sky.

[0259] In method step S535, an image analysis of the data of the image 110 and optionally the hemispherical irradiance 120 recorded with the radiation sensor unit 12 is carried out similarly to the method step S235 described above.

[0260] In method step S538, the third machine learning model is used to extract features from the hemispheric irradiance 120 and from the features extracted in method step S535.

[0261] In method step S538, the third machine learning model is applied to the measurement data of the hemispheric irradiance 120, which was analyzed together with the information from the camera image 110 in method step S535. This results in the vector data set being less complex than the data set passed to the second machine learning model in methods S100, S200, S300, and S400. This reduces the training effort for the second machine learning model. At the same time, potential overfitting can be at least partially avoided.

[0262] In process step S540, the extracted data from process step S530 and process step S538 of the hemispheric irradiance 120 are concatenated, and a vector data set is generated. Features of the data set contain data from S530 and S538.

[0263] In method step S550, the second machine learning model is applied to the vector data set to determine the diffuse irradiance 122 in the horizontal plane and / or the direct irradiance 124.

[0264] It is also possible to combine the S100, S200, S300, S400, and S500 methods differently than shown in the figures. Furthermore, the components of each machine learning model can be weighted. This involves applying a special training procedure for the machine learning models.

[0265] The machine learning models can be trained differently, meaning different training strategies and / or different training datasets are used. Different algorithms are also used for the three machine learning models.

[0266] The training process for the first machine learning model can be supervised or unsupervised. If the first machine learning model is trained supervised, it can include the steps and strategies listed below and use data that uses one or more datasets from the list below:

[0267] - Data on diffuse irradiance in the horizontal plane, for example from solar trackers,

[0268] - data from a shaded radiation sensor unit 12, in particular pyranometer,

[0269] - direct normal radiation of a sun-tracking pyrheliometer,

[0270] - regular input data, for example images from a cloud camera,

[0271] - regular input data, for example data of the hemispheric irradiance, intermediate results of an evaluation with physical models, data of a radiation sensor unit 12, in particular pyranometer.

[0272] When the first machine learning model is trained unsupervised, images 110 from camera 14, in particular from a cloud camera, are used. For example, existing images 110 from cloud camera 14 or existing images from any cloud camera can be used. In particular, image features in the region of the sun 54 are extracted.

[0273] Training can be carried out using at least one unsupervised or self-supervised training approach, for example using a DeepCluster v2 method according to CARON, Mathilde et al., in particular with a number of k=30 clusters / groups.

[0274] From the image features around the sun 54, or from the area of ​​the sun 54, effects such as lens refraction and image saturation can be detected and then taken into account when analyzing the currently recorded images 110. Thus, a spectral correction of the recorded images 110 can be performed. The second machine learning model can be trained in a supervised manner. Depending on the implementation of the method, the second machine learning model can, in particular, correct measurement errors from a cloud camera-based radiation measurement using provided image features and intermediate results from the cloud-based radiation measurement and the hemispheric irradiance 120. This utilizes data from both a hemispheric radiation measurement from the camera 14 and the radiation sensor unit 12.This results in better and increased accuracy of the diffuse irradiance 122 data in the horizontal plane and the direct irradiance 124 data obtained from the hemispherical irradiance 120 and the camera image 110.

[0275] In method S100, the second machine learning model uses the image features and the hemispheric irradiance 120 to estimate the diffuse irradiance 122 in the horizontal plane and the direct irradiance 124. In method S100, the second machine learning model thus performs the analysis of the contribution of the diffuse irradiance 122 in the horizontal plane and the direct irradiance 124 to the hemispheric irradiance 120. Thus, the second machine learning model has taken over the task of the decomposition model or the split model used in the prior art.

[0276] The second machine learning model, for example, is a machine learning model known as a multilayer perceptron. It can also be a so-called RNN (recurrent neural network), a so-called LSTM (long-short-term memory), or so-called transformer models.

[0277] The training of the first machine learning model therefore aims to extract image features that support the tasks of the second machine learning model. For the third machine learning model, which is applied in method S500 in method step S538, the same training data as for the second machine learning model can, in principle, be used. The third machine learning model can, for example, be a multilayer perceptron.

[0278] For example, reference data from more than one year and from multiple locations under very different atmospheric conditions can be used as training data for the first machine learning model and the second machine learning model. Atmospheric conditions include: sky cloud cover, opacity and optical thickness of the atmosphere, and the position of the sun.

[0279] The training data is filtered so that all conditions are represented to a similar extent. For example, averages over one minute (one-minute averages) can be used as a time resolution.

[0280] High-quality measurements of diffuse irradiance in the horizontal plane and direct irradiance obtained from the measurements and measures listed below can be used as reference data:

[0281] - a two-axis solar tracker accurately tracks instruments and shadow balls or shadow shields to the sun. The solar tracker is precisely leveled horizontally to obtain accurate measurements at all positions of the sun.

[0282] - A pyrheliometer tracking the sun 54 measures the direct radiation in a plane oriented perpendicular to the sun 54 and obtains data on the direct-normal irradiance; - A sun sensor is attached to the solar tracker and connected to the solar tracker's control system. The sun sensor contributes to reducing small errors in the tracking of the solar tracker; a shaded radiation sensor unit 12, in particular a pyranometer, measures the diffuse irradiance in the horizontal plane. The radiation sensor unit 12, in particular a pyranometer, complies with the ISO 9060:2018 standard with accuracy class A (spectrally flat). A ventilation and heating unit is used to reduce fogging due to dew, pollution, and temperature influences. During the measurement, the radiation sensor unit 12 is mounted on the solar tracker and leveled horizontally.The shadow ball or shadow aperture provides the radiation sensor unit 12, in particular a pyranometer. The shadow ball or shadow aperture is adjusted to provide an aperture angle of 2.5° around the sun 54 with an inclination angle of 1°.

[0283] - An unshaded horizontally leveled radiation sensor unit 12, in particular an unshaded pyranometer, is used to measure the hemispheric irradiance 120 in the horizontal plane (GHI) in order to detect measurement errors of the pyrheliometer, which measures the direct normal radiation 124, and the shaded radiation sensor unit 12, in particular the pyranometer, which measures the diffuse irradiance 122 in the horizontal plane. A consistency check is performed on the measured values ​​according to the formula: DNI*sin(sun elevation) + DHI = GHI. The unshaded radiation sensor unit 12, in particular the unshaded pyranometer, meets the same requirements as the shaded radiation sensor unit 12, in particular the shaded pyranometer; - During the working week, a daily cleaning and inspection of the measurement setup and on-site inspection are usually carried out, which allows measurement errors to be detected promptly.Contamination of the sensors of the radiation sensor unit 12 and the camera 14, and in particular of the pyrheliometer and the shaded pyranometer, is detected and remedied in a timely manner;.

[0284] On-site check of the leveling of all sensors (radiation sensor unit 12 and camera 14), tracking of the shadow ball and the shadow aperture and the pyrheliometer;

[0285] - On-site inspection of the wiring and the glass dome of the radiation sensor unit 12 and the window of the pyrheliometer and all other instruments for damage and / or wear;

[0286] - regular calibration of all instruments and a data logger used;

[0287] - semi-automatic test for data control, as described for example in the literature: Geuder, N. et al., Energy Procedia, 2015, Volume 69, pp. 1989-1998.

[0288] The hemispheric irradiance 120 is used because it is very often available in solar applications without additional measurement technology. The diffuse irradiance in the horizontal plane and the direct irradiance are components of the hemispheric irradiance 120.

[0289] Thus, the hemispheric irradiance of 120 provides an indication of the diffuse irradiance in the horizontal plane and the direct irradiance in each case. The pyrheliometer used above complies with ISO 9060:2018 with accuracy class A. The pyrheliometer is mounted on the solar tracker and accurately tracks the sun.

[0290] The images 110 of the camera 14, in particular a cloud camera, for example a surveillance camera with a fisheye lens, are used for training.

[0291] For this purpose, cameras are selected whose images exhibit minimal lens refraction effects, which can, if possible, do not require a protective dome, whose images have the highest possible dynamic range, and whose exposure settings are consistent with the image data. Furthermore, image correction may be omitted under certain circumstances.

[0292] The dynamic range of a surveillance camera's image sensor is typically 10 bits per color channel. Alternatively, it may be possible to capture exposure time series within a few seconds. This high dynamic range is advantageous because it allows the brightest and darkest areas of the sky to be captured, ensuring neither underexposure nor overexposure.

[0293] Data automatically stored in the image data typically includes: exposure time, color temperature, analog gain, digital gain, and gamma correction. This data, which is part of the captured image data, is referred to as metadata.

[0294] All measuring devices are synchronized in time, for example, via a so-called NTP protocol, which maintains a time offset of a few seconds. NTP stands for Network Time Protocol. If transformed camera images from the sun's region 54 are used as input to the first machine learning model, as in the S300 method, for example, only image regions with high information content are passed to the first machine learning model.

[0295] If image features from the camera images 110 and the hemispheric irradiance 120 are used for training, an underestimation of radiation due to too many saturated pixels in the digital image 110 can be avoided.

[0296] Knowledge of the solar elevation provides a general influence on solar radiation measurements, especially on irradiance.

[0297] Figure 10 shows a detailed implementation of method S500 according to another embodiment. First, images 110 and values ​​of the hemispheric irradiance 120 are input as input parameters of the camera 14 and the radiation sensor unit 12. The input parameters 120 may also contain tabular values ​​from previous measurements. The input parameters are subjected to various method steps S538 and S530 for modification and analysis. S538 may, for example, be implemented as a multilayer perceptron with a single hidden layer. The analyzed data sets are combined, in particular concatenated, into a common data set in method step S540. The common data set represents a data vector or vector data set.

[0298] The optional method step S535 for image analysis of the hemispheric irradiance 120 and / or the image 110 of the camera 14, shown in the embodiment in Figure 9, is omitted in the embodiment shown in Figure 10. The image transformation step S515 can optionally be applied. The data vector after method step S540 is then transferred to the second machine learning model S550 for analysis. The desired components of the hemispheric radiation 120 are extracted using the second machine learning model. The extracted components are the diffuse irradiance 122 in the horizontal plane and / or the direct irradiance 124. In method step S550, the hidden layers of the machine learning model are shown in summary. Two hidden layers 555 and 556 are shown as examples in the illustration in Figure 10.However, it may also be provided to use more or fewer hidden layers in process step S550.

[0299] Method steps S530 and S516 summarize various substeps of the first machine learning model. Method steps S530 and / or S516, for example, relate to convolution steps of the data used in the CNN algorithm. Method steps S530 and / or S516 may also include method steps that involve pooling the mean values, taking maximum values ​​(maximum pooling) and / or minimum values ​​(minimal pooling) into account. Following these method steps, a multilayer perceptron, particularly one with a single hidden layer, may be used, for example, in S538, to reduce the number of extracted image features.

[0300] The architecture is standard. For example, an architecture known as ResNet-18 is used. ResNet stands for Residual Neural Network. A residual neural network is a deep learning model in which the weight layers learn residual functions with respect to the layer inputs. A residual neural network is a network with jump connections that perform identity mappings and are merged with the layer outputs by summing them. The number "18" denotes the number of layers. Figure 11 shows parameters used to train the first machine learning model. Figure 11 is a representation created using data from fastai's OneCycleScheduler class, which is described at https: / / fastai1.fast.ai / callbacks.one_cycle.html#OneCycleScheduler.This training strategy is called a one-cycle policy. The one-cycle policy is a learning strategy applied over one learning cycle and allows for faster training of a model. After the first cycle, the learning rate can be applied repeatedly to obtain better results.

[0301] The left part of Figure 11 shows a curve 600. Curve 600 describes a progression of a learning rate 630 plotted as a function of the number of iterations 620. At the beginning of the process of training the first machine learning model, the learning rate 630 increases at a low iteration rate, and at the end of the process of training the first machine learning model, the learning rate 630 decreases again.

[0302] In the right part of Figure 11, a curve 640 of an impulse 650 is plotted as a function of the number of iterations 620.

[0303] The momentum is a parameter that is used to update the weights. The momentum accumulates the gradients from the last iteration steps, which enables faster convergence and the overcoming of saddle points and local minima. The momentum 640 decreases to a minimum after a relatively high value at the beginning of the process and increases again at the end of the process for training the first machine learning model. In addition to the structures and textures of the image 110 of the sky 50, intermediate results of a method for determining the irradiance of solar radiation and / or its components in any plane, in particular a plane inclined to a horizontal plane, can be used as features. The method uses physical models that take physical conditions and assumptions into account. The intermediate results can be one of the features listed below:

[0304] - values ​​of diffuse irradiance in the horizontal plane,

[0305] - direct radiation values,

[0306] - values ​​of the irradiance in any plane, wherein the arbitrary plane is different from the plane in which the radiation sensor unit 12 and / or the camera 14 are arranged,

[0307] - Estimates of the irradiance in the red channel, the green channel and / or the blue channel.

[0308] The method is described in detail in WO 2021 / 219570 A1. The method comprises the following method steps, wherein one or more of the method steps are applied to obtain the intermediate results. The raw data used can be measurement data from the camera 14 and the radiation sensor unit 12 or measurement data recorded at other locations and with other cameras and / or radiation sensor units and stored, for example, in a database or table.

[0309] The red-green-blue (RGB) color channels of the sky image are summed in a weighted manner. The weighting of the channels ensures the camera's sensitivity is as uniform as possible in the visible wavelength range. This gray value is multiplied by a broadband correction to account for radiation at wavelengths outside the camera's measurement range. Using a geometric internal and external calibration standard for cloud cameras, a sky region (azimuth and zenith angle) is specified for each pixel of the camera image. This results in an estimate of the sky's radiance distribution.

[0310] Analogous to the radiance distribution, a luminance distribution is calculated. For this purpose, the RGB color channels are weighted according to the sensitivity of the human eye before summation. Integration of the luminance distribution across all angular ranges yields a measured value of the illuminance. The illuminance output by the camera and the illuminance calculated from the camera image are compared. The radiance distribution is scaled according to the ratio of the two values ​​to compensate for any influence of the camera control on the camera's sensitivity.

[0311] The area of ​​the solar disk is masked. For an evaluated sensor plane, each sky region in the radiance image is weighted according to a projection into the plane. Integrating the radiance distribution across all sky regions within the field of view of the inclined plane yields the diffuse irradiance from the sky for the respective plane.

[0312] The horizontal diffuse irradiance in the plane of the radiation sensor unit is calculated accordingly. The direct normal irradiance (DNI) is calculated by comparing it with the horizontal hemispheric irradiance measured by a pyranometer and taking into account the current position of the sun. To correct for refraction effects in the camera lens, the initial estimate of the diffuse irradiance in the plane of the radiation sensor unit and all other calculated diffuse irradiances are reduced by a portion of the DNI (lens refraction correction). The correction is then added to the direct irradiance in the plane of the radiation sensor unit. The DNI is then recalculated.

[0313] The GTI in an evaluated plane is ultimately determined from a direct component, a diffuse component from the sky, and a component reflected from the ground. The DNI is projected into the evaluated plane and thus yields the direct component. The diffuse irradiance is calculated from the camera image for this plane as described above. The reflected component is determined as GHI multiplied by the albedo of the background and the term

[0314] (1 - cos(angle of inclination of the inclined plane to the horizontal)) / 2.

[0315] According to the specified method, in order to convert the hemispherical irradiance of the solar radiation determined by the radiation sensor unit in the plane of the radiation sensor unit into the irradiance and / or their respective components, direct irradiance, diffuse irradiance, irradiance reflected on the ground, in the horizontal plane and / or in the plane inclined to the horizontal plane, at least one of the quantities of irradiance reflected on the ground, and / or diffuse irradiance, and / or the position of the sun during the radiation measurement, and / or a sensor-specific correction factor, which in particular includes lens parameters of the camera, can be used.

[0316] In this case, factors such as the position of the sun that influence the radiation measurement can be taken into account during the conversion. Furthermore, according to the specified method for converting measured values ​​from the camera, at least one of the variables from a ratio of broadband radiation to the portion of the radiation recorded by the camera, and / or a spectral intensity of RGB channels of the camera, and / or an internal and / or external calibration of the camera, and / or an inclination and orientation of the camera sensor, and / or the position of the sun during the radiation measurement and / or a camera sensitivity that is determined from an illuminance of the camera and / or the spectral sensitivity of the RGB channels and / or recording settings and / or the RGB camera image and / or the internal and / or external calibration of the camera can be used. In this case, factors such as the position of the sun that influence the radiation measurement can be taken into account during the conversion.

[0317] The individual steps for determining the hemispheric irradiance of solar radiation and / or its respective components, direct irradiance, diffuse irradiance, ground-reflected irradiance, in the horizontal plane and in the plane inclined to the horizontal plane are described.

[0318] The radiation reflected from the ground is determined using albedo, inclination and orientation of the inclined plane, as well as a measured value of the hemispheric irradiance in the plane of the radiation sensor unit.

[0319] The direct irradiance in the plane of the radiation sensor unit is determined by subtracting the measured value of the diffuse irradiance, evaluated for the plane of the radiation sensor unit, from the hemispherical irradiance in the plane of the radiation sensor unit. The diffuse irradiance in the plane of the radiation sensor unit is determined as described in more detail below. The direct irradiance is then determined by inverting the projection into the plane of the radiation sensor unit using the position of the sun calculated from the location and time.

[0320] The direct normal radiation is multiplied by a correction factor, which includes, in particular, the lens parameters of the camera used. This results in a lens refraction correction.

[0321] The direct irradiance in the plane of the radiation sensor unit and the lens refraction correction are added, and the projection into the plane of the radiation sensor unit is inverted, taking into account the calculated position of the sun. This direct normal irradiance is then projected into the horizontal and / or inclined plane, taking into account the inclination and orientation of the horizontal and / or inclined plane, to obtain the corrected measured value of the direct irradiance in this plane.

[0322] The lens refraction correction is subtracted from the diffuse irradiance evaluated for the inclined or horizontal plane as described above. This results in a corrected measured value of the diffuse irradiance in the respective inclined or horizontal plane.

[0323] The hemispheric irradiance in the horizontal and / or inclined plane can then be determined by summing the radiation reflected from the ground, the direct radiation as a component of the hemispheric irradiance in the horizontal and / or inclined plane, and the diffuse irradiance as a component of the hemispheric irradiance, evaluated for the horizontal and / or inclined plane, and corrected. The determination of the diffuse irradiance originating from the sky in the horizontal or inclined plane, particularly in the plane of the radiation sensor unit, is summarized below. First, a broadband correction factor is determined from the ratio of broadband radiation to the portion registered by the camera using the daylight spectrum and the spectral sensitivity of the camera's RGB channels.

[0324] For this purpose, weights of the RGB channels are determined according to the inverse sensitivity using the camera's recording settings.

[0325] This allows the weighted RGB channels of the camera image to be summed.

[0326] The summed RGB channels are then multiplied by the determined broadband correction factor.

[0327] This allows angular areas of the sky to be assigned to camera image pixels using the camera's internal and / or external calibration values.

[0328] These image areas are then weighted according to the projection into the horizontal and / or inclined plane.

[0329] In parallel, the angular range of the field of view of the horizontal and / or inclined plane can be determined from the inclination and orientation of this plane, while the angular range of the solar disk is determined from the location and time of day.

[0330] This allows the angular range of the solar disk to be excluded from the angular range of the field of view of the horizontal and / or inclined plane.

[0331] The image areas can then be integrated across the field of view of the horizontal or inclined plane. The diffuse irradiance in the horizontal or inclined plane, particularly in the plane of the radiation sensor unit, can then be calculated by multiplying it by the previously determined camera sensitivity correction factor.

[0332] The calculation of the camera sensitivity correction factor based on a comparison of the illuminance output by the camera and the calculated illuminance can be done as described below.

[0333] First, weights corresponding to the sensitivity of each RGB channel of the camera used are determined using the spectral sensitivity of the RGB channels and the camera's recording settings.

[0334] Weights are then determined based on human perception.

[0335] Using these weights, RGB channels from the RGB camera image can then be summed up in a weighted manner.

[0336] In parallel, angular areas of the sky are assigned to image pixels of the camera using the camera's internal and / or external calibration values.

[0337] The weighted RGB channels are then integrated over the hemisphere of the sky.

[0338] This allows the correction factor for camera sensitivity to be determined by calculating the ratio of the camera's illuminance to the integrated weighted RGB camera image. The method comprises evaluating the camera image using the physical camera model, and obtaining a preliminary measurement of the diffuse irradiance in the horizontal or inclined plane. This preliminary measurement of the diffuse irradiance or a direct irradiance calculated therefrom is incorporated as a feature by the second machine learning model, and / or a direct irradiance is determined using the camera model and the measurement of the hemispheric irradiance, and this parameter is incorporated as a feature by the second machine learning model.

[0339] Furthermore, the camera image can be evaluated using the physical camera model, whereby circumsolar radiation and / or illuminance or radiation information per color channel and / or a ratio of the illuminances or radiation information for two different color channels and / or a number of saturated image pixels is determined using the camera model. These parameters are incorporated as features by the second machine learning model. Circumsolar radiation refers to the radiation emitted from an area directly surrounding the solar disk.

[0340] It can be advantageous to determine the circumsolar radiation from the camera image, since the radiation from the immediate vicinity of the sun's disk contributes to the DNI. In particular, the circumsolar diffuse radiation, which is a result of the scattering of radiation in the atmosphere, can contribute to the DNI. It can also be advantageous to know the illuminance per color channel, since the distribution of the illuminance across the color channels reflects the spectral distribution of the solar spectrum. It can be advantageous to determine the number of saturated image pixels from the camera image, as this allows a statement to be made about the intensity of the radiation hitting the camera, in particular its distribution in the camera image.

[0341] The procedural steps carried out are summarized below:

[0342] (i) observed sky areas are assigned to image pixels; and / or

[0343] (ii) a situation-adapted broadband correction is applied, which is a function of irradiance calculated over the different color channels of the camera, in particular the diffuse irradiance; and / or

[0344] (iii) the radiance is integrated or summed in different parts of the sky, in particular to determine circumsolar radiation or diffuse irradiance in the horizontal plane; and / or

[0345] (iv) intensity values ​​of the colour channels or parameters derived therefrom are summed in a weighted manner; and / or

[0346] (v) image metadata is included to compensate for any influence of the camera’s exposure control; and / or

[0347] (vi) a calibration of the camera is carried out using the hemispherical irradiance measurements (120) of the radiation sensor unit.

[0348] In addition, the following procedural steps can be carried out

[0349] (i) an estimate of the sky radiance from the camera image based on a camera model can be obtained;

[0350] (ii) and / or the camera model may take into account the spectral sensitivity of the camera per color channel or color matching functions adapted to the color space used;

[0351] (iii) and / or an intrinsic and external geometric calibration of the camera can be used. Reference numerals

[0352] 5 Device

[0353] 10 measuring units

[0354] 12 Radiation sensor unit

[0355] 14 Camera

[0356] 16 Measurement data acquisition unit

[0357] 18 Evaluation unit

[0358] 19 computers

[0359] 20 Data processing device

[0360] 22 housings

[0361] 24 Sensor of the radiation sensor unit 12

[0362] 26 Field of view of the radiation sensor unit 12 and field of view of the sensor

[0363] 28 Camera housing

[0364] 30 Camera sensor

[0365] 32 Camera field of view 14 and sensor field of view 30

[0366] 34 distance

[0367] 36 horizontal plane

[0368] 38 Level of the radiation sensor unit 12 and level of the sensor 24

[0369] 40 Camera plane 14 and sensor plane 30

[0370] 42 Elevation, Inclination

[0371] 43 Elevation

[0372] 44 angles, angle of inclination

[0373] 46 Level of the radiation sensor unit 12 and the sensor 24

[0374] 48 Level of the camera 14 and the sensor 30

[0375] 50 Heavens

[0376] 52 Cloud

[0377] 54 Sun

[0378] 100 procedures

[0379] 110 image taken with the camera

[0380] 120 hemispheric irradiance

[0381] 122 diffuse irradiance

[0382] 124 direct irradiance 530 hidden layer

[0383] 538 hidden layer

[0384] 555 hidden layer

[0385] 556 hidden layer

[0386] 600 Curve of the learning rate as a function of iterations 620

[0387] 620 iterations

[0388] 630 learning rate

[0389] 640 Curve of the impulse plotted as a function of iterations 620

[0390] 650 pulse

[0391] S100 procedure

[0392] S130 Extracting features using the first machine learning model

[0393] S140 Concatenation, merging into a common data set

[0394] S150 Determining the diffuse irradiance in the horizontal plane and / or the direct irradiance using the second machine learning model

[0395] S200 procedure

[0396] S230 Extracting features using the first machine learning model

[0397] S235 Image analysis

[0398] S240 Concatenation, merging into a common data set

[0399] S250 Determining the diffuse irradiance in the horizontal plane and / or the direct irradiance using the second machine learning model

[0400] S300 procedure

[0401] S315 Image transformation

[0402] S330 Extracting features using the first machine learning

[0403] Model

[0404] S335 Image analysis

[0405] S340 Concatenation, merging to a common data set S350 Determination of the diffuse irradiance in the horizontal plane and / or the direct irradiance using the second machine learning model

[0406] S400 procedure

[0407] S415 Image transformation

[0408] S430 Extracting features using the first machine learning model

[0409] S440 Concatenation, merging into a common data set

[0410] S450 Determining diffuse irradiance in the horizontal plane and / or direct irradiance using the second machine learning model

[0411] S500 procedure

[0412] 5515 Image transformation

[0413] 5516 process steps, for example folding steps

[0414] S530 Extracting features using the first machine learning model

[0415] S535 Image Analysis

[0416] S538 Processing features from image analysis using the third machine learning model

[0417] S540 Concatenation, merging into a common data set

[0418] S550 Determining the diffuse irradiance in the horizontal plane and / or the direct irradiance using the second machine learning model

Claims

Claims 1. A method (S100, S200, S300, S400, S500) for determining at least one component of a hemispherical irradiance (120) of solar radiation in any plane, wherein the at least one component comprises a diffuse irradiance (122) and / or a direct irradiance (124), comprising the following steps: (i) determining measurement data of the hemispherical irradiance (120) with a radiation sensor unit (12) in a field of view (26) above a plane (46) of the radiation sensor unit (12); (ii) capturing an image (110) of the sky (50) with a camera (14) in a field of view (32) above a plane (48) of the camera (14); (iii) extracting (S130, S230, S330, S430, S530) features, in particular structures, from the image (110) of the sky (50) by means of a first machine learning model and generating a result data set; (iv) merging (S140, S240, S340, S440, S540) the measurement data of the radiation sensor unit (12) and the result data set of the camera (14) into a common data set; (v) determining (S150, S250, S350, S450, S550) the at least one component of the hemispherical irradiance (120) in the arbitrary plane, in particular the diffuse irradiance (122) and / or the direct irradiance (124), from the data set by means of a second machine learning model.

2. The method according to claim 1, wherein features are extracted in an image analysis (S235, S335, S535) using the image (110) of the camera (14), wherein a physical camera model is applied.

3. Method according to one of the preceding claims, wherein the image (110) of the camera (14) is evaluated via the physical camera model and wherein a preliminary measurement of the diffuse irradiance (122) in the horizontal or in an inclined plane, and wherein this preliminary measurement of the diffuse irradiance or a direct irradiance calculated therefrom is included as a feature by the second machine learning model, and / or wherein a direct irradiance is determined with the aid of the camera model and the measurement of the hemispherical irradiance, and wherein this parameter is included as a feature by the second machine learning model.

4. Method according to one of the preceding claims, wherein the image (110) of the camera (14) is evaluated via the physical camera model, and wherein with the aid of the camera model a circumsolar radiation and / or an illuminance or radiation information per color channel and / or a ratio of the illuminances or radiation information for two different color channels and / or a number of saturated image pixels is determined, and wherein these parameters are included as features by the second machine learning model.

5. Method according to one of the preceding claims, wherein (i) image pixels are assigned to observed areas of the sky; and / or (ii) a situation-adapted broadband correction is applied, which is a function of a radiation calculated over the various color channels of the camera (14), in particular the diffuse irradiance; and / or (111) the radiance in different parts of the sky is integrated or summed, in particular to determine circumsolar radiation or diffuse irradiance in the horizontal plane; and / or (iv) intensity values of the colour channels or parameters derived therefrom are summed in a weighted manner; and / or (vi) image metadata is included to compensate for any influence of the camera's exposure control (14); and / or (vii) a calibration of the camera (14) is carried out using the measurements of the hemispherical irradiance (120) of the radiation sensor unit (12).

6. Method according to one of the preceding claims, wherein (i) an estimate of the radiance of the sky is determined from the image (110) of the camera (14) based on the camera model; and / or (ii) the camera model takes into account the spectral sensitivity of the camera (14) per color channel or color matching functions adapted to the color space used; and / or (iii) an intrinsic and external geometric calibration of the camera (14) is used.

7. The method according to any one of the preceding claims, wherein the second machine learning model is different from the first machine learning model, wherein the first machine learning model is trained with first training data sets, wherein the second machine learning model is trained with second training data sets.

8. The method according to any one of the preceding claims, wherein in the image analysis (S235, S335, S535) using the image (110) of the camera (14) features are extracted from the measurement data of the radiation sensor unit (12).

9. Method according to one of the preceding claims, wherein in the image analysis (S235, S335, S535) using the image (110) of the camera (14) features are derived from the intensities of the image pixels using a number of saturated pixels to extract features.

10. Method according to one of the preceding claims, wherein at least one sum and / or weighted sum of the intensities of the image pixels and / or at least one ratio of sums and / or weighted sums of the intensities of the image pixels are included.

11. Method according to one of the preceding claims, wherein the features extracted from the image (110) of the camera (14) and / or from the measurement data of the radiation sensor unit (12) are transferred to a third machine learning model and processed by means of the third machine learning model (S538) and a radiation sensor unit data set is generated, wherein the radiation sensor unit data set and the result data set of the camera (14) are merged (S540) before the second machine learning model is applied (S550).

12. Method according to one of the preceding claims, wherein an image transformation (S315, S415, S515) of the image (110) of the sky (50) takes place before the application of the first machine learning model.

13. Method according to one of the preceding claims, wherein the first machine learning model and / or the second machine learning model is trained with reference data of at least one of the components of the hemispheric irradiance (120), in particular the diffuse irradiance (122) in the horizontal and / or inclined plane, in particular from solar trackers, and / or the direct irradiance (124), in particular from a pyrheliometer tracking the sun, and / or with regular input data, in particular with images (110) of the camera (14), data of the hemispheric irradiance (120) and / or intermediate results, which were obtained in particular from the application of a physical model.

14. The method according to any one of the preceding claims, wherein the first machine learning model is pre-trained using at least one of the following steps: (i) Use of publicly available weights, (ii) Training using at least one unsupervised or self-supervised training approach, (iii) extending data with Gaussian blurs and / or distorting image colors and / or mirroring the images (110), (iv) Training with images (110) of the sky (50) taken by the camera (14).

15. Method according to one of the preceding claims, wherein a convolutional neural algorithm, in particular a convolutional neural network, is used in the first machine learning model.

16. The method according to any one of the preceding claims, wherein at least one algorithm from a multilayer perceptron (MLP), random forest algorithm, recurrent neural network, long short term memory algorithm, transformer model algorithm, k-nearest neighbor (k-NN) algorithm, and / or support vector machine algorithm is used in the second machine learning model.

17. Method according to one of the preceding claims, wherein the first training data sets and / or the second training data sets are filtered such that different atmospheric conditions are represented to a similar extent.

18. Method according to one of the preceding claims, wherein the first machine learning model, the second machine learning model and / or the third machine learning model are trained in a supervised manner, wherein input data and / or correct reference data from at least one of the Components of the hemispheric irradiance (120) are used.

19. Method according to one of the preceding claims, wherein the first machine learning model, the second machine learning model and the third machine learning model are trained together, in particular supervised, in particular after a respective individual training.

20. The method according to any one of the preceding claims, wherein image features from the image (110) of the sky (50) using estimated values of a red channel, and / or a blue channel, and / or a green channel from the measurement data of the image (110) of the sky (50) are used as additional input data of the third machine learning model or the merging (S140, S240, S340, S440, S540).

21. Device (5) for carrying out a method (S100, S200, S300, S400, S500) for determining at least one component of a hemispheric irradiance (120) of solar radiation in any plane, according to one of the preceding claims, wherein the at least one component comprises a diffuse irradiance (122) and / or a direct irradiance (124), comprising at least one radiation sensor unit (12), a camera (14), and an evaluation unit (18) which is provided for evaluating measurement data of the radiation sensor unit (12) and the camera (14), wherein the radiation sensor unit (12) determines the irradiance of solar radiation (120) in a field of view (26) of 180° above a plane (38, 46), wherein the plane (38, 46) is a horizontal plane or an inclined plane, wherein the camera (14) acquires an image (110) of the sky (50) in a field of view (32) of 180° above a plane (40, 48),wherein the plane (40, 48) is a horizontal plane or an inclined plane, wherein the evaluation unit (18) has at least a first machine learning model and a second machine learning model, wherein the evaluation unit (18) is at least designed to carry out the following steps: (i) extracting (S130) features, in particular structures, from the image (110) of the sky (50) by means of a first machine learning model and generating a result data set; (ii) merging (S140) the measurement data of the radiation sensor unit (12) and the result data set of the camera (14) into a common data set; (iii) determining (S150) the at least one component of the hemispherical irradiance (120) in the arbitrary plane, in particular the diffuse irradiance (122) and / or the direct irradiance (124), from the data vector by means of a second machine learning model.

22. The device according to claim 21, wherein features are extracted in an image analysis (S235, S335, S535) using the image (110) of the camera (14), wherein a physical camera model is applicable.

23. Device according to claim 21 or 22, wherein the measurement data of the camera (14) contain angle-resolved radiation information.

24. A method for training a first machine learning model for extracting features, in particular structures, from an image (110) of the sky (50), comprising at least one of the following steps: (i) determining or using data sets of images (110) of the sky (50) taken with a camera (14), in particular a cloud camera; (ii) training the machine learning model with the data sets of images (110) of the sky (50); (iii) Training the first machine learning model with known images (110) of the sky (50) taken at other locations.

25. The method according to claim 24, wherein pre-training is used with at least one of the following steps: (i) Use of publicly available weights; (ii) training using at least one unsupervised or self-supervised training approach; (iii) augmenting data with Gaussian blurs and / or distorting image colors and / or mirroring the images; (iv) Training with images (110) of the sky (50) taken by the camera (14).

26. A computer program for determining at least one component of a hemispheric irradiance (120) of solar radiation in any plane, comprising instructions which, when the program is executed by a computer (19), cause the computer (19) to carry out the steps of a method (S100, S200, S300, S400, S500) for determining at least one component of a hemispheric irradiance of solar radiation according to one of claims 1 to 20.

27. Data processing device (20) for a device (5) for determining at least one component of a hemispherical irradiance (120) of solar radiation in any plane, according to one of claims 21 to 23, at least comprising a measurement data acquisition unit (16), an evaluation unit (18) and a computer (19).

28. A trained first machine learning model for a method (S100, S200, S300, S400, S500) for determining at least one component of a hemispheric irradiance (120) of solar radiation in any plane according to one of claims 1 to 20, for extracting Features, in particular structures, from an image (110) of the sky (50) which is trained according to a method according to claim 24 or 25.

29. A trained second machine learning model for a method (S100, S200, S300, S400, S500) for determining at least one component of a hemispheric irradiance (120) of solar radiation in any plane according to one of claims 1 to 20, for determining at least one component of a hemispheric irradiance (120) of solar radiation in any plane, which has been trained in a supervised manner, in particular using input data and reference data, in particular comprising a multilayer perceptron algorithm.

Images