Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Neural Network NMPC for Computationally Tractable Optimal Power Management of Hybrid Electric Vehicle

Suyong Park, Duc Giap Nguyen, Jinrak Park, Dohee Kim, Jeong Soo Eo, Kyoungseok Han

TL;DR

The proposed DNN-MPC closely approximates the NMPC performance while substantially reducing the computational burden, and its practical utility is shown through its application in optimizing the energy management of hybrid electric vehicles (HEVs).

Abstract

This study presents a method for deep neural network nonlinear model predictive control (DNN-MPC) to reduce computational complexity, and we show its practical utility through its application in optimizing the energy management of hybrid electric vehicles (HEVs). For optimal power management of HEVs, we first design the online NMPC to collect the data set, and the deep neural network is trained to approximate the NMPC solutions. We assess the effectiveness of our approach by conducting comparative simulations with rule and online NMPC-based power management strategies for HEV, evaluating both fuel consumption and computational complexity. Lastly, we verify the real-time feasibility of our approach through process-in-the-loop (PIL) testing. The test results demonstrate that the proposed method closely approximates the NMPC performance while substantially reducing the computational burden.

Deep Neural Network NMPC for Computationally Tractable Optimal Power Management of Hybrid Electric Vehicle

TL;DR

The proposed DNN-MPC closely approximates the NMPC performance while substantially reducing the computational burden, and its practical utility is shown through its application in optimizing the energy management of hybrid electric vehicles (HEVs).

Abstract

This study presents a method for deep neural network nonlinear model predictive control (DNN-MPC) to reduce computational complexity, and we show its practical utility through its application in optimizing the energy management of hybrid electric vehicles (HEVs). For optimal power management of HEVs, we first design the online NMPC to collect the data set, and the deep neural network is trained to approximate the NMPC solutions. We assess the effectiveness of our approach by conducting comparative simulations with rule and online NMPC-based power management strategies for HEV, evaluating both fuel consumption and computational complexity. Lastly, we verify the real-time feasibility of our approach through process-in-the-loop (PIL) testing. The test results demonstrate that the proposed method closely approximates the NMPC performance while substantially reducing the computational burden.
Paper Structure (14 sections, 10 equations, 10 figures, 3 tables)

This paper contains 14 sections, 10 equations, 10 figures, 3 tables.

Table of Contents

  1. Introduction
  2. SYSTEM MODELING AND PROBLEM FORMULATION
  3. Control-oriented Model
  4. Nonlinear MPC Design
  5. NMPC Control Result
  6. Approximated NMPC using Neural Network
  7. DNN Training Procedure
  8. Training Validation
  9. Test Results
  10. Test Environments
  11. Simulation Results
  12. Process-in-the-Loop Experiments
  13. Limitations
  14. Conclusion

Figures (10)

  • Figure 1: Power flows in the parallel and mild HEVs: $\dot{m}_f$ is fuel consumption; $P_b$ the battery power; $T$ the torque; $w$ the rotational speed; and subscripts $e, m, d, \text{and }, w$ are engine, motor, demand, and wheel respectively.
  • Figure 2: Multiple speed profiles obtained by the driving simulator from Hakjeong-dong to Kyungpook National University in Daegu (9.6 km)
  • Figure 3: Cost-to-go for the average driving outcomes: (a) DP cost-to-go value and (b) approximated $\lambda_k$ from Eq. \ref{['app_lambda']}.
  • Figure 4: Simulation results: (a) speed profile, (b) SOC trajectory, (c) cumulated fuel consumption, and (d) power distribution ratio.
  • Figure 5: Engine operation points of the rule-based and NMPC methods.
  • ...and 5 more figures