Forecasting Coalmine Gas Concentration Based on RBF Neural Network

The characteristic and learning algorithm of the RBF Neural Network (RBFN) were analyzed. A model with 2 inputs and 1 output was proposed to forecast the gas concentration. The value of sensors on key measuring points under a coalmine were chosen as training samples and testing samples. Then simulations were carried out with Matlab7.0. The results show that the training and testing root mean squared error were 0.0036 and 0.0016 respectively. Finally, simulation of BP Neural Network (BPNN) was also done as a comparison, which indicates that the RBFN has better performance, thus, the model is creditable and feasible.