Transmission line icing prediction based on data driven algorithm and LS-SVM

Serious icing on the transmission line leads to great damage to the safety production of the grid. Owing to the inability of mathematical models describing line icing and meteorological condition to present the relation between ice thickness and micro meteorological condition plus the complexity and low accuracy of the intelligent icing prediction algorithm based on the global data, on-line prediction cannot be realized. In view of the above problems, the concept of data driven is adopted to propose an icing prediction model based on the data driven algorithm and least squares-support vector machine (LS-SVM). The icing sample data are dealt with as vectors. This method makes optimization selection of icing historical data based on the k-vector nearest neighbors (k-VNN) method and realizes fast modeling with LS-SVM with the advantages of small sample, fast training and strong generalization ability. The case study shows the effectiveness and correctness of the proposed method. ©2014 State Grid Electric Power Research Institute Press