Research On Day-ahead Electricity Price Forecasting Method In Electricity Market With High Proportion Of Renewable Energy

The development of electricity marketization is the development trend of power system construction in various countries around the world.The day-ahead market is the main channel for spot trading of electricity.At present,renewable energy power generation such as wind energy and photovoltaics is developing rapidly,and its participation in day-ahead market transactions is increasing.Due to its considerable randomness and extremely low marginal cost of power generation,the fluctuation of day-ahead market price becomes quite severe,which greatly increases the difficulty of forecasting.The accurate prediction of electricity price has also become an important demand for power generation enterprises,electricity enterprises and power market regulators.How to improve the accuracy of electricity price forecasting in the electricity market with high proportion of renewable energy participation has become an urgent problem to be solved.To this end,the main research contents and conclusions of this thesis are as follows:(1)The characteristics of electricity price in electricity market with high proportion of renewable energy are analyzed,and the influencing factors of electricity price are analyzed.The correlation between different factors and electricity price is judged by calculating the Pearson coefficient.(2)Considering that electricity price forecasting will use a large amount of data,in which irrelevant data will cause over-fitting in model training,which will not only reduce the forecasting speed,but also affect the accuracy of forecasting.Aiming at the problem of data redundancy,a similar day screening method based on RF-TOPSIS-GRA is proposed.The RF method is used to calculate the variable importance score of each feature and obtain its objective weight coefficient.The features with higher scores were selected as the consideration factors for the screening of similar days,and the similar days were screened by TOPSIS-GRA method.After screening similar days,the redundant information is effectively eliminated,and a data set with strong correlation with the day to be predicted is obtained.(3)Electricity price has the characteristics of volatility and peak.If the fluctuating signal is directly input into the forecasting model,the forecasting accuracy will be low.According to the idea of ’ decomposition-prediction-reconstruction ’,the CEEMD method is used to decompose the input price series to obtain a series of intrinsic mode functions and a residual component,which are used as the input of the LSSVM regression prediction model.The simulation results show that this method can effectively reduce the prediction complexity and improve the prediction accuracy.(4)When LSSVM forecasting model is used to forecast electricity price,different choices of parameters in the model will have a great impact on forecasting performance.This thesis presents an ISSA algorithm to optimize the parameters of prediction model.The performance test results show that the ISSA algorithm has the best convergence speed and convergence accuracy compared with other algorithms.The simulation results show that the proposed ISSA-LSSVM forecasting model can effectively improve the forecasting accuracy of electricity price.