基于灰色RBF神经网络瓦斯涌出量预测

Gas Emission Prediction Based on Grey RBF Neural Network

影响瓦斯涌出量的各种因素往往存在高度非线性,神经网络适合处理高度非线性数据,但样本数据随机性往往容易影响神经网络模型的预测精度,所以将改进的灰色模型引入进来弱化样本的随机性。又考虑到瓦斯预测上,RBF神经网络比BP神经网络更有优势,通过做差值将改进的灰色GM(1,1)模型和RBF神经网络结合起来,建立灰色RBF神经网络模型。仿真实验证明,灰色RBF神经网络模型比RBF神经网络模型的预测精度更高,达到了较理想的预测效果。

Various factors of gas emission are highly nonlinear and neural network is suitable for the highly nonlinear data processing. However, the randomness of sample data always affects the prediction accuracy of neural network models, so an improved grey model was introduced to reduce the randomness.In gas prediction, RBF neural network has more advantages than BP network. Combined with improved grey model G M（1,1）and RBF neural network by calculating difference, a grey-RBF neural network model was established. The simulation experiment proved that the grey-RBF neural network had higher prediction accuracy than the RBF model, which has achieved a better prediction effect.