基于变权重RBF组合模型的煤与瓦斯突出强度预测

Prediction of Coal and Gas Outburst Intensity Based on Variable Weight RBF Combination Model

为更准确预测煤与瓦斯突出强度,在组合算法和径向基函数(RBF)神经网络的基础上,建立变权重RBF组合模型。首先,选取最具代表性的3种单项模型:BP神经网络、支持向量回归机(SVR)、免疫遗传算法(IGA),分别建模后对样本序列进行预测,并重构预测结果数据。以重构后的预测序列为输入层,突出强度为输出层,对变权重RBF组合模型进行训练,获得各单项模型的动态权值,从而建立动态变权重RBF组合模型,最后对突出强度进行预测。结果表明:变权重RBF组合模型预测结果的平均相对误差为2.621 2%,优于各单项模型、定权重组合模型以及数据不重构组合模型。

It was held by the authors that in order to predict the intensity of coal and gas outburst more accurately, a variable weight RBF combination model should be built and used on basis of combination algorithm and RBF neural network. For this end, the most representative three kinds of single model were selected which were BP neural network, support vector regression （SVR） machine, IGA-BP. They were used to predict the sample sequence respectively after modeled, and made prediction results were used for data reconstruction. The predicted sequences after reconstruction were used as input layer and outburst intensity as output layer to train variable weight RBF combination model. The dynamic variable weight combination of RBF model has been built through obtaining dynamic weight of each single model, and then to predict outburst intensity. The results show that average relative error of the prediction results by model is 2.621 2%, much better than that by each single model, the weight combination model and data not refactoring combination model.