Research On Transmission Line Icing Prediction Based On Support Vector Machine

Overhead transmission lines are an important part of the power grid,but in alpine regions,overhead transmission lines are affected by ice cover,which is one of the important reasons for the safe and stable operation of the power grid.Over-icing can cause wire dancing,line breakage,tower collapse and even casualties.Therefore,the study of transmission line ice prediction has important theoretical significance and practical engineering value.This paper firstly describes the background and significance of the research on transmission line icing,analyzes the mechanism of line icing,introduces the conditions of icing formation,types and influencing factors,and also conducts literature analysis on the status of line icing research at home and abroad,analyzes the monitoring technology research of icing and icing prediction model research;secondly,designs the icing monitoring information collection device,which can realize the icing parameters such as ambient temperature,ambient humidity,ambient wind speed information,icing prediction model research,The second is the design of an ice monitoring information acquisition device,which can realize the acquisition of ice parameters such as ambient temperature,ambient humidity,ambient wind speed information,tension and inclination angle,and the testing of each component unit.At present,there are mainly neural network models and other machine learning models for transmission line ice-cover prediction model research,because the training of neural network needs a large number of sample data support,and there is a model construction complex training slow prediction accuracy is affected by the data set,so for the above shortcomings proposed a support vector machine regression(Support Vector Regression.Therefore,a support vector regression(SVR)based transmission line ice coverage prediction model is proposed to address the above disadvantages,which can significantly reduce the dependence on data and is more suitable for small sample data prediction.Then further proposes to optimize the hyperparameters and regularization constants of SVR using Beluga whale optimization(BWO)algorithm,and proposes the following improvements for BWO:improvement of nonlinear balance factor based on Sigmoid function,population preference strategy,improved whale colony learning strategy,and elite learning strategy.Secondly,in order to verify the performance of Beluga whale optimization(IBWO),we combine Particle Swarm Optimization(PSO),Grey Wolf Optimization(GWO),Improved Grey Wolf Optimization The algorithms are compared and analyzed with Particle Swarm Optimization(PSO),Grey Wolf Optimization(GWO),Improved Grey Wolf Optimization(IGWO)and BWO,and 17 test functions are used for the optimization test.Finally,the line ice thickness prediction model based on IBWO-SVR is established and compared with the SVR models of other algorithms to further verify the superiority of the proposed model.