Prediction Method of Dissolved Gas Concentration in Transformer Oil Based on LSTSVR Model

To improve fault prediction for power transformer, a new model of the least squares double support vector regression machine (LSTSVR) model is adopted to predict the concentration and development trend of dissolved gases in transformer oil. This method is based on the double support vector regression machine, uses the idea of least squares and replaces the inequality constraints in the traditional support vector regression machine with equality constraints, which reduces the computational complexity and makes the prediction more accurate. Simulation results show that the proposed LSTSVR prediction model is feasible and dependable. The predicted values of LSTSVR model fit well with the measured values by the foregoing analysis. Compared with LSSVR, the MPE, MRE and RMSE of LSSVR are much less than those of the least squares support vector regression machine (LSSVR) prediction model. Besides decreasing prediction error and improving forecasting precision greatly, it also provides a new way to solve other data forecasting problems in power system.