基于PCA和PSO-ELM的煤与瓦斯突出软测量研究

Research on the Soft Sensor of Coal and Gas Outburst Based on PCA and PSO-ELM

针对煤与瓦斯突出预测效率和准确率不高这一问题,提出将主成分分析(PCA)法与改进的极端学习机(PSO-ELM)相结合的方法对煤与瓦斯突出进行预测。根据某煤矿地质动力区划方法,在划分活动断裂,岩体应力计算等工作基础上获取影响突出的相关数据;通过主成分分析法对原始数据进行降维处理,消除变量间的线性相关性;利用粒子群算法(PSO)对极端学习机(ELM)的输入权值和隐层阈值进行优化,建立PSO-ELM预测模型,将提取的主成分作为该模型的输入,煤与瓦斯突出强度作为模型输出。实验结果表明,该方法的预测精度高、结构简化,具有较强的泛化性能力强。

In order to solve the problems of low efficiency and accuracy of the coal and gas outburst prediction,in the paper, primary component analysis （ PCA ） combined with improved extreme learning machine （ PSO-ELM ） method for prediction of the coal and gas outburst is proposed. According to a coal mine geology dynamic division method,prominent influenced relevant data is acquired by the basic work of divisions of active faults and rock mass stress calculation. Through the primary component analyze method to reduce the dimension of the original data, eliminate the linear correlation volume. Using particle swarm optimization（ PSO） to optimize the input weights and hidden layer threshold of extreme learning machine （ ELM ） , establish PSO-ELM prediction model, treat the extractive principal components as the input of the prediction model,the outburst intensity of coal and gas as the model output. The results show that the method has high accuracy of the prediction, simplification of the model structure and strong generalization performance.