Prediction method, device and equipment for generation power of photovoltaic power station and storage medium

Abstract

The invention relates to the technical field of new energy power generation, and particularly provides a photovoltaic power station generation power prediction method and device, equipment and a storage medium, and the method comprises the steps: obtaining the current dust data and the current cloud cluster feature point data set of an upstream region of a photovoltaic power station; obtaining the current generation power of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; predicting target sand and dust data of an area where the photovoltaic power station is located based on the current sand and dust data and the current cloud cluster feature point data set; and predicting the target power generation power of the photovoltaic power station based on the target dust data, the meteorological data and the current power generation power. According to the embodiment, the generation power of the photovoltaic power station is predicted by using the meteorological data, the cloud cluster feature point data set, the dust features and the current generation power, so that the accuracy of prediction of the generation power of the photovoltaic power station is effectively improved.

Description

Technical Field

[0001] The present application relates to the field of renewable energy power generation technology, and in particular to a method, device, equipment and storage medium for predicting power generation of a photovoltaic power station. Background Art

[0002] Western China is rich in photovoltaic power generation resources, but sandstorms occur frequently, which has a serious impact on the power generation of photovoltaic power stations. In particular, as the photovoltaic grid-connected capacity continues to increase, the large-scale low power of photovoltaic power generation caused by large-scale sandstorms has an increasingly serious impact on the power grid. Therefore, accurate and reliable prediction of the occurrence of sandstorms and the output of photovoltaic power stations under sandstorms can help power grid dispatchers better deal with low power generation and ensure the reliability and stability of power supply.

[0003] In the prediction of photovoltaic power generation, the similar day theory is used to divide the training samples into four weather types: sunny, cloudy, rainy and rainy. The singular spectrum analysis is used to decompose and reconstruct the training samples of each weather type, extract the trend characteristics in the time series, and design a back propagation prediction model with time delay characteristics. In addition, the historical data of photovoltaic power stations are first classified according to seasons, and then clustered according to day types, and then the corresponding power prediction model is established. The above studies set clustering indicators and use clustering algorithms to aggregate the characteristics of samples at the data level, which are often difficult to correspond to weather types one by one.

[0004] In the related technologies, in terms of prediction of power generation of photovoltaic power stations, existing research mostly focuses on regular weather or some special weather types. How to accurately predict the impact of sandstorm weather conditions on the power generation of photovoltaic power stations is an urgent problem to be solved. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides a method, device, equipment and storage medium for predicting the power generation of a photovoltaic power station, which uses meteorological data, cloud feature point data set, dust characteristics, and current power generation to predict the power generation of the photovoltaic power station, thereby effectively improving the accuracy of the power generation prediction of the photovoltaic power station.

[0006] In a first aspect, the present application provides a method for predicting the power generation of a photovoltaic power station, the method comprising: obtaining current dust data and a current cloud feature point data set in an upstream area of ​​the photovoltaic power station; obtaining the current power generation of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; predicting target dust data in the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; predicting the target power generation of the photovoltaic power station based on the target dust data, the meteorological data and the current power generation.

[0007] In the second aspect, the present application provides a device for predicting the power generation of a photovoltaic power station, the device comprising: an acquisition module for acquiring current dust data and a current cloud feature point data set in the upstream area of ​​the photovoltaic power station; acquiring the current power generation of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; a dust data prediction module for predicting target dust data in the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; a power generation prediction module for predicting the target power generation of the photovoltaic power station based on the target dust data, the meteorological data and the current power generation.

[0008] In a third aspect, the present application provides a device for predicting the power generation of a photovoltaic power station, the device comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by one or more processors, the one or more processors implement the method for predicting the power generation of a photovoltaic power station as in the first aspect above.

[0009] In a fourth aspect, the present application provides a storage medium, which may be a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting the power generation of a photovoltaic power station as in the first aspect above.

[0010] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, a method for predicting the power generation of a photovoltaic power station as described in any one of the first aspects above is implemented.

[0011] Compared with the prior art, the technical solution provided by the embodiments of the present application has the following advantages:

[0012] The embodiment of the present application provides a method, device, equipment and storage medium for predicting the power generation of a photovoltaic power station. The method includes: obtaining current dust data and current cloud feature point data set of the upstream area of ​​the photovoltaic power station; obtaining the current power generation of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; predicting the target dust data of the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; predicting the target power generation of the photovoltaic power station based on the target dust data, meteorological data and current power generation. Since the meteorological data, cloud feature point data set, dust characteristics and current power generation are used to predict the power generation of the photovoltaic power station in this embodiment, the accuracy of the power generation prediction of the photovoltaic power station is effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0015] Figure 1 A schematic diagram of a process flow of a method for predicting power generation of a photovoltaic power station provided in an embodiment of the present application;

[0016] Figure 2 A schematic diagram of the structure of a weather forecast model provided in an embodiment of the present application;

[0017] Figure 3 A schematic diagram of the structure of a weather forecast model provided in an embodiment of the present application;

[0018] Figure 4 A schematic diagram of a flow chart of a method for training a weather prediction model provided in an embodiment of the present application;

[0019] Figure 5 is a flow chart of a weather prediction model training method provided in an embodiment of the present application;

[0020] Figure 6 is a flow chart of a power prediction model training method provided in an embodiment of the present application;

[0021] Figure 7 A schematic diagram of the structure of a photovoltaic power station power generation prediction device provided in an embodiment of the present application;

[0022] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the scheme of the present application will be further described below. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only part of the embodiments of the present application, rather than all of the embodiments.

[0025] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0026] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0027] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0028] The following is a detailed description of the method for predicting the power generation of a photovoltaic power station provided by the present application in conjunction with the accompanying drawings and specific implementation methods.

[0029] Figure 1 This is a flow chart of a method for predicting the power generation of a photovoltaic power station in an embodiment of the present application. This embodiment can be applied to predicting the output of photovoltaic power stations in dusty areas. The method can be executed by a device for predicting the power generation of a photovoltaic power station. The device for predicting the power generation of a photovoltaic power station can be implemented in software and / or hardware. The device for predicting the power generation of a photovoltaic power station can be configured in an electronic device.

[0030] like Figure 1 As shown, the method for predicting the power generation of a photovoltaic power station provided in the embodiment of the present application mainly includes steps S101-S104.

[0031] S101. Obtain current dust data and a current cloud feature point dataset in an upstream area of ​​a photovoltaic power station.

[0032] A photovoltaic power station refers to a power station that uses solar panels to convert sunlight into electrical energy. The photovoltaic power station in the embodiment of the present application is mainly deployed in areas prone to sandstorms. In the process of predicting the power generation of the photovoltaic power station, the influence of sandstorms on the power generation needs to be considered.

[0033] The upstream area of ​​the photovoltaic power station can be understood as the area between the origin of the sandstorm and the photovoltaic power station. Specifically, the sandstorm begins to form at the origin and is carried by the wind to the downstream area. Before the sandstorm reaches the photovoltaic power station, the area it passes through can be called the upstream area of ​​the photovoltaic power station.

[0034] The current dust data can be understood as the data related to the sandstorm collected by the meteorological station in the upstream area at the current moment. The current moment can be understood as the moment when the power generation of the photovoltaic power station is predicted. Furthermore, the current dust data mainly includes the concentration of inhalable particulate matter, which refers to particles with a particle size of less than 10 microns, also known as PM10. Since inhalable particulate matter persists in the ambient air for a long time and has a great impact on the power generation of the photovoltaic power station, the concentration of PM10 is selected as the current dust data.

[0035] A cloud feature point refers to a pixel in a cloud image, which includes information describing the cloud attributes at that location. For example, the cloud attribute information corresponding to a pixel includes at least one or more of the following: reflectivity, brightness temperature, aerosol concentration, water vapor content, ozone content, etc. of the pixel. A cloud feature point dataset can be understood as a dataset consisting of multiple pixels and the cloud attribute information corresponding to each pixel.

[0036] Since meteorological stations are set up in various places, the meteorological stations set up in the upstream area of ​​the photovoltaic power station are determined by the location of each meteorological station. The meteorological stations set up in the upstream area are mainly used to detect meteorological data in the upstream area. A communication connection is established with the collection equipment of the meteorological station, and the meteorological data of each meteorological station is obtained through the communication connection, and the concentration of PM10 is screened out from the meteorological data as the current dust data.

[0037] Alternatively, establish a communication connection with the collection device of the weather station, and when it is necessary to predict the power generation power of the photovoltaic power station, send a data request to the collection device, where the data request is used to instruct the collection device to feedback the concentration of PM10, and receive the concentration of PM10 fed back by the collection device as the current dust data.

[0038] The current dust data includes multiple PM10 concentrations, specifically, the number of PM10 concentrations is less than or equal to the number of weather stations in the upstream area. In other words, the PM10 concentration values ​​collected by each weather station are used as the current dust data. For example: the upstream area includes 4 weather stations, then the current dust data includes 4 PM10 concentrations. Due to the influence of bad weather or special events, individual weather stations are allowed to fail to work normally, resulting in their inability to collect and transmit PM10 concentrations. At this time, the number of PM10 concentrations is less than the number of weather stations in the upstream area. In the case where all weather stations can normally collect and transmit PM10 concentrations, the number of PM10 concentrations is equal to the number of weather stations in the upstream area, and the PM10 concentration corresponds to the weather station one by one. In addition, in the embodiment of the present application, only the method of obtaining the current dust data is described, not limited.

[0039] Satellite cloud image data is obtained, and the satellite cloud image data is processed to obtain a cloud feature point data set.

[0040] In a possible implementation, obtaining a current cloud feature point dataset in an upstream region of a photovoltaic power station includes: obtaining initial satellite cloud image data of the upstream region of the photovoltaic power station, wherein the initial satellite cloud image data is satellite cloud image data of one band; performing geometric correction, radiation calibration, and data normalization on the initial satellite cloud image data to obtain a current cloud feature point dataset.

[0041] Specifically, satellite remote sensing data can be obtained from an open source satellite remote sensing data service website. The satellite remote sensing data usually includes cloud image data of multiple bands, and cloud image data of one band can be obtained as initial cloud image data. The band can be a visible light band with a wavelength of 0.47 microns.

[0042] After the initial satellite cloud image data is determined, the initial cloud image data is processed by geometric correction, radiation calibration and data normalization.

[0043] 1) Geometric correction

[0044] Some systematic or non-systematic factors can cause geometric distortion of satellite cloud image data during the imaging process. Systematic geometric distortion is caused by the sensor itself, and this part of the correction has been completed through the sensor model at the satellite ground receiving station. However, non-systematic geometric distortion is caused by multiple factors. In order to ensure the accuracy of sand and dust data prediction, non-systematic geometric distortion needs to be corrected.

[0045] The geometric correction process includes geocoding, which is mainly to correct the data to a unified geographic coordinate system.

[0046] Meteorological satellites generally use nominal projection, and their L1 data and products do not contain the two data sets of "Longitude" and "Latitude", so the FY-4A satellite needs to use a different correction method. The official website of the National Meteorological Satellite Center provides the FY-4A data row and column number and longitude and latitude lookup table. This file is a nominal grid file in RAW format. The file is filled in from north to south by row and from west to east by column. Each grid contains 16 bytes, the first 8 bytes are longitude values, and the last 8 bytes are latitude values. The lookup table file can be used to perform geometric correction on the FY-4A L1 data and cloud detection products.

[0047] 2) Radiation calibration

[0048] After geometric correction, the L1-level data and cloud detection products are projected onto the corresponding equal longitude and latitude grids. Next, the L1-level data needs to be radiometrically calibrated.

[0049] Radiation calibration is the process of converting the meaningless digital numbers (DN) collected by satellites into meaningful physical quantities. The DN value is a quantified value of the radiation received by the satellite. Its size is related to the quantization depth and it has no units. The meaningful physical quantities refer to the reflectivity of the visible light and short-wave infrared channels and the brightness temperature data of the infrared band.

[0050] In the L1 level data, a corresponding calibration table is provided for the image data layer of each channel. The radiation calibration process is realized by simply taking the DN value corresponding to a certain position in the image data layer as an index and then finding the reflectivity or brightness temperature value corresponding to the index position in the calibration table.

[0051] 3) Data Normalization

[0052] If the values ​​stored in the data vary greatly, not processing the data will have a significant impact on the model results. Therefore, in the field of deep learning, data needs to be normalized before entering the network model to eliminate the impact of differences in value ranges. Data normalization mainly scales the data to a certain ratio so that its value range is within a specific area, which is conducive to comprehensive analysis of the data.

[0053] The minimum-maximum normalization method is used in this application. After the initial cloud map data is processed by radiation calibration, the value data range in the channel is large, so the minimum value of the data is taken as 0 and the maximum value is taken as 1, and the data is mapped to the interval [0,1] by the calculation method of the following formula (1).

[0054]

[0055] Among them, m represents the longitude value, n represents the latitude value, and X m,n Indicates the actual value corresponding to the location of longitude value m and latitude value n, X max Indicates the maximum value within a channel, X min Indicates the minimum value within a channel. X' m，n Indicates the normalized value corresponding to the location of longitude value m and latitude value n.

[0056] S102: Acquire the current power generation of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located.

[0057] The current power generation of a photovoltaic power station can be understood as the power generation of all photovoltaic units in the photovoltaic power station that are working normally at the current moment. Furthermore, the average value of the power generation within a period of time can be obtained as the current power generation to avoid inaccurate prediction results caused by fluctuations in power generation and improve the accuracy of the predicted power generation. The period of time can be a period of time including the current moment. For example, the current power generation can be the average value of the power generation within 1 hour at the current moment.

[0058] Meteorological data mainly refers to data that characterizes weather conditions, including: temperature, precipitation, wind speed, wind direction, air pressure, humidity, visibility, etc.

[0059] The current meteorological data of the area where the photovoltaic power station is located can be obtained from a public meteorological platform, or from the collection equipment of the meteorological station in the area where the photovoltaic power station is located, which is not specifically limited in the embodiments of the present application.

[0060] S103: predicting target dust data in the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set.

[0061] The target dust data refers to the predicted dust data of the area where the photovoltaic power station is located after a certain period of time, wherein the certain period of time refers to the time from the current moment to the time when the dust obtains the photovoltaic power station.

[0062] A weather prediction model is trained in advance, the current dust data and the current cloud feature point data set are input into the weather prediction model, and the weather prediction model is used to predict the target dust data in the area where the photovoltaic power station is located.

[0063] In a possible implementation, target dust data in the area where the photovoltaic power station is located is predicted based on current dust data and a current cloud feature point data set, including: inputting the current dust data and the current cloud feature point data set into a pre-trained weather prediction model; using the weather prediction model to process the current dust data and the current cloud feature point data set to obtain the target dust data in the area where the photovoltaic power station is located; wherein the weather prediction model is a convolutional neural network (CNN)-gated recurrent unit (GRU).

[0064] The weather prediction model uses the CNN-GRU network, which is a deep learning model that combines the characteristics of CNN and GRU and is used to process sequence data with spatiotemporal relationships. The input of the weather prediction model is the PM10 concentration sequence collected by multiple meteorological stations in the upstream area selected according to the dust movement trajectory, as well as the cloud map feature point dataset covering all upstream areas. CNN is used to extract the spatiotemporal features of the PM10 concentration sequence, such as the shape, distribution and movement direction of the dust. The PM10 concentration sequence is passed through a series of convolutional layers and pooling layers to gradually extract features and reduce data dimensions. GRU is used to capture the time series information in the input data, especially the changing trend of dust weather. GRU controls the transmission and forgetting of information through a gating mechanism, and effectively handles long-term dependencies. The output layer is a fully connected layer that integrates the features extracted by the CNN and GRU layers and outputs the prediction results of dust weather, which are the target dust data.

[0065] The CNN network consists of an input layer, a convolutional neural layer, a downsampling layer, a fully connected layer, and an output layer. The convolutional neural network layer, the downsampling layer, and the fully connected layer are collectively referred to as hidden layers.

[0066] In the embodiment of the present application, the PM10 concentration sequence is preliminarily downsampled by the above-mentioned CNN network, and the sampling result is input into the GRU network. The calculation method of the GRU network is as follows:

[0067] GRU is a special temporal neural network that combines the forget gate and the input gate into a single update gate. The network first obtains two gated states through the last transmitted state ht-1 and the input xt of the current node. After obtaining the gated signal, the "reset gate" is first used to obtain the data rt after the "reset", and then rt is concatenated with the current memory content ht'. After activating the "update gate" zt, the final gated recurrent unit output content is obtained. The calculation process of GRU can be expressed as follows.

[0068] Among them, W(z), U(z), W(r), U(r) are the weight parameters of the input gates, σ is the sigmoid activation function, and tanh is the tanh activation function.

[0069] z t =σ(W (z) x t +U (z) h t-1 ) (2)

[0070] r t =σ(W(r) x t +U (r) h t-1 ) (3)

[0071] h t '=tanh(Wx t +r t ⊙Uh t-1 ) (4)

[0072] h t =z t ⊙h t-1 +(1-z t )⊙h t ' (5)

[0073] In a possible implementation, the weather prediction model includes an image information encoding layer, a dust information encoding layer and a decoding layer, and the dust information encoding layer includes a CNN network and a GRU network; the weather prediction model is used to process the current dust data and the current cloud feature point data set to obtain the target dust data in the area where the photovoltaic power station is located after the dust delay period, including: using the CNN network to extract the spatiotemporal features in the current dust data; using the GRU network to extract the time series features in the current dust data; fusing the spatiotemporal features with the time series features to obtain dust features; using the image information encoding layer to perform information analysis on the current cloud feature point data set to obtain cloud features; after merging the dust features and the cloud features, they are processed by the decoding layer to obtain the target dust data.

[0074] Take the upstream area including 4 meteorological stations, where 4 PM10 concentration values ​​can be collected, as an example. Figure 2 As shown, the dust information encoding layer includes 4 dust information analysis layers, each of which includes a CNN network and a GRU network connected in sequence. The 4 current PM10 concentration values ​​are respectively input into the corresponding dust information analysis layers to perform temporal feature and spatiotemporal feature analysis, and the various dust information analysis layers are fused to obtain the dust features.

[0075] The convolution layer in the image information decoding layer is used to parse the information of each channel of the current cloud map feature point data set, and then the pooling layer is used to parse the information of each pixel point to obtain the cloud cluster features.

[0076] After the dust features and cloud features are merged, they are input into the decoding layer. After decoding by GRU and ReGRU in the decoding layer, the target dust data, that is, the predicted PM10 concentration value, is output.

[0077] S104: predicting the target power generation of the photovoltaic power station based on the target dust data, meteorological data and current power generation.

[0078] In a possible implementation, the target dust data, meteorological data and current power generation are input into a power prediction model to perform power prediction, thereby obtaining the target power generation of the photovoltaic power station, wherein the target power generation is the power generation of the photovoltaic power station predicted after a period of time.

[0079] For example, if the current power generation is today's actual power generation, then the target power generation may be the predicted power generation for tomorrow.

[0080] In one possible implementation, a target power generation power of a photovoltaic power station is predicted based on target dust data, meteorological data and current power generation power, including: inputting the target dust data, meteorological data and current power generation power into a pre-trained power prediction model; using the power prediction model to process the current dust data and the current cloud feature point data set to obtain the target power generation power of the photovoltaic power station; wherein the power prediction model is a GRU network including an attention mechanism.

[0081] The power prediction model proposes an improved CNN-GRU neural network combined with an Attention layer. The main advantage of this network is that an Attention layer is added between the CNN network and the GRU layer to adjust the network's attention to different features. The neural network structure of the power prediction model is as follows: Figure 3 shown.

[0082] The current meteorological data, target dust data, and current power generation are used as inputs to the power prediction model, and the inputs are combined with the CNN-GRU network of the Attention layer to extract time series information and build a power prediction model. This network structure can make full use of the target dust data and the target dust data of downstream points, effectively capture spatiotemporal information and important features, and thus achieve high-precision prediction of power generation under dust weather.

[0083] In the embodiment of the present application, the input features are preliminarily downsampled through the above-mentioned CNN network, and the obtained results are input into the Attention layer. The core idea of ​​the Attention layer is to introduce attention weights on the input sequence to evaluate the importance of different input information, which helps the model to pay more attention to the input information with a higher correlation with the expected output and enhance the accuracy of the model.

[0084] A time series with a span of n [x 1 ,x 2 ,...,x n ] as an example, the attention weight vector α t and the weight vector x t The calculation rules of are shown in equations (6)-(8). (α)Weight matrix; ⊙ is the Hadamard operation; σ is the Sigmoid activation function.

[0085] α t =σ(W (α) x t +b α )(7)

[0086]

[0087] x t ′=α t ⊙x t =[α t1 x t1 ,α t2 x t2 ,...,α tn x tn ] (9)

[0088] In the formula, α ti is the attention weight of the i-th dimension feature at time t; x ti is the i-th dimension feature at time t.

[0089] Based on the output data of the above Attention layer, the data is input into the GRU network to extract the timing information, and then decoded in the decoding layer to output the target power generation.

[0090] The embodiment of the present application provides a method for predicting the power generation of a photovoltaic power station, the method comprising: obtaining current dust data and a current cloud feature point data set in the upstream area of ​​the photovoltaic power station; obtaining the current power generation of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; predicting the target dust data of the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; predicting the target power generation of the photovoltaic power station based on the target dust data, meteorological data and the current power generation. Since the meteorological data, cloud feature point data set, dust characteristics and current power generation are used to predict the power generation of the photovoltaic power station in this embodiment, the accuracy of the power generation prediction of the photovoltaic power station is effectively improved.

[0091] Based on the above embodiments, the present application provides a method for training a weather prediction model. Figure 4 As shown, the weather model training method provided in the embodiment of the present application mainly includes the following process:

[0092] S201. Obtain first sample dust data and a sample cloud feature point data set.

[0093] The first sample dust data and the sample cloud feature point data set are both obtained from historical data of the upstream area of ​​the photovoltaic power station, and the historical data includes: historical meteorological data and historical satellite cloud image data. The span of the historical data is greater than or equal to 1 year.

[0094] The photovoltaic power station of this embodiment is a city along the dust movement track. In a specific embodiment, the historical data of XX year of four weather stations in the area where sandstorms frequently occur are obtained, wherein the historical data is obtained at intervals of 1 hour, and the data of 24 points are obtained every day.

[0095] Specifically, the above meteorological data specifically includes: PM10 concentration value, and may also include: air quality index (AQI), etc.; illustratively, the embodiment of the present application selects the concentration value of PM10 concentration as the sample dust data. The first sample dust data refers to any concentration value of PM10 concentration.

[0096] The satellite cloud image data has one band, and in this embodiment, satellite cloud image data of one band is selected, which is a visible light band with a wavelength of 0.47 microns. The time interval is 1 hour.

[0097] The satellite cloud image data at each moment is processed to obtain the cloud cluster feature points at that moment to construct a satellite cloud cluster feature point dataset. The specific steps are as follows:

[0098] 1) Geometric correction

[0099] Some systematic or non-systematic factors can cause geometric distortion in the remote sensing satellite during the imaging process. Systematic geometric distortion is caused by the sensor itself, and this part of the correction has been completed through the sensor model at the satellite ground receiving station. However, non-systematic geometric distortion is caused by multiple factors, and this part of the correction needs to be completed in the application of satellite data. Therefore, geometric correction is a necessary preprocessing step for any remote sensing satellite image before application. Geometric correction also includes geocoding, which is mainly to correct the data to a unified geographic coordinate system.

[0100] The FY-4A satellite uses a nominal projection method, and its L1-level data and products do not contain the two data sets of "Longitude" and "Latitude", so the FY-4A satellite needs to use a different correction method. The official website of the National Meteorological Satellite Center provides a lookup table of FY-4A data row and column numbers and longitude and latitude. This file is a nominal grid file in RAW format. The file is filled in rows from north to south and columns from west to east. Each grid contains 16 bytes, the first 8 bytes are longitude values, and the last 8 bytes are latitude values. This lookup table file can be used to perform geometric correction on FY-4A's L1-level data and cloud detection products.

[0101] 1-2-2) Radiation calibration

[0102] After geometric correction, the L1 data and cloud detection products of the FY-4A satellite are projected onto the corresponding equal longitude and latitude grids. Next, the L1 data needs to be radiometrically calibrated.

[0103] Radiation calibration is the process of converting the meaningless DN value collected by the satellite into meaningful physical quantities. The DN value is a quantified value of the radiation received by the satellite. Its size is related to the quantification depth and it has no unit. The meaningful physical quantities refer to the reflectivity of the visible light and short-wave infrared channels and the brightness temperature data of the infrared band.

[0104] In the L1-level data of the FY-4A satellite, a corresponding calibration table is provided for the image data layer of each channel. The radiation calibration process is realized by simply taking the DN value corresponding to a certain position in the image data layer as an index and then finding the reflectivity or brightness temperature value corresponding to the index position in the calibration table.

[0105] 1-2-3) Data normalization

[0106] If the values ​​stored in the data vary greatly, not processing the data will have a significant impact on the model results. Therefore, in the field of deep learning, data needs to be normalized before entering the network model to eliminate the impact of differences in value ranges. Data normalization mainly scales the data to a certain ratio so that its value range is within a specific area, which is conducive to comprehensive analysis of the data.

[0107] The present invention adopts the minimum-maximum normalization method. The research data used is the visible light band of the FY-4A satellite multi-channel imager. After the radiation calibration process, the value data range in the channel is large. Therefore, the present invention takes the minimum value of the data as 0 and the maximum value as 1, and maps the data to the [0,1] interval through the calculation method of the following formula.

[0108] S202: Determine a dust delay time, wherein the dust delay time refers to the time interval between the time when the first sample dust data is collected and the time when the dust reaches the photovoltaic power station.

[0109] The dust delay time is calculated by formula (10):

[0110]

[0111] Wherein, h is the dust delay duration, i is the serial number of the weather station, N is the total number of weather stations, x is the longitude value of the photovoltaic power station, y is the latitude value of the photovoltaic power station, and x is the latitude value of the photovoltaic power station. i represents the longitude value of the ith weather station, y is the latitude value of the ith weather station, and k is the seasonal variation coefficient.

[0112] In this embodiment, four weather stations are used as an example. The sandstorm delay duration h is calculated by the longitude and latitude of each weather station and the photovoltaic power station. Specifically, the longitude and latitude coordinates (x1, y1), (x2, y2), (x3, y3), (x4, y4) of the four weather stations in the upstream area and the longitude and latitude coordinates (x, y) of the photovoltaic power station are used to calculate the sandstorm delay duration h by formula (11).

[0113]

[0114] Wherein, k is a coefficient that varies with the region and season, and is taken as 2.47 in this embodiment.

[0115] S203, obtaining second sample dust data, wherein the second sample dust data is dust data corresponding to a collection time after a dust delay time interval between the collection time of the first sample dust data.

[0116] Among them, the second sample sand and dust data is also the concentration value of PM10. The difference between the second sample sand and dust data and the first sample sand and dust data lies in the different collection times. Specifically, the collection time of the first sample sand and dust data is t1, and the sand and dust delay time is Δt, then the collection time of the second sample sand and dust data is t2=t1+Δt. For example: the collection time of the first sample sand and dust data is 13 o'clock, and the sand and dust delay time is 4 hours, then the sand and dust data collected at 17 o'clock is used as the second sample sand and dust data corresponding to the first sample sand and dust data. In other words, the collection time of the second sample sand and dust data is separated from the collection time of the first sample sand and dust data by a sand and dust delay time.

[0117] S204: Using the first sample dust data and the sample cloud feature point data set as input data, and using the second sample dust data as supervision data, the CNN-GRU model is trained to obtain a weather prediction model.

[0118] In this embodiment, the weather prediction model is trained based on the first sample dust data, the sample cloud cluster characteristic point data set, and the second sample dust data obtained in the above steps to obtain a trained weather prediction model.

[0119] In a specific embodiment, the data of the training samples comes from the data of a power station throughout 2021. The batch size is set to 8, the epoch is set to 50, the learning rate is set to 0.001, and the loss function is set to MAE and RMSE. The model is trained using the training set. If the maximum number of iterations (k>K) is reached, the iteration is terminated and the network parameters are output; otherwise, k=k+1 is set and the iteration continues.

[0120] Figure 5is a flow chart of another weather prediction model training method provided in an embodiment of the present application; Figure 5 As shown, the weather prediction model training method provided in the embodiment of the present application mainly includes the following process:

[0121] S301. Obtain the PM10 concentration value and initial satellite cloud image data in the upstream area.

[0122] S302: preprocess, normalize, and serialize the initial satellite cloud image data to obtain a cloud image feature point data set.

[0123] S303, input the PM10 concentration value and cloud map feature point data set into the convolutional neural network.

[0124] S304, input the features output by the convolutional neural network into the gated recurrent network.

[0125] S305: Determine whether the target accuracy or the maximum number of iterations is reached. If not, adjust the model parameters and return to execute S303. If yes, execute S306.

[0126] S306: Obtain a trained weather forecast model and output a predicted PM10 concentration value.

[0127] Figure 6 is a flow chart of a power prediction model training method provided in an embodiment of the present application; Figure 6 As shown, the power prediction model training method provided in the embodiment of the present application mainly includes the following process:

[0128] S401. Obtain predicted PM10 concentration values ​​and meteorological data of the photovoltaic power station.

[0129] S402: Preprocess, normalize, and serialize the meteorological data to obtain processed meteorological data.

[0130] S403, input the predicted PM10 concentration value and meteorological data into the convolutional neural network.

[0131] S404: Input the features output by the convolutional neural network into the attention layer and the gated recurrent network.

[0132] S405: Determine whether the target accuracy or the maximum number of iterations is reached. If not, adjust the model parameters and return to execute S403. If yes, execute S406.

[0133] S406: Obtain a trained power prediction model and output the predicted power.

[0134] Figure 7FIG. 1 is a schematic diagram of a photovoltaic power station power generation prediction device in an embodiment of the present application. Figure 7 As shown, the photovoltaic power station power generation prediction device 70 provided in the embodiment of the present application mainly includes: an acquisition module 71, a dust data prediction module 72 and a power generation prediction module 73.

[0135] Among them, the acquisition module 71 is used to obtain the current dust data and the current cloud feature point data set of the upstream area of ​​the photovoltaic power station; to obtain the current power generation power of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; the dust data prediction module 72 is used to predict the target dust data of the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; the power generation prediction module 73 is used to predict the target power generation power of the photovoltaic power station based on the target dust data, meteorological data and current power generation power.

[0136] The embodiment of the present application provides a device for predicting the power generation of a photovoltaic power station, which is used to execute the following process: obtaining the current dust data and the current cloud feature point data set of the upstream area of ​​the photovoltaic power station; obtaining the current power generation of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; predicting the target dust data of the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; predicting the target power generation of the photovoltaic power station based on the target dust data, meteorological data and the current power generation. Since the meteorological data, cloud feature point data set, dust characteristics and current power generation are used to predict the power generation of the photovoltaic power station in this embodiment, the accuracy of the power generation prediction of the photovoltaic power station is effectively improved.

[0137] In a possible implementation, the acquisition module 71 is specifically used to obtain initial satellite cloud image data of the upstream area of ​​the photovoltaic power station, wherein the initial satellite cloud image data is satellite cloud image data of one band; the initial satellite cloud image data is geometrically corrected, radiometrically calibrated, and data normalized to obtain the current cloud feature point data set.

[0138] In one possible implementation, the dust data prediction module 72 is specifically used to input the current dust data and the current cloud feature point data set into a pre-trained weather prediction model; use the weather prediction model to process the current dust data and the current cloud feature point data set to obtain the target dust data in the area where the photovoltaic power station is located; wherein the weather prediction model is a convolutional neural network-gated recurrent network CNN-GRU.

[0139] In one possible implementation, the weather prediction model includes an image information encoding layer, a dust information encoding layer and a decoding layer, the dust information encoding layer includes a CNN network and a GRU network; the dust data prediction module 72 is specifically used to use the CNN network to extract the spatiotemporal features in the current dust data; use the GRU network to extract the time series features in the current dust data; fuse the spatiotemporal features with the time series features to obtain dust features; use the image information encoding layer to perform information analysis on the current cloud feature point data set to obtain cloud features; after merging the dust features and the cloud features, they are processed by the decoding layer to obtain the target dust data.

[0140] In a possible implementation, it also includes a model training module for training a weather prediction model, specifically, obtaining first sample dust data and a sample cloud feature point data set; determining the dust delay duration, wherein the dust delay duration refers to the time interval between the collection time of the first sample dust data and the time when the dust reaches the photovoltaic power station; obtaining second sample dust data, wherein the second sample dust data is the dust data corresponding to the collection time after the dust delay duration at the collection time of the first sample dust data; using the first sample dust data and the sample cloud feature point data set as input data, and the second sample dust data as supervision data, to train the CNN-GRU model to obtain a weather prediction model.

[0141] In a possible implementation, the dust delay duration is calculated using the following formula:

[0142]

[0143] Where, h is the dust delay duration, i is the serial number of the weather station, N is the total number of weather stations, x is the longitude of the photovoltaic power station, y is the latitude of the photovoltaic power station, and x is the latitude of the photovoltaic power station. i represents the longitude value of the ith meteorological station, y is the latitude value of the ith meteorological station, and k is the seasonal variation coefficient.

[0144] In one possible implementation, the power generation prediction module 73 is specifically used to input the target dust data, meteorological data and current power generation into a pre-trained power prediction model; use the power prediction model to process the current dust data and the current cloud feature point data set to obtain the target power generation of the photovoltaic power station; wherein the power prediction model is a GRU network including an attention mechanism.

[0145] The photovoltaic power station power generation prediction device provided in the embodiment of the present application can execute the photovoltaic power station power generation prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0146] Figure 8is a schematic diagram of the structure of an electronic device provided in this embodiment. The electronic device may include a photovoltaic power station power generation prediction device, such as Figure 8 As shown, the electronic device 800 includes a processor 810, a memory 820, an input device 830, and an output device 840; the number of the processor 810 in the electronic device can be one or more. Figure 8 A processor 810 is taken as an example; the processor 810, the memory 820, the input device 830 and the output device 840 in the electronic device can be connected via a bus or other means. Figure 8 The example of connecting through bus is taken in the following.

[0147] The memory 820, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the photovoltaic power station power generation prediction method in the embodiment of the present invention. The processor 810 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 820, that is, the photovoltaic power station power generation prediction method provided in the embodiment of the present invention is realized.

[0148] The memory 820 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 820 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 820 may further include a memory remotely arranged relative to the processor 810, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0149] The input device 830 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the electronic device, and may include a keyboard, a mouse, etc. The output device 840 may include a display device such as a display screen.

[0150] This embodiment also provides a storage medium containing computer executable instructions, which, when executed by a computer processor, are used to implement the method for predicting the power generation of a photovoltaic power station provided in the embodiment of the present invention.

[0151] Of course, the storage medium containing computer executable instructions provided in an embodiment of the present invention is not limited to the method operations described above, and can also execute related operations in the photovoltaic power station power generation prediction method provided in any embodiment of the present invention.

[0152] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0153] It is worth noting that in the embodiment of the above-mentioned photovoltaic power station power generation prediction device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0154] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0155] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting power generation of a photovoltaic power station, characterized in that: The method comprises: Obtain the current dust data and current cloud feature point dataset in the upstream area of ​​the photovoltaic power station; Acquire the current power generation of the photovoltaic power station and the current meteorological data of the area where the photovoltaic power station is located; Predicting target dust data in the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; The target power generation of the photovoltaic power station is predicted based on the target dust data, the meteorological data and the current power generation.

2. The method according to claim 1, characterized in that Obtain the current cloud feature point dataset in the upstream area of ​​the photovoltaic power station, including: Acquire initial satellite cloud image data of the upstream area of ​​the photovoltaic power station, wherein the initial satellite cloud image data is satellite cloud image data of one band; The initial satellite cloud image data is subjected to geometric correction, radiation calibration and data normalization processing to obtain the current cloud cluster feature point data set.

3. The method according to claim 1, characterized in that Predicting target dust data in the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set includes: Inputting the current dust data and the current cloud feature point data set into a pre-trained weather prediction model; Processing the current dust data and the current cloud feature point data set using the weather forecast model to obtain target dust data for the area where the photovoltaic power station is located; Among them, the weather prediction model is a convolutional neural network CNN-gated recurrent network GRU.

4. The method according to claim 3, characterized in that The weather prediction model includes an image information encoding layer, a dust information encoding layer and a decoding layer, and the dust information encoding layer includes a CNN network and a GRU network; The current dust data and the current cloud feature point data set are processed using the weather forecast model to obtain target dust data of the photovoltaic power station area after the dust delay time, including: Extracting the spatiotemporal features of the current dust data using the CNN network; Extracting time series features in the current dust data using the GRU network; The spatiotemporal features are integrated with the time series features to obtain dust features; Using the image information coding layer to perform information analysis on the current cloud feature point data set to obtain cloud features; After the dust features and the cloud features are combined and processed by the decoding layer, target dust data is obtained.

5. The method according to claim 3, characterized in that: The training process of the weather prediction model includes: Acquire the first sample dust data and the sample cloud feature point data set; Determine a dust delay time, wherein the dust delay time refers to the time interval between the time when the first sample dust data is collected and the time when the dust reaches the photovoltaic power station; Acquire second sample dust data, wherein the second sample dust data is dust data corresponding to a collection time after a dust delay time interval between the collection time of the first sample dust data; The first sample dust data and the sample cloud feature point data set are used as input data, and the second sample dust data is used as supervision data to train the CNN-GRU model to obtain a weather prediction model.

6. The method according to claim 5, characterized in that The dust delay duration is calculated by the following formula: Wherein, h is the dust delay duration, i is the serial number of the weather station, N is the total number of weather stations, x is the longitude value of the photovoltaic power station, y is the latitude value of the photovoltaic power station, and x is the latitude value of the photovoltaic power station. i represents the longitude value of the ith meteorological station, y is the latitude value of the ith meteorological station, and k is the seasonal variation coefficient.

7. The method according to claim 1, characterized in that Predicting the target power generation power of the photovoltaic power station based on the target dust data, the meteorological data and the current power generation power includes: Inputting the target dust data, the meteorological data and the current power generation into a pre-trained power prediction model; Processing the current dust data and the current cloud feature point data set using the power prediction model to obtain a target power generation power of the photovoltaic power station; The power prediction model is a GRU network including an attention mechanism.

8. A device for predicting power generation of a photovoltaic power station, characterized in that: The device comprises: An acquisition module is used to acquire current dust data and current cloud feature point data set in the upstream area of ​​the photovoltaic power station; acquire current power generation of the photovoltaic power station and current meteorological data of the area where the photovoltaic power station is located; A dust data prediction module, used to predict target dust data in the area where the photovoltaic power station is located based on the current dust data and the current cloud feature point data set; The power generation prediction module is used to predict the target power generation of the photovoltaic power station based on the target dust data, the meteorological data and the current power generation.

9. An electronic device, characterized in that: The device comprises: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting the power generation of a photovoltaic power station according to any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for predicting the power generation of a photovoltaic power station as described in any one of claims 1 to 7 is implemented.

Images