The Analysis Of Special Short-Term Load Forecasting Based On Local Wave

This topic is the main content of the science and technology project "The Study and Analysis of the Electrified Railway Short-term Load Forecasting in Xinzhou", hosted by Shanxi Electric Power Company. The electrified railway load, as an important factor influencing the local load forecasting precision, is considered as a huge difficulty for the local work of the load forecasting. In view of the whole network load and electrical railway load of Xinzhou power grid, the analysis of load characteristics is analyzed. while the study of special load short-term forecasting is proposed for the electrified railway load. It may provide a theoretical basis for planning generation and transmission scheme of the power grid.Firstly, with the composition of the special load as the starting point in this paper, the load data of the area is EMD decomposed and restructured according to the local wave method. Through the time-frequency analysis and approximate entropy calculation of the components, the type of the decomposed component is determined. On the basis of the volatility and regularity, the probabilistic forecasting model based on the error statistical analysis and Extreme Learning Machine model are used to predict the components. This provides the theoretical support for the local wave prediction model.Secondly, according to the actual demand of electric power company, the load characteristics of electric railway is analyzed. Based on this evidence, the short-term load forecasting of electrified railway is carried out using the probabilistic models, which reduced from the traditional model at13.23%to8.57%now. It can effectively solve the problem of strong impact and untidy of the electrified railway load, provide important data protection for improving prediction accuracy of the power grid. At the same time, the traditional deterministic load forecasting research work is extended to obtain a probabilistic prediction interval at a certain confidence level, which provide a theoretical basis for the staff to make reasonable decisions.Finally, with the actual situation in Xinzhou, the short-term load forecasting of this area is performed by using the local wave prediction model, the average relative error is2.2%which has reached the related requirements for short-term load forecasting of the State Grid. In addition, the short-term load forecasting of the same historical data is carried out by the global wave prediction model (ELM model and BP neural network model). Among them, the average relative error of ELM model is2.66%, and the BP neural network model is5.67%. Comparative analysis of the results showed that the local wave prediction model which has better prediction capability is feasible.The project has passed the acceptance and identification of the relevant departments, and has been put into practice in the local power company with good results. This subject with a certain promotional value may also study and analysis the special load which has the similar characteristics.