Study On The Model Of Forecasting Icing Growth On Transmission Lines

The frequent occurring of global extreme climate has brought the state grid in ourcountry more and more serious harm caused by exacerbating and extending icing disasters.Sponsored by the national natural science fund project "the theoretical and experimentalstudy of transmission conductor icing prediction model”, this paper selected the study on themodel of forecasting icing growth on transmission lines as the subject. Through the study ofthis subject, the prediction accuracy of ice thickness can be improved, it also providesnecessary theoretical basis for ice thickness prediction, and through it we can instantly knowvery well about the conductor icing condition on the transmission lines, which has importantpractical significance and guiding function on the stable and reliable operation of power grid.Firstly, in this paper, the causes, conditions and process of the formation of conductoricing are analyzed, the types of conductor icing is also introduced, at the same time, the mainfactors that influenced the transmission line icing and the harm of transmission line icing areboth studied in this paper.Secondly, the fuzzy system and the neural network are both a kind of simulation tohuman’s intelligence; both of them can establish a non-linear model when they are given asystem’s input/output data. For the complexity and the nonlinear structure characteristics ofthe conductor icing, and according to the related principles and learning algorithm in fuzzysystem and the neural network, the ice growth predicting model based on the fuzzy systemand the neural network are respectively built, then the validity of two methods are proved bythe MATLAB simulation analysis and the forecast results. Prediction results show that onlydepend upon the fuzzy system or the neural network which can’t achieve very good predictioneffect, therefore, a prediction model for icing growth based on fuzzy neural networks is builtin this paper, which combined the advantages of fuzzy theory and neural network. This modelhas a higher reliability since it overcomes the shortcomings that the fuzzy system itself has noadaptive ability and the neural network can’t express the human brain reasoning function verywell.Lastly, through comparing the simulation results of the four kinds of growth predictionmodel built in this paper, the results show that: the prediction accuracy of the icing growthprediction model based on the fuzzy neural network is highest and the error between theactual ice thickness and its predicted ice thickness is less than2.5mm; the pred iction accuracy of the RBF network comes second and its prediction error is less than4mm; incontrast, the prediction accuracy of BP network is the worst and its prediction error is lessthan6mm. So the fuzzy neural network is proved to be the most suitable prediction modelfor conductor icing growth on transmission lines. And analysis results truly reflect the on-siteicing conditions of power transmission lines.