基于LGWO-LSTSVR的电解质温度预测  

Prediction of Electrolytic Temperature Based on LGWO-LSTSVR

作　　者：徐辰华 何俊隆[2] XU Chenhua;HE Junlong(School of Automation,Guangdong Polytechnic Normal University,Guangzhou 510080;School of Electrical Engineering,Guangxi University,Nanning 530004)

机构地区：[1]广东技术师范大学自动化学院,广州510080 [2]广西大学电气工程学院,南宁530004

出　　处：《计算机与数字工程》2025年第2期572-577,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:62073090);广西重点研发项目(编号:2018AB67003);校级科研项目人才专项(编号:2021SDKYA118)资助。

摘　　要：为了解决铝电解过程的复杂性导致铝电解槽的电解质温度难以直接测量的问题,论文采用改进灰狼优化算法(LGWO)与最小二乘孪生支持向量回归机(LSTSVR)建立电解质温度预测模型。首先利用最小二乘孪生支持向量回归机建立电解质温度预测模型;然后针对最小二乘孪生支持向量回归机的结构参数选取不佳的情况,利用改进的灰狼算法对最小二乘孪生支持向量回归机的结构参数进行参数寻优,利用Logistic混沌策略,位置更新策略和高斯变异策略来提高GWO算法的全局优化能力,优化最小二乘孪生支持向量回归机以获得较高预测精度;最后利用广西某铝厂实际铝电解生产数据对LGWO-LSTSVR模型进行验证,实验结果表明,LGWO-LSTSVR的电解质温度预测模型具有较好的预测效果,能更准确地预测电解质温度。To address the challenge of direct temperature measurement in aluminum electrolysis cells caused by process com⁃plexity,this paper proposes an electrolyte temperature prediction model that integrates an improved grey wolf optimization algorithm(LGWO)with least squares twin support vector regression(LSTSVR).Initially,the LSTSVR method is employed to establish a fun⁃damental electrolyte temperature prediction model.Subsequently,to resolve the issue of poor parameter selection in LSTSVR's struc⁃tural configuration,the enhanced grey wolf optimization algorithm is implemented for parameter optimization.The LGWO algorithm incorporates three strategic improvements,which are Logistic chaotic mapping,enhanced position updating mechanisms,and Gaussian mutation operators,to strengthen its global optimization capability.This optimized algorithm is then applied to refine the LSTSVR parameters for improved prediction accuracy.Finally,the proposed LGWO-LSTSVR model is validated using actual pro⁃duction data from an aluminum electrolysis plant in Guangxi,China.Experimental results demonstrate that the LGWO-LSTS⁃VR-based electrolyte temperature prediction model achieves superior predictive performance with enhanced accuracy,providing an effective solution for temperature monitoring in aluminum electrolysis processes.

关 键 词：电解质温度预测 灰狼优化算法 最小二乘孪生支持向量回归机 混沌映射 

分 类 号：P731.11[天文地球—海洋科学]