A New Combination Forecasting Model for Concentration Prediction of Dissolved Gases in Transformer Oil
Xiangjun Zeng · Proceedings of the CSEE · 2008
To improve fault prediction for power transformer,a new model of combination forecasting with optimal weights was proposed.Four existing methods namely Gray theory,BP neural network,Genetic Algorithm and Kalman filtering arithmetic were adopted synthetically to forecast the concentration and development trend of dissolved gases in transformer oil.Each optimal weight of the four methods was calculated firstly according to the principle of least error sum of square and their prediction error ratios.Then,the optimal combination forecasting model was formed based on the optimal weights.In terms of the model,concentration of dissolved gases in transformer oil would be obtained.In this way,the advantages of the four methods were concentrated and a maximum forecasting precision will be gained.Simulation results show that the proposed forecasting algorithm is feasible and dependable.Besides decreasing prediction error and improving forecasting precision greatly,it also provides a new way to solve other data forecasting problems in power system.