GM-PSO based dynamically forecasting and compensation of liquid steel temperature

Accurately forecasting the temperature of liquid steel directly affects the quality of steel and continuous casting operation. To solve the problem of real time temperature forecasting and compensation of dynamic liquid steel temperature measuring system, a forecasting model based on the over-damped second-order system theory is developed applying the collected raw data of the liquid steel´s temperature, in which a grey model based particle swarm optimization (GM-PSO) is presented for solving both the small size samples problem of the raw data and excluding the unreliable data. Then both the raw data and the generated data are employed to conduct the model parameter identification. Simulation results show that the proposed method performance better than the traditional genetic algorithms based ones in considerably improving both the forecasting accuracy for small sample and compensation accuracy.