我国是世界最大的煤炭生产国和消费国，特别是在我国的华北地区，煤炭储量非常丰富，但由于华北地区的水文地质条件复杂，煤炭实际生产过程中事故频发，特别是煤层底板突水事故，一旦发生往往会造成较为严重的人员伤亡和财产损失。因此煤层底板突水预测已经成为煤矿安全生产领域研究的重点。巨野煤田红旗煤矿是典型的华北型煤田，其主要可采煤层3煤层平均厚度5.48 m，实际生产过程中受底板突水威胁严重，在矿井的建设及生产过程中多次出现底板突水。为了对3煤层进行底板突水预测，在分析收集红旗煤矿相关矿井水文地质资料的基础上，选取断裂分维值、取心率、隔水层厚度、单位涌水量、渗透系数、底板含水层总厚度、承压含水层水压共7个因素，作为进行底板突水预测的主要影响因素;以现场实际数据为输入样本，通过灰狼优化算法(Grey Wolf Optimizer,GWO)得到Elman神经网络优化的最佳权重和阈值，分别为18.7482和0.014435，之后建立相应的GWO-Elman神经网络底板突水预测模型;在此基础上通过测试样本输入模型验证，结果准确率达到100%，再用熵值法确定权重的脆弱性指数法进行对比，证明神经网络模型准确度更高，可以用于工程实际。最后，利用所建立的神经网络模型对2个未开采工作面进行了底板突水预测，将预测结果指导矿井实际安全生产。

China is the world’ s largest coal producer and consumer,especially in the North China region of China,where coal reserves are very rich. However,due to the complex hydrogeological conditions in North China,accidents occur frequently during the actual coal production,especially water inrush accidents in coal floor,which often result in serious casualties and coal losses. Therefore,the prediction of coal floor water inrush has become the focus of research in the field of coal mine safe production. Hongqi Coal Mine of Juye Coalfield is a typical North China coal field. Its main recoverable coal seam is No. 3 coal seam,with an average thickness of 5. 48 m. The coal seam is seriously threatened by floor water inrush in the actual production process,and some floor water inrush accidents have occurred many times during mine construction and production. In order to predict the water inrush from the bottom of the No. 3 coal seam,based on the analysis and collection of relevant hydrogeological data of the Hongqi coal mine,seven factors are selected,including the fractal dimension of the fracture,the heart rate,the thickness of the water barrier,the amount of water inflow,the permeability coefficient,the total thickness of the aquifer of the floor,and the water pressure of the confined aquifer,these seven factors are used as the main influencing factors for the prediction of water inrush from the floor. Taking the actual field data as input samples,the best weights and thresholds of Elman neural network optimization are obtained by Grey Wolf Optimizer (GWO),which are 18. 748 2 and 0. 014 435,respectively. Then the corresponding GWO-Elman neural network prediction model for floor water inrush is established. On this basis,the test sample input model is verified,and the accuracy rate is 100% . The entropy method is used to determine the weight of the vulnerability index method for comparison. The neural network model is more accurate and can be used in engineering practice. Finally,the established neural network model is used to predict the water inrush from the floor of the two unmined working faces,and the prediction results will guide the actual safety production of the mine.