Research On Probability Prediction Of Wind Power Based On Deep Learning

Wind power prediction is helpful for the scheduling of the power grid.Compared with the wind power point prediction methods,the wind power probability prediction can obtain the prediction interval of the wind power at a given confidence level,and it can also calculate the probability density function of the predicted value of the wind power,namely The probability that the predicted value falls within the prediction interval at the current moment.With the development of the wind power industry,the accuracy of wind power point forecasting needs to be further improved.A single wind power forecast result can no longer meet the needs of today’s electricity market.In response to the above problems,this paper introduces deep learning to study the probability prediction of wind power.In order to improve the wind power point prediction accuracy and interval prediction performance,and shorten the training time of the long short-term memory network,a wind power interval prediction method based on the new forget gate long short-term memory network and gaussian process regression is proposed.By coupling the forget gate of the long short-term memory network and the current information state,a new forget gate long short-term memory network is constructed,and a short-term wind power interval prediction model based on the new forget gate long short-term memory network and the gaussian process regression is designed.Wind power and forecast interval at the moment.Based on the actual measurement data of the wind farm,the results show that the new forget gate long short-term memory network can shorten the training time,the new forget gate long short-term memory network and the gaussian process regression hybrid model compared with the gaussian process regression model,the new forget gate long short-term memory network,the gated recurrent unit model and the long short-term memory network model,this method can not only obtain high point prediction accuracy,but also provide interval prediction performance.In order to improve the prediction accuracy and probability prediction performance of wind power points,a wind power probability prediction method based on quantile regression new forget gate long short-term memory network combined with kernel density estimation is proposed.Construct quantile regression new forget gate long short-term memory network hybrid model framework,according to the predicted value of wind power at different quantile points,obtain the point predicted value and prediction interval of wind power power,and solve the prediction through the kernel density estimation of the cosine kernel function The probability density function of the value.Based on the actual measurement data of the wind farm,the results show that the quantile regression new forget gate long short-term memory network compared with the quantile regression minimal gated memory network,the quantile regression long short-term memory network,and the quantile regression gated recurrent unit model,this method can effectively improve the point prediction accuracy and probability prediction performance.