Research Of The Icing Prediction Model Of Transmission Line Based On Data Mining

In recent years, frequent continuous low temperature, snow and other weather phenomena caused icing to occur frequently in some areas. Icing disaster affected the stable operation of transmission lines seriously, which caused great harm. Currently, the icing on the monitoring system has been relatively mature, but technique on icing forecasts and early warnings are not sound and mature which cannot be applied in practice. Therefore, the study of transmission line icing thickness prediction model and icing level forecast model for future icing thickness and icing grade prediction has some theoretical and practical significance.First, the paper analyzed the icing mechanism and researched meteorological factors’ impact for icing. Meteorological factors included ambient temperature, humidity and ambient wind speed.Secondly, C4.5 algorithm have a visual decision tree and production rule,easy to understand and explain, and a relatively high accuracy rate to make up for the fuzzy inference system. C4.5 algorithms can deal with categorical attributes, it just so happens that fuzzy inference system input and output values are fuzzy which can be viewed as categorical attribute features. And fuzzy inference system also remedies the shortcomings that C4.5 algorithm output attributes cannot be numeric attributes. Therefore, this article combined C4.5 algorithm with fuzzy inference system to form a new inference algorithm,namely, C4.5- fuzzy inference algorithm; and the article proposed a new transmission line icing forecast model based on this algorithm: C4.5- fuzzy inference prediction model. The model was validated and compared with fuzzy inference system prediction model, verification and comparison of results showed that the absolute error of forecast ice thickness less than the actual ice thickness, and the part of less is 2mm, and it is smaller than the all fuzzy system prediction model. It shows that the model is feasible and has higher accuracy;Third, it proposed method of divided ice thickness grade, and designed transmission lines icing forecast model based on C4.5 algorithm. The step is:data preprocessing, the ice thickness is divided into several levels in certain range, using C4.5 algorithm to generate decision tree and generating rules, at last, icing level prediction model were formed. Using cross validation method to verify the performance of the model and comparing the experimental results with the Naive Bayes(NB) and estimation techniques based on nuclear naive Bayes(NBK), the results shows that the model is reliable and has good performance, it can be used for transmission line icing level forecast.