CN-116701868-B - Short-term wind power section probability prediction method

CN116701868BCN 116701868 BCN116701868 BCN 116701868BCN-116701868-B

Abstract

The invention relates to the technical field of power system operation and planning, in particular to a short-term wind power section probability prediction method, which utilizes hidden information in deep learning mining data and nonlinear characteristics in a wind power sequence to generate a prediction probability interval, selects a nonlinear weight method to improve the optimization performance of a particle swarm algorithm, namely an IPSO algorithm solves part of problems existing in the traditional algorithm, improves convergence rate, and then a CNN-LSTM hybrid algorithm is selected by the hybrid artificial intelligence algorithm to construct an IPSO-CNN-LSTM algorithm prediction model based on combination of SVM and fractional regression, short-term wind power probability prediction is completed after training, wherein the CNN network can extract potential characteristics of the CNN network from sample data by using convolution kernels, long-term components can be captured by the LSTM of the long-term memory network, gradient disappearance and explosion of the existing partial algorithm are avoided, and the wind power probability prediction efficiency is improved.

Inventors

GAO PENG
Cai Jianuo
LUO YI
ZHANG YINGHONG
CHEN JINLONG
GAO CHENG
MO CHOU
LIU SHIQI
XU JINYONG

Assignees

桂林电子科技大学
桂林智工科技有限责任公司

Dates

Publication Date: 20260505
Application Date: 20230607

Claims (5)

1. The short-term wind power section probability prediction method is characterized by comprising the following steps of: step1, collecting historical meteorological data and processing to obtain an initial data set of wind power probability prediction; step 2, screening and supplementing the initial data set to obtain a wind power data set, and carrying out normalization treatment; dividing the data of the wind power data set into a training set and a testing set, constructing an IPSO-CNN-LSTM algorithm prediction model based on SVM and fractional regression, and training and predicting; Dividing the obtained data into two groups of training sets and test sets, constructing an IPSO-CNN-LSTM algorithm prediction model based on combination SVM and quantile regression in the training sets through historical data, carrying out prediction after training through the training sets, carrying out quantization analysis on an output side of the model to obtain short-term wind power upper and lower boundary values in quantile form under a preset confidence level and a to-be-predicted daily prediction sequence, and finally comparing the short-term wind power upper and lower boundary values with a preset test set; the construction process of the IPSO-CNN-LSTM algorithm prediction model based on combining SVM and fractional regression comprises the following steps: an IPSO algorithm is selected to optimize the inertia weight omega; Combining the CNN network with the LSTM network, extracting potential characteristics of the CNN network from sample data by using the CNN network, and capturing long-term components by using the LSTM network; reflecting the nonlinear condition of the data by using QRNN model, obtaining the prediction probability interval under different confidence degrees, and further analyzing the obtained data; Adopting an average absolute error MAE and a root mean square error RMSE, and taking a section coverage rate PICP and a section average bandwidth PINAW as evaluation indexes of model accuracy; And 4, adjusting model parameters according to the prediction result and the error until the result is close to the test set, and completing short-term wind power probability prediction.
2. The short-term wind power segment probability prediction method of claim 1, The process of acquiring the initial data set of wind power probability prediction by collecting historical meteorological data processing is specifically to acquire regional historical meteorological data, process the initial meteorological data and wind power conditions by inquiring local logs, record information of weather stations and record information of other weather monitoring systems, and acquire the initial data set of wind power probability prediction according to the wind speed fluctuation quantity of the numerical weather forecast NWP prediction result of the historical prediction power and the wind farm.
3. A short-term wind power segment probability prediction method as defined in claim 2, The data of the numerical weather forecast NWP adopts selected annual meteorological data of a certain wind power plant, and the parameters comprise wind speed, wind direction, temperature and air pressure.
4. A short-term wind power segment probability prediction method according to claim 3, The method comprises the steps of screening and supplementing an initial data set to obtain a wind power data set, and carrying out normalization processing, namely screening weather factors with the highest correlation with wind power, processing abnormal data parts after screening relevant data, screening the whole data set after screening into a plurality of condition subsets through a clustering algorithm or a manual method, carrying out supplement processing on abnormal data in each subset through a cleaning method or an interpolation method, and carrying out normalization processing on the obtained wind power data set.
5. The short-term wind power segment probability prediction method of claim 4, And screening meteorological factor data by adopting a pearson coefficient, and reserving the meteorological data with the strongest correlation to perform subsequent prediction, wherein the clustering algorithm comprises a DBSCAN algorithm and a K-means algorithm, and the manual method is to manually process the abnormal data.

Description

Short-term wind power section probability prediction method Technical Field The invention relates to the technical field of operation and planning of power systems, in particular to a short-term wind power probability prediction method. Background With the development of social technology, the role of energy is increasing, and the rapid development is accompanied by continuous consumption of traditional fossil energy, and renewable energy starts to enter into the research of students. In the power industry, the traditional energy structure is optimized, the renewable energy duty ratio is increased, the grid-connected efficiency is improved, the method has become an important research direction of global energy development, and the characteristics of cleanliness, abundant reserves and the like of wind energy are favored by students. However, the instability and uncertainty of wind speed also enable the power generation efficiency of the wind power plant to generate larger fluctuation, so that wind power dispatching is difficult to be performed in advance, the influence caused by the uncertainty of wind power can be effectively relieved through accurate power prediction, and the safety operation of a power system is guaranteed. At present, the new trend of wind power prediction development is an artificial intelligence method, but the traditional single-point prediction model cannot quantify wind power irregularity and uncertainty, the shallow learning model cannot completely extract deep nonlinear characteristics in a wind power sequence, and a single prediction model cannot capture the change rule in the wind power sequence, so that a satisfactory prediction effect is achieved. There is therefore a need in the art for new solutions to this problem. Disclosure of Invention The invention aims to provide a short-term wind power probability prediction method, and aims to solve the technical problems that hidden information in a wind power data set subsequence is difficult to mine by a traditional point prediction method and a change rule in a wind power sequence is difficult to capture by a single prediction model. In order to achieve the above purpose, the invention provides a short-term wind power probability prediction method, which comprises the following steps: step1, collecting historical meteorological data and processing to obtain an initial data set of wind power probability prediction; step 2, screening and supplementing the initial data set to obtain a wind power data set, and carrying out normalization treatment; Dividing the data of the wind power data set into a training set and a testing set, constructing an IPSO-CNN-LSTM algorithm prediction model based on combination of SVM and quantile regression, and training and predicting; And 4, adjusting model parameters according to the prediction result and the error until the result is close to the test set, and completing short-term wind power probability prediction. Optionally, the process of acquiring the initial data set of wind power probability prediction by collecting historical meteorological data is specifically to acquire regional historical meteorological data, process the initial meteorological data and wind power conditions by inquiring local logs, record information of weather stations and record information of other weather monitoring systems, and acquire the initial data set of wind power probability prediction according to the wind speed fluctuation quantity of the numerical weather forecast NWP prediction result of the historical prediction power and the wind power field. Optionally, the data of the numerical weather forecast NWP adopts selected annual weather data of a certain wind farm, and the parameters include wind speed, wind direction, temperature and air pressure. Optionally, screening and supplementing the initial data set to obtain a wind power data set, and carrying out normalization processing, specifically screening weather factors with the strongest correlation with wind power, processing abnormal data parts after screening relevant data, screening the whole screened data set into a plurality of condition subsets through a clustering algorithm or a manual method, carrying out supplementation processing on abnormal data in each subset by adopting a cleaning method or an interpolation method, and finally carrying out normalization processing on the obtained wind power data set. Optionally, the pearson coefficient is adopted to screen meteorological factor data, the meteorological data with the strongest correlation is reserved for subsequent prediction, the clustering algorithm comprises a DBSCAN algorithm and a K-means algorithm, and the manual method is to manually process abnormal data. Optionally, in the executing process of the step 3, the obtained data is divided into two groups of a training set and a testing set, an IPSO-CNN-LSTM algorithm prediction model based on combination of SVM and quantile regression is cons