Probability Prediction Method of Short-Term Electricity Price Based on Quantile Neural Network Model

Abstract

Aiming at the inaccuracy of short-term electricity price forecasting in competitive power markets, a probabilistic short-term electricity price forecasting method based on the quantile neural network model is proposed. First, a method for selecting electricity price similarity based on comprehensive influencing factors is designed to select the forecast data set with similar characteristics to the forecast date. The similar daily quantile regression algorithm is then combined with the generalized dynamic fuzzy neural network to construct a quantile neural network electricity price model for obtaining the predicted daily electricity price condition quantile. Finally, the kernel density function is used to convert the predicted daily electricity price condition quantile into the predicted probability density curve to realize short-term electricity price probability prediction. The data of the electricity market of the city of Dayton, Ohio in the United States is used as an example. The experimental results demonstrate that the proposed method can effectively improve the accuracy of short-term electricity price forecasting.