Model identification of thermal object based on smooth support vector regression

Usually, all sorts of parameters in real time system will be changed at any moment, so the traditional methods of system identification such as statistics and neural network are not very fit for this kind of complicated dynamic system. Based on the arithmetic of support vector machine (SVM), the smooth method of SVM was put forward and used in regression analysis for system state model, which is obviously superior to neural network methods in system identification. The arithmetic of smooth support vector regression (SSVR) was achieved with Matlab 6.5, and used in model identification of state in turbine system. The result of simulation indicated that SSVR has faster operation speed and higher precision, and is very fit for the complicated turbine state model system, which effectively extends the application of support vector machine.