基于混沌粒子群神经网络的瓦斯浓度预测

Chaotic PSO-RBFNN in coal mine gas concentration prediction

通过定量法确定瓦斯浓度数据具有混沌特性,计算瓦斯序列的延迟时间和最优嵌入维数并对其相空间重构。在混沌分析的基础上结合人工神经网络技术,针对传统RBFNN模型参数确定的问题,提出通过粒子群算法对网络参数优化,建立了CT—PSO—RBFNN预测模型。利用实际煤矿监测数据对提出的模型训练预测,并与其他3种模型横向对比,得出性能排序为CT—PSO—RBFNN>T—PSO—RBFNN>CT—RBFNN>T—RBFNN。结果证明,CT—PSO—RBFNN模型预测精度高、预测误差小、性能稳定,能够为瓦斯灾害的预报预警提供一定技术支持。

Gas disaster is the serious threat to coal mine safety,the accurate prediction of coal mine gas concentration is one effective method avoiding the occurrence of coal mine gas disasters.This paper determined the chaotic characteristic of gas concentration sequence by quantitative method,calculated the embedding dimension and optimal delay time.Combined the nonlinear analysis and artificial neural network,proposed to optimize the parameters of RBFNN by PSO algorithm and build CT-PSO-RBFNN prediction model.This paper compared three other models by simulation experiment,their performances ranking was CT-PSO-RBFNN,CT-RBFNN,T-PSORBFNN,T-RBFNN.The experiment result demonstrated the performance of CT-PSO-RBFNN with stable application,high accuracy and low errors which could be applied in coal mine safety such as gas concentration prediction.