基于AQPSO-LSSVM的煤焦油加氢产物凝点预测

Condensate Prediction of Coal Tar Hydrogenation Products Based on AQPSO-LSSVM

采用最小二乘支持向量机(LSSVM)技术,以煤焦油加氢产物馏程数据为输入变量,建立煤焦油加氢产物凝点预测模型。该模型引入自适应量子行为粒子群算法(AQPSO)对最小二乘支持向量机的核宽度参数和正则化参数进行优化选择,提高凝点预测的速度和精度。结果表明,AQPSO-LSSVM模型的最优适应度值为3.8017,凝点的预测值与实验值的相关系数值为0.95835,误差绝对值低于标准样本占95.71%,从预测精度和泛化性能上均优于所比较的PSO-LSSVM模型和QPSO-LSSVM模型,对石化产品凝点的预测具有较高的准确性和可行性。

Using the least squares support vector machine(LSSVM)technology and taking the distillation range data of coal tar hydrogenation products as the input variable,a condensate prediction model of coal tar hydrogenation products was established.The model introduces an adaptive quantum behavioral particle group algorithm(AQPSO)to optimize the kernel width and regularization parameters of least squares support vector machines,so as to improve the speed and accuracy of condensate prediction.The results show that the optimal fitness value of the AQPSO-LSSVM model is 3.8017,the correlation coefficient value and experimental value is 0.95835,the error absolute value is below the standard sample accounted for 95.71%.The prediction accuracy and generalization performance are better than the compared PSO-LSSVM model and QPSO-LSSVM model,which has high accuracy and feasibility for the condensate prediction of petrochemical products.