铁道科学与工程学报2025，Vol.22Issue(10)：4674-4686,13.DOI:10.19713/j.cnki.43-1423/u.T20250111
考虑参数失配的PMSM级联神经网络预测电流控制
Cascaded neural network predictive current control for PMSM considering parameter mismatch
摘要
Abstract
The performance of predictive current control(PCC)for permanent magnet synchronous motors(PMSMs)heavily relies on the matching degree between controller parameters and motor parameters.However,due to the strong nonlinearity,high electromagnetic coupling,and limited available differential equations in PMSM dynamics,traditional PCC methods struggle to achieve online correction for parameter mismatches.To address this issue,this paper proposed a cascaded neural network-based predictive current control method.Specifically,the limitations of traditional PCC under parameter mismatches were first analyzed,and a deviation equation for the response current under parameter mismatches was established.Subsequently,a neural network-based deviation parameter identifier was designed.Unlike traditional neural network parameter estimation methods,the proposed identification strategy satisfies the Lyapunov stability condition,ensuring theoretical stability guarantees.Furthermore,based on the analysis of response current deviations,a cascaded online parameter identifier for PMSM was constructed.Additionally,to address the lack of stability guidance in traditional PCC,a closed-loop PCC scheme incorporating deviation parameter identification was proposed,and its stability was thoroughly analyzed.Finally,simulation comparisons with traditional PCC strategies demonstrated that the proposed control framework effectively mitigates the impact of parameter mismatches on PMSM control system performance,significantly enhancing control robustness.Experimental validation on a Dspace hardware-in-the-loop platform confirmed that the proposed identification method can quickly and accurately lock parameter deviation values under sudden parameter mismatches.Moreover,the proposed predictive current control law provides more precise current responses and significantly reduces the adverse effects of parameter mismatches compared to traditional methods.In conclusion,the proposed cascaded neural network-based predictive current control method for PMSMs,which considers parameter mismatches,fully leverages the performance of PMSM systems and achieves efficient control under multi-parameter transients.关键词
永磁同步电机/参数失配/在线辨识/预测控制/神经网络Key words
permanent magnet synchronous motor(PMSM)/parameter mismatch/online identification/predictive control/neural network分类
动力与电气工程引用本文复制引用
程翔,瞿少成,何静..考虑参数失配的PMSM级联神经网络预测电流控制[J].铁道科学与工程学报,2025,22(10):4674-4686,13.基金项目
国家自然科学基金资助项目(62173137) （62173137）
中央高校基本科研业务费专项资金资助项目(CCNU22JC011,CCNU24CG008) （CCNU22JC011,CCNU24CG008）
湖北省技术创新计划项目(2023BAB049) （2023BAB049）
湖南省自然科学基金资助项目(2023JJ60232) （2023JJ60232）
湖南省教育厅优秀青年项目(23B1018) （23B1018）
