Irradiation calculation method, device and equipment

Abstract

The present application discloses an irradiation calculation method, device and equipment. Among them, the method includes: obtaining satellite cloud map feature data of a first area; obtaining first ground meteorological monitoring data of the first area; obtaining second ground meteorological monitoring data of a second area around the first area; inputting the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiation calculation model, and obtaining a target irradiation value of the first area through the output of the irradiation calculation model. The present application solves the technical problem of large result errors in irradiation calculation in the related art.

Description

Technical Field

[0001] This application relates to the technical field of photovoltaic power generation, and more particularly, to a method, device, and equipment for irradiance calculation. Background Art

[0002] In a photovoltaic power generation system, the output power of the system depends to a large extent on the solar irradiance received by the solar panels. Therefore, to accurately predict the output power of a photovoltaic power station, it is necessary to accurately calculate the irradiance value in the area of the photovoltaic power station.

[0003] Currently, the industry mainly calculates based on satellite data. However, in many cases, the surface irradiance changes continuously and instantaneously in real time. Due to the limitations of the spatio-temporal resolution of satellite data and the superposition of the irradiance deviation between high altitude and the surface, the irradiance calculation model based solely on satellite data has a large error in complex weather conditions.

[0004] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0005] Embodiments of this application provide a method, device, and equipment for irradiance calculation to at least solve the technical problem of large errors in the results of irradiance calculation in related technologies.

[0006] According to one aspect of the embodiments of this application, a method for irradiance calculation is provided, including: obtaining satellite cloud map feature data of a first area; obtaining first ground meteorological monitoring data of the first area; obtaining second ground meteorological monitoring data of a second area surrounding the first area; inputting the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiance calculation model, and outputting a target irradiance value of the first area through the irradiance calculation model.

[0007] Optionally, obtaining satellite cloud map feature data of a first area includes: obtaining the satellite cloud map of the first area; inputting the satellite cloud map into a pre-trained target neural network, and outputting the satellite cloud map feature data, where the satellite cloud map feature data at least includes: the type, thickness, moving speed, and moving direction of the cloud layer, and the longitude and latitude of the first area.

[0008] Optionally, obtaining the first ground meteorological monitoring data of the first region includes: obtaining the first ground meteorological monitoring data monitored by the environmental monitoring equipment within the first region, where the first ground meteorological monitoring data at least includes: the wind speed, wind direction, temperature, humidity, and air pressure of the first region; obtaining the second ground meteorological monitoring data of the second region surrounding the first region includes: obtaining the second ground meteorological monitoring data monitored by the environmental monitoring equipment within the second region, where the second ground meteorological monitoring data at least includes: the wind speed, wind direction, temperature, humidity, air pressure, and irradiance value of the second region.

[0009] Optionally, the training process of the irradiance calculation model includes: obtaining multiple sets of sample data collected at multiple preset times, where each set of the sample data includes: the satellite cloud image feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiance value of the center point of the first region collected at the same preset time; training a target neural network model based on the multiple sets of sample data to obtain the irradiance calculation model, where the target neural network model at least includes: an attention layer, a recurrent neural network layer, and a fully connected layer, and the recurrent neural network layer includes one of the following: a long short-term memory network, a gated recurrent unit.

[0010] Optionally, for any set of sample data, perform normalization processing on the sample data; input the satellite cloud image feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiance value of the center point of the first region corresponding to the sample data into the attention layer to obtain a first output matrix, where the irradiance value of the center point of the first region is used as a sample label; input the first output matrix into the recurrent neural network layer to obtain a second output matrix; input the second output matrix into the fully connected layer to obtain a prediction result; adjust the model parameters of the target neural network model based on the sample label and the prediction result; perform iterative training on the target neural network model based on the multiple sets of sample data to determine the target model parameters and obtain the irradiance calculation model.

[0011] Optionally, determine a first input feature based on the satellite cloud map feature data, determine a second input feature based on the first ground meteorological monitoring data, determine a third input feature based on the second ground meteorological monitoring data, use the irradiance value of the center point of the first region as a sample label, and construct an input matrix; copy the input matrix through the attention layer to obtain a first matrix, a second matrix, and a third matrix; perform a dot product calculation on the first matrix and the transposed matrix of the second matrix to obtain a first result; scale the first result based on a preset scaling factor to obtain a second result; perform a softmax process on the second result to obtain a third result; perform a tensor multiplication calculation on the third result and the third matrix to obtain the first output matrix.

[0012] Optionally, determine a first preset weight corresponding to the satellite cloud map feature data, a second preset weight corresponding to the first ground meteorological monitoring data, and a third preset weight corresponding to the second ground meteorological monitoring data; determine the first input feature based on the first preset weight and the satellite cloud map feature data; determine the second input feature based on the second preset weight and the first ground meteorological monitoring data; determine the third input feature based on the third preset weight and the second ground meteorological monitoring data.

[0013] Optionally, sequentially input the multiple groups of sample data into the target neural network model for iterative training, construct a target loss function, and dynamically optimize the model parameters of the target neural network model based on the optimization algorithm of gradient descent to obtain the target model parameters, where the target loss function includes one of the following: mean square error loss function, cross entropy loss function.

[0014] Optionally, the first region is the region where the solar panels of the photovoltaic power station are located. After obtaining the target irradiance value of the first region through the irradiance calculation model, predict the target output power of the photovoltaic power station in a future target time period based on the target irradiance value, where the target output power is used to evaluate the reliability of the photovoltaic power station connected to the power grid.

[0015] According to another aspect of the embodiments of the present application, there is also provided an irradiance calculation device, including: a first acquisition module for acquiring satellite cloud map feature data of a first region; a second acquisition module for acquiring first ground meteorological monitoring data of the first region; a third acquisition module for acquiring second ground meteorological monitoring data of a second region surrounding the first region; a calculation module for inputting the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiance calculation model, and outputting the irradiance value of the first region through the irradiance calculation model.

[0016] According to another aspect of the embodiments of the present application, a non-volatile storage medium is further provided. The non-volatile storage medium includes a stored program, wherein when the program runs, it controls the device where the non-volatile storage medium is located to execute the above-mentioned irradiation calculation method.

[0017] According to another aspect of the embodiments of the present application, an irradiation calculation device is further provided, including: a memory and a processor. Among them, a computer program is stored in the memory, and the processor is configured to execute the above-mentioned irradiation calculation method through the computer program.

[0018] In the embodiments of the present application, first, the satellite cloud map feature data of the first area, the first ground meteorological monitoring data of the first area, and the second ground meteorological monitoring data of the second area surrounding the first area are obtained. Then, the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data are input into a pre-trained irradiation calculation model, and the target irradiation value of the first area is obtained through the output of the irradiation calculation model. Among them, the irradiation calculation model comprehensively calculates the irradiation value based on satellite observation data and ground monitoring data. It takes into account the advantages of large satellite observation coverage and high ground monitoring accuracy at the same time, can realize the mutual correction of satellite observation data and ground monitoring data, effectively improve the irradiation calculation accuracy, and thus solves the technical problem of large result errors in the related art when performing irradiation calculation. Description of the Drawings

[0019] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0020] Figure 1 is a flowchart of an irradiation calculation method according to an embodiment of the present application;

[0021] Figure 2 is a schematic diagram of sample data collection according to an embodiment of the present application;

[0022] Figure 3 is a schematic diagram of the structure of an irradiation calculation device according to an embodiment of the present application. Detailed Embodiments

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present application.

[0024] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of this application are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0025] Embodiment 1

[0026] Generally, surface irradiation is real-time continuous and instantaneously changing. Due to the spatio-temporal resolution limitations of satellite data and the influence of the superposition of the irradiation deviation between high altitude and the surface, the error of irradiance calculation based solely on satellite data is relatively large in complex weather conditions; while the irradiation data monitored by ground equipment is simple and accurate, but the coverage range of a single-point device is small. If a large area of multi-point devices need to be installed, the cost is relatively high, and the maintenance of multi-point devices is difficult in the later stage.

[0027] To solve the above problems, the embodiments of the present application provide a scheme for comprehensively calculating irradiance values based on satellite observation data and ground monitoring data, which can realize the mutual correction of satellite observation data and ground monitoring data. By taking advantage of the advantages of ground monitoring, which is close to the source and has high accuracy, it can effectively improve the accuracy of irradiance calculation. At the same time, taking into account the advantage of large satellite observation coverage, only a small number of irradiance acquisition devices need to be installed on the ground to obtain accurate irradiance data for a large target area, reducing the hardware procurement and later manual maintenance costs.

[0028] Specifically, the embodiments of the present application provide an irradiance calculation method. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0029] Figure 1 is a schematic flowchart of an optional irradiance calculation method according to an embodiment of the present application, as Figure 1 shown, this method at least includes steps S102 - S108, where:

[0030] Step S102, obtain the satellite cloud map feature data of the first area.

[0031] Step S104: Obtain the first ground meteorological monitoring data of the first region.

[0032] Step S106: Obtain the second ground meteorological monitoring data of the second region around the first region.

[0033] Step S108: Input the satellite cloud image feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiation calculation model, and output the target irradiation value of the first region through the irradiation calculation model.

[0034] Generally, the purpose of irradiation calculation is to provide reference data for the power generation revenue of a newly built photovoltaic power station in the target region by evaluating the regional irradiation level, or to quantify the power generation efficiency and personnel operation and maintenance level of an existing power station. Therefore, the above-mentioned first region mainly refers to the region where the solar panels of the photovoltaic power station are located. In some optional embodiments of the present application, when calculating the irradiation value of the first region, sample data of the first region at different times can be collected first for training the irradiation calculation model.

[0035] Specifically, multiple groups of sample data collected at multiple preset times can be obtained first. Each group of sample data includes: satellite cloud image feature data, first ground meteorological monitoring data, second ground meteorological monitoring data, and the irradiation value at the center point of the first region collected at the same preset time. Train a target neural network model based on the multiple groups of sample data to obtain an irradiation calculation model. The target neural network model at least includes: an attention layer, a recurrent neural network layer, and a fully connected layer. The recurrent neural network layer includes one of the following: Long Short-Term Memory Network (LSTM), Gate Recurrent Unit (GRU).

[0036] When obtaining the satellite cloud image feature data of the first region, the satellite cloud image of the first region can be obtained first, such as obtaining the satellite cloud image transmitted by the Fengyun meteorological satellite, and then inputting the satellite cloud image into a pre-trained target neural network to output the satellite cloud image feature data. The target neural network is mainly used to extract cloud features. The satellite cloud image feature data at least includes: the type, thickness, moving speed, moving direction of the cloud layer, the longitude and latitude of the first region, especially the longitude and latitude of the center point position of the first region, and other data.

[0037] When obtaining the ground meteorological monitoring data, considering the influence of the surrounding environment on the irradiation value of the first region, in addition to obtaining the first ground meteorological monitoring data of the first region, the second ground meteorological monitoring data of the second region around the first region also needs to be obtained.

[0038] Specifically, when obtaining the first ground meteorological monitoring data of the first region, the first ground meteorological monitoring data monitored by environmental monitoring devices (such as irradiance monitors) within the first region can be obtained. Among them, the first ground meteorological monitoring data at least includes: data such as wind speed, wind direction, temperature, humidity, and air pressure in the first region.

[0039] When obtaining the second ground meteorological monitoring data of the second region surrounding the first region, the surrounding region within a preset range (such as within 80 km) from the first region can be determined as the second region, and the second ground meteorological monitoring data monitored by environmental monitoring devices within the second region can be obtained. Among them, the second ground meteorological monitoring data at least includes: data such as wind speed, wind direction, temperature, humidity, air pressure, and irradiance value in the second region.

[0040] At the same time, the irradiance value at the center point of the first region can be obtained through the irradiance monitor at the center point of the first region, and this value can be regarded as the true value of the irradiance value of the first region and used as a sample label.

[0041] Figure 2 Fig. shows a schematic diagram of sample data acquisition. Among them, by obtaining the satellite cloud image transmitted by the satellite and performing feature extraction, the satellite cloud image feature data of the first region is obtained; the first ground meteorological monitoring data of the first region is obtained through the irradiance monitor 1 in the first region; the second ground meteorological monitoring data of the second region is obtained through the irradiance monitors 2-9 in the surrounding second region. It can be seen that the number of irradiance monitors required for obtaining ground meteorological monitoring data is not large, and only uniform distribution is needed.

[0042] When collecting the above sample data, the collection time and collection frequency can also be set. For example, a target time period is determined, and the sample data is collected once every 20 s within this target time period, that is, multiple groups of sample data corresponding to multiple preset moments are obtained.

[0043] After that, the target neural network model can be iteratively trained based on multiple groups of sample data to obtain the final irradiance calculation model.

[0044] In some optional embodiments of the present application, a target neural network model combined with GRU + Attention can be built based on pytorch. Among them, pytorch is an open-source Python deep learning framework, which can achieve powerful GPU acceleration and support dynamic neural networks; GRU is a variant of LSTM, mainly including an update gate and a reset gate. Compared with LSTM, GRU can better capture the dependencies with larger intervals in time series data, and the structure is simpler. Therefore, the recurrent neural network layer preferably selects GRU; Attention is used to improve the interpretability of the target neural network model.

[0045] Taking advantage of the accuracy and training efficiency advantages of GRU in calculating sequence data, and using Attention to mine the potential relationships between different meteorological data, the input feature weights of the target neural network model can be made more reasonable, thus realizing the mutual correction between satellite observation data and ground monitoring data.

[0046] During the model training process, for any set of sample data, first perform normalization processing on this set of sample data, and then input the satellite cloud map feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiation value of the first regional center point corresponding to this set of sample data into the attention layer Attention to obtain the first output matrix. Among them, the irradiation value of the first regional center point is used as the sample label.

[0047] Optionally, in Attention, the first input feature can be determined based on the satellite cloud map feature data, the second input feature can be determined based on the first ground meteorological monitoring data, the third input feature can be determined based on the second ground meteorological monitoring data, and the irradiation value of the first regional center point is used as the sample label to construct the input matrix.

[0048] Among them, considering that the satellite data and the ground monitoring data have different influence degrees on the final irradiation value, different weights can be set for them. Specifically, the first preset weight corresponding to the satellite cloud map feature data, the second preset weight corresponding to the first ground meteorological monitoring data, and the third preset weight corresponding to the second ground meteorological monitoring data can be determined first; then the first input feature is determined based on the first preset weight and the satellite cloud map feature data; the second input feature is determined based on the second preset weight and the first ground meteorological monitoring data; the third input feature is determined based on the third preset weight and the second ground meteorological monitoring data. Usually, the first preset weight is greater than the second preset weight is greater than the third preset weight.

[0049] After that, the first input feature can be recorded as module data 1, the second input feature can be recorded as module data 2, the third input feature can be recorded as module data 3, and the irradiation value of the first regional center point is used as the sample label, recorded as Y-L, to construct the input matrix tensor, and its format is [[module data 1, module data 2, module data 3], Y-L]; then the input matrix tensor is copied through the attention layer Attention to obtain the first matrix Q, the second matrix K, and the third matrix V; the dot product calculation is performed on the first matrix Q and the transposed matrix k of the second matrix K to obtain the first result; based on the preset scaling factor (where d is the length of the first matrix Q) scale the first result to obtain the second result; perform softmax processing on the second result to obtain the third result; perform tensor multiplication calculation on the third result and the third matrix to obtain the first output matrix Attention out, the specific calculation formula is as follows:

[0050]

[0051] After that, the first output matrix is input into the recurrent neural network layer, that is, input into the GRU, to obtain the second output matrix. Usually, a dropout function is nested in the GRU layer, and the specific ratio can be selected as 0.5; then the second output matrix is input into the fully connected layer to obtain the prediction result; then the model parameters of the target neural network model are adjusted based on the sample labels and the prediction results; the target neural network model is iteratively trained based on multiple groups of sample data to determine the final target model parameters and obtain the irradiation calculation model.

[0052] When performing iterative training, a target loss function can be constructed, such as the commonly used mean square error loss function MSE or the cross entropy loss function Cross Entropy, and the model parameters of the target neural network model are dynamically optimized based on the optimization algorithm of gradient descent to obtain the final target model parameters. For example, the target neural network model can be iteratively trained for 500 rounds, and the model parameters are dynamically optimized through the commonly used Adam optimization algorithm. After the model training is completed, a model parameter file is generated to obtain the final irradiation calculation model.

[0053] After obtaining the irradiation calculation model, when it is necessary to calculate the irradiation value of the first area at the target time, only the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data of the first area at the target time need to be obtained and input into the trained irradiation calculation model, and the target irradiation value of the first area at the target time can be obtained. Since this data is obtained by mutually correcting satellite observation data and ground monitoring data, it has higher accuracy, and at the same time, the required number of ground environmental monitoring hardware devices is not high.

[0054] After obtaining the target irradiation value of the first area, the target output power of the photovoltaic power station in the future target time period can be predicted based on this target irradiation value. Through this target output power, the reliability of the photovoltaic power station connected to the grid can be evaluated, providing a basis for the photovoltaic power station to be connected to the grid.

[0055] In the embodiments of the present application, first, satellite cloud map feature data of a first region, first ground meteorological monitoring data of the first region, and second ground meteorological monitoring data of a second region surrounding the first region are obtained. Then, the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data are input into a pre-trained irradiance calculation model, and the target irradiance value of the first region is obtained through the output of the irradiance calculation model. Among them, the irradiance calculation model comprehensively calculates the irradiance value based on satellite observation data and ground monitoring data. It takes into account both the advantages of large satellite observation coverage and high ground monitoring accuracy, and can realize the mutual correction of satellite observation data and ground monitoring data, effectively improving the irradiance calculation accuracy, thereby solving the technical problem of large result errors in irradiance calculation in related technologies.

[0056] Embodiment 2

[0057] According to the embodiments of the present application, an irradiance calculation device for implementing the above irradiance calculation method is also provided. As Figure 3 shown, the device at least includes a first acquisition module 30, a second acquisition module 32, a third acquisition module 34, and a calculation module 36, where:

[0058] The first acquisition module 30 is used to acquire satellite cloud map feature data of the first region.

[0059] The second acquisition module 32 is used to acquire the first ground meteorological monitoring data of the first region.

[0060] The third acquisition module 34 is used to acquire the second ground meteorological monitoring data of the second region surrounding the first region.

[0061] The calculation module 36 is used to input the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiance calculation model, and obtain the irradiance value of the first region through the output of the irradiance calculation model.

[0062] In some alternative embodiments of the present application, the above-mentioned first region mainly refers to the region where the solar panels of the photovoltaic power station are located. When calculating the irradiance value of the first region, sample data of the first region at different times can be collected first for training the irradiance calculation model.

[0063] Specifically, when acquiring the satellite cloud map feature data of the first region, the first acquisition module can first acquire the satellite cloud map of the first region, such as acquiring the satellite cloud map transmitted by the Fengyun meteorological satellite, and then input the satellite cloud map into a pre-trained target neural network to output the satellite cloud map feature data. Among them, the target neural network is mainly used to extract cloud layer features, and the satellite cloud map feature data at least includes: the type, thickness, moving speed, moving direction of the cloud layer, the longitude and latitude of the first region, especially the longitude and latitude of the center point position of the first region, and other data.

[0064] When obtaining the first surface meteorological monitoring data of the first region, the second obtaining module may obtain the first surface meteorological monitoring data monitored by environmental monitoring devices (such as irradiance monitors) within the first region. Among them, the first surface meteorological monitoring data at least includes: data such as wind speed, wind direction, temperature, humidity, and air pressure in the first region.

[0065] When obtaining the second surface meteorological monitoring data of the second region around the first region, the third obtaining module may first determine that the surrounding region within a preset range (such as within 80 km) from the first region is the second region, and obtain the second surface meteorological monitoring data monitored by environmental monitoring devices within the second region. Among them, the second surface meteorological monitoring data at least includes: data such as wind speed, wind direction, temperature, humidity, air pressure, and irradiance value in the second region.

[0066] Optionally, the irradiance calculation device further includes a model training module, which is used to obtain multiple groups of sample data collected at multiple preset times. Among them, each group of sample data includes: satellite cloud map feature data, first surface meteorological monitoring data, second surface meteorological monitoring data, and irradiance value of the center point of the first region collected at the same preset time; based on the multiple groups of sample data, train a target neural network model to obtain an irradiance calculation model. Among them, the target neural network model at least includes: an attention layer, a recurrent neural network layer, and a fully connected layer. Among them, the recurrent neural network layer includes one of the following: long short-term memory network LSTM, gated recurrent unit GRU.

[0067] Among them, the irradiance calculation device further includes a fourth obtaining module, which is used to obtain the irradiance value of the center point of the first region through the irradiance monitor at the center point of the first region. This value can be regarded as the true value of the irradiance value of the first region and used as a sample label.

[0068] When the model training module collects sample data, it can set the collection time and collection frequency. For example, determine a target time period, and collect sample data once every 20 s within this target time period, that is, obtain multiple groups of sample data corresponding to multiple preset times.

[0069] After obtaining the sample data, the model training module can perform iterative training on the target neural network model based on the multiple groups of sample data to obtain the final irradiance calculation model.

[0070] In some optional embodiments of the present application, a target neural network model of GRU+Attention combination can be built based on pytorch. Among them, pytorch is an open source Python deep learning framework that can achieve powerful GPU acceleration and support dynamic neural networks; GRU is a variant of LSTM, mainly including update gates and reset gates. Compared with LSTM, GRU can better capture the dependencies with larger intervals in time series data, and the structure is simpler, so GRU is preferred for the recurrent neural network layer; Attention is used to improve the interpretability of the target neural network model.

[0071] By taking advantage of GRU’s accuracy and training time for sequential data calculations, and using Attention to explore the potential relationships between different meteorological data, the input feature weights of the target neural network model can be made more reasonable, thereby achieving mutual correction between satellite observation data and ground monitoring data.

[0072] During the model training process, for any group of sample data, the group of sample data is first normalized, and then the satellite cloud image feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiance value of the center point of the first area corresponding to the group of sample data are input into the attention layer Attention to obtain the first output matrix, in which the irradiance value of the center point of the first area is used as the sample label.

[0073] Optionally, in Attention, the first input feature can be determined based on the satellite cloud image feature data, the second input feature can be determined based on the first ground meteorological monitoring data, and the third input feature can be determined based on the second ground meteorological monitoring data. The irradiance value of the center point of the first area is used as a sample label to construct an input matrix.

[0074] Among them, considering the different degrees of influence of satellite data and ground monitoring data on the final irradiance value, different weights can be set for them. Specifically, the first preset weight corresponding to the satellite cloud image feature data, the second preset weight corresponding to the first ground meteorological monitoring data, and the third preset weight corresponding to the second ground meteorological monitoring data can be determined first; then the first input feature is determined based on the first preset weight and the satellite cloud image feature data; the second input feature is determined based on the second preset weight and the first ground meteorological monitoring data; the third input feature is determined based on the third preset weight and the second ground meteorological monitoring data. Usually, the first preset weight is greater than the second preset weight and the third preset weight.

[0075] After that, the first input feature can be denoted as module data 1, the second input feature as module data 2, the third input feature as module data 3, and the irradiation value at the center point of the first region as the sample label, denoted as Y-L. An input matrix tensor is constructed, and its format is [[module data 1, module data 2, module data 3], Y-L]; then, the input matrix tensor is copied through the attention layer Attention to obtain the first matrix Q, the second matrix K, and the third matrix V; the dot product of the first matrix Q and the transposed matrix k of the second matrix K is calculated to obtain the first result; the first result is scaled based on the preset scaling factor √d (where d is the length of the first matrix Q) to obtain the second result; the second result is processed by softmax to obtain the third result; the tensor multiplication of the third result and the third matrix is calculated to obtain the first output matrix Attention_out, and the specific calculation formula is as follows:

[0076]

[0077] After that, the first output matrix is input into the recurrent neural network layer, that is, input into GRU, to obtain the second output matrix. Usually, a dropout function is nested in the GRU layer, and the specific ratio can be selected as 0.5; then, the second output matrix is input into the fully connected layer to obtain the prediction result; then, the model parameters of the target neural network model are adjusted based on the sample label and the prediction result; the target neural network model is iteratively trained based on multiple sets of sample data to determine the final target model parameters and obtain the irradiation calculation model.

[0078] When performing iterative training, a target loss function can be constructed, such as the commonly used mean squared error loss function MSE or the cross-entropy loss function Cross Entropy, and the model parameters of the target neural network model are dynamically optimized based on the optimization algorithm of gradient descent to obtain the final target model parameters. For example, the target neural network model can be iteratively trained for 500 rounds, and the model parameters are dynamically optimized through the commonly used Adam optimization algorithm. After the model training is completed, a model parameter file is generated to obtain the final irradiation calculation model.

[0079] After obtaining the irradiation calculation model, when it is necessary to calculate the irradiation value of the first region at the target time, only the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data of the first region at the target time need to be obtained and input into the trained irradiation calculation model, and the target irradiation value of the first region at the target time can be obtained. Since this data is obtained by mutually correcting satellite observation data and ground monitoring data, it has higher accuracy, and at the same time, the required number of ground environmental monitoring hardware devices is not high.

[0080] Optionally, the irradiation calculation device further includes a prediction module, which can predict the target output power of the photovoltaic power station within a future target time period based on the target irradiation value after obtaining the target irradiation value of the first area. The reliability of the photovoltaic power station connected to the power grid can be evaluated through the target output power, providing a basis for the photovoltaic power station to connect to the power grid.

[0081] Embodiment 3

[0082] According to an embodiment of the present application, a non-volatile storage medium is further provided. The non-volatile storage medium includes a stored program, wherein when the program runs, it controls the device where the non-volatile storage medium is located to execute the irradiation calculation method in Embodiment 1.

[0083] According to an embodiment of the present application, an irradiation calculation device is further provided. The device includes: a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the irradiation calculation method in Embodiment 1 through the computer program.

[0084] Specifically, when the program runs, it executes the following steps: obtaining satellite cloud map feature data of the first area; obtaining first ground meteorological monitoring data of the first area; obtaining second ground meteorological monitoring data of a second area around the first area; inputting the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiation calculation model, and obtaining the target irradiation value of the first area through the output of the irradiation calculation model.

[0085] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0086] In the above embodiments of the present application, the descriptions of the respective embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0087] In several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0088] The unit described as a separating component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place or distributed over multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0089] In addition, each functional unit in various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0090] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present application. The aforementioned storage medium includes: USB flash drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), mobile hard disks, magnetic disks or optical discs and other various media that can store program codes.

[0091] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. An irradiation calculation method, characterized in that, it includes: Obtain satellite cloud map feature data of a first area, wherein the satellite cloud map feature data at least includes: the type, thickness, moving speed, and moving direction of the cloud layer, and the longitude and latitude of the first area; Obtain first ground meteorological monitoring data of the first area, wherein the first ground meteorological monitoring data at least includes: the wind speed, wind direction, temperature, humidity, and air pressure of the first area; Obtain second ground meteorological monitoring data of a second area surrounding the first area, wherein the second ground meteorological monitoring data at least includes: the wind speed, wind direction, temperature, humidity, air pressure, and irradiation value of the second area; Input the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiation calculation model, and obtain the target irradiation value of the first area through the output of the irradiation calculation model, wherein the irradiation calculation model is a neural network model including an attention layer, a recurrent neural network layer, and a fully connected layer; The training process of the irradiation calculation model includes: obtaining multiple groups of sample data collected at multiple preset times, training a target neural network model based on the multiple groups of sample data to obtain the irradiation calculation model. For any group of sample data, perform normalization processing on the sample data; input the satellite cloud map feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiation value of the center point of the first area corresponding to the sample data into the attention layer to obtain a first output matrix, wherein the irradiation value of the center point of the first area is used as a sample label; input the first output matrix into the recurrent neural network layer to obtain a second output matrix; input the second output matrix into the fully connected layer to obtain a prediction result; adjust the model parameters of the target neural network model based on the sample label and the prediction result; perform iterative training on the target neural network model based on the multiple groups of sample data to determine the target model parameters and obtain the irradiation calculation model.

2. The method according to claim 1, characterized in that, obtaining satellite cloud map feature data of a first area includes: Obtain the satellite cloud map of the first area; Input the satellite cloud map into a pre-trained target neural network, and output to obtain the satellite cloud map feature data.

3. The method according to claim 1, characterized in that: Each group of the sample data includes: the satellite cloud map feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiation value of the center point of the first area collected at the same preset time; The recurrent neural network layer includes one of the following: long short-term memory network, gated recurrent unit.

4. The method according to claim 3, characterized in that, inputting the satellite cloud map feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiation value of the center point of the first area corresponding to the sample data into the attention layer to obtain a first output matrix, includes: Determine the first input feature based on the satellite cloud image feature data, determine the second input feature based on the first ground meteorological monitoring data, determine the third input feature based on the second ground meteorological monitoring data, and use the irradiation value at the center point of the first region as the sample label to construct an input matrix; Copy the input matrix through the attention layer to obtain a first matrix, a second matrix, and a third matrix; Perform a dot product calculation on the first matrix and the transposed matrix of the second matrix to obtain a first result; Scale the first result based on a preset scaling factor to obtain a second result; Perform a softmax process on the second result to obtain a third result; Perform a tensor multiplication calculation on the third result and the third matrix to obtain the first output matrix.

5. The method according to claim 4, wherein, Determining the first input feature based on the satellite cloud image feature data, determining the second input feature based on the first ground meteorological monitoring data, and determining the third input feature based on the second ground meteorological monitoring data includes: Determine a first preset weight corresponding to the satellite cloud image feature data, a second preset weight corresponding to the first ground meteorological monitoring data, and a third preset weight corresponding to the second ground meteorological monitoring data; Determine the first input feature based on the first preset weight and the satellite cloud image feature data; Determine the second input feature based on the second preset weight and the first ground meteorological monitoring data; determine the third input feature based on the third preset weight and the second ground meteorological monitoring data.

6. The method according to claim 3, wherein, Iteratively training the target neural network model based on the multi-group sample data to determine the target model parameters includes: Sequentially inputting the multi-group sample data into the target neural network model for iterative training, constructing a target loss function, and dynamically optimizing the model parameters of the target neural network model based on the optimization algorithm of gradient descent to obtain the target model parameters, wherein the target loss function includes one of the following: mean square error loss function, cross-entropy loss function.

7. The method according to claim 1, wherein, The first region is the region where the solar panels of the photovoltaic power station are located. After the target irradiation value of the first region is output through the irradiation calculation model, the method further includes: Predict the target output power of the photovoltaic power station within a future target time period based on the target irradiation value, wherein the target output power is used to evaluate the reliability of the photovoltaic power station connected to the power grid.

8. The method according to claim 1, wherein, Obtaining the first ground meteorological monitoring data of the first region includes: obtaining the first ground meteorological monitoring data monitored by the environmental monitoring equipment within the first region; Obtaining the second ground meteorological monitoring data of the second region surrounding the first region includes: obtaining the second ground meteorological monitoring data monitored by the environmental monitoring equipment within the second region.

9. An irradiation calculation device, wherein, Comprising: A first acquisition module, configured to acquire satellite cloud map feature data of a first region, where the satellite cloud map feature data at least includes: the type, thickness, moving speed, and moving direction of clouds, and the longitude and latitude of the first region; A second acquisition module, configured to acquire first ground meteorological monitoring data of the first region, where the first ground meteorological monitoring data at least includes: the wind speed, wind direction, temperature, humidity, and air pressure of the first region; A third acquisition module, configured to acquire second ground meteorological monitoring data of a second region surrounding the first region, where the second ground meteorological monitoring data at least includes: the wind speed, wind direction, temperature, humidity, air pressure, and irradiation value of the second region; A calculation module, configured to input the satellite cloud map feature data, the first ground meteorological monitoring data, and the second ground meteorological monitoring data into a pre-trained irradiation calculation model, and output the irradiation value of the first region through the irradiation calculation model, where the irradiation calculation model is a neural network model including an attention layer, a recurrent neural network layer, and a fully connected layer; A training module, configured to acquire multiple sets of sample data collected at multiple preset times, and train a target neural network model based on the multiple sets of sample data to obtain the irradiation calculation model, including: for any set of sample data, performing normalization processing on the sample data; inputting the satellite cloud map feature data, the first ground meteorological monitoring data, the second ground meteorological monitoring data, and the irradiation value of the center point of the first region corresponding to the sample data into the attention layer to obtain a first output matrix, where the irradiation value of the center point of the first region is used as a sample label; inputting the first output matrix into the recurrent neural network layer to obtain a second output matrix; inputting the second output matrix into the fully connected layer to obtain a prediction result; adjusting the model parameters of the target neural network model based on the sample label and the prediction result; performing iterative training on the target neural network model based on the multiple sets of sample data to determine target model parameters, and obtaining the irradiation calculation model.

10. A non-volatile storage medium, characterized in that, the non-volatile storage medium includes a stored program, where, when the program runs, it controls the device where the non-volatile storage medium is located to execute the irradiation calculation method according to any one of claims 1 to 8.

11. An irradiation calculation device, characterized in that, it includes: a memory and a processor, where a computer program is stored in the memory, and the processor is configured to execute the irradiation calculation method according to any one of claims 1 to 8 through the computer program.

Images