CN113723712A - Wind power prediction method, system, device and medium - Google Patents

Wind power prediction method, system, device and medium Download PDF

Info

Publication number
CN113723712A
CN113723712A CN202111201545.5A CN202111201545A CN113723712A CN 113723712 A CN113723712 A CN 113723712A CN 202111201545 A CN202111201545 A CN 202111201545A CN 113723712 A CN113723712 A CN 113723712A
Authority
CN
China
Prior art keywords
wind power
neural network
data
prediction
distribution
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.)
Granted
Application number
CN202111201545.5A
Other languages
Chinese (zh)
Other versions
CN113723712B (en
Inventor
汪运
李意芬
邹润民
张菲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Electric Power Research Institute Co Ltd CEPRI
Central South University
Original Assignee
China Electric Power Research Institute Co Ltd CEPRI
Central South University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China Electric Power Research Institute Co Ltd CEPRI, Central South University filed Critical China Electric Power Research Institute Co Ltd CEPRI
Priority to CN202111201545.5A priority Critical patent/CN113723712B/en
Publication of CN113723712A publication Critical patent/CN113723712A/en
Application granted granted Critical
Publication of CN113723712B publication Critical patent/CN113723712B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • 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 embodiment of the disclosure provides a wind power prediction method, a wind power prediction system, a wind power prediction device and a medium, which belong to the technical field of data processing and specifically comprise the following steps: acquiring historical wind power data of an area to be predicted; preprocessing historical wind power data; obtaining multi-time scale characteristics; establishing a t-distribution probability neural network according to the multi-time scale characteristics; obtaining a plurality of initial models; respectively inputting the verification set into each initial model, and taking the initial model with the optimal output result as a prediction model; and inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted. According to the scheme, the multi-time scale features are extracted by utilizing the one-dimensional convolutional neural network and the two-way long and short memory network, uncertainty of wind power prediction is described by utilizing student t distribution, the uncertainty of future wind power is accurately predicted by constructing the deep student t probability neural network, and the prediction efficiency and accuracy are improved.

Description

Wind power prediction method, system, device and medium
Technical Field
The embodiment of the disclosure relates to the technical field of data processing, in particular to a wind power prediction method, a wind power prediction system, a wind power prediction device and a wind power prediction medium.
Background
At present, wind energy plays a crucial role in global energy supply as a clean, environment-friendly and inexhaustible renewable energy source, however, when large-scale wind power is incorporated into a power grid, the inherent randomness and intermittency of the wind power bring huge challenges to power grid maintenance, power grid planning, power system scheduling and the like. In addition, accurate wind power prediction is also beneficial to efficient operation of an energy market, great improvement of power transmission capacity, reasonable planning of a wind power system and the like. Accurate wind power prediction is an effective method for solving the above problems. And multi-scale information contained in the wind power data is considered. Different information is often contained in data of different scales. High resolution data often contains much information reflecting detail fluctuation, while low resolution data often contains long-term fluctuation information. However, the existing prediction method usually realizes prediction by a complex and huge amount of calculation, or predicts future wind power by using only information in certain resolution data, and cannot describe the uncertainty of wind power fluctuation in detail.
Therefore, a wind power prediction method with high efficiency and precision is needed.
Disclosure of Invention
In view of this, embodiments of the present disclosure provide a wind power prediction method, system, device, and medium, which at least partially solve the problems in the prior art that the calculation amount is large, and the prediction efficiency and the prediction accuracy are poor.
In a first aspect, an embodiment of the present disclosure provides a wind power prediction method, including:
acquiring historical wind power data of an area to be predicted;
preprocessing the historical wind power data to obtain a training set, a verification set and a test set;
extracting feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-short memory network and fusing the feature data to obtain multi-time scale features;
establishing a t-distribution probability neural network according to the multi-time scale features;
adjusting the number of neurons in the t-distribution probability neural network to obtain a plurality of initial models;
respectively inputting the verification set into each initial model, and taking the initial model with the optimal output result as a prediction model;
and inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted.
According to a specific implementation manner of the embodiment of the present disclosure, the step of preprocessing the historical wind power data to obtain a training set, a verification set and a test set includes:
dividing the historical wind power data into three disjoint subsets;
and carrying out normalization processing on the three subsets, and using the three subsets as the training set, the verification set and the test set.
According to a specific implementation manner of the embodiment of the present disclosure, the step of extracting and fusing feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-short memory network to obtain a multi-time scale feature includes:
converting the data in the training set to different time scales by using an average pooling method;
extracting spatial features from the input of different time scales by using the one-dimensional convolutional neural network;
extracting time characteristics from the space characteristics by using the bidirectional long and short memory network;
automatically weighting the time characteristics to obtain effective characteristics;
and fusing the effective features and the corresponding time scales to obtain the multi-time scale features.
According to a specific implementation manner of the embodiment of the present disclosure, the step of establishing a t-distribution probabilistic neural network according to the multi-time scale feature includes:
establishing a student t distribution probability density function;
obtaining a distribution parameter corresponding to the student t distribution probability density function according to the multi-time scale features and the corresponding data in the training set;
and establishing the t-distribution probability neural network according to the association of the distribution parameters and the multi-time scale features.
According to a specific implementation manner of the embodiment of the present disclosure, the step of adjusting the number of neurons in the t-distribution probabilistic neural network to obtain a plurality of initial models includes:
and adjusting the number of neurons in the t-distribution probability neural network according to a preset value range, and training by using an Adam optimizer to obtain a plurality of initial models.
According to a specific implementation manner of the embodiment of the present disclosure, the step of inputting the test set into the prediction model to obtain the wind power prediction result of the area to be predicted includes:
inputting the test set into the prediction model, and acquiring a distribution parameter corresponding to each data in the test set;
obtaining a confidence interval according to a preset confidence and a distribution parameter corresponding to each data in the test set;
and taking the confidence interval as a wind power prediction result of the area to be predicted.
In a second aspect, an embodiment of the present disclosure provides a wind power prediction system, including:
the acquisition module is used for acquiring historical wind power data of an area to be predicted;
the preprocessing module is used for preprocessing the historical wind power data to obtain a training set, a verification set and a test set;
the extraction module is used for extracting and fusing feature data on different time scales in the training set by utilizing a one-dimensional convolutional neural network and a two-way long-short memory network to obtain multi-time scale features;
the establishing module is used for establishing a t-distribution probability neural network according to the multi-time scale features;
the adjusting module is used for adjusting the number of neurons in the t-distribution probability neural network to obtain a plurality of initial models;
the verification module is used for respectively inputting the verification set into each initial model and taking the initial model with the optimal output result as a prediction model;
and the prediction module is used for inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted.
In a third aspect, an embodiment of the present disclosure further provides an electronic device, where the electronic device includes:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the wind power prediction method of the first aspect or any implementation manner of the first aspect.
In a fourth aspect, the present disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the wind power prediction method in the first aspect or any implementation manner of the first aspect.
In a fifth aspect, the present disclosure also provides a computer program product, where the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, where the computer program includes program instructions, and when the program instructions are executed by a computer, the computer is caused to execute the wind power prediction method in the foregoing first aspect or any implementation manner of the first aspect.
The wind power prediction scheme in the embodiment of the disclosure includes: acquiring historical wind power data of an area to be predicted; preprocessing the historical wind power data to obtain a training set, a verification set and a test set; extracting feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-short memory network and fusing the feature data to obtain multi-time scale features; establishing a t-distribution probability neural network according to the multi-time scale features; adjusting the number of neurons in the t-distribution probability neural network to obtain a plurality of initial models; respectively inputting the verification set into each initial model, and taking the initial model with the optimal output result as a prediction model; and inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted.
The beneficial effects of the embodiment of the disclosure are: according to the scheme, the multi-time scale features are extracted by utilizing the one-dimensional convolutional neural network and the two-way long and short memory network, uncertainty of wind power prediction is described by utilizing student t distribution, the uncertainty of future wind power is accurately predicted by constructing the deep student t probability neural network, and the prediction efficiency and accuracy are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings needed to be used in the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present disclosure, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a wind power prediction method provided in an embodiment of the present disclosure;
fig. 2 and fig. 3 are schematic diagrams of confidence intervals of different data sets of a wind power prediction method according to an embodiment of the present disclosure;
fig. 4 is a schematic structural diagram of a wind power prediction system provided in the embodiment of the present disclosure;
fig. 5 is a schematic view of an electronic device provided in an embodiment of the present disclosure.
Detailed Description
The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
The embodiments of the present disclosure are described below with specific examples, and other advantages and effects of the present disclosure will be readily apparent to those skilled in the art from the disclosure in the specification. It is to be understood that the described embodiments are merely illustrative of some, and not restrictive, of the embodiments of the disclosure. The disclosure may be embodied or carried out in various other specific embodiments, and various modifications and changes may be made in the details within the description without departing from the spirit of the disclosure. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
It is noted that various aspects of the embodiments are described below within the scope of the appended claims. It should be apparent that the aspects described herein may be embodied in a wide variety of forms and that any specific structure and/or function described herein is merely illustrative. Based on the disclosure, one skilled in the art should appreciate that one aspect described herein may be implemented independently of any other aspects and that two or more of these aspects may be combined in various ways. For example, an apparatus may be implemented and/or a method practiced using any number of the aspects set forth herein. Additionally, such an apparatus may be implemented and/or such a method may be practiced using other structure and/or functionality in addition to one or more of the aspects set forth herein.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present disclosure, and the drawings only show the components related to the present disclosure rather than the number, shape and size of the components in actual implementation, and the type, amount and ratio of the components in actual implementation may be changed arbitrarily, and the layout of the components may be more complicated.
In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects may be practiced without these specific details.
At present, wind energy plays a crucial role in global energy supply as a clean, environment-friendly and inexhaustible renewable energy source, however, when large-scale wind power is incorporated into a power grid, the inherent randomness and intermittency of the wind power bring huge challenges to power grid maintenance, power grid planning, power system scheduling and the like. In addition, accurate wind power prediction is also beneficial to efficient operation of an energy market, great improvement of power transmission capacity, reasonable planning of a wind power system and the like. Accurate wind power prediction is an effective method for solving the above problems. And multi-scale information contained in the wind power data is considered. Different information is often contained in data of different scales. High resolution data often contains much information reflecting detail fluctuation, while low resolution data often contains long-term fluctuation information. However, the existing prediction method usually realizes prediction by a complex and huge amount of calculation, or predicts future wind power by using only information in certain resolution data, and cannot describe the uncertainty of wind power fluctuation in detail.
The embodiment of the disclosure provides a wind power prediction method, which can be applied to a wind power prediction process of a wind power generation scene.
Referring to fig. 1, a flow diagram of a wind power prediction method provided in the embodiment of the present disclosure is shown. As shown in fig. 1, the method mainly comprises the following steps:
s101, acquiring historical wind power data of an area to be predicted;
in specific implementation, when wind power of a certain future time is required to be predicted, the wind power can be used as the area to be predicted, and then historical wind power data of the area to be predicted can be obtained from a background database.
S102, preprocessing the historical wind power data to obtain a training set, a verification set and a test set;
in specific implementation, considering that the historical wind power data contains a large amount of data and is different data, the historical wind power data can be preprocessed to reduce the complexity of the historical wind power data, and then the preprocessed data are divided into the training set, the verification set and the test set.
S103, extracting and fusing feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-short memory network to obtain multi-time scale features;
in specific implementation, the existing prediction method is considered, and multi-scale information contained in wind power data is often ignored in the wind power prediction process. Different information is often contained in data of different scales. High resolution data often contains much information reflecting detail fluctuation, while low resolution data often contains long-term fluctuation information. In the process of wind power prediction, the future wind power is usually predicted only by using information in certain resolution data. For example, in the process of predicting the wind power of 10 minutes in the future by constructing a prediction model for the wind power data at intervals of 10 minutes, only the information contained in the 10-minute data is often utilized.
Each data in the training set can be processed by utilizing the one-dimensional convolutional neural network and the two-way long and short memory network based on the attent i on mechanism, and the feature data of each data in the training set on different time scales is extracted and fused to obtain the multi-time scale features, so that more fluctuation information and detail fluctuation can be obtained, and the subsequent prediction is accurate.
S104, establishing a t-distribution probability neural network according to the multi-time scale features;
in specific implementation, in consideration of the fact that compared with a traditional prediction model, the probability neural network does not directly predict the study object, but predicts the parameters of the probability distribution, such as the uncertainty of each sample point in the gaussian probability neural network is assumed to be described by the gaussian distribution, so as to predict the mean value and the variance of the gaussian distribution, and the negative log-likelihood function is used as a loss function to optimize the whole probability network. However, the complex uncertainty of the Gaussian distribution during wind power prediction is difficult to be drawn, so that the performance of the Gaussian distribution has certain limitation, a t-distribution probability neural network can be established according to the multi-time scale features after the multi-time scale features are obtained, the uncertainty is drawn by using three-parameter student t-distribution, and the uncertainty of the wind power is predicted by constructing the student t-probability neural network.
S105, adjusting the number of neurons in the t-distribution probability neural network to obtain a plurality of initial models;
in specific implementation, considering that an Attention-based two-way long-short memory network is adopted to fully depict the complex relationship between the multi-time scale features obtained in the steps and the prediction target, the t-distribution probability neural network is further constructed to predict the uncertainty of each sample, and aiming at the parameter of the number of the neurons in the two-way long-short memory network, the number can be adjusted by setting different numbers, so that a plurality of initial models are obtained.
S106, respectively inputting the verification set into each initial model, and taking the initial model with the optimal output result as a prediction model;
after obtaining a plurality of initial models, the verification set may be respectively input into each of the initial models, an optimal model is selected by observing the prediction performance of the models in different numbers of neurons on the verification set, and the initial model with the optimal output result is used as the prediction model.
And S107, inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted.
In specific implementation, after the prediction model is obtained, the test set is input into the prediction model, the uncertainty distribution of each data in the test set is obtained through the prediction model, and then the wind power prediction result corresponding to the area to be predicted is obtained.
According to the wind power prediction method provided by the embodiment, the multi-time scale features are extracted by utilizing the one-dimensional convolutional neural network and the two-way long and short memory network, the uncertainty of wind power prediction is described by utilizing student t distribution, the deep student t probability neural network is constructed to accurately predict the uncertainty of future wind power, and the prediction efficiency and accuracy are improved.
On the basis of the above embodiment, step S102 is to preprocess the historical wind power data to obtain a training set, a verification set, and a test set, and includes:
dividing the historical wind power data into three disjoint subsets;
and carrying out normalization processing on the three subsets, and using the three subsets as the training set, the verification set and the test set.
For example, after the historical wind power data is obtained, the historical wind power data may be divided into three disjoint subsets, and considering the convenience of model training, all data may be normalized according to the following formula:
Figure BDA0003305127250000081
wherein x and
Figure BDA0003305127250000082
representing the raw data and the data after normalization, xminAnd xmaxRespectively representing the minimum and maximum values in the raw data.
On the basis of the foregoing embodiment, in step S103, extracting and fusing feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-and-short memory network to obtain a multi-time scale feature, including:
converting the data in the training set to different time scales by using an average pooling method;
extracting spatial features from the input of different time scales by using the one-dimensional convolutional neural network;
extracting time characteristics from the space characteristics by using the bidirectional long and short memory network;
automatically weighting the time characteristics to obtain effective characteristics;
and fusing the effective features and the corresponding time scales to obtain the multi-time scale features.
For example, the raw power input data may be first converted to a different time scale using an average pooling method. Given a d-dimensional continuous input feature { x with time resolution of t1,x2,…,xd-1,xdAnd f, if the average pooling window size is m and the step length is m, rounding up the sequence length d/m after average pooling, namely
Figure BDA0003305127250000091
The averaged pooled sequences can be represented as
Figure BDA0003305127250000092
Wherein each z isiIs the mean of m adjacent data points of time resolution t, i.e.
Figure BDA0003305127250000093
Figure BDA0003305127250000094
If d/m is a positive integer, the original sequence can be just converted into low-resolution data with the length of d/m according to the formula; if d/m is a non-positive integer, to obtain a length of
Figure BDA0003305127250000095
The original sequence is complemented by 0 until the sequence length reaches
Figure BDA0003305127250000096
When m is 1, the averaged pooled sequence is the same as the original sequence. Sequences with the same or lower original sequence resolution t, namely sequences with different time scales can be obtained by the average pooling method.
Then, the one-dimensional convolutional neural network is utilized to extract spatial features from the input with different time scales, the convolutional neural network is widely applied to feature extraction in various tasks, wherein the most important part is convolutional operation, and the convolutional operation in the one-dimensional convolutional neural network can be expressed as:
Figure BDA0003305127250000097
wherein the content of the first and second substances,
Figure BDA0003305127250000098
represents the output of the jth neuron of the first convolutional layer,
Figure BDA0003305127250000099
which represents the kernel of the convolution,
Figure BDA00033051272500000910
a bias term is represented as a function of,
Figure BDA00033051272500000911
representing convolution operation, M representing the number of neurons in the l-1 th convolutional layer, and g (·) representing an activation function, e.g., a ReLU activation function may be employed to avoid gradient vanishing and speed up network convergence. In addition, 2 layers of convolution layers can be adopted to perform feature extraction on data with different time scales, the size of convolution kernels in each layer of convolution is 2 x 1, and the number of convolution kernels is 32.
And then, the two-way long and short memory network is used for further extracting time characteristics from the space characteristics, so that the output of the two-way long and short memory network contains both space information and time information. In addition, because a large number of convolution kernels in the convolutional neural network cause the feature dimensionality extracted by the convolutional neural network to be too high, the bidirectional long and short memory network can also map the feature dimensionality to a lower dimensionality, and further the calculation complexity is reduced. Compared with the traditional long and short memory network model, the bidirectional long and short memory network is provided with two long and short memory modules, namely a forward long and short memory module and a backward long and short memory module. Not only the past information but also the past information can be utilized in the bidirectional long and short memory networkFuture information characterizes the current state. Assume that the output of the forward direction long/short memory module corresponding to the ith time step input is
Figure BDA0003305127250000101
The output of the backward long-short memory module is
Figure BDA0003305127250000102
The output of the bidirectional long/short memory network corresponding to the ith time step input can be expressed as
Figure BDA0003305127250000103
And
Figure BDA0003305127250000104
splicing of, i.e.
Figure BDA0003305127250000105
Wherein h isiRepresents the output of the BiLSTM and,
Figure BDA0003305127250000106
representing a stitching operation.
And then, an attention mechanism can be utilized to automatically weight the time step output result of each bidirectional long-short memory network to obtain key useful information. The first step is to use a layer of perceptron pairs h with weight matrix W and bias vector biMapping is carried out, and the mapped features are expressed as thetai(ii) a The second step uses the softmax function and θiNormalizing the weight betai(ii) a And thirdly, calculating the weighted comprehensive output information V. The calculation formula can be expressed as:
θi＝tanh(Whi+b)，
Figure BDA0003305127250000107
Figure BDA0003305127250000108
wherein, tanh (·) represents a hyperbolic tangent function, and an attention mechanism automatically weights output characteristics of different time steps, and finally outputs the effective characteristics.
Finally, useful features of the mining on different time scales are fused. After data on different time scales are obtained, V is taken as a characteristic extracted from input data with t as a time scaletThen the fused features obtained at different time scales can be expressed as
Figure BDA0003305127250000109
Wherein
Figure BDA00033051272500001010
Representing the fused multi-time scale features.
Further, the step S104 of establishing a t-distribution probabilistic neural network according to the multi-time scale feature includes:
establishing a student t distribution probability density function;
obtaining a distribution parameter corresponding to the student t distribution probability density function according to the multi-time scale features and the corresponding data in the training set;
and establishing the t-distribution probability neural network according to the association of the distribution parameters and the multi-time scale features.
For example, assuming that the variable g satisfies a three-parameter student t-distribution, its probability density function can be expressed as
Figure BDA0003305127250000111
Wherein p (-) represents a probability density function, τ >0 represents degrees of freedom, ψ is a position parameter, Γ (-) represents a gamma function, and s >0 represents a scale parameter.
Giving N corresponding data in the training set, wherein the input is the multi-time scale feature obtained by the steps
Figure BDA0003305127250000112
The output is yiThe likelihood function L based on the uncertainty assumption of the student's t-distribution can be expressed as
Figure BDA0003305127250000113
Wherein
Figure BDA0003305127250000114
And
Figure BDA0003305127250000115
respectively represent the input of the ith sample
Figure BDA0003305127250000116
Corresponding distribution parameters. The Loss function Loss of the final student's t probabilistic neural network can be expressed as a negative log-likelihood function L, i.e.
Figure BDA0003305127250000117
An Attention-based bidirectional long and short memory network can be adopted to fully describe the complex relationship between the multi-time scale features obtained in the steps and the prediction target, and then the t-distribution probability neural network is constructed to predict the uncertainty of each sample.
Optionally, in step S105, adjusting the number of neurons in the t-distribution probabilistic neural network to obtain a plurality of initial models, including:
and adjusting the number of neurons in the t-distribution probability neural network according to a preset value range, and training by using an Adam optimizer to obtain a plurality of initial models.
In specific implementation, considering that the wind power data are all greater than or equal to 0, the model outputs taui、ψiAnd siParameter limits of student t distribution must be met, so that the outputs of the models are all required to be greater than 0, and then a softplus activation function needs to be used at the last full-connection layer of the network to enable the whole network to be in useSatisfies the above-described restriction conditions. For example, the number of neurons in the t-distribution probability neural network can be adjusted according to a preset value range, and an Adam optimizer is used for training to obtain a plurality of initial models.
On the basis of the foregoing embodiment, step S107 includes inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted, where the wind power prediction result includes:
inputting the test set into the prediction model, and acquiring a distribution parameter corresponding to each data in the test set;
obtaining a confidence interval according to a preset confidence and a distribution parameter corresponding to each data in the test set;
and taking the confidence interval as a wind power prediction result of the area to be predicted.
In specific implementation, the test set may be input into the prediction model, a distribution parameter corresponding to each data in the test set is obtained, uncertainty distribution of each data in the test set is further obtained, a plurality of different confidence levels may be preset, a plurality of confidence intervals are obtained according to the distribution parameters, and then the wind power prediction results of the region to be predicted, which are obtained by using the different confidence intervals as the wind power prediction results, are shown in fig. 2 and fig. 3.
The present solution will be described below with reference to a specific embodiment, and the present disclosure uses wind power data provided by two energy companies (TransnetBW and Amprion) to verify the validity of the proposed model. The sampling frequency of the data is 5 minutes, and the sampling interval is from 8 months and 23 days in 2019 to 22 days in 9 months and 2020. The first 60% of the samples were used as training samples, and the proportion of both the validation set and the test set was 20%. If the wind power at the future 10 minutes can be predicted in advance by using data at 5 minutes, 10 minutes, 15 minutes and 20 minutes, the average pooling window size m is 1,2,3 and 4.
Then 4 evaluation indices, namely Pinball Loss (PL), Winkler Score (WS), interval coverage (PICP) and normalized interval width (PINAW), were used. To evaluate effectiveness, the following comparative model Quantile Regression (QR), the LSTM model based on mean PL (PL-LSTM), the wind power prediction model based on multi-scale information and a deep Gaussian probability neural network (MSDNN-G), the deep Gaussian probability neural network model without considering multi-scale information (SSDDNN-G), and the deep student T-distribution probability neural network model without considering multi-scale information (SSDDNN-T) are considered. For convenience of description, the prediction model to which this disclosure relates is named MSDNN-T.
At 85%, 90% and 95% confidence, the results of the wind power predictions of the different models on the data set Amprion are shown in table 1, and the results of the wind power predictions of the different models on the data set transmetbw are shown in table 2.
Figure BDA0003305127250000131
TABLE 1
Figure BDA0003305127250000132
TABLE 2
As can be seen from Table 1, the PICP of all models, except the SSDDNN-G and PL-LSTM models, is above a given confidence level, indicating that the resulting prediction interval is reliable. For example, the PICP of SSDDNN-G is less than a given confidence at 90% and 95% confidence, and the PICP of PL-LSTM is less than a given confidence at 95% confidence. From the span width index PINAW, the span width is relatively small because the PICP value of SSDDNN-G is relatively small, i.e. the coverage is minimum. From the two indexes of PL and WS, the minimum PL and the maximum WS are obtained respectively in the disclosure, and the comparison model is superior to other 5 comparison models. In addition, according to the comparison between MSDNN-T and MSDNN-G and the comparison between SSDDNN-G and SSDDNN-T, the model based on the student T distribution is superior to the model based on the Gaussian distribution, and the student T distribution can more accurately depict the uncertainty of wind power prediction. As can be seen from the comparison of MSDNN-T with SSDDNN-T and the comparison of MSDNN-G with SSDDNN-G, the model performance considering the multi-time scale features is better, indicating the effectiveness of the multi-time scale features.
As can be seen in table 2, the PICP values for most models are greater than a given confidence level. From the PINAW index, the PICP of the MSDNN-G model is smaller, and the interval width is smaller than that of other models. From PL and WS indexes, the comparison of the SSDDNN-G, SSDDNN-T, MSDDNN-G and the MSDNN-T four models can also find that the utilization of the multi-time scale features is beneficial to the improvement of the model performance and the uncertainty of wind power prediction can be more accurately described by the T distribution of students.
Corresponding to the above method embodiment, referring to fig. 4, the embodiment of the present disclosure further provides a wind power prediction system 40, including:
the obtaining module 401 is configured to obtain historical wind power data of an area to be predicted;
a preprocessing module 402, configured to preprocess the historical wind power data to obtain a training set, a verification set, and a test set;
an extracting module 403, configured to extract and fuse feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-short memory network to obtain a multi-time scale feature;
an establishing module 404, configured to establish a t-distribution probabilistic neural network according to the multi-time scale feature;
an adjusting module 405, configured to adjust the number of neurons in the t-distribution probability neural network to obtain multiple initial models;
a verification module 406, configured to input the verification set into each initial model, and use the initial model with the optimal output result as a prediction model;
and the prediction module 407 is configured to input the test set into the prediction model to obtain a wind power prediction result of the area to be predicted.
The system shown in fig. 4 can correspondingly execute the content in the above method embodiment, and details of the part not described in detail in this embodiment refer to the content described in the above method embodiment, which is not described again here.
Referring to fig. 5, an embodiment of the present disclosure also provides an electronic device 50, including: at least one processor and a memory communicatively coupled to the at least one processor. The storage stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the wind power prediction method in the foregoing method embodiment.
The disclosed embodiments also provide a non-transitory computer readable storage medium storing computer instructions for causing the computer to execute the wind power prediction method in the foregoing method embodiments.
The disclosed embodiments also provide a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, cause the computer to perform the wind power prediction method in the aforementioned method embodiments.
Referring now to FIG. 5, a schematic diagram of an electronic device 50 suitable for use in implementing embodiments of the present disclosure is shown. The electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., car navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. The electronic device shown in fig. 5 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 5, electronic device 50 may include a processing means (e.g., central processing unit, graphics processor, etc.) 501 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)502 or a program loaded from a storage means 508 into a Random Access Memory (RAM) 503. In the RAM 503, various programs and data necessary for the operation of the electronic apparatus 50 are also stored. The processing device 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input/output (I/O) interface 505 is also connected to bus 504.
Generally, the following devices may be connected to the I/O interface 505: input devices 506 including, for example, a touch screen, touch pad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 507 including, for example, a Liquid Crystal Display (LCD), speakers, vibrators, and the like; storage devices 508 including, for example, magnetic tape, hard disk, etc.; and a communication device 509. The communication means 509 may allow the electronic device 50 to communicate with other devices wirelessly or by wire to exchange data. While the figures illustrate an electronic device 50 having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication means 509, or installed from the storage means 508, or installed from the ROM 502. The computer program performs the above-described functions defined in the methods of the embodiments of the present disclosure when executed by the processing device 501.
It should be noted that the computer readable medium in the present disclosure can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In contrast, in the present disclosure, a computer readable signal medium may comprise a propagated data signal with computer readable program code embodied therein, either in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.
The computer readable medium may be embodied in the electronic device; or may exist separately without being assembled into the electronic device.
The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to perform the steps associated with the method embodiments.
Alternatively, the computer readable medium carries one or more programs which, when executed by the electronic device, enable the electronic device to perform the steps associated with the method embodiments.
Computer program code for carrying out operations for aspects of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present disclosure may be implemented by software or hardware.
It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.
The above description is only for the specific embodiments of the present disclosure, but the scope of the present disclosure is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present disclosure should be covered within the scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.

Claims (9)

1. A wind power prediction method is characterized by comprising the following steps:
acquiring historical wind power data of an area to be predicted;
preprocessing the historical wind power data to obtain a training set, a verification set and a test set;
extracting feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-short memory network and fusing the feature data to obtain multi-time scale features;
establishing a t-distribution probability neural network according to the multi-time scale features;
adjusting the number of neurons in the t-distribution probability neural network to obtain a plurality of initial models;
respectively inputting the verification set into each initial model, and taking the initial model with the optimal output result as a prediction model;
and inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted.
2. The method according to claim 1, wherein the step of preprocessing the historical wind power data to obtain a training set, a validation set, and a test set comprises:
dividing the historical wind power data into three disjoint subsets;
and carrying out normalization processing on the three subsets, and using the three subsets as the training set, the verification set and the test set.
3. The method according to claim 1, wherein the step of extracting and fusing feature data on different time scales in the training set by using a one-dimensional convolutional neural network and a two-way long-short memory network to obtain multi-time scale features comprises:
converting the data in the training set to different time scales by using an average pooling method;
extracting spatial features from the input of different time scales by using the one-dimensional convolutional neural network;
extracting time characteristics from the space characteristics by using the bidirectional long and short memory network;
automatically weighting the time characteristics to obtain effective characteristics;
and fusing the effective features and the corresponding time scales to obtain the multi-time scale features.
4. The method of claim 3, wherein the step of building a t-distribution probabilistic neural network from the multi-time scale features comprises:
establishing a student t distribution probability density function;
obtaining a distribution parameter corresponding to the student t distribution probability density function according to the multi-time scale features and the corresponding data in the training set;
and establishing the t-distribution probability neural network according to the association of the distribution parameters and the multi-time scale features.
5. The method of claim 1, wherein the step of adjusting the number of neurons in the t-distribution probabilistic neural network to obtain a plurality of initial models comprises:
and adjusting the number of neurons in the t-distribution probability neural network according to a preset value range, and training by using an Adam optimizer to obtain a plurality of initial models.
6. The method according to claim 4, wherein the step of inputting the test set into the prediction model to obtain the wind power prediction result of the area to be predicted comprises:
inputting the test set into the prediction model, and acquiring a distribution parameter corresponding to each data in the test set;
obtaining a confidence interval according to a preset confidence and a distribution parameter corresponding to each data in the test set;
and taking the confidence interval as a wind power prediction result of the area to be predicted.
7. A wind power prediction system, comprising:
the acquisition module is used for acquiring historical wind power data of an area to be predicted;
the preprocessing module is used for preprocessing the historical wind power data to obtain a training set, a verification set and a test set;
the extraction module is used for extracting and fusing feature data on different time scales in the training set by utilizing a one-dimensional convolutional neural network and a two-way long-short memory network to obtain multi-time scale features;
the establishing module is used for establishing a t-distribution probability neural network according to the multi-time scale features;
the adjusting module is used for adjusting the number of neurons in the t-distribution probability neural network to obtain a plurality of initial models;
the verification module is used for respectively inputting the verification set into each initial model and taking the initial model with the optimal output result as a prediction model;
and the prediction module is used for inputting the test set into the prediction model to obtain a wind power prediction result of the area to be predicted.
8. An electronic device, characterized in that the electronic device comprises:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the wind power prediction method of any of the preceding claims 1-6.
9. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the wind power prediction method of any of the preceding claims 1-6.
CN202111201545.5A 2021-10-15 2021-10-15 Wind power prediction method, system, equipment and medium Active CN113723712B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111201545.5A CN113723712B (en) 2021-10-15 2021-10-15 Wind power prediction method, system, equipment and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111201545.5A CN113723712B (en) 2021-10-15 2021-10-15 Wind power prediction method, system, equipment and medium

Publications (2)

Publication Number Publication Date
CN113723712A true CN113723712A (en) 2021-11-30
CN113723712B CN113723712B (en) 2023-08-18

Family

ID=78686057

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111201545.5A Active CN113723712B (en) 2021-10-15 2021-10-15 Wind power prediction method, system, equipment and medium

Country Status (1)

Country Link
CN (1) CN113723712B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180223812A1 (en) * 2017-02-07 2018-08-09 International Business Machines Corporation Reducing curtailment of wind power generation
CN112365040A (en) * 2020-11-03 2021-02-12 哈尔滨工业大学 Short-term wind power prediction method based on multi-channel convolution neural network and time convolution network
CN112365033A (en) * 2020-10-26 2021-02-12 中南大学 Wind power interval prediction method, system and storage medium
CN112598180A (en) * 2020-12-21 2021-04-02 北京华能新锐控制技术有限公司 Distributed regional wind power prediction method
CN113409307A (en) * 2021-07-15 2021-09-17 中南大学 Image denoising method, device and medium based on heterogeneous noise characteristics

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180223812A1 (en) * 2017-02-07 2018-08-09 International Business Machines Corporation Reducing curtailment of wind power generation
CN112365033A (en) * 2020-10-26 2021-02-12 中南大学 Wind power interval prediction method, system and storage medium
CN112365040A (en) * 2020-11-03 2021-02-12 哈尔滨工业大学 Short-term wind power prediction method based on multi-channel convolution neural network and time convolution network
CN112598180A (en) * 2020-12-21 2021-04-02 北京华能新锐控制技术有限公司 Distributed regional wind power prediction method
CN113409307A (en) * 2021-07-15 2021-09-17 中南大学 Image denoising method, device and medium based on heterogeneous noise characteristics

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
SAROHA, SUMIT 等: ""A Review and Evaluation of Current Wind Power Prediction Technologies"", 《WSEAS TRANSACTIONS ON POWER SYSTEMS》, vol. 15, pages 1 - 12 *
张浩: ""风电功率时空不确定性预测方法研究"", 《中国博士学位论文全文数据库工程科技Ⅱ辑》, no. 1, pages 042 - 1348 *
李卓 等: ""基于IDSCNN-AM-LSTM组合神经网络超短期风电功率预测方法"", 《高电压技术》, vol. 48, no. 6, pages 2117 - 2127 *

Also Published As

Publication number Publication date
CN113723712B (en) 2023-08-18

Similar Documents

Publication Publication Date Title
EP3757900A1 (en) Time series prediction with confidence estimates using sparse recurrent mixture density networks
WO2018209576A1 (en) Systems and methods for digital route planning
Shang et al. Short-term load forecasting based on PSO-KFCM daily load curve clustering and CNN-LSTM model
WO2023005120A1 (en) Energy consumption prediction method and apparatus for building, and computer device and storage medium
CN111582559A (en) Method and device for estimating arrival time
CN111291715B (en) Vehicle type identification method based on multi-scale convolutional neural network, electronic device and storage medium
CN114330935B (en) New energy power prediction method and system based on multiple combination strategies integrated learning
CN112668238A (en) Rainfall processing method, device, equipment and storage medium
CN116684330A (en) Traffic prediction method, device, equipment and storage medium based on artificial intelligence
CN112418291A (en) Distillation method, device, equipment and storage medium applied to BERT model
CN115034497A (en) Multi-site daily water level prediction method and device, electronic equipment and computer medium
Ye et al. Multi-year ENSO forecasts using parallel convolutional neural networks with heterogeneous architecture
CN114972877A (en) Image classification model training method and device and electronic equipment
CN117079148A (en) Urban functional area identification method, device, equipment and medium
CN115391746B (en) Interpolation method, interpolation device, electronic device and medium for meteorological element data
CN113723712B (en) Wind power prediction method, system, equipment and medium
EP4040421B1 (en) Method and apparatus for predicting traffic data and electronic device
CN115225520B (en) Multi-mode network flow prediction method and device based on meta-learning framework
CN111161238A (en) Image quality evaluation method and device, electronic device, and storage medium
CN115293889A (en) Credit risk prediction model training method, electronic device and readable storage medium
CN114970357A (en) Energy-saving effect evaluation method, system, device and storage medium
CN111339446B (en) Interest point mining method and device, electronic equipment and storage medium
CN115700548A (en) Method, apparatus and computer program product for user behavior prediction
CN111178630A (en) Load prediction method and device
CN113516315B (en) Wind power generation power interval prediction method, device and medium

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
GR01 Patent grant
GR01 Patent grant