基于GA-SVM的矿井涌水量预测

Mine water inflow prediction based on GA-SVM

矿井涌水量的准确预测对预防矿山透水事故的发生至关重要,提出利用GA优化的SVM模型(GA-SVM)来实现矿井涌水量的短期准确预测。该方法利用GA的自动寻优功能寻找SVM的最佳参数,提高了预测的准确率。首先,利用微熵率法求矿井涌水量时间序列的最佳嵌入维数和延迟时间,进行相空间重构。其次,采集义煤集团千秋煤矿2011—2015年实际涌水量的时间序列,利用GA-SVM模型对最后12组数据进行预测,其预测平均绝对百分比误差仅为0.92%,最大相对误差为2.62%。最后,与PSO-SVM和BP神经网络预测进行对比,结果表明GA-SVM优化模型适用于矿井涌水量的预测并且预测精度较高。

The accurate prediction of mine water inflow is very important to prevent mine water inrush accident. In this paper, the SVM model optimized by GA（GA-SVM） was put forward to realize the short term and accurate prediction of mine water inflow. In this method the automatic optimization function of GA was used to find the optimal parameters of SVM, which can improve the accuracy of prediction. Firstly, the optimal embedding dimension and delay time of mine water inflow were obtained by using the entropy ratio method, then the phase space was reconstructed. Secondly, the actual time series of water inflow from 2011—2015 in Qianqiu coal mine of Yima Coal Group Company were collected. GA-SVM model was used to predict the final 12 sets of data, the mean absolute percentage error was only 0.92%, the maxmum relative error was 2.62%. Finally, compared with the PSO-SVM and BP neural network method, the prediction results show that the proposed GA-SVM optimization model is suitable for the prediction of mine water inflow and the prediction accuracy is higher.