Study On Power Load Forecasting Method Based On Ant Colony Algorithm

Power load forecasting is the important basis for the production of electricity, there is no guarantee in any method that can get satisfied result under any circumstances. Combination forecasting can make use of information provided by single forecasting models. We got combination forecasting model with an appropriate combination form to improve the forecasting accuracy. Based on previous studies, the author use combination forecasting and ant colony algorithm to solve power load forecasting, the combination forecasting model for middle-term and long-term power load forecasting has been set up in this paper.There are two keys which are the most important in combination forecasting. First is the selection of single forecasting model. Another one is how to solve weighted average coefficients. The author selected method of load density, method of power elasticity coefficient, move average, index smoothness, regression, grey forecasting, etc. as single forecasting model. This article formulated objective function according to least squares criterion to solve the weighted average coefficients and select the single forecasting model by the weighted average coefficients. We eliminate the single forecasting model with smallest weighted averaged coefficient. Then the filtration of single forecasting model is qualitative and quantitative.The objective function is a non-liner program which can't be solved by a universal method. Ant colony algorithm mainly solved the problem of discrete combination optimization. The author improves ant colony algorithm to solve the problem of consecutive optimization. The author solves weighted average coefficients by improved ant colony algorithm.In this paper, the author combines the combination forecasting and ant colony algorithm to solve the problem that the weight coefficient is determined difficultly in the combination forecasting. These made power load forecasting become more reasonable. It has instructional significance on power load forecasting. We proved its practicability and veracity by applying it to Fu Ling. The results show this method has some value of application.