基于CPSO-RVM的蒸发过程在线预测模型

Online prediction model of evaporation process based on CPSO-RVM

针对密闭、长流程的蒸发过程由于待浓缩溶液黏度高或腐蚀性强、设备易结垢、工况变化复杂等原因引起的在线预测模型难以建立的问题,提出了一种基于混沌粒子群优化相关向量机(CPSO-RVM)的预测模型。基于贝叶斯学习框架构建了蒸发过程相关向量机预测模型,克服模型对核函数类型的限制和数据敏感性,在此基础上利用混沌粒子群算法对预测模型的核函数进行优化,获得计算量小、泛化性能优的在线预测模型。某厂实际蒸发过程生产数据的算例表明,在存在新蒸汽和原液干扰、设备结垢的整个清洗周期内,CPSO-RVM模型都能获得很好的预测效果,并且比偏最小二乘回归模型(PLSR)和最小二乘支持向量机模型(LSSVM)精度更高,能为实际蒸发过程的在线控制提供参考。

Considering that the output concentrations of long-flow evaporation processes are hard to be measured online due to high viscosity and strong corrosion of the solution and the complex operation conditions,this paper proposes a new online prediction model based on chaotic particle swarm optimization and relevance vector machine（CPSO-RVM）.In the prediction model,correlation vector machine is built under Bayesian learning framework,which overcomes the data sensitivity and the limitation of choosing kernel function types.On this basis,a chaotic particle swarm optimization algorithm is used to optimize the kernel function parameters of the prediction model.Through it,the online prediction model with small computation burden and good generalization performance is obtained.Experimental results on an actual aluminates evaporation process show that CPSO-RVM has better predictive effect than the partial least squares regression and the least square support vector machine in the presence of steam and liquid interference in the whole cleaning cycle of the evaporation process.The proposed algorithm also provides method for online concentration measurement of other actual evaporation processes.