The Development On Intelligent Recognition Software Of The Icing Thickness Of Transmissionline

The great harm to the grid and huge losses to the lives and property of people caused by large frozing of the south In early 2008.The main reason is the thickness of transmission line icing is too large,while ice thickness has not been predicted.To this end,the research of smart identification model for ice thickness of transmission was selected by the paper based on technology project of grid company of Shanxi province named "the measures analysis and application to reduce the amount of transmission line icing".Through the research of this model,the necessary theoretical foundation of the forecast for the icing thickness was provided,the accuracy of ice thickness prediction of transmission line was improved and the direct and potential hazards caused by icing on the grid was reduced,so the project had a good theoretical and practical significance.Firstly,the paper studied the main factors which effected the ice thic kness of transmission line.The main factors concluded temperature,humidity, rainfall,wind,direction of wind,wire suspension height,altitude,geography and so on.And the results of research showed that temperature,humidity,rainfall, wind,direction of wind were the main factors which effected ice thickness of the transmission line.Secondly,two prediction models were established on the MATLAB software platform on the base of neural networks,which were respectively based on generalized regression neural network (GRNN) and ELMAN neural network,then conducted simulation the two models.The results of simulation showed that the error of GRNN prediction model was 0.0018 and the error of ELMAN neural network prediction model was 0.0631.Based on th rerults that the GRNN model was better than the ELMAN neural network model on prediction precision,so GRNN model was better suited to predict the thickness of transmission line icing than ELMAN model.Thirdly,based on the disadvantages of this paper neural networks included too dependent on the initial value,the phenomenon of existed learn, easy to fall into local minimum during training,the paper introduced the support vector machine (SVM) theory into the ice thickness prediction model,and built SVM-based prediction model on the MATLAB software platform.Then based on the parameter of SVM were difficult to be selected,genetic algorithm (GA) and particle swarm optimization (PSO) were introduced into the SVM model,GA-SVM and PSO-SVM model were established.The parameter optimization results showed that the optimal parameter combination (c, g) were respectively (6.2389,1.6113) and (9.3845,0.01),while conducted simulation to SVM prediction model token advantage of the parameters.The simulation results showed that, the error of GA-SVM model was 0.00109729 and the error of PSO-SVM model was 0.00147842,so GA-SVM model was better suited to predict the thickness of transmission line icing than PSO-SVM model.Fourthly,wavelet neural network (WNN) model was built on MATLAB platform based on the conbination of wavelet theory and neural network algorithm through brought the wavelet algorithmto into the training process of neural networks,and the results of simulation showed that,the error of WNN model was 0.2618.Finally,the five models are analyzed and compared by the paper,then the results showed that,the prediction error of GA-SVM model was the most little and the accuracy was highest,therefore GA-SVM model was most suitable for forecasting the ice thickness of transmission line.