Research Of Combined Forecasting Model In Short-Term Electric Power Load Forecasting

Short-term electric power load forecasting usually forecasts the load indicator of a month, a week or a day, as well as the load of 24 hours in one day. To improve the level of power load forecasting technique could help the related departments to deal with the operation mode of electricity grid and the unit maintenance scheduling better, save coal, oil and reduce electricity costs, make a reasonable plan of power supply construction as well. As a result of that, the economic benefit and social benefit of power system will improve. Electric power load forecasting has become one of the main content of a modern power system management.Combined forecasting is a kind of forecasting method which uses the predicting outcomes of some forecast methods with appropriate weights to calculate the weighted average. The core of this paper is to use combined forecasting to improve the accuracy of short-term electric power load forecasting. The main research contents are as follows: Firstly, this paper outlines the background and significance of this research, summarizes the method classification and research focus of short-term power load forecasting. Secondly, the principle of combined forecasting and the weight distribution of the various methods are reviewed in detail, and the characteristics and influencing factors of short-term load forecasting are analyzed deeply. Thirdly, on the basis of analyzing and selecting the data pre-processing algorithm and load forecasting algorithm, the combined forecasting model is built by three stages of virtual forecasting—methods' parameter optimization, filtering of methods and weight optimization. There is a deep study of relevant factors on the impact of short-term load forecasting and a relationship database is established through relationship derivation. Fourthly, on the background of Liaoning electric power load forecasting system which is based on this combined forecasting model, this paper introduces the function modules, architecture and characteristics of this system, goes through model construction, forecasting and outcomes evaluation by the way of case application, and proposes an optimization model of this model.Combined forecasting model is not a simple method of piling up, but a methods combination dynamically built by virtual forecasting before the real forecasting. It displays a variety of internal relations of the forecasted objects by the methods in the combination, and redistributes the weight scientifically in accordance with the virtual forecasting effect, thereby compensates for the lack of a single method, while keeping the advantages of each method. Practice has proved that, combined forecasting model presented by this paper could improve the accuracy of electric power load forecasting which also has a certain theoretical and application value.