基于PSO-Lasso算法的电站湿法脱硫出口SO_(2)浓度预测  

Prediction of SO_(2)Concentration at WFGD Outlet of Power Station Based on PSO-Lasso Algorithm

作　　者：郑茗友 王伟[1] 赵文杰[1] 王建峰 郄英杰 ZHENG Mingyou;WANG Wei;ZHAO Wenjie;WANG Jianfeng;QIE Yingjie(School of Electrical and Electronic Engineering,North China Electric Power University,Baoding 071003,China;Shanxi Zhangshan Power Generation Co.,Ltd.Changzhi 046021,China)

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003 [2]山西漳山发电有限责任公司,山西长治046021

出　　处：《华北电力大学学报（自然科学版）》2022年第1期112-117,共6页Journal of North China Electric Power University：Natural Science Edition

摘　　要：为了能够简便地研究机组运行中的过程参数与脱硫效率的关系,提出了一种基于套索(Lasso,Least absolute shrinkage and selection operator)算法的预测方法。结合脱硫系统现场的实际数据,通过相关性分析筛选影响脱硫出口的主要因素。相比于传统的回归算法,创新性的使用粒子群算法(PSO)确定提取出的主要参数的函数形式,最后使用Lasso算法确定最终的预测模型,该模型在简化运算复杂度的同时保证了预测精度。训练数据的选取使用正交化法则,保证训练数据的丰富性与有效性。测试结果表明,模型预测误差为2.23 mg/m^(3),能够反映在工况变化下脱硫出口浓度的对应关系也有助于脱硫系统的优化控制,具有一定的研究与应用价值。In order to easily study the relationship between process parameters and the desulfurization efficiency in unit operation,this paper proposes a new method based on Lasso(Least absolute shrinkage and selection operator)algorithm.Combined with the actual data of the desulfurization system site,the main factors affecting the FGD outlet were screened by correlation analysis.Compared with the traditional regression algorithm,the particle swarm algorithm(PSO)was innovatively used to determine the function form of the extracted main parameters.Finally we used the Lasso algorithm to determine the final prediction model,which simplifies the operational complexity and ensures the prediction accuracy.The training data were selected using the orthogonalization rule so that the richness and effectiveness are ensured.The test results show that the prediction error of the model is 2.23 mg/m^(3),which can reflect the correspondence of desulfurization outlet concentration under the change of working conditions and is helpful for the optimal control of desulfurization system,and has certain value of research and application.

关 键 词：WFGD Lasso回归 粒子群算法 出口SO_(2)浓度 预测模型 

分 类 号：TK39[动力工程及工程热物理—热能工程]