基于集成学习的改进灰色瓦斯浓度序列预测*

Improved grey prediction of gas concentration sequence based on integrated learning

摘要:

为有效提高煤矿瓦斯浓度动态预测精度，基于微分方程理论和最小二乘法，从灰色预测模型静态灰色作用量出发，优化灰色作用量，推导幂指数型灰色作用量的改进灰色瓦斯浓度预测算法，推导基于集成学习不同灰色作用量幂指数型灰色瓦斯预测模型，进而研究吉林八连城长期和短期瓦斯浓度监控数据预测精度。结果表明：瓦斯浓度时间序列近似线性时，基于集成学习的改进灰色瓦斯浓度预测算法优于传统灰色瓦斯浓度预测算法，使瓦斯浓度预测值和实际值的均方根误差降低，均方根差最大降低2.25%。研究结果可有效提瓦斯浓度预测精度。

Abstract:

In order to effectively improve the dynamic prediction accuracy of coal mine gas concentration,based on the theory of differential equations and the least square method,starting from the static gray action of the gray prediction model,the gray action is optimized,and an improved grey gas concentration prediction algorithm of power exponential gray action is derived.Based on the ensemble learning of the power exponential gray gas prediction model of different gray effects,the prediction accuracy of the long-term and short-term gas concentration monitoring data in Balian city of Jilin is studied.The results show that when the gas concentration time series is approximately linear,the improved gray gas concentration prediction algorithm based on integrated learning,is better than the traditional gray gas concentration prediction algorithm,which reduces the root mean square error between the predicted value and the actual value of the gas concentration,and maximizes the root mean square difference decrease by 2.25%.The research results can effectively improve the accuracy of gas concentration prediction.