基于PSO-SVR的取水泵组优化调度方法研究  

Optimal scheduling of water intake pump unit based on PSO-SVR

作　　者：王冬生 李智轩 渠赛赛 张雪[2] 蒋福春 曹勇[2] WANG Dongsheng;LI Zhixuan;QU Saisai;ZHANG Xue;JIANG Fuchun;CAO Yong(School of Automation and School of Artificial Intelligence.Nanjing University of Posts and Telecom munications,Nanjing 210023,China;Suzhou Water Supply Co.,Ltd.,Suzhou 215002,China)

机构地区：[1]南京邮电大学自动化学院人工智能学院,南京210023 [2]苏州市自来水有限公司,苏州215002

出　　处：《给水排水》2022年第8期143-150,共8页Water & Wastewater Engineering

基　　金：国家自然科学基金(52170001);国家水体污染控制与治理科技重大专项(2012ZX07403-001);江苏省水利科技项目(2020056);江苏省研究生科研与实践创新计划项目(SJCX20_0254)。

摘　　要：针对自来水厂取水泵组用电消耗量大、节能效果差的问题,将粒子群优化(PSO)算法和支持向量回归(SVR)算法相结合,提出了一种基于PSO-SVR的取水泵组调度方法。该方法使用SVR算法对往年节能的泵组调度工况数据进行学习,并结合PSO算法对SVR参数进行寻优,训练出准确的泵水效果预测模型。使用模型对不同环境下不同泵组搭配的泵水效果进行预测,并设计排序算法,依据泵组预测单位泵水能耗、预测管网压力差等预测结果排序,生成下一时段泵组搭配推荐表,并渲染在Web交互系统中,方便调度工程师进行决策。通过试验验证,相比传统调度方法,能够在每日泵组搭配切换次数不超过2次、管网压力波动小于0.006 MPa、泵水量满足需求的情况下,实现单耗平均降低3.92%。Aiming at the problem of high energy consumption and poor energy efficiency in water works,A new scheduling method of water pump group based on PSO-SVR is proposed,which combines particle swarm optimization(PSO)algorithm and support vector regression(SVR)algorithm.This method uses SVR algorithm to learn the previous energy-saving pump group scheduling data,and combines PSO algorithm to train an accurate pump water effect prediction model.The model is used to predict the pump water effect of different pump groups in different environments,and the sorting algorithm is designed to generate the pump group matching recommendation table according to the predicted results.These results are rendered in the web interactive system,which is convenient for scheduling engineers to make decisions.The scheme can reduce the energy consumption by an average of 3.92%under the conditions of no more than 2 times of pump group switching times per day,less than 0.06 MPa fluctuation of pipe network pressure and meeting the demand of pump water.

关 键 词：水泵组 优化调度 支持向量回归 粒子群优化 效果分析 

分 类 号：TU991[建筑科学—市政工程]