基于卡尔曼滤波及神经网络的瓦斯涌出量预测

Prediction of gas emission based on Kalman filter and neural network

为解决当前瓦斯涌出量预测过程中存在的因影响因素过多、预测指标与瓦斯涌出量之间非线性关系及其自身时变性等特点而导致预测精度降低的问题,采用SPSS因子分析法对瓦斯涌出量影响因素进行分析降维,将得到的预测指标由构建的BP神经网络与卡尔曼滤波相结合的瓦斯涌出量预测模型进行预测。研究结果表明:采用因子分析的方法能够有效筛选瓦斯涌出量影响因素,并得到了预测指标,降低了预测模型预算复杂度;经过BP神经网络与卡尔曼滤波耦合瓦斯涌出量预测模型,其预测精度明显高于直接采用神经网络模型预测的结果,预测性能明显改善,其平均误差仅为2.75%,表明所采取的瓦斯涌出量预测方法是可行和有效的。

In order to solve the problem of low prediction accuracy caused by too many influencing factors,non-linear relationship between prediction index and gas emission and its own time-varying in the process of gas emission prediction,SPSS factor analysis method was used to analyze the influencing factors of gas emission and reduce the dimension,and the prediction index was predicted by the gas emission prediction model combining BP neural network and Kalman filter.The analysis results show that the factor analysis method can effectively screen the influencing factors of gas emission,obtain the prediction index,and reduce the complexity of the prediction model.Using the gas emission prediction model combining BP neural network and Kalman filter,the prediction accuracy is obviously higher than that of the neural network model,and the average prediction error is only 2.75%,which shows that the gas emission prediction method is feasible and effective.