Research On Early Warning Fusion Method Of Transmission Line Ice Coating Based On DS Evidence Theory

The main characteristics of Yunnan-Guizhou Plateau are complex topography and variable climate.Those lead to serious icing on power transmission line in winter.But transmission line icing process is a time series which faced with uncertainty,mutability,multi-dimension,and nonlinear.It is difficult to find the mathematical model.For the problem of uncertainty and unpredictability in Icing prediction model as statistics type,mechanism and intelligent computational,we establish the information fusion system of transmission line icing warning,and present the model of transmission line icing warning fusion based on DS evidence theory.The main findings are as follows:The excessive micro meteorological factors during icing and the limitations of raw data result in some problems,such as a complex prediction model and the prediction inaccurate results.This paper designs an icing warning fusion model which based on original prediction model.The model includes data preprocessing,icing prediction and decision fusion.The method of data dimension reduction based on PCA simplifies the model of icing process.And this paper also illustrates a method based on smooth filtering which is to be denoise of icing monitoring data.Then this paper proposes the icing prediction qualitative model.According to the relevant literature,we design three icing warning model of multivariate linear regression,fuzzy reasoning and support vector.In this paper,we propose the icing warning fusion model based on DS evidence theory for the varied icing prediction method and differences of characters.In the decision fusion,the various icing warning results as the evidence to fusion according the DS combination rule.Then we can get the final fusion with the judgment method.The arresting characteristics of the model are better results of nonlinear fitting,the lower requirement of integrity of the sample data and the weaker influence of subjective factors.The experiment shows that this model has strong fault tolerance and improves the precision of prediction results,reduces unknown.