Wind power prediction method based on continuous time slice clustering and support vector machine (SVM) modeling

Abstract

The invention discloses a wind power prediction method based on continuous time period clustering and support vector machine modeling, which includes the following steps: ① unsupervised clustering of similar days throughout the year according to wind characteristics; Based on the clustering results, the whole year is divided into n consecutive time periods, and each time period is clustered and classified according to the frequency of various days in each time period and the wind characteristics of the connected time period; ③Use SVM to model the time period of the same category in step ②, and use it for the prediction of the same time in the following years. Since the annual continuous time period clustering method is added on the basis of daily similarity, both daily similarity and time continuity are considered, which greatly improves the similarity of training samples in the prediction model and wind power prediction Compared with the traditional method, the relative error of power prediction is reduced by 7.2%, so that the prediction accuracy of wind power reaches 83.96%.

Description

technical field [0001] The invention relates to a wind power prediction method, in particular to a wind power prediction method for performing cluster analysis and support vector machine modeling on the actual data of a wind farm. Background technique [0002] As an intermittent energy source, due to its randomness and uncontrollability, the amplitude of output power varies greatly and the frequency is unstable, which has a great impact on the power grid. With the increase of wind power installed capacity, the proportion of wind power grid connection is gradually increasing, so it is very important to predict the output power of wind farms. [0003] Wind power prediction methods mainly include physical methods [ and statistical methods. The physical method does not require a large amount of historical data, but it is generally difficult to model. It needs to analyze and study various conditions of the geographical location of the wind farm, and is suitable for new wind far...

AI Technical Summary

Problems solved by technology

These two methods only consider the similarity law of the day from the horizontal, and do not consider the influence of the continuity of the date on the power prediction from the vertical. Since the atmospheric motion is a long-term continuous and gradual process, the commonly used above-mentioned statistical prediction methods do not take both into account The similarity and continuous change of wind speed lacks effective classification of modeling data, so the similarity of training samples in the prediction model and the accuracy of prediction are not ideal, which cannot meet the requirements of wind power grid integration

[0006] In addition, when clustering similar days throughout the year, the traditional K-means algorithm, the total number of categories C is determined, through continuous calculation, the position of the center of category C is adjusted to achieve optimal classification, but due to wind Uncertainty and randomness, the optimal value of the total number C of categories cannot be determined artificially before classification, and the C values ​​of different wind farms may not be the same, so the traditional K-means algorithm has certain limitations

Images