煤调湿系统蒸汽消耗量的多模型建模研究.pdf

煤调湿系统蒸汽消耗量的多模型建模研究摇 王建华,等 煤调湿系统蒸汽消耗量的多模型建模研究 Researchonthe Multi鄄model Modeling forVapor Consumptionof Coal Moisture Control System 王建华 摇 罗雷涛 摇 耿佳灿1 2 2 (上海应用技术学院电气与电子工程学院 ,上海摇 201418;1 华东理工大学化工过程先进控制与优化技术教育部重点实验室 ,上海摇 200237)2 摘摇 要: 煤调湿工艺是炼焦过程的关键技术,建立煤调湿系统的蒸汽消耗量软测量模型对节约生产成本、减少环境污染和提高焦炭 质量具有促进作用。 针对某炼铁厂煤调湿装置的非线性、强耦合、工况波动大等特点,提出了一种多模型建模方法。 对模糊核C均值 聚类引入了基于密度的聚类中心初始化方法和聚类数目自适应策略,并将其用于生产工况最优划分;使用最小二乘支持向量机对每 个子工况进行数据驱动建模,并利用贝叶斯证据框架优化最小二乘支持向量机的超参数。 检验结果表明,所提出的煤调湿系统蒸汽 消耗量多模型具有良好的跟踪性能与较高的预测精度。 关键词: 煤调湿摇 模糊核C均值聚类摇 最小二乘支持向量机(LSSVM)摇 贝叶斯证据框架摇 多模型 中图分类号: TP391+.9摇 摇 摇 摇 文献标志码: A Abstract: Coal moisture control technology is the key technique of the coking process. Establishing the soft sensing model for vapor consumption in coal moisture control system may save production costs,reduce environmental pollution and improve the quality of coke. In accordance with the features of the coal moisture control equipment in certain ironworks, e. g. , nonlinearity, strong coupling and large fluctuations of working condition,the multi鄄model modeling method is proposed. The density鄄based clustering center initialization method and number of clustersadaptivestrategy areintegratedintothefuzzy C鄄meansclusteringtogetoptimaldivisionoftheproductionworkingconditions. The least squares support vector machines(LSSVM) is applied to conduct data鄄driven modeling for each sub鄄working conditions, and the parameters of the LSSVM is optimized by adopting Bayesian evidence framework. The checked results indicate that the proposed multi鄄model possesses good tracking performance and higher prediction accuracy. Keywords: Coal moisture control摇 Fuzzy kernel C鄄means clustering摇 Least squares support vector machines(LSSVM)摇 Bayesian evidence framework摇 Multi鄄model (sup