基于K-means聚类与核极限学习机的浆液循环泵优化
王庆福，余程，崔科杰，竹小锋，蒋红辉，吴可泽，黄建伟，朱宇超
（浙江浙能兰溪发电有限责任公司，浙江 金华 321110）
摘要： 作为湿法烟气脱硫系统主要耗电设备，浆液循环泵的运行优化，对降低电厂电耗、提高电厂经济效益具有重
要意义。针对当前火电厂浆液循环泵运行情况，提出一种基于历史运行数据、采用聚类和分类相结合的浆液循环泵运行
优化方法。首先，采用K-means 聚类算法和基于层次分析法与熵权法加权融合的组合评价方法，对不同工况下浆液循环
泵历史运行数据自动寻优；后以最优历史操作记录训练核极限学习机模型，并使用粒子群算法进行寻优，以实时运行数
据为输入给出浆液循环泵组合运行策略。以一电厂实测数据进行的仿真实验表明，浆液循环泵运行优化模型能够明显降
低循环泵系统电耗，对火电厂节能减排工作具有参考意义。
关键词： 浆液循环泵；聚类算法；核极限学习机；运行优化
中图分类号： TK229     文献标志码： A      文章编号： 2095-0802-(2022)11-0059-05
Optimization of Slurry Circulation Pump Based on K-means Clustering and Kernel Extreme
Learning Machine
WANG Qingfu, YU Cheng, CUI Kejie, ZHU Xiaofeng, JIANG Honghui, WU Keze, HUANG Jianwei,
ZHU Yuchao
(Zheneng Lanxi Power Generation Co., Ltd. of Zhejiang, Jinhua 321110, Zhejiang, China)
Abstract: As the main power-consuming equipment of wet flue gas desulfurization system, the operation optimization of slurry
circulation pump is of great significance to reduce the power consumption and improve the economic efficiency of power plants.
Aiming at the current situation of slurry circulation pump operation in thermal power plants, this paper proposed a slurry
circulation pump operation optimization method based on historical operation data, using a combination of clustering and
classification. Firstly, K-means clustering algorithm and a comprehensive evaluation method based on the weighted fusion of
hierarchical analysis and entropy method were used to automatically optimize the slurry circulation pump historical operation data
under different operating conditions; subsequently, a kernel extreme learning machine model was trained with the optimal
historical operation records, and a particle swarm algorithm was used to optimize the combined slurry circulation pump operation
strategy with real-time operation data as input. The simulation experiments on the actual measurement data of a power plant show
that the slurry circulation pump operation optimization model can significantly reduce the electricity consumption of the circulation
pump system, which is of referential significance to the energy saving and emission reduction of thermal power plants.

 Key words: slurry circulating pump; clustering algorithm; kernel extreme learning machine; operation optimization