小波神经网络煤矿井下涌水量影响因素挖掘与预测

合理预测矿井涌水量,及时有效制订排水方案,是保证井下人员安全生产及经济开采的需要。根据煤矿井下涌水量影响因素的不确定性,选取的平均影响值通过误差反向传播网络,确定涌水量各影响因素相关程度,实现涌水量影响因素的筛选。采用小波方法对涌水量进行分析,发现其相同的季节大致有相同的走势。依据小波中的时间聚合因子,建立了基于时间序列的小波神经网络短时预测模型,对某矿涌水量进行分析。结果表明:涌水量预测最大绝对误差为4.179 7m3/d,最大相对误差为2.079%,最小绝对误差为0.000 5 m3/d,最小相对误差为0.000 2%;平均相对误差绝对值为0.482 3 m3/d,平均相对误差为0.285 7%,正确率平均达到了99.714 3%。预测指标满足了矿井涌水量排水的要求。

Excavation and prediction of factors affecting water inflow in mine in wavelet neural-networks

Reasonable prediction of mine discharge and timely and effective development of the drainage plan presuppose ensuring the safety of underground mine production and economic needs.Aimed at addressing uncertainty factors of underground coal water inflow,this paper proposes a method designed for determining relevance of the water yield of the factors using affection of the value combining average error back propagation network to achieve the screening water yield factor.The paper introduces the study on the water yield by wavelet and finds roughly the same trend of the water yield in the same season.This method builds on fact that wavelet in the time factor can compensate for polymerization in the presence of artificial neural network input sync issues limit the instantaneous and involves development of short-term prediction model for wavelet neural network based on time series.Data analysis shows maximum absolute error of 4.179 7 m3/d,the maximum relative error of 2.079 percent,the smallest absolute error of 0.000 5 m3/d,the smallest relative error of 0.000 2%;average relative absolute error for the 0.482 3 m3/d,the average relative error of 0.285 7%,the average correct rate reached 99.714 3%.Prediction indexes meet the requirements of the mine drainage.