煤矿瓦斯涌出量动态预测的PCA-MFOA-GRNN模型及应用

摘要

针对煤矿瓦斯涌出受许多因素的影响，为了克服瓦斯涌出中存在的复杂的非线性关系，从而实现稳定、可靠、精确的对煤矿综采工作面瓦斯涌出量进行动态预测，提出了主成分分析法（PCA）结合改进的果蝇算法（MFOA）优化GRNN的绝对瓦斯涌出量的预测手段。运用PCA算法对原始输入数据降维；并且对果蝇算法中的Si函数增加一个跳脱参数B，避免局部最优因子对预测模型的干扰；将MFOA算法对GRNN的平滑因子σ进行优化；将PCA结果作为模型的输入，建立了PCA-MFOA-GRNN算法的回采工作面瓦斯涌出量动态预测模型，结合实际矿井瓦斯涌出量监测的相关数据检验该模型，并将该模型的预测结果与未修正的FOA-GRNN算法、CIPSO-ENN算法、BP神经网络预测、Elman网络预测结果进行对比，结果表明：该预测模型对GRNN的参数优化后得到的预测模型较其他预测模型有更强的泛化能力和更高的预测精度。%For coal mine gas emission is affected by many factors,in order to overcome the complex nonlinear rela⁃tionship exists in gas emission,achieving stable,reliable and accurate dynamic prediction for the absolute amount of gas emission in fully mechanized face,then proposing the PCA and improved drosophila algorithm (MFOA) opti⁃mizing GRNN to predict the absolute amount of gas emission. Use the PCA algorithm to raw input data dimension re⁃duction and adding a trip parameter B in Si function in drosophila algorithm,to avoid local optimum factor influenc⁃ing forecasting model. Using MFOA algorithm to optimize the smoothing factor σof GRNN,establishing dynamic forecasting model for the amount of gas emission in mining face based on PCA-MFOA-GRNN algorithm,then com⁃bining with the actual mine monitoring data of gas emission to verify the model,and comparing the model predic⁃tions with FOA-GRNN algorithm uncorrected、CIPSO-ENN algorithm、BP neural network predictions and Elman network predictions,results show that:The forecasting model after improved drosophila algorithm optimizing GRNN parameter has stronger generalization ability and higher prediction accuracy than other forecasting models.