Study On The Analysis And Forecasting Methods For High Randomness Electric Load

The continuous development of random sequence prediction algorithm andtheory provides a solid theoretical foundation for load forecasting. Electric loadseries is a kind of random sequence. The theory and method of random sequencetechnology can be also applied to the high stochastic power load forecasting in thesmart grid environment.This paper introduces the theoretical basis and basic algorithm of randomsequences forecasting, including the definition of the random sequence prediction,statistical description, basic principle, classification, basic procedure, forecastanalysis, the commonly used prediction method and its advantages anddisadvantages.Then separately analysis the randomness of wind power, bus load andhigh-speed rail from the regularity of the typical daily load and week load curve, andsummarizes their characteristics.This paper analyzes the output power characteristics of wind power andinfluencing factors such as different heights of wind speed, wind direction. Focusingon the characteristic of wind power includes fluctuation, intermittency andinfluenced by meteorological factors such as wind speed, short-term wind powerforecasting of the echo state network model based on genetic optimization is putforward. Example show that the forecasting accuracy of the method is superior toother methods.Bus load historical data preprocessing and the selection of similar days are thepremise and foundation for bus load forecasting. Aiming at the characteristic thatbus load characteristic is varied with different load types, a chaos optimizationcombination forecasting model based on multiple artificial neural networks ispresented. The model has better adaptability and robustness in different types of busload forecasting.Determined value forecasting method has lower accuracy in high speed rail loadforecasting, which exists obvious limitations. Point-to-point times ratio method havebetter forecast effect than the average method, but the accuracy of both is nothigh.Aiming at high speed rail load characterized stochastic fluctuations,nonparametric probabilistic interval forecasting method is put forward. It achievesthe prediction results of intervals with different confidence and alleviates the influence on power grid caused by the high speed rail impact load more scientificallyand reasonably.Thesis research results show that one by one analysis the load characteristics ofthe different types’ high randomness power load, and establish the correspondingprediction model, can improve the accuracy of power load forecasting.