Scholarlink

<p>Fuel cell electric vehicles (FCEVs) typically require an additional energy source to provide dynamic compensation, which makes the energy management strategy (EMS) critical for FCEVs. Considering the existing EMS issues generally treated as nonlinear optimization problems, this paper proposes a fast nonlinear model predictive control (NMPC) energy management scheme based on driving pattern classification, which provides improvements in velocity prediction and alleviates the computational burden of the NMPC controller. Specifically, an approach to classify driving patterns based on density is designed, then Gaussian process regression (GPR) velocity predictors are built separately under each pattern. Based on the classified patterns and built predictors, the k-nearest neighbor (kNN) algorithm is utilized for real-time driving pattern recognition and makes velocity forecasts accordingly. Meanwhile, a parametric approach is employed to reduce the dimensions of the variables to be solved in the optimization problem, thereby alleviating the optimization burden. Simulation results for a fixed roadway show that the proposed velocity predictor improves the accuracy by 6.5–14.7% compared to the predictor based on the K-means method, by 10.5–26.4% compared to the unclassified predictor. Moreover, fast NMPC can reduce computing burden by more than 85%, while the optimization costs are comparable to normal NMPC.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Fast NMPC Energy Management Scheme for Fuel Cell Electric Vehicles based on Driving Pattern Classification

  • 期刊论文

  • 发表日期：

    2025年3月13日

查看全文
  • Xuan Liu,
  • Jinwu Gao,
  • Shengyan Hou,
  • Runzi Lin,
  • Hong Chen

摘要

Fuel cell electric vehicles (FCEVs) typically require an additional energy source to provide dynamic compensation, which makes the energy management strategy (EMS) critical for FCEVs. Considering the existing EMS issues generally treated as nonlinear optimization problems, this paper proposes a fast nonlinear model predictive control (NMPC) energy management scheme based on driving pattern classification, which provides improvements in velocity prediction and alleviates the computational burden of the NMPC controller. Specifically, an approach to classify driving patterns based on density is designed, then Gaussian process regression (GPR) velocity predictors are built separately under each pattern. Based on the classified patterns and built predictors, the k-nearest neighbor (kNN) algorithm is utilized for real-time driving pattern recognition and makes velocity forecasts accordingly. Meanwhile, a parametric approach is employed to reduce the dimensions of the variables to be solved in the optimization problem, thereby alleviating the optimization burden. Simulation results for a fixed roadway show that the proposed velocity predictor improves the accuracy by 6.5–14.7% compared to the predictor based on the K-means method, by 10.5–26.4% compared to the unclassified predictor. Moreover, fast NMPC can reduce computing burden by more than 85%, while the optimization costs are comparable to normal NMPC.