Ultra-short-term wind power prediction method based on long-short-term memory neural network

Abstract

The invention provides an ultra-short-term wind power prediction method based on a long-short-term memory neural network. Firstly, historical sample data and prediction day sample data are obtained, then correlation coefficients of all factors in historical days and corresponding factors in prediction days are calculated, then weights of all influence factors in wind power influence factors are calculated, similarity between the historical days and the prediction days is calculated, and similar sample data is obtained to serve as training data; and finally, the wind power of the prediction dayis predicated by adopting an LSTM model. The training data is screened, and the data with high similarity with the sample data of the prediction day is selected as the training data, so that the similarity of the modeling data is enhanced, the accuracy of the model is improved, the LSTM model is adopted as the training model, and the purpose of fully considering the time sequence and nonlinearityof the wind power can be achieved.

Description

technical field [0001] The invention relates to the technical field of wind power forecasting, in particular to an ultra-short-term wind power forecasting method based on a long-short-term memory neural network. Background technique [0002] With the depletion of conventional energy, the development and utilization of renewable energy has become a current research hotspot. As one of the most abundant resources in renewable energy, wind energy has a very broad application prospect. However, as an intermittent energy source, wind power has the characteristics of randomness and uncontrollability. When wind power is connected to the grid on a large scale, it will have a certain impact on the stability, adequacy and economy of the power system. The ultra-short-term forecasting of wind power is not only beneficial to alleviate the pressure of peak regulation and frequency regulation of the power system, but also helps grid dispatchers to formulate plans, arrange backup reasonabl...

AI Technical Summary

Problems solved by technology

In order to fully exploit the natural information contained in the wind power data, the historical data should be screened and classified before the forecasting model training, which is conducive to enhancing the similarity of the modeling data and thus improving the accuracy of the model. However, the current forecasting methods There is little mention of reasonable handling of training samples

Images