摘要： 在大型火电厂烧煤锅炉的运行过程中，受热面的积灰是降低锅炉运行效率和安全性的一个重要原因。目前，主要采用高压空气或者蒸汽把积灰吹掉。吹灰的困难在于确定应该何时吹灰，该文提出一种基于时序聚类的新方法，从经过预处理的锅炉历史数据中抽取出代表吹灰的模式，建立吹灰模型，用来预测吹灰时间。该方法在应用到一个900 MW的超临界锅炉上时，表现出较好的效果。

关键词: k-均值聚类, 时序聚类, 吹灰, 预测

Abstract: Ash deposition on heat transfer surfaces is a primary cause of reducing operating efficiency and safety in coal-fired boilers. Among all available means for managing ash deposition, sootblowing is the primary controllable factor in practice. A new time series clustering based method is set forth for predicting when to sootblow. In this method, sootblowing patterns, which represent the characteristics of sootblowing, are picked out by time series clustering from historical data that are preprocessed, then a model is built to predict when to sootblow. The performance of this method on a 900 MW supercritical boiler is good.

Key words: k-means clustering, time series clustering, sootblowing, prediction

中图分类号: 

• TP391