基于模糊最小二乘支持向量机的蒸发过程预测控制研究

Predictive Control Study for Evaporation Process Based on Fuzzy Least Square Support Vector Machine

针对氧化铝蒸发过程的多变量、非线性和大滞后特点及不同时间和空间样本数据不同的特征,提出了基于末位淘汰机制的混沌粒子群算法的综合加权模糊最小二乘支持向量机蒸发过程预测控制方法。用变异混沌粒子群算法对模型预测控制进行滚动优化,计算出最优控制序列。以某氧化铝厂蒸发过程生产数据进行实验验证分析,结果表明:模型预测结果中相对误差小于8%的样本达到93.9%,出口浓度稳定在240g/L附近,其控制性能得到显著改善,同时也起到了降低能耗的目的。

Aiming at the characteristics of multivariable, nonlinearity, large time-delay in alumina evaporation process and the different features of various temporal and spatial samples, a predictive control strategy combining the fuzzy least square support vector machine （LS-SVM） with weight factor and chaotic particle swarm optimization （CPSO） with last out mechanism was proposed. To achieve rolling optimization in predictive control, a CPSO algorithm with last out mechanism was introduced to calculate the control sequence. The experimental verification analysis was performed using the industrial production data from evaporation process of an alumina plant. The results show that percentage of the samples with prediction relative error less than 8% was up to 93.9%, while the outlet concentration was stabilized at about 240 g/L, and the prediction control performance is greatly improved and plays a role in reducing energy consumption.