Research On Mid-long Term Load Forecasting Based On Semi-parametric Statistical Model

The power industry is a major basic industry in the energy field of the state.Electricity for China’s economic construction, national security, and social stabilityhas a crucial role. With the rapid development of power industry, the acceleratedpace of the national smart grid, the study of theory and technology of power loadforecasting has become important increasingly. High-precision power loadforecasting has a direct and significant economic benefits and social benefits, forthe economic optimization of the development of power projects to develop theeconomy of electricity at reasonable deployment plan, control of power systemeconomic operation, the electricity market analysis, protection of production andpower consumption.The complex characteristics of long-term power load determine its impactfactor of uncertainty, there are many factors that are difficult to quantify, such as:macroeconomic policy, economic development. These factors often existcorrelation. Therefore, in order to grasp the trend of the electricity load demand, weneed to consider the complexity of influencing factors, and overcome thecollinearity between the complex parameter set to the more scientific andreasonable description of the power load characteristics.There are shortcomings of traditional methods for the medium-term andlong-term power load forecasting. In this paper will introduce the semi-parametricstatistical theory for the work of medium-and long-term load forecasting. We builda semi-parametric model based on EEMD, semi-parametric model based onpunishment of least squares, and the model were used in medium-and long-termload forecasting. Finally, verify the model through simulation experiments, on theone hand, to verify the feasibility and availability of the semi-parametric model; theother hand, verify that the model has good potential applications. The maincontents are as follows:（1）Comparative analysis the inadequate and scope of technology forecast,trend extrapolation, regression analysis, the characteristics of time series, the grayprediction method and a new theory of forecasting techniques. pointed out that thesemi-parametric model in the field of medium-and long-term power loadforecasting certain advantages.（2）Analysis of the medium-term electricity load characteristics to build apredictive model based on EEMD of semi-parametric model to ensure improvedhandling multi-scale medium-term load sequence, not only to overcome theshortcomings of the traditional model,but also to use of the ARMA modelhigh-frequency noise sequence extrapolation capacity, increase the scope ofapplication of the semi-parametric model.（3）Analysis of long-term power load characteristics, semi-parametric prediction model was constructed based on the punishment of least squaresestimates, and to overcome the traditional forecasting methods suited to the lack ofsubstantial epitaxial forecast to meet the long-term power load forecastingrequirements.