摘要
为了解决矿井涌水量预测难题,在Grid-Search_PSO优化SVM参数的基础上,采用SVM非线性回归预测法,对大海则煤矿1999~2008年7月份的矿井涌水量进行了预测。分析对比SVM回归预测法和ARIMA时间序列预测法预测结果的数据误差,发现SVM回归法预测值与实测值之间的偏差比ARIMA时间序列法要小很多。可见在影响矿井涌水量各种因素值具备的情况下,SVM非线性回归预测所建立的模型能够更准确地预测矿井的涌水量,在矿井安全生产中具有很大的应用价值。
In order to solve the problem of mine inflow prediction, based on optimized support vector machine parameters with grid search algorithm and particle swarm optimization algorithm support vector machine parameters,by the method of using the support vector machine nonlinear regression prediction,forecast mine inflow in July from 1999 to 2008 in Dahaize coal mine. Comparative analysis error data of predicted results between support vector machine regression prediction and ARIMA-time-series,found that predicted values deviation of support vector regression is much smaller than that of the ARIMA-time-series method. It is obvious that support vector machine nonlinear regression prediction model can more accurately predict the mine water inflow and has great application value in the mine safety production under the condition of enough affecting mine inflow factors value.
引文
[1]窦贤明,杨永国,徐伟伟,等.基于遗传算法和BP神经网络的矿井涌水量预测[J].中国煤炭地质,2009,21(10):37-38,66.
[2]陈永新,徐东晶,林宁,等.多项式曲线拟合在矿井涌水量预测中的应用[J].采矿技术,2013,13(4):75-77,93.
[3]黄存捍,冯涛,王卫军,等.基于分形和支持向量机矿井涌水量的预测[J].煤炭学报,2010,35(5):806-810.
[4]王健峰,张磊,陈国兴,等.基于改进的网格搜索法的SVM参数优化[J].应用科技,2012,39(3):28-31.
[5]施龙青,谭希鹏,王娟,等.基于PCA_Fuzzy_PSO_SVC的底板突水危险性评价[J].煤炭学报,2015,40(1):167-171.
[6]臧大进,刘增良,曹云峰,等.基于粒子群优化支持向量机的矿井涌水量预测[J].凯里学院学报,2010,28(6):26-29.
[7]谭大国.BP神经网络在矿井涌水量预测中的应用[J].制造业自动化,2015,37(5):66-68.
[8]王金国,江洪清,高永奎.基于MATLABR的矿井涌水量神经网络预测方法及应用[J].煤炭技术,2004,23(7):67-68.
[9]宫晨.基于ARIMA模型的拉萨降水量时间序列分析与预测[J].北华航天工业学院学报,2011,21(5):16-19.