Research On Rational Self-Adaptive Extreme Learning Machine For Electricity Price Forecast

The electricity marketization has become a development trend in the world,and the effective forecast of electricity price plays an important role in the trend.The price of electricity is not only the signal of supply and demand in the electricity market,but also the economic lever to control the electricity market transactions.Reasonably according to market demand to determine the corresponding price directly affect the power market can operate normally.Therefore,how to accurately predict the future market price based on the historical data of the power market is of great significance to all participants in the market.This paper firstly introduces the research background,research significance and research status at home and abroad,secondly it studies and analyzes prediction and related machine learning algorithms,including regression methods,neural networks,SVM algorithms,and thirdly the basic differential evolution and extreme learning machines and their variants have been studied.It was found that although they are widely used in electricity price forecasting,they still have certain defects.In view of the complicated time series of electricity prices,various methods of Extreme Learning Machine(ELM)have been identified as effective prediction methods.However,in the high-dimensional space,the evolutional Extreme Learning Machine is very time-consuming and cannot converge to the best region under the mere reliance on the random search method.At the same time,due to the complex functional relationship,it seems difficult to find some useful mathematical information from the objective function of evolutional Extreme Learning Machine to guide the optimization exploration.Therefore,this paper presents a new similar Differential Evolution(DE)Algorithm to enhance the E-ELM,so as to more accurately and reliably predict the electricity price.The approximate model used to generate the DE experiment vector is the key mechanism that allows simpler mathematical mapping to replace the original but complex functional relationships within a small area.Therefore,the evolutional Extreme Learning Machine whose evolutionary process is often guided by a reasonable search direction is faster and more robust than that just supported by random search methods.Finally,this paper uses some standard test functions to test the proposed algorithm.Experimental results show that the proposed algorithm can effectively improve the performance of evolutionary extreme learning machine.In addition,the proposed algorithm in the price of electricity forecast,compared with other algorithms,using the proposed algorithm in a certain period of time to obtain higher accuracy.