HHT-based Analysis Of Influencing Factors Of Electrical Power Load And Forecasting

High-precision electrical load forecasting is an important basis for the safe operation of electrical power system, and also an important guarantee for the economic operation of electrical power system. There for, improving the precision of load forecasting plays a decisive role in electrical power system management. Short-term load forecasting is much instructive to production control of power grid while efficient forecasting methods help to improve the precision of forecasting. This thesis studies important issues existing in short-term load forecasting of power supply enterprises in China, such as screening of influencing factors, analysis of load characteristics, and selection of forecasting model etc. It proposes a combination forecasting model based on Hilbert-Huang Transform and the analysis of influencing factors of electrical power load.Hilbert-Huang Transform is a completely new signal processing algorithm used for processing non-linear and non-stationary signals. It first uses EMD to decompose the original sequence into the sum of IMF components which are at different frequencies and are relatively stationary. After that, conduct Hilbert transform to each component to conclude the time-frequency characteristic of components. It is a completely self-adaptive time-frequency analysis method.This thesis makes an in-depth analysis of the research status and development trend of short-term load forecasting technology at home and abroad and discusses the existing electrical power load forecasting method and its characteristics. It delves into the historical load of grid and meteorological data etc, and sums up the relationship between various characteristic index and various influencing factors based on an analysis of the characteristic index of load.First, preprocess the historical load and meteorological data and use EMD to decompose prediction sample sequence into stationary subsequence of different frequencies. Then use Hilbert transform to calculate the stationary subsequence to conclude the frequency response of each component. The subsequence is more predictable than the original sequence. Based on the analysis of the characteristics of each component and adaptability of the forecasting model, choose RBF neural network model to make a prediction about high frequency component, BP neural network model to make a prediction about intermediate frequency component, time series model to make a prediction about low frequency component. The influence of meteorological factor is taken into consideration in the forecasting model of intermediate frequency component and high frequency component. Add the forecast result of each component together to achieve the final predicted value. Thus a new combination model is obtained.With the practical load data of somewhere in Henan province in 2012 as the forecasting sample, this thesis uses the forecasting model to make a prediction about the load value of the 96 time points in the 24 hours of a day, and contrast it with the actual load value. The prediction accuracy reaches a high level. The result shows that the HHT-based combination forecasting model is conductive to realize the complementary advantages of models. It is of greater rationality.