The Application Research Of Gene Expression Programming In Short-Term Electricity Price Forecasting

The electricity marketization have become the main trend of power industry reformation in the world of today. Electricity price is the core content in the electricity market environment, accurate electricity price forecasting for all participants is of great significance. How the electricity market in accordance with the relevant historical data to forecast the future market clearing price accurately is a important subject and the research focus of the electricity market.Firstly, based on large amount of related research and relevant documents, this paper discusses the merits and shortcomings of some current forecasting methods. And then deeply introduce the formation mechanism, the influence factors and the characteristics of the electricity price in the electricity market, and get ready for modeling.The main content of this paper is short-term electricity price forecasting, by using the methods of the load forecasting for reference, an attempt to apply the GP and GEP to the short-term electricity price forecasting, thus it establishes the following three models, in comparison to the traditional short-term electricity price forecasting, the accuracy of the prediction has improved : (l)the GP model. In the process of modeling, the correlation analysis technique is used to determine the input factors, moreover, similarity searching technique is applied to build the training samples. so this model can improve the accuracy and rate of the prediction. (2)the GEP model. Compared with the GP, GEP has the higher efficiency. Combined with the own characteristics of the electricity price, it buildtwo short-term electricity price models based on GEP--the sliding windowforecasting model and the considering load forecasting model. And then uses the relevant data to test this model, it has achieved a better result. (3)the Wavelet-GEP model. In view of the advantage of processing the non-stationary time series by using the wavelet analysis technique, Firstly, by use of wavelet analysis, the history sequence of the price is decomposed and reconstructed, and the general picture sequence in low frenquency band and the detail sequence in high frequency band are obtained.then, on this basis the different GEP models for each sub-series are established respectively in diveed period. Finally, the forecasting results of sub-series are directly added and taken as the final forecasted result. And then uses the relevant data to test the models, the accuracy of the prediction has improved obviously in the periods of the larger fluctuation. This paper describes in detail the three modeling principle, and then uses the relevant data to test these models, it has achieved a better result. It demonstrates that these models are feasible and effective.Finally, this paper gives a summary and outlook, and points out the issues that require further reaserch.