球磨机制粉过程煤粉粒度FCM-SVRs软测量建模

FCM-SVRs-Based Soft Sensor Modeling of Granularity of Pulverizing Coal During Ball Milling

根据多个模型相加可以提高整体预测精度和鲁棒性的思想,提出了一种基于模糊C均值聚类(FCM)算法的煤粉粒度多最小二乘支持向量机回归(MLS-SVRs)软测量模型.采用变长度染色体的遗传算法同时优化模糊聚类数和聚类中心,每种聚类子集用LS-SVRs进行局部模型的建立和训练,再用模糊聚类后产生的隶属度将各子模型的输出加权求和得到最后软测量结果.仿真结果表明该软测量模型具有更好的泛化结果和预测精度,可以满足煤粉制备过程实时控制的在线软测量要求.

Inspired by the idea that combining several models can improve the prediction accuracy and robustness on the whole,a multiple least squares-support vector machine regressor（MLS-SVRs） soft sensing modeling of the granularity of pulverized coal is proposed based on fuzzy C-means（FCM） clustering algorithm.Genetic algorithm based on sizable chromosome is introduced to optimize the number of fuzzy clustering and cluster centers.Then,the whole training data set is divided into several clusters with different centers by FCM algorithm and each subset is trained by LS-SVRs,and the degrees of membership resulting from fuzzy clustering are used to weight and summarize the outputs of all submodels,thus giving the finial soft sensing outcome.Simulation results showed that the proposed model is effective in the granularity prediction and meets the requirement of the on-line soft sensor for real-time optimization control in pulverized coal production process.