Insulator ESDD forecasting under complex climate conditions on the basis of LSSVM

The paper, based on least squares support vector machines (LS-SVM), builds an intellectual prediction model whose input variables are temperature (T), relative humidity (H), wind velocity (WV), air pressure (P), rainfall(R) and equal salt deposit density (ESDD) measured one day before and the output variable is ESDD observed on the same day of the climate data, which are all provided by ¿optical sensor system for the ESDD monitoring of transmission equipment¿ (OSSEMTE). In this model, the non-sensitive loss function is subrogated by quadratic loss function and the inequality constraints are substituted by equality constraints. Consequently, quadratic programming problem is simplified as the problem of solving linear equation groups, and the SVM algorithm is realized by least squares method. Through grid search method, the optimal parameters of LS-SVM are selected automatically, which has improved the speed and accuracy of the forecasting. Compared to the BP simulated results, the predicted ESDD of the model are closer to the on-line measured ones. Therefore, the model presented supplies a feasible thread for the computerization of pollution distribution map of power network.