基于FM度量的自适应K-Means聚类的工业生产运行基准挖掘

Industrial Production Operation Benchmark Mining for Adaptive K-Means Clustering Based on FM Metrics

针对工业生产传统运行优化方法计算复杂,以及优化目标往往很难达到稳态的问题,提出基于FM度量的自适应K-Means聚类的工业生产运行基准挖掘方法。首先,以运行负荷等变量作为筛选标准,提出基于方差的稳态判别算法,筛选出历史数据中的稳态工况,并根据实际运行参数,对稳态工况进行细分;其次,由于K-Means算法需要预先设定K值,提出基于FM度量的自适应算法确定K值进行K-Means聚类,并利用能耗指标确定最优的聚类中心;最后,采用某实际生产企业的历史运行数据进行模型验证。

Aiming at the problems that the traditional operation optimization method of industrial production is complicated in calculation and the optimization objective is often difficult to achieve steady state,an industrial production operation benchmark mining method for adaptive K-Means clustering based on FM metric is proposed.Firstly,with the variables such as operating load and so on as the screening criteria,a variance-based steady-state discriminant algorithm is presented to screen out the steady-state working conditions in the existing data and subdivide the steady-state working conditions according to the actual operating parameters;Secondly,since K-Means algorithm needs to preset K value,an adaptive algorithm based on FM metric is given to determine K value for K-Means clustering,and the energy consumption index is used to determine the optimal clustering center;Finally,the existed operation data of an actual production enterprise is used to carry out the model validation.