Research On Short-term Wind Power Prediction Based On Error Compensation Neural Network Model

How to improve the accuracy of wind power forecasting has always been a pressing issue in the field of wind power forecasting.This is because a higher wind power prediction accuracy provides a scientific basis for grid connection and grid dispatching and also saves wind farms from fines due to lower wind power prediction accuracy.It is a problem of both economic and social research value.With the rapid development of society,people’s demand for energy is increasing,accompanied by a series of ecological problems such as serious air pollution.The question of how to balance economic development with environmental concerns is an urgent and difficult one.Clean energy,which can reduce pollution,is gaining more and more attention,as reflected in the fact that the proportion of clean energy in the overall energy mix is increasing year by year,and the development of carbon peaking and carbon neutral strategies proposed in the National 13 th Five-Year Plan has made it clear that clean energy will prevail in the future.As a clean energy source,wind power is favoured for its wide distribution,ease of access and low cost.However,the development of wind power in China is facing many challenges,as the development of wind power in China started late and there are still technical gaps compared to developed countries in Europe and the US,especially in wind power forecasting technology.Due to the strong support of the national policy,the wind power prediction accuracy is not high,the consequence is that a large number of wind power projects around the lack of scientific planning led to abandoned wind and abandoned power and other serious problems,has seriously hindered the development of the domestic wind power industry.Therefore,how to improve the accuracy of wind power prediction has become an urgent problem to be solved.This paper focuses on the core problem of how to effectively improve the accuracy of short-term wind power forecasting.The main core point and innovation is the use of Bayesian dynamic linear model to model the prediction error data of the neural network model and then to achieve error compensation in order to improve the accuracy of wind power prediction.For the selection of the preliminary prediction model the principle of this paper is general,which also indirectly reflects the universality of this error compensation method.Therefore,for the preliminary wind power prediction model we use the most applied and mature neural network model in the field of short-term wind power prediction,and give the prediction effect of BP neural network and RBF neural network through specific experiments.The RBF neural network,which has the better prediction effect among the two,is finally adopted as the preliminary prediction model.The data used in this paper are historical data collected from a wind farm in Inner Mongolia,based on which short-term predictions of the future wind power of individual wind turbines in the wind farm for 8.25 hours are made.An overview of the basic wind power generation process is given,followed by a statistical analysis of the wind farm parameters.This is followed by data processing of the historical data to build a neural network prediction model based on the historical data for the short-term prediction of wind power.The prediction error is then analysed and a Bayesian dynamic linear prediction model is built to predict the prediction error.The experimental results show that the method achieves good prediction results.