Photovoltaic power prediction method and system based on deep learning and multi-source data fusion driving

Abstract

The invention discloses a photovoltaic power prediction method and system based on deep learning and multi-source data fusion driving, and the method comprises the following steps: collecting photovoltaic power output, multi-source meteorological data and a multi-wavelength channel satellite cloud picture of a target region, and forming a data sample library of a prediction model; the invention provides a bidirectional dynamic residual mechanism suitable for a long short-term memory neural network and a self-attention convolution long short-term memory neural network, and the problems of gradient explosion and insufficient long-term time sequence feature modeling are effectively improved; an improved self-attention convolution long-short-term memory neural network is used for predicting a satellite cloud picture at a prediction moment, and a prediction model can fully capture space-time dynamic information in a historical cloud picture sequence under the guidance of a future cloud picture; the method can be used for actual distributed photovoltaic power prediction, generalization and robustness of the prediction method are improved, and safe and stable operation of a power system is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic power generation, and specifically relates to a photovoltaic power prediction method and system based on deep learning and multi-source data fusion drive. Background Art

[0002] The rapid development of new energy is an inevitable trend under the current energy transformation and policy background. Driven by the dual forces of cost reduction and increased sustainability requirements, the installed capacity of distributed photovoltaic power generation has grown rapidly in recent years.

[0003] The volatility of photovoltaic power generation is manifested in that even if the output forecast value at a certain moment is accurate, the photovoltaic output may still fluctuate violently due to cloud disturbances within a smaller time scale, affecting the stability of the power system, power quality and other aspects. Therefore, the demand for accurate power prediction is becoming increasingly urgent. In the existing technology, the current research and application of photovoltaic power prediction technology has problems such as single data, insufficient information, low sample quality, and missing main feature information. In addition, due to the uncertainty of atmospheric movement, photovoltaic power series, especially photovoltaic power generation series, have strong volatility and strong noise, which greatly reduces its measurability.

[0004] To this end, we propose a photovoltaic power prediction method and system driven by deep learning and multi-source data fusion to solve the problems mentioned in the above background technology. Summary of the invention

[0005] The purpose of the present invention is to provide a photovoltaic power prediction method and system based on deep learning and multi-source data fusion drive to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a photovoltaic power prediction method based on deep learning and multi-source data fusion drive, comprising the following steps:

[0007] S1. Collect the photovoltaic power output, multi-source meteorological data, and multi-wavelength channel satellite cloud images of the target area to form a data sample library for the prediction model. The collected data sequence is:

[0008] I=[i t-n ,i t-n+1 ,...,i t ]

[0009] P = [p t-n ,p t-n+1 ,...,p t ]

[0010] W=[w t+1 ,w t+2 ,...,w t+m ];

[0011] S2. Clean the photovoltaic power output and multi-source meteorological data, and use the pyramid algorithm to fuse the multi-wavelength channel satellite cloud images. Use the multi-level pyramid algorithm to fuse the satellite cloud images of visible light, near infrared, infrared light, water vapor and other spectral bands to obtain a fused satellite cloud image that can fully represent the cloud information.

[0012] S3. Based on the historical multi-step satellite cloud images, the improved self-attention convolutional long short-term memory neural network is used to predict the satellite cloud images at the future prediction time. The spatial distribution characteristics of the predicted cloud images and the spatiotemporal dynamic characteristics of the historical cloud images are respectively proposed through the two-dimensional convolutional neural network and the three-dimensional convolutional neural network model. The size of the historical satellite cloud images is N, H, W, and C; N is the number of historical cloud image frames, H and W are the height, width and number of channels of the cloud image respectively, and C is the length of the historical satellite cloud image sequence used. Among them, the continuous historical satellite cloud images are taken as input, and the improved self-attention convolutional long short-term memory neural network is used for image prediction, and then the satellite cloud images at the prediction time are obtained. The definition of the improved self-attention convolutional long short-term memory neural network is:

[0013]

[0014] S4. Use wavelet transform algorithm to decompose the photovoltaic power sequence, remove noise, and obtain inherent characteristic components. The process can be expressed as:

[0015]

[0016] Based on the improved long short-term memory neural network to capture the temporal characteristics of the inherent feature components, the process can be described as follows:

[0017]

[0018] Use multi-layer perceptron to extract deep features of multi-source weather forecast data:

[0019]

[0020] S5. Taking the spatiotemporal characteristics of the historical cloud image, the spatial characteristics of the predicted cloud image, the time series characteristics of the historical power, and the deep meteorological characteristics as the input of the multi-layer perceptron, the photovoltaic power prediction result O is obtained. The process can be described as follows:

[0021]

[0022] O=[o t+1 ,o t+2 ,...,o t+m ];

[0023] Among them, O represents the multi-step prediction result, [o t+1,o t+2 ,...,o t+m ] is the photovoltaic power prediction result at each moment from t+1 to t+m;

[0024] S6. Based on real-time new data, the prediction model is optimized and corrected in real time using the incremental update method.

[0025] Preferably, in the formula S1: I, P, and W are satellite cloud images, photovoltaic power, and weather forecast data, respectively, i t is the satellite cloud image set of multi-wavelength channels at time t, n is the length of the historical sequence, and m is the length of the predicted sequence.

[0026] Preferably, the photovoltaic power and multi-source weather forecast in S2 are subjected to data cleaning, including outlier screening and data interpolation, using the box plot method to identify outliers in the original data and delete them, querying missing values ​​in the time series direction, and filling the missing values ​​through the K nearest neighbor algorithm.

[0027] Preferably, in the formula S3: wherein x t ,h t-1 is the input value at time t and the output value at time t-1; For x t Updated features obtained through the self-attention module; h t-1 The updated value obtained by the dynamic residual module and the self-attention module in turn; c t , c t-1 represents the unit state at time t and time t-1; BDR is the dynamic residual module; SA is the self-attention memory module; i t , f t , g t They are input gate, forget gate and output gate respectively; W xi , W hi , b i , W xf , W hf , b f , W xc , W hc , b c , W xg , W hg , b g is the learning parameter; tanh is the hyperbolic tangent function, σ is the activation function, represents the Hadamard product; h t is the cell state at time t.

[0028] Preferably, the satellite cloud image features at the prediction time are extracted through a two-dimensional convolutional neural network model to obtain the spatial distribution features of the predicted cloud image, and the process can be expressed as:

[0029]

[0030] in, To predict the spatial distribution characteristics of the cloud image, f 2DCNN is a two-dimensional convolutional neural network model, [i t+1 ,i t+2 ,...,i t+m ] is the predicted satellite cloud image sequence.

[0031] Preferably, the feature extraction of the historical satellite cloud image sequence is performed through a three-dimensional convolutional neural network model to obtain the spatiotemporal dynamic features of the cloud layer, and the process can be described as follows:

[0032]

[0033] in, is the spatiotemporal dynamic characteristics of the historical cloud sequence, f 3DCNN is a three-dimensional convolutional neural network model, [i t-n ,i t-n+1 ,...,i t ] is a historical satellite cloud image sequence.

[0034] Preferably, in S4, p denoised is the denoised photovoltaic power sequence, t is the time variable, represents the sum of all scales s and all frequency bands k, W k,s is the wavelet coefficient after threshold processing, k represents the frequency band of the wavelet, s represents the scale parameter, ψ k,s is the wavelet function corresponding to scale s and frequency band k;

[0035] is the time series characteristic of the historical photovoltaic power series, f IM-LSTM For the improved long short-term memory neural network model, [p t-n ,p t-n+1 ,...,p t ] is the historical photovoltaic power series;

[0036] is the deep meteorological characteristics of multi-source meteorological forecast data, [w t+1 ,w t+2 ,...,w t+m ] is the weather forecast data corresponding to each forecast time from t+1 to t+m, f MLP It is a multi-layer perceptron model.

[0037] Photovoltaic power prediction system based on deep learning and multi-source data fusion, including:

[0038] The data acquisition module collects photovoltaic power output, multi-source meteorological data, and multi-wavelength channel satellite cloud images of the target area to form a data sample library for the prediction model;

[0039] The data preprocessing module cleans the photovoltaic power output and multi-source meteorological data, and uses the pyramid algorithm to fuse multi-wavelength channel satellite cloud images;

[0040] The feature extraction module uses an improved self-attention convolutional long short-term memory neural network to predict satellite cloud images at future forecast times based on historical multi-step satellite cloud images. The spatial distribution characteristics of the forecast cloud images and the spatiotemporal dynamic characteristics of the historical cloud images are respectively obtained through two-dimensional convolutional neural network and three-dimensional convolutional neural network models. The photovoltaic power sequence is decomposed using the wavelet transform algorithm to obtain the inherent characteristic components. The time series characteristics of the inherent characteristic components are captured based on the improved long short-term memory neural network. The deep features of multi-source meteorological forecast data are extracted using a multi-layer perceptron.

[0041] The power prediction module uses the spatiotemporal features of the historical cloud map, the spatial features of the predicted cloud map, the time series features of the historical power, and the deep meteorological features as the input of the multi-layer perceptron to obtain the multi-step prediction results;

[0042] The incremental update module optimizes and corrects the prediction model in real time based on real-time new data using the incremental update method.

[0043] A computer device comprising:

[0044] one or more processors;

[0045] Memory;

[0046] And one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, and when the programs are executed by the processors, the steps of the photovoltaic power prediction method based on deep learning and multi-source data fusion are implemented as described.

[0047] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the photovoltaic power prediction method based on deep learning and multi-source data fusion drive.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] (1) The present invention proposes a bidirectional dynamic residual mechanism suitable for long short-term memory neural network and self-attention convolutional long short-term memory neural network, which effectively improves the problems of gradient explosion and insufficient modeling of long-term temporal features.

[0050] (2) The present invention uses an improved self-attention convolutional long short-term memory neural network to predict the satellite cloud images at the prediction time. Guided by the future cloud images, the prediction model can fully capture the spatiotemporal dynamic information in the historical cloud image sequence.

[0051] (3) The present invention can be used for actual distributed photovoltaic power prediction, improve the generalization and robustness of the prediction method, and ensure the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the photovoltaic power prediction method of the present invention;

[0053] Figure 2 A schematic diagram of a self-attention convolutional long short-term memory neural network used in the present invention;

[0054] Figure 3 A schematic diagram of the bidirectional dynamic residual mechanism proposed by the present invention;

[0055] Figure 4 A schematic diagram of a two-dimensional convolutional neural network used in the present invention;

[0056] Figure 5 It is a schematic diagram of fitting the prediction results using the proposed method in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] See also Figure 1-5 , the present invention provides a technical solution:

[0059] The photovoltaic power prediction method based on deep learning and multi-source data fusion driving includes the following steps:

[0060] S1. Collect the photovoltaic power output, multi-source meteorological data, and multi-wavelength channel satellite cloud images of the target area to form a data sample library for the prediction model. The collected data sequence is:

[0061] I=[i t-n ,it-n+1 ,...,i t ]

[0062] P = [p t-n ,p t-n+1 ,...,p t ]

[0063] W=[w t+1 ,w t+2 ,...,w t+m ];

[0064] S2. Clean the photovoltaic power output and multi-source meteorological data, and use the pyramid algorithm to fuse the multi-wavelength channel satellite cloud images. Use the multi-level pyramid algorithm to fuse the satellite cloud images of visible light, near infrared, infrared light, water vapor and other spectral bands to obtain a fused satellite cloud image that can fully represent the cloud information.

[0065] S3. Based on the historical multi-step satellite cloud images, the improved self-attention convolutional long short-term memory neural network is used to predict the satellite cloud images at the future prediction time. The spatial distribution characteristics of the predicted cloud images and the spatiotemporal dynamic characteristics of the historical cloud images are respectively proposed through the two-dimensional convolutional neural network and the three-dimensional convolutional neural network model. The size of the historical satellite cloud image is N, H, and W; N is the number of historical cloud image frames, H and W are the height, width and number of channels of the cloud image respectively, and C is the length of the historical satellite cloud image sequence used. Among them, the continuous historical satellite cloud images are taken as input, and the improved self-attention convolutional long short-term memory neural network is used for image prediction, and then the satellite cloud image at the prediction time is obtained. The definition of the improved self-attention convolutional long short-term memory neural network is:

[0066]

[0067] S4. Use wavelet transform algorithm to decompose the photovoltaic power sequence, remove noise, and obtain inherent characteristic components. The process can be expressed as:

[0068]

[0069] Based on the improved long short-term memory neural network to capture the temporal characteristics of the inherent feature components, the process can be described as follows:

[0070]

[0071] Use multi-layer perceptron to extract deep features of multi-source weather forecast data:

[0072]

[0073] S5. Taking the spatiotemporal characteristics of the historical cloud image, the spatial characteristics of the predicted cloud image, the time series characteristics of the historical power, and the deep meteorological characteristics as the input of the multi-layer perceptron, the photovoltaic power prediction result O is obtained. The process can be described as follows:

[0074]

[0075] O=[o t+1 ,o t+2 ,...,o t+m ];

[0076] Among them, O represents the multi-step prediction result, [o t+1 ,o t+2 ,...,o t+m ] is the photovoltaic power prediction result at each moment from t+1 to t+m;

[0077] S6. Based on real-time new data, the prediction model is optimized and corrected in real time using the incremental update method.

[0078] In S1, the photovoltaic power output, multi-source meteorological data, and multi-wavelength channel satellite cloud images of the target area are collected to form a data sample library for the prediction model. The collected data sequence is:

[0079] I=[i t-n ,i t-n+1 ,...,i t ]

[0080] P = [p t-n ,p t-n+1 ,...,p t ]

[0081] W=[w t+1 ,w t+2 ,...,w t+m ]

[0082] Among them, I, P, and W are satellite cloud images, photovoltaic power, and weather forecast data, respectively. t is the satellite cloud image set of multi-wavelength channels at time t, n is the length of the historical sequence, and m is the length of the predicted sequence;

[0083] In S2, the present invention performs data cleaning on photovoltaic power and multi-source weather forecasts, including outlier screening and data interpolation; uses a box plot method to identify outliers in the original data and deletes them; queries missing data values ​​along the time series direction and fills the missing values ​​using a K-nearest neighbor algorithm;

[0084] The satellite cloud images of visible light, near infrared, infrared light, water vapor and other spectral bands are fused using a multi-level pyramid algorithm to obtain a fused satellite cloud image that can fully represent cloud information.

[0085] In S3 of the present invention, the sizes of the historical satellite cloud images are N, H, W, and C; N is the number of frames of the historical cloud images, H and W are the height, width, and number of channels of the cloud images, respectively, and C is the length of the sequence of historical satellite cloud images used; Figure 2 As shown in the figure, continuous historical satellite cloud images are used as input, and the improved self-attention convolutional long short-term memory neural network is used for image prediction to obtain the satellite cloud image at the predicted time. The improved self-attention convolutional long short-term memory neural network is defined as:

[0086]

[0087] Among them, x t ,h t-1 is the input value at time t and the output value at time t-1; For x t Updated features obtained through the self-attention module; h t-1 The updated value obtained by the dynamic residual module and the self-attention module in turn; c t , c t-1 represents the unit state at time t and time t-1; BDR is the dynamic residual module; SA is the self-attention memory module; i t , f t , g t They are input gate, forget gate and output gate respectively; W xi , W hi , b i , W xf , W hf , b f , W xc , W hc , b c , W xg , W hg , b g is a learnable parameter; tanh is the hyperbolic tangent function, σ is the activation function, represents the Hadamard product; h t is the unit state at time t; Figure 3 As shown, the specific implementation process of the bidirectional dynamic residual mechanism can be expressed by the following formula:

[0088]

[0089] Among them, v is the residual information transmitted vertically, z is the residual information transmitted horizontally, e is the number of layers of the convolutional long short-term memory neural network used, t is the length of the time series, tanh is the hyperbolic tangent function, σ is the activation function, represents the Hadamard product, W h1 , Wv1 、b v1 , W h2 , W v2 、b v2 , W h3 , W z1 、b z1 , W h4 , W z2 、b z2 is a learnable parameter;

[0090] The satellite cloud image features at the prediction time are extracted through the 2DCNN model to obtain the spatial distribution characteristics of the predicted cloud image. The process can be described as follows:

[0091]

[0092] in, To predict the spatial distribution characteristics of the cloud image, f 2DCNN For the 2DCNN model, [i t+1 ,i t+2 ,...,i t+m ] is the predicted satellite cloud image sequence; Figure 4 As shown, the 2DCNN model can be expressed by the following formula:

[0093]

[0094] in, is the input of layer l, is the cth convolution kernel of the lth layer, a (l-1) is the two-dimensional convolution output feature map of layer l-1, * represents the convolution operation, is the bias of the cth convolution kernel, is the activation value of the cth feature map of the lth layer at position (x, y), and f is the mapping function;

[0095] The feature extraction of historical satellite cloud image sequences is carried out through the 3DCNN model to obtain the spatiotemporal dynamic characteristics of the cloud layer. The process can be described as follows:

[0096]

[0097] in, is the spatiotemporal dynamic characteristics of the historical cloud sequence, f 3DCNN For the 3DCNN model, [i t-n ,i t-n+1 ,...,i t ] is a historical satellite cloud image sequence; the 3DCNN model can be expressed by the following formula:

[0098]

[0099] in, is the activation value of the cth feature map of the lth layer at position (x, y, n), is the input of the lth layer;

[0100] In S4, the present invention uses a wavelet transform algorithm to decompose the photovoltaic power sequence, remove noise, and obtain inherent characteristic components. The process can be expressed as:

[0101]

[0102] Among them, p denoised is the denoised photovoltaic power sequence, t is the time variable, represents the sum of all scales s and all frequency bands k, W k,s is the wavelet coefficient after threshold processing, k represents the frequency band of the wavelet, s represents the scale parameter, ψ k,s is the wavelet function corresponding to scale s and frequency band k; the selection process of wavelet coefficients is:

[0103]

[0104] Among them, W original is the coefficient obtained by the original wavelet transform; θ is the threshold used to determine which coefficients are considered noise;

[0105] Based on the improved long short-term memory neural network to capture the temporal characteristics of the inherent feature components, the process can be described as follows:

[0106]

[0107] in, is the time series characteristic of the historical photovoltaic power series, f IM-LSTM For the improved long short-term memory neural network model, [p t-n ,p t-n+1 ,...,p t ] is the historical photovoltaic power series;

[0108] Use multi-layer perceptron to extract deep features of multi-source weather forecast data:

[0109]

[0110] in, is the deep meteorological characteristics of multi-source meteorological forecast data, [w t+1 ,w t+2 ,...,w t+m ] is the weather forecast data corresponding to each forecast time from t+1 to t+m, f MLP It is a multi-layer perceptron model;

[0111] In S5, the present invention uses the spatiotemporal dynamic features of the historical cloud map, the spatial features of the predicted cloud map, and the historical power time series features as the input of the multi-layer perceptron to obtain the photovoltaic power prediction result O. The process can be expressed as follows:

[0112]

[0113] O=[o t+1 ,o t+2 ,...,o t+m ]

[0114] In S6, the present invention optimizes and corrects the prediction model in real time based on the real-time newly added data using an incremental update method, thereby ensuring that the prediction model can be continuously improved over time and better adapted to new data and environmental changes.

[0115] In order to verify the performance of the method proposed in the present invention, the following experiments were conducted in the embodiment. The data set used is from. And the model performance is evaluated by quantitative error indicators. The satellite cloud images and photovoltaic power data used in this study are from a photovoltaic power station in China (2017-2018). 80% of the data set is used for training and 20% for testing. Three error analysis indicators are used in this embodiment: mean deviation error (MAE), root mean square error (RMSE), and determination coefficient (R 2 ), defined as follows:

[0116]

[0117]

[0118] Where N represents the total number of test samples, y i and Represent the predicted value and actual value of the i-th sample, y i Represents the mean of the actual values.

[0119] In order to intuitively reflect the prediction effect, Figure 5 The photovoltaic power generation prediction curve obtained by the method of the present invention 1 hour in advance is shown. It can be seen that the method of the present invention can achieve accurate prediction of photovoltaic power. In order to further evaluate the prediction performance of the prediction model, a total of four comparison models are set, which can be represented as S1, S2, S3, and S4. S1 uses Transformer to extract time series features, S2 uses ConvLSTM to extract the spatiotemporal features of cloud images, and S3 uses SA-ConvLSTM to extract the feature information of satellite cloud images. In addition, Table 1 shows the error indicators of multi-step prediction of each model. It can be seen that compared with the comparison model, the prediction result of the model constructed based on the present invention has smaller error and better performance.

[0120] Table 1 Comparison of prediction results of different models

[0121]

[0122] Photovoltaic power prediction system based on deep learning and multi-source data fusion, including:

[0123] The data acquisition module collects photovoltaic power output, multi-source meteorological data, and multi-wavelength channel satellite cloud images of the target area to form a data sample library for the prediction model;

[0124] The data preprocessing module cleans the photovoltaic power output and multi-source meteorological data, and uses the pyramid algorithm to fuse multi-wavelength channel satellite cloud images;

[0125] The feature extraction module uses an improved self-attention convolutional long short-term memory neural network to predict satellite cloud images at future forecast moments based on historical multi-step satellite cloud images; the spatial distribution characteristics of the forecast cloud images and the spatiotemporal dynamic characteristics of the historical cloud images are respectively obtained through 2DCNN and 3DCNN models; the photovoltaic power sequence is decomposed using the wavelet transform algorithm to obtain the inherent characteristic components; the time series characteristics of the inherent characteristic components are captured based on the improved long short-term memory neural network; and the deep features of multi-source meteorological forecast data are extracted using a multi-layer perceptron;

[0126] The power prediction module uses the spatiotemporal features of the historical cloud map, the spatial features of the predicted cloud map, the time series features of the historical power, and the deep meteorological features as the input of the multi-layer perceptron to obtain the multi-step prediction results;

[0127] The incremental update module optimizes and corrects the prediction model in real time based on the real-time new data using the incremental update method;

[0128] It should be understood that the photovoltaic power prediction system based on deep learning and multi-source data fusion in the embodiment of the present invention can implement all the technical solutions in the above-mentioned method embodiment, and the functions of its various functional modules can be specifically implemented according to the methods in the above-mentioned method embodiments. The specific implementation process can refer to the relevant description in the above-mentioned embodiments, which will not be repeated here.

[0129] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the photovoltaic power prediction method based on deep learning and multi-source data fusion as described above are implemented.

[0130] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the photovoltaic power prediction method based on deep learning and multi-source data fusion drive as described above are implemented.

[0131] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices (systems), computer equipment or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0132] The present invention is described with reference to a flowchart of a method according to an embodiment of the present invention. It should be understood that each process in the flowchart and a combination of processes in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A device that specifies functions in a process or multiple processes.

[0133] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple processes.

[0134] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 The steps of a specified function in a process or multiple processes.

[0135] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A photovoltaic power prediction method based on deep learning and multi-source data fusion, characterized in that: The following steps are involved: S1. Collect the photovoltaic power output, multi-source meteorological data, and multi-wavelength channel satellite cloud images of the target area to form a data sample library for the prediction model. The collected data sequence is: I=[i t-n ,i t-n+1 ,...,i t ] P=[p t-n ,p t-n+1 ,...,p t ] In=[in t+1 ,In t+2 ,...,In t+m ]; S2. Clean the photovoltaic power output and multi-source meteorological data, and use the pyramid algorithm to fuse the multi-wavelength channel satellite cloud images. Use the multi-level pyramid algorithm to fuse the satellite cloud images of visible light, near infrared, infrared light, water vapor and other spectral bands to obtain a fused satellite cloud image that can fully represent the cloud information. S3. Based on the historical multi-step satellite cloud images, the improved self-attention convolutional long short-term memory neural network is used to predict the satellite cloud images at the future prediction time. The spatial distribution characteristics of the predicted cloud images and the spatiotemporal dynamic characteristics of the historical cloud images are respectively proposed through the two-dimensional convolutional neural network and the three-dimensional convolutional neural network model. The size of the historical satellite cloud images is N, H, W, and C; N is the number of historical cloud image frames, H and W are the height, width and number of channels of the cloud image respectively, and C is the length of the historical satellite cloud image sequence used. Among them, the continuous historical satellite cloud images are taken as input, and the improved self-attention convolutional long short-term memory neural network is used for image prediction, and then the satellite cloud images at the prediction time are obtained. The definition of the improved self-attention convolutional long short-term memory neural network is: S4. Use wavelet transform algorithm to decompose the photovoltaic power sequence, remove noise, and obtain inherent characteristic components. The process can be expressed as: Based on the improved long short-term memory neural network to capture the temporal characteristics of the inherent feature components, the process can be described as follows: Use multi-layer perceptron to extract deep features of multi-source weather forecast data: S5. Take the spatiotemporal characteristics of the historical cloud map, the spatial characteristics of the predicted cloud map, the time series characteristics of the historical power, and the deep meteorological characteristics as the input of the multi-layer perceptron to obtain the photovoltaic power prediction result O. The process is described as follows: O=[o t+1 ,o t+2 ,...,o t+m ]; Among them, O represents the multi-step prediction result, [o t+1 ,o t+2 ,...,o t+m ] is the photovoltaic power prediction result at each moment from t+1 to t+m; S6. Based on real-time new data, the prediction model is optimized and corrected in real time using the incremental update method.

2. The photovoltaic power prediction method based on deep learning and multi-source data fusion drive according to claim 1 is characterized in that: In the formula S1, I, P, and W are satellite cloud images, photovoltaic power, and weather forecast data, respectively. t is the satellite cloud image set of multi-wavelength channels at time t, n is the length of the historical sequence, and m is the length of the predicted sequence.

3. The photovoltaic power prediction method based on deep learning and multi-source data fusion drive according to claim 1 is characterized in that: The photovoltaic power and multi-source weather forecast in S2 are cleaned, including outlier screening and data interpolation. The outliers in the original data are identified and deleted using the box plot method. Missing values ​​are queried along the time series direction and filled by the K nearest neighbor algorithm.

4. The photovoltaic power prediction method based on deep learning and multi-source data fusion drive according to claim 1 is characterized in that: In the S3 formula: where x t ,h t-1 is the input value at time t and the output value at time t-1; For x t Updated features obtained through the self-attention module; h t-1 The updated value obtained by the dynamic residual module and the self-attention module in turn; c t , c t-1 represents the unit state at time t and time t-1; BDR is the dynamic residual module; SA is the self-attention memory module; i t , f t , g t They are input gate, forget gate and output gate respectively; W xi , W hi , b i , W xf , W hf , b f , W xc , W hc , b c , W xg , W hg , b g is the learning parameter; tanh is the hyperbolic tangent function, σ is the activation function, represents the Hadamard product; h t is the cell state at time t.

5. The photovoltaic power prediction method based on deep learning and multi-source data fusion drive according to claim 1 is characterized in that: The satellite cloud image features at the prediction time are extracted through a two-dimensional convolutional neural network model to obtain the spatial distribution characteristics of the predicted cloud image. The process can be described as follows: in, To predict the spatial distribution characteristics of the cloud image, f 2DCNN is a two-dimensional convolutional neural network model, [i t+1 ,i t+2 ,...,i t+m ] is the predicted satellite cloud image sequence.

6. The photovoltaic power prediction method based on deep learning and multi-source data fusion drive according to claim 1 is characterized in that: The feature extraction of historical satellite cloud image sequences is carried out through a three-dimensional convolutional neural network model to obtain the spatiotemporal dynamic characteristics of the cloud layer. The process can be described as follows: in, is the spatiotemporal dynamic characteristics of the historical cloud sequence, f 3DCNN is a three-dimensional convolutional neural network model, [i t-n ,i t-n+1 ,...,i t ] is a historical satellite cloud image sequence.

7. The photovoltaic power prediction method based on deep learning and multi-source data fusion drive according to claim 1 is characterized in that: The S4 p denoised is the denoised photovoltaic power sequence, t is the time variable, represents the sum of all scales s and all frequency bands k, W k,s is the wavelet coefficient after threshold processing, k represents the frequency band of the wavelet, s represents the scale parameter, ψ k,s is the wavelet function corresponding to scale s and frequency band k; is the time series characteristic of the historical photovoltaic power series, f IM-LSTM For the improved long short-term memory neural network model, [p t-n ,p t-n+1 ,...,p t ] is the historical photovoltaic power series; is the deep meteorological characteristics of multi-source meteorological forecast data, [w t+1 ,w t+2 ,...,w t+m ] is the weather forecast data corresponding to each forecast time from t+1 to t+m, f MLP It is a multi-layer perceptron model.

8. A photovoltaic power prediction system based on deep learning and multi-source data fusion drive, characterized in that: include: The data acquisition module collects photovoltaic power output, multi-source meteorological data, and multi-wavelength channel satellite cloud images of the target area to form a data sample library for the prediction model; The data preprocessing module cleans the photovoltaic power output and multi-source meteorological data, and uses the pyramid algorithm to fuse multi-wavelength channel satellite cloud images; The feature extraction module uses an improved self-attention convolutional long short-term memory neural network to predict satellite cloud images at future forecast times based on historical multi-step satellite cloud images. The spatial distribution characteristics of the forecast cloud images and the spatiotemporal dynamic characteristics of the historical cloud images are respectively obtained through two-dimensional convolutional neural network and three-dimensional convolutional neural network models. The photovoltaic power sequence is decomposed using the wavelet transform algorithm to obtain the inherent characteristic components. The time series characteristics of the inherent characteristic components are captured based on the improved long short-term memory neural network. The deep features of multi-source meteorological forecast data are extracted using a multi-layer perceptron. The power prediction module uses the spatiotemporal features of the historical cloud map, the spatial features of the predicted cloud map, the time series features of the historical power, and the deep meteorological features as the input of the multi-layer perceptron to obtain the multi-step prediction results; The incremental update module optimizes and corrects the prediction model in real time based on real-time new data using the incremental update method.

9. A computer device, characterized in that: include: one or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, and when the programs are executed by the processors, the steps of the photovoltaic power prediction method based on deep learning and multi-source data fusion driving as described in any one of claims 1-7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the photovoltaic power prediction method based on deep learning and multi-source data fusion driving as described in any one of claims 1 to 7 are implemented.

Images