Wind power forecast method based on continuous period clustering

Abstract

The invention relates to the field of machine learning and wind power generation, and particularly relates to a wind power forecast method based on continuous period clustering. The wind power forecast method comprises the steps that an Elman neural network and a support vector machine are used as a forecast model to perform iterative forecasting on the basis of a similar day forecast method so asto determine the similar period length; the similarity measure standard is determined through combination of the power vector and the meteorological information according to the similar period lengththrough a two-stage search strategy, and the optimal similar period set is found in the historical data; and a wind power forecast model is created based on the Elman neural network, and the obtainedoptimal similar period set acts as the training data to perform iterative computation through the wind power forecast model so as to complete wind power forecasting of the future periods. The meteorological factor is introduced on the basis of the similar day forecast method, and the clustering-classifying-based similar period selection strategy is adopted so that the optimal similar period set can be rapidly searched and the forecast precision and accuracy can be enhanced.

Description

technical field [0001] The invention relates to the fields of machine learning and wind power generation, in particular to a wind power prediction method based on continuous period clustering. Background technique [0002] With the reduction of world energy, the research, development and utilization of renewable new energy has become a top priority. With the advantages of wide range, renewable and non-polluting, wind energy has gradually become the most promising energy source. However, natural wind is random and intermittent, and the centralized connection of large-scale wind power will pose a threat to the safe and stable operation of the power grid. Wind power forecasting is an effective way to solve this problem. [0003] Commonly used wind power prediction methods include physical and statistical methods. The physical method does not rely on the historical data of the wind farm, but only requires detailed physical information of the wind farm and digital weather foreca...

AI Technical Summary

Problems solved by technology

However, the unsupervised clustering method is highly dependent on the sample, and it is easy to generate too many classifications, and the prediction accuracy is difficult to guarantee; by subdividing similar days into "similar time periods", first look for wind power with similar changes in the 12 hours before the prediction time The curve is used as the "baseline segment", and then the daily feature vectors that are similar to the changes in the 12h after the prediction time are searched for as the "prediction segment". Does not take into account the influence of reference power curves and meteorological characteristic values

Images