Forecasting of coal seam gas content by using support vector regression based on particle swarm optimization

Accurately forecasting coal seam gas content is important for coal mine safety and energy production, but it is quite difficult and complicated due to the nonlinear characteristics of gas content and lack of available observed data set. Recently, support vector regression (SVR) is being proved an effective tool for solving nonlinear regression problem with small sample set, because of its nonlinear mapping capabilities. Nevertheless, it has also been proved that the prediction precision of SVR is highly dependent of SVR parameters, which usually are determined empirically or by lots of time-consuming trials. In present works, we introduced particle swarm optimization (PSO) serving as a method for pre-selecting SVR parameters. PSO is motivated by social behaviors of organisms. It not only has strong global searching capability, but also is very easy to implement. Based on SVR and PSO algorithms, we proposed a forecasting model of coal seam gas content. Where, an SVR model with Radial Basis Function (RBF) kernel was used to facilitate the forecasting, and PSO is employed to optimize the hyper-parameters of SVR model. Afterward, a procedure was put forward for forecasting coal seam gas content, and a data set observed from a coal mine in China was used to test the performance of proposed PSO-SVR model, which was compared with Artificial Neural Network (ANN) model and normal SVR model. The experimental results show that the PSO-SVR model can achieve greater forecasting accuracy than the ANN model and the normal SVR model, especially under the circumstances of limited samples.