Research On Short-term Electricity Price Forecasting In Smart Grid

Electricity price forecasting is the core of the whole power market. Electricity price fluctuation influences the flow and distribution of all kinds of resources in the power market, reflects the relationship between supply and demand of power commodity in the short term. Accurate short-term electricity price forecasting has important significance, it will help market participants to formulate reasonable competition strategy and achieve the minimum revenue with the maximum risk. It is an important work of the participants and one of the problems to be solved in the power market short-term electricity price trading.This paper is based on the historical electricity price data in the power market, founding the volatility and periodic characteristics and analyzing the influence factors of electricity price. It has a significant role for improving the electricity price forecasting accuracy to fully grasp the electricity characteristics and influence factors.Short-term electricity price forecasting ensures the maximum benefit of all the participants. Due to the market clearing price has strong randomness and volatility, and Hilbert Huang transform(HHT) has advantages over processing the non-linear and non-stationary signal, it can get the signal’s time-frequency distribution characteristics. So HHT has been widely applied in the field of power system, such as the detection of power quality, harmonic analysis, and so on. And HHT method is very suitable to apply in the electricity price forecasting research, therefore this paper proposes a combination forecasting model of electricity price based on HHT. For the model mixing effect of HHT method, using the finite difference method to improve it. After pretreatment of the original electricity price data, using the difference method of empirical mode decomposition, the electricity price sequence is decomposed into several intrinsic mode function components and a remainder, then according to the different frequency components to establish the different respectively forecasting model, and we add up the prediction results of each component to obtain the final prediction value. The model uses the actual data of PJM(Pennsylvania- New Jersey- Maryland) power market in the United States to test. Using respectively the actual data to test all the models built with each single algorithm, the simulation results indicate that the short-term price forecasting model established in the paper has higher prediction accuracy.