Abstract: This paper made in-depth research and analysis about gas pressure relief range of different seams' overlying near layers in part coal mines of Luliang,then proposed the model of the BP neural network by the genetic algorithm in terms of forecasting overlying strata damage extent.Through the actual comparison of training prediction caving zone and fractured zone in exploitation of twenty mines with the old and new model,the optimized network model showed faster convergence,less error and better fitting effect.