相空间重构和混沌遗传神经网络融合的矿井涌水量预测研究  

Mine Water Inflow Prediction Research Based on Phase Space Reconstruction and Chaotic Genetic Neural Network

作　　者：邓高[1] 杨珊[1] 江时雨 

机构地区：[1]中南大学资源与安全工程学院,长沙410083

出　　处：《世界科技研究与发展》2016年第5期973-978,共6页World Sci-Tech R&D

基　　金：国家自然科学基金青年基金(51404305);中国博士后科学基金(144760);湖南省科技计划项目(2015RS4060)资助

摘　　要：为了更准确预测矿井涌水量变化,有效防治矿山水害,本文提出利用相空间重构和混沌遗传神经网络相结合的方法预测矿井涌水量。选用C-C算法确定嵌入维数和延迟时间,通过对时间序列进行相空间重构来判断涌水量时间序列的混沌特性。为避免BP神经网络极易陷入局部解的问题,采用遗传算法对混沌神经网络进行参数优化,构建混沌遗传神经网络预测模型。将构建的模型应用于某矿山-100 m水平巷道涌水量的预测,在理论预测时长内预测最大误差为3.38%,表明该方法能够反映短期内矿井涌水量变化的趋势,相比单纯的混沌BP神经网络预测模型,预测精度有所提高,可为矿山企业的灾害防治提供科学的参考依据。In order to predict the mine water inflow accurately,and control water gushing-out effectively,the method of combing the phase space reconstruction with chaotic genetic neural network had been proposed. In order to determine the chaotic characteristics,the phase space of the mine water inflow time series was reconstructed after getting correlation dimension and delay-time with C-C algorithms. To avoid the BP neural network common problem of trapped into a local solution,the genetic algorithm was used to optimize BP network parameters and construct the chaotic genetic neural network model. This model was presented to predict the mine inflow of a-100 m drift,the maximum forecast error was 3. 38% in the theoretical forecast length. Compared with pure chaotic BP neural network prediction model,the prediction results show that the proposed method can reflect the mine water flow in the short time,improve the accuracy of rating prediction effectively and provide scientific reference for disaster prevention of mining enterprises.

关 键 词：矿井涌水量 混沌理论 C-C算法 遗传算法 BP神经网络 

分 类 号：TD742[矿业工程—矿井通风与安全] TP183[自动化与计算机技术—控制理论与控制工程]