CN116152206A - Photovoltaic output power prediction method, terminal equipment and storage medium - Google Patents
Photovoltaic output power prediction method, terminal equipment and storage medium Download PDFInfo
- Publication number
- CN116152206A CN116152206A CN202310166515.8A CN202310166515A CN116152206A CN 116152206 A CN116152206 A CN 116152206A CN 202310166515 A CN202310166515 A CN 202310166515A CN 116152206 A CN116152206 A CN 116152206A
- Authority
- CN
- China
- Prior art keywords
- image
- output power
- images
- photovoltaic output
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 239000013598 vector Substances 0.000 claims abstract description 53
- 238000003066 decision tree Methods 0.000 claims abstract description 14
- 230000007246 mechanism Effects 0.000 claims abstract description 14
- 230000003287 optical effect Effects 0.000 claims description 43
- 238000004590 computer program Methods 0.000 claims description 19
- 230000015654 memory Effects 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 10
- 239000000443 aerosol Substances 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 238000013136 deep learning model Methods 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 8
- 238000010248 power generation Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000001556 precipitation Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000002834 transmittance Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 230000003595 spectral effect Effects 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001704 evaporation Methods 0.000 description 1
- 230000008020 evaporation Effects 0.000 description 1
- 238000005206 flow analysis Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/10—Image enhancement or restoration by non-spatial domain filtering
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention relates to a photovoltaic output power prediction method, terminal equipment and a storage medium, wherein the method comprises the following steps: collecting satellite images and extracting cloud motion characteristics and cloud position characteristics; extracting global information and local information of visible and invisible wave band images, and extracting global feature vectors and local feature vectors; weighting and summing the four feature vectors based on an attention mechanism to obtain enhanced image features; based on historical meteorological data, predicting rainfall probability at each moment through a decision tree model to obtain a predicted value; and carrying out weighted summation on the cloud motion characteristics, the cloud position characteristics, the image characteristics and the rainfall probability predicted value of the current moment image and the photovoltaic output power value of the current moment based on the attention mechanism, and obtaining the photovoltaic output power value predicted value of the next moment through an LSTM network. The invention further improves the accuracy of the deep learning model in the field of photovoltaic prediction.
Description
Technical Field
The present invention relates to the field of photovoltaic power generation, and in particular, to a photovoltaic output power prediction method, a terminal device, and a storage medium.
Background
Currently, the main energy sources on the earth are still fossil energy sources, but fossil energy sources are exhausted all the time, so that new energy sources including photovoltaic energy sources are very actively explored in various countries, and are an important direction of new energy source research. Photovoltaic energy is available anywhere in the world and is not polluting during operation, being a completely green energy source. Therefore, how to increase the utilization rate of the photovoltaic energy is a very important issue to be studied, and at present, since the photovoltaic energy is greatly influenced by factors such as weather, installation angle, position and the like, and seasonal factors, it is difficult to comprehensively consider all the influencing factors in the current photovoltaic prediction.
Disclosure of Invention
In order to solve the problems, the invention provides a photovoltaic output power prediction method, terminal equipment and a storage medium.
The specific scheme is as follows:
a photovoltaic output power prediction method comprising the steps of:
s1: acquiring satellite images, obtaining cloud motion characteristics corresponding to the satellite images at all times through an optical flow method, and obtaining cloud position characteristics corresponding to the satellite images at all times through an image characteristic extraction model;
s2: selecting visible wave band images and invisible wave band images from satellite images, and respectively extracting global information and local information of the two wave band images to obtain global images and local images corresponding to the two wave band images;
s3: the global image and the local image corresponding to the two types of wave band images are respectively processed through a convolutional neural network to obtain global feature vectors and local feature vectors corresponding to the two types of wave band images;
s4: based on the global feature vector and the local feature vector corresponding to the two types of wave band images, the image features of the satellite images are calculated through the following formula
α pi ＝MLP(Wh p ,WI i )
α p j＝MLP(Wh p ,WI j )
wherein ,
representing a concatenation of 4 feature vectors, σ (·) representing the activation function, β pi Weight coefficient representing the ith feature vector, W p Representing a second parameter matrix, I k Represents the kth eigenvector, I i Representing the ith feature vector, I j The j-th feature vector is represented, i, k and j each represent the sequence number of the feature vector, i E [1,4 ]]，j∈[1,4]And j is not equal to i, k is E [1,4 ]]，α pi Representing the fusion feature of the ith feature vector, alpha pj The fusion feature of the j-th feature vector is represented, MLP (·) represents the neural network layer, W represents the first parameter matrix, h p Representing a photovoltaic output power characteristic vector;
s5: based on historical meteorological data, predicting rainfall probability at each moment through a decision tree model to obtain a predicted value;
s6: and carrying out weighted summation on the cloud motion characteristics, the cloud position characteristics, the image characteristics and the rainfall probability predicted value of the current moment image and the photovoltaic output power value of the current moment based on the attention mechanism, and obtaining the photovoltaic output power value predicted value of the next moment through an LSTM network.
Furthermore, the image feature extraction model adopts a VGG network model.
Further, the satellite image acquisition process comprises the following steps: on the basis of satellite remote sensing data, the aerosol optical thickness data is synthesized and then converted into satellite images.
Further, global information is obtained through positive fourier transform and inverse fourier transform, and local information is obtained through high-pass filtering.
Further, in step S4, the historical weather data includes a plurality of dimensional features, and by calculating the information gain of each feature, the feature with the largest information gain is selected as the root node of the decision tree, and the number of branches of the tree and the number of values of the feature to which the root node belongs.
The photovoltaic output power prediction terminal equipment comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein the steps of the method according to the embodiment of the invention are realized when the processor executes the computer program.
A computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the method described above for embodiments of the present invention.
The invention adopts the technical scheme and has the beneficial effects that:
(1) The influence of position perception in satellite images is innovatively explored, the thought of dynamic and static combination is adopted, optical flow analysis is used for learning motion characteristics at different moments, and rich image characteristics of satellite images under static conditions are fully considered.
(2) An attention mechanism is introduced to learn global and local information of different band satellite clouds.
(3) Better feature representation is achieved through the ideas of image enhancement and denoising.
(4) The influence of the meteorological factors is also taken into account to integrate the meteorological factors into the photovoltaic power prediction task to better assist the time series prediction task.
Drawings
Fig. 1 is a schematic diagram of a satellite image sequence optical flow calculation process according to an embodiment of the invention.
Fig. 2 shows an image in the visible light band and an image in the invisible light band in the first embodiment of the present invention.
FIG. 3 is a schematic diagram of a decision tree model according to an embodiment of the present invention.
Fig. 4 is a flowchart of a method according to a first embodiment of the invention.
Fig. 5 is a schematic diagram of a model overall frame according to an embodiment of the present invention.
Detailed Description
For further illustration of the various embodiments, the invention is provided with the accompanying drawings. The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate embodiments and together with the description, serve to explain the principles of the embodiments. With reference to these matters, one of ordinary skill in the art will understand other possible embodiments and advantages of the present invention.
The invention will now be further described with reference to the drawings and detailed description.
Embodiment one:
the embodiment provides a time sequence prediction model integrating time, space and meteorological conditions, which is named STMLSTM. Similar to a long-term and short-term memory network structure, the position, thickness and movement of the cloud are considered in the continuous updating process, various weather indexes in history and image data of different wave bands are considered, the importance of different factors is balanced by introducing a self-focusing mechanism, and finally future photovoltaic output is predicted. The modules of the model are described below.
(1) Cloud motion feature extraction module
The trend of movement of the cloud is very important for photovoltaic prediction, and the occlusion of the cloud can lead to a decrease in photovoltaic efficiency. By extracting the motion characteristics of the cloud, the motion trend of the cloud can be predicted, so that the prediction of the photovoltaic change trend is indirectly assisted.
The cloud motion feature extraction module is used for mining cloud motion information on the satellite image sequence and providing a prediction basis for subsequent photovoltaic prediction. Because the optical flow method can find the offset of pixel points among different frames of the image sequence, and can calculate the correlation among adjacent frames, obtain the motion information of the object in the adjacent frames. Therefore, in this embodiment, the optical flow method is used to extract the motion characteristics of the satellite images, and the satellite images are arranged in time order to form a satellite image sequence. The optical flow method utilizes the change of pixels of the satellite image sequence in the time domain and the correlation between adjacent frames to find the corresponding relation between the previous frame and the current frame, and calculates the motion information of the cloud between the adjacent frames.
The optical flow method detects the motion of the cloud, namely, analyzes a vector field formed by the motion vectors of the pixels stored in the satellite images. When the cloud in the satellite image does not move, the optical flow field in the image should be uniform; when the cloud starts to move, the optical flow vector formed by the moving cloud is inconsistent with the optical flow vector in the satellite image background, and the optical flow field is not continuous and uniform, so that the movement and the position of the cloud can be detected. Thus, the result of the optical flow calculation contains information of the structure of the cloud and the movement of the cloud itself. The advantage of the optical flow method in this case is that it does not require a priori knowledge nor the features of the artificially constructed image.
The core of the cloud motion feature extraction module in the embodiment is a Horn-Schunck optical flow method, which is a global method, and the dense optical flow field of the satellite image is estimated by calculating the optical flow of each pixel in the image. The algorithm converts the problem of solving the optical flow value of the optical flow field into the problem of solving the minimum value of the image energy function by constructing an energy function for the image. Given a sequence of satellite images P (x, y, t), the optical flow field F (x, y) between the image sequences can be converted into optical flow components m (x, y) and n (x, y) that are resolved in the x, y directions, respectively. For satellite images we assume the following energy functions
wherein ,Px 、P y 、P t Representing the derivatives of the image sequence in the x, y, t dimensions, respectively, (P) x m+P y n+P t ) 2 The gray scale variation factor of the image is represented,
a smoothing limiting factor that limits the rate of change of the optical flow components m, n is represented.
The cloud motion feature extraction module can perform optical flow calculation on an input 8-bit depth satellite image sequence, output a 24-bit depth optical flow image sequence, and convert the 24-bit depth optical flow image sequence into the 8-bit depth optical flow sequence, so that subsequent feature extraction is facilitated. The module selects satellite images in a time series with a window size of 2 using a sliding window. This is to calculate the optical flow between two successive moments. The goal of the calculation is to minimize the energy function E, which is a generalized polar problem, solved by the euler-lagrangian equation. After successive iterations, when the error is less than a predetermined threshold, a final optical flow magnitude is output.
An example of a satellite image sequence optical flow calculation process is shown in fig. 1. T1 and T2 are two satellite images at the time of T1 and T2 respectively, and T1 and T2 are continuous time periods. The following optical flow image is the result of the optical flow calculation between the two images T1, T2, we calculate the optical flow between every two moments of the satellite image sequence. (2) Cloud location (static) feature extraction module
For ground photovoltaic solar power plants, the solar transmittance on the ground photovoltaic solar power plant is an influencing factor for influencing the photovoltaic power generation efficiency. Therefore, researching the specific position of the cloud and the thickness of the cloud in the geographical area of the power station is beneficial to accurately predicting the photovoltaic. To study the cloud location and thickness information, an aerosol optical thickness image was synthesized using satellite telemetry data in this example.
The earth's atmosphere may be considered an aerosol. The optical thickness of an aerosol is one of the important optical properties of an aerosol, and represents the light transmittance per unit section of a vertical atmosphere column in the atmosphere, which is also widely referred to as the atmospheric transmittance. The larger the aerosol optical thickness, the lower the atmospheric transmittance and the lower the ground light intensity. In this embodiment, based on satellite remote sensing data, aerosol optical thickness data is integrated and converted into satellite image data.
In order to obtain more image features while controlling the number of parameters and improve the response rate of the photovoltaic prediction system, the VGG network model is selected in this embodiment to extract the position features from the satellite images. The VGG network has deeper hierarchy than the common CNN, but the computation complexity and the complexity of the model structure are reduced due to the application of the small convolution kernel and the pooling layer. In the VGG network used in this embodiment, the input satellite image sequence is a 2401×2401RGB image sequence with a fixed size, and it is necessary to convert the satellite image with 32 bit depth into an 8 bit depth image in advance. During training, the size of the modulo convolution kernel is 3, and the moving step length is 1 pixel. The number of channels in the convolutional layer increases from 64 to 512, passing through each two-dimensional pooling layer. After a series of convolution layers, the data is compressed into one-dimensional data. And then adding a full communication layer to obtain the cloud position feature vector of the satellite image. The nonlinear activation functions of all hidden layers are set to ReLU throughout the convolutional network to make the network train faster.
(3) Attention-based spectrum learning module
The satellite image includes both an invisible band and a visible band. The invisible band in this example selects an infrared image greater than 760nm and having a wavelength of 0.86 μm. The visible light band selects an image with a wavelength of 0.64 μm. As shown in fig. 2, the left side is an image in the visible light band, and the right side is an image in the invisible light band.
In this embodiment, a fourier transform method is used to perform spectral processing on images in two types of bands. The method decomposes the image into sine and cosine components. It converts the image from the spatial domain to the frequency domain. The formula for fourier transforming the image is shown in formula 2.
F(u,v)＝∫∫f(x,y)e -i(ux+vy) dxdy (2)
Where x, y represents the pixel coordinates of the spatial domain, u, v represents the pixel coordinates of the frequency domain.
The pixels of the image are discrete, finite. Thus, the actual fourier transform of the image is performed using a discrete fourier transform. Because the integral in equation 2 becomes a summation. Through the Fourier forward transformation and the Fourier inverse transformation, the enhanced image information of different wave bands can be obtained, namely the global information.
The importance of the edge information should also be considered for images of both types of bands. Thus, high pass filtering is used in this embodiment to enhance the sharp details in the image, resulting in local information of the image. The high pass filter attenuates low frequency signals while allowing high frequency signals to pass. The inverse fourier transform moves the zero frequency component to the center region of the image. Thus, the diffusion frequency from the central region outwards is from low to high. In this method we mask the low frequency signal in the central surrounding part, leaving the high frequency signal.
Global information and local information can better help us learn the features of an image. And inputting images corresponding to the global information and the local information into a convolutional neural network (such as VGG 16) for feature learning to obtain four feature vectors, namely a feature vector visg of the global visible image, a feature vector visl of the local visible image, a feature vector invisg of the global invisible image and a feature vector invisl of the local invisible image.
Four feature vectors are derived for each time satellite image. Here, a focus mechanism is introduced in this embodiment to focus on the importance level between different feature vectors. The importance of different band information, as well as the importance of global and local information, can be better revealed by learning the attention mechanism.
For ease of notation, we here use the set i= { I1, I2, I3, I4} to represent the four eigenvectors that were spectrally processed at time t, where I1 to I4 correspond to visg, visl, invsg and invsl above, respectively. Simultaneously, photovoltaic power data per hour are processed to obtain a photovoltaic power characteristic vector h p The size is 1×n, where N is the number of photovoltaic power acquisitions per hour. We employ an attention mechanism for each image feature vector. The first step is to learn the photovoltaic power eigenvector h using a multilayer perceptron p And a single feature vector I i Nonlinear relationship between the two. As shown in formula 3.
α pi ＝MLP（Wh p ,WI i )
α pj ＝MLP(Wh p ,WI j ) (3)
wherein ,αpi Representing the fusion feature of the ith feature vector, alpha pj Representing the fusion feature of the jth feature vector, MLP (·) representing the neural network layer, W representing the first parameter matrix, and I determined by learning i Representing the ith feature vector, I j Represents the j-th feature vector, i.e. [1,4 ]]，j∈[1,4]And j+.i.
Thus, for the photovoltaic power eigenvector h p We can calculate the weight coefficient beta of images in different wave bands pi As shown in formula 4.
Wherein σ (·) represents the activation function, β pi Weight coefficient representing the ith feature vector, W p Representing a second parameter matrix, I k Represents the kth eigenvector, h t Representing the final representation of the image features of the satellite image obtained by the spectral method based on the attention mechanism. The photovoltaic power prediction task can be well assisted by learning global and local information of images in different wave bands.
(4) Meteorological feature learning
Weather data of a photovoltaic power plant is used to predict whether or not it will rain in the future, and rainfall has a significant effect on photovoltaic power generation. However, historical weather data contains many irrelevant parameters, such as wind speed, wind direction, humidity, temperature, etc. In order to find precipitation patterns from a number of uncorrelated weather data, a decision tree model is used in this embodiment to extract precipitation patterns from each of the historical weather parameters. The extracted precipitation pattern is used to predict the amount of precipitation for a period of time in the future. The decision tree model performs well when certain attributes of the dataset are missing. They can handle irrelevant features well and can train quickly without special hardware acceleration, which is critical when handling large data sets. The decision tree model is the best choice for the scenario. This is because weather history data often has many missing and previously unrelated features and photovoltaic predictions have a need for quick response.
The embodiment provides a rainfall probability prediction module, which can collect various meteorological data at historical time and predict the rainfall probability of the next day. The core of the prediction algorithm is an ID3 algorithm based on a decision tree. The rainfall prediction can be converted into a problem of classifying the probability of future rainfall, and the rainfall prediction system is assumed to have a sample space of (F, R), wherein F is the number of samples, and m features are included. In this module the number of features is 6, temperature, evaporation, wind direction, wind speed, humidity and air pressure, respectively. R is the number of categories to be classified by the classification system, and n categories are included, and the probability of raining in the daytime is represented in the module. For a rainfall probability classification system, possible choices in the system are D1, D2, …, dn, and the probability of each class occurrence is P (D1), P (D2), …, P (Dn). The entropy of the rainfall probability classification system is shown in formula 6.
Furthermore, in order to calculate the impact of a certain feature of the classification system on the system, the concept of conditional entropy is introduced. The conditional entropy in the classification system is defined as the information entropy of the sample when the one-dimensional feature F is fixed and represented. As H (D|F) is characteristic F= (F) 1 ,f 2 ，…，f n ) And when the condition entropy of the rainfall probability classification system is fixed, the condition entropy is shown as a formula 7.
Wherein the sample features F in the classification system have probabilities P i Taking f i When the characteristic F takes a fixed value F i At the time H (d|f=f i ) Is the conditional information entropy.
The entropy change of a rainfall probability classification prediction system is defined as the information gain specific to a certain feature. That is, the information gain is the difference between the information entropy of a system with features and the information entropy of a system without features. For example, the information gain of the feature F in the rainfall probability classification system is gain (F, R) =h (D) -H (d|f). The calculation process is to continuously calculate the information gain of each feature, and then select the feature with the largest information gain, which is the basic principle of the rainfall probability classification system.
In the decision tree calculation based on the information entropy, firstly, the information gain of each feature is calculated, and the feature with the largest information gain is selected as the root node of the decision tree. The number of branches of the tree is the number of values of the characteristic to which the root node belongs. In the classification process, each branch generates a new subset of data that iterates through until all subsets of data are classified into the same class. In this module, a plurality of weather factors are used as the characteristics of the classification system, a decision tree model with depth of 3 is used, and rainfall is predicted by using historical weather data, and the tree model is shown in fig. 3.
(5) Time series prediction module
In dealing with problems related to the event time axis, such as photovoltaic power prediction, there is a certain correlation between the time series data, which cannot be solved by the conventional neural network. A Recurrent Neural Network (RNN) solves the problems associated with time spans by forming a connection matrix between neurons, recursively generating updated prediction data continually. However, when the interval between the relevant information and the current predicted position becomes very large, the RNN loses the ability to learn information connected over such a distance, the root cause of which is a problem of gradient disappearance or gradient divergence. The photovoltaic output prediction time range in the embodiment is wider, the task cannot be completed by RNN, and a core iteration module of Long Short-term Memory (LSTM) is adopted in the prediction module, so that the problem of insufficient Long-term prediction precision in the photovoltaic is solved.
Similar to RNN, LSTM has a repeated recursive structure, but since LSTM innovatively proposes a gate structure, sequence information is selectively extracted, reducing information redundancy and gradient vanishing. Thus, LSTM has the ability to solve the medium-long time series problem. The key to the gate structure is the dot product kernel sigmoid activates the neural layer. In the LSTM core module, there are three gate structures for filtering and controlling information, namely an input gate, a forget gate and an output gate.
Based on the above model, a photovoltaic output power prediction method is proposed herein, as shown in fig. 4, and includes the following steps:
s1: acquiring satellite images, obtaining cloud motion characteristics corresponding to the satellite images at all times through an optical flow method, and obtaining cloud position characteristics corresponding to the satellite images at all times through an image characteristic extraction model;
s2: selecting visible wave band images and invisible wave band images from satellite images, and respectively extracting global information and local information of the two wave band images to obtain global images and local images corresponding to the two wave band images;
s3: the global image and the local image corresponding to the two types of wave band images are respectively processed through a convolutional neural network to obtain global feature vectors and local feature vectors corresponding to the two types of wave band images;
s4: calculating the enhanced image features of the satellite images by using an attention mechanism based on the global feature vectors and the local feature vectors corresponding to the two types of wave band images;
s5: based on historical meteorological data, predicting rainfall probability at each moment through a decision tree model to obtain a predicted value;
s6: and carrying out weighted summation on the cloud motion characteristics, the cloud position characteristics, the image characteristics and the rainfall probability predicted value of the current moment image and the photovoltaic output power value of the current moment based on the attention mechanism, and obtaining the photovoltaic output power value predicted value of the next moment through an LSTM network.
The method can be used for simultaneously processing the historical photovoltaic output power data, the satellite image data and the historical meteorological data, and improves the accuracy of photovoltaic prediction.
The present embodiment improves upon conventional LSTM models, adding time, space and weather modules that will update with the time series of the LSTM model. In addition, since the time granularity of the various data is not uniform, a uniform time granularity of 15 minutes was selected in this embodiment. For data with time granularity greater than 15 minutes, the time granularity of various data is unified to 15 minutes by using a filling method, and the filling method uses the average value of two values with the latest time to fill.
Referring to the overall model block diagram shown in fig. 5, in step S1, satellite images at time t-1 and time t are processed to obtain an optical flow image at time t, then the optical flow image is extracted by an independent image feature extraction model to obtain 512-dimensional features of the cloud, and then the features of the dimensions are accumulated to obtain a cloud motion feature value Tt at time t. Similarly, another independent image feature extraction model processes the satellite image at the moment t into 512-dimensional sparse features, and adds the 512-dimensional sparse features to obtain a position feature value St of the cloud at the moment t. In steps S2-S4, spectral learning is performed on the satellite image At time t, global information and local information of images with two types of wavelengths At time t are obtained, and weight learning is performed by combining an attention mechanism and a photovoltaic power feature vector, so that image features At (At is h) At time t are obtained t The accumulated results of the dimension features). In step S5, all weather data from time 0 to time t are input, and a predicted value Mt of whether or not the weather is rainy at time t is output. In step S6, the attention mechanism is applied to Tt, st, at, mt to pay attention to the photovoltaic output power vector xt at the time t, and the predicted photovoltaic output power value at the time t is obtained by summing the weights and inputting the sum to the LSTM module. It is noted that when t is smaller than the current time t0, we input historical photovoltaic power generation amount data, and when t is larger than t0, input is the predicted value of the photovoltaic power generation amount at the last moment. And taking the implicit output ht of each layer of the model as the implicit input of the next layer of the model, and iterating in this way to finally obtain the photovoltaic output power prediction sequence of the next period.
The embodiment of the invention further improves the accuracy of the deep learning model in the field of photovoltaic prediction, and overcomes the defect of lack of response to uncertainty disturbance in the traditional photovoltaic prediction.
Embodiment two:
the invention also provides a photovoltaic output power prediction terminal device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the steps in the method embodiment of the first embodiment of the invention are realized when the processor executes the computer program.
Further, as an executable scheme, the photovoltaic output power prediction terminal device may be a computing device such as a desktop computer, a notebook computer, a palm computer, and a cloud server. The photovoltaic output power prediction terminal device may include, but is not limited to, a processor, a memory. It will be appreciated by those skilled in the art that the above-described composition structure of the photovoltaic output power prediction terminal device is merely an example of the photovoltaic output power prediction terminal device, and does not constitute limitation of the photovoltaic output power prediction terminal device, and may include more or fewer components than the above, or combine some components, or different components, for example, the photovoltaic output power prediction terminal device may further include an input/output device, a network access device, a bus, and the like, which is not limited by the embodiment of the present invention.
Further, as an implementation, the processor may be a central processing unit (Central Processing Unit, CPU), other general purpose processor, digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like, which is a control center of the photovoltaic output power prediction terminal equipment, and various interfaces and lines are used to connect various parts of the entire photovoltaic output power prediction terminal equipment.
The memory may be used to store the computer program and/or module, and the processor may implement various functions of the photovoltaic output power prediction terminal device by running or executing the computer program and/or module stored in the memory, and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, at least one application program required for a function; the storage data area may store data created according to the use of the cellular phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart Media Card (SMC), secure Digital (SD) Card, flash Card (Flash Card), at least one disk storage device, flash memory device, or other volatile solid-state storage device.
The present invention also provides a computer readable storage medium storing a computer program which when executed by a processor implements the steps of the above-described method of an embodiment of the present invention.
The module/unit integrated with the photovoltaic output power predicting terminal device may be stored in a computer readable storage medium if implemented in the form of a software functional unit and sold or used as a separate product. Based on such understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of each of the method embodiments described above. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a software distribution medium, and so forth.
While the invention has been particularly shown and described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (7)
1. The photovoltaic output power prediction method is characterized by comprising the following steps of:
s1: acquiring satellite images, obtaining cloud motion characteristics corresponding to the satellite images at all times through an optical flow method, and obtaining cloud position characteristics corresponding to the satellite images at all times through an image characteristic extraction model;
s2: selecting visible wave band images and invisible wave band images from satellite images, and respectively extracting global information and local information of the two wave band images to obtain global images and local images corresponding to the two wave band images;
s3: the global image and the local image corresponding to the two types of wave band images are respectively processed through a convolutional neural network to obtain global feature vectors and local feature vectors corresponding to the two types of wave band images;
s4: based on the global feature vector and the local feature vector corresponding to the two types of wave band images, the image features of the satellite images are calculated through the following formula
α pi ＝MLP(Wh p ,WI i )
α pj ＝MLP(Wh p ,WI j )
wherein ,
representing a concatenation of 4 feature vectors, σ (·) representing the activation function, β pi Weight coefficient representing the ith feature vector, W p Representing a second parameter matrix, I k Represents the kth eigenvector, I i Representing the ith feature vector, I j The j-th feature vector is represented, i, k and j each represent the sequence number of the feature vector, i E [1,4 ]]，j∈[1,4]And j is not equal to i, k is E [1,4 ]]，α pi Representing the fusion feature of the ith feature vector, alpha pj The fusion feature of the j-th feature vector is represented, MLP (·) represents the neural network layer, W represents the first parameter matrix, h p Representing a photovoltaic output power characteristic vector;
s5: based on historical meteorological data, predicting rainfall probability at each moment through a decision tree model to obtain a predicted value;
s6: and carrying out weighted summation on the cloud motion characteristics, the cloud position characteristics, the image characteristics and the rainfall probability predicted value of the current moment image and the photovoltaic output power value of the current moment based on the attention mechanism, and obtaining the photovoltaic output power value predicted value of the next moment through an LSTM network.
2. The photovoltaic output power prediction method according to claim 1, characterized in that: the image feature extraction model adopts a VGG network model.
3. The photovoltaic output power prediction method according to claim 1, characterized in that: the satellite image acquisition process comprises the following steps: on the basis of satellite remote sensing data, the aerosol optical thickness data is synthesized and then converted into satellite images.
4. The photovoltaic output power prediction method according to claim 1, characterized in that: global information is obtained by fourier positive and inverse fourier transforms, and local information is obtained by high pass filtering.
5. The photovoltaic output power prediction method according to claim 1, characterized in that: in the step S4, the historical meteorological data includes features of multiple dimensions, and by calculating the information gain of each feature, the feature with the largest information gain is selected as the root node of the decision tree, and the number of branches of the tree and the number of values of the feature to which the root node belongs.
6. The photovoltaic output power prediction terminal device is characterized in that: comprising a processor, a memory and a computer program stored in the memory and running on the processor, which processor, when executing the computer program, carries out the steps of the method according to any one of claims 1 to 5.
7. A computer-readable storage medium storing a computer program, characterized in that: the computer program implementing the steps of the method according to any one of claims 1 to 5 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310166515.8A CN116152206A (en) | 2023-02-27 | 2023-02-27 | Photovoltaic output power prediction method, terminal equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310166515.8A CN116152206A (en) | 2023-02-27 | 2023-02-27 | Photovoltaic output power prediction method, terminal equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116152206A true CN116152206A (en) | 2023-05-23 |
Family
ID=86357994
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310166515.8A Pending CN116152206A (en) | 2023-02-27 | 2023-02-27 | Photovoltaic output power prediction method, terminal equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116152206A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116706903A (en) * | 2023-08-07 | 2023-09-05 | 深圳航天科创泛在电气有限公司 | Photovoltaic power generation amount prediction method, device, equipment and medium |
-
2023
- 2023-02-27 CN CN202310166515.8A patent/CN116152206A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116706903A (en) * | 2023-08-07 | 2023-09-05 | 深圳航天科创泛在电气有限公司 | Photovoltaic power generation amount prediction method, device, equipment and medium |
| CN116706903B (en) * | 2023-08-07 | 2024-01-26 | 深圳航天科创泛在电气有限公司 | Photovoltaic power generation amount prediction method, device, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Han et al. | A deep learning method for bias correction of ECMWF 24–240 h forecasts | |
| CN111738329B (en) | Land use classification method for time series remote sensing images | |
| US10996374B1 (en) | Short-term weather forecasting using artificial intelligence and hybrid data | |
| CN111310666B (en) | High-resolution image ground feature identification and segmentation method based on texture features | |
| CN109001736B (en) | Radar echo extrapolation method based on deep space-time prediction neural network | |
| CN112446419A (en) | Time-space neural network radar echo extrapolation forecasting method based on attention mechanism | |
| CN111161199A (en) | Spatial-spectral fusion hyperspectral image mixed pixel low-rank sparse decomposition method | |
| CN112017178A (en) | Remote sensing image region change detection method based on double-temporal difference image method | |
| JP2022534890A (en) | Image processing method and apparatus, electronic equipment and storage medium | |
| CN116152206A (en) | Photovoltaic output power prediction method, terminal equipment and storage medium | |
| CN113935249A (en) | Upper-layer ocean thermal structure inversion method based on compression and excitation network | |
| CN115393690A (en) | Light neural network air-to-ground observation multi-target identification method | |
| Zhao et al. | Mastgn: Multi-attention spatio-temporal graph networks for air pollution prediction | |
| Jin et al. | Deep learning for seasonal precipitation prediction over China | |
| CN117233869B (en) | Site short-term wind speed prediction method based on GRU-BiTCN | |
| Sari et al. | Daily rainfall prediction using one dimensional convolutional neural networks | |
| CN113989612A (en) | Remote sensing image target detection method based on attention and generation countermeasure network | |
| Kim et al. | Time series prediction of tropical storm trajectory using self-organizing incremental neural networks and error evaluation | |
| Zheng et al. | Hybrid scheme of kinematic analysis and lagrangian koopman operator analysis for short-term precipitation forecasting | |
| KR102218179B1 (en) | System and method for emulating radiative transfer parameterization of korea local analysis and prediction system | |
| Jing et al. | Time series land cover classification based on semi-supervised convolutional long short-term memory neural networks | |
| CN113378731B (en) | Green space water system vector extraction method based on convolutional neural network and energy constraint | |
| CN116776935B (en) | Improved MICN-based photovoltaic power prediction method | |
| Yao et al. | Multivariate deep learning for reconstruction of spatial missing climate data | |
| Lian et al. | A sequence-to-sequence based multi-scale deep learning model for satellite cloud image prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |