Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep-MPC: A DAGGER-Driven Imitation Learning Strategy for Optimal Constrained Battery Charging

Jorge Espin, Dong Zhang, Daniele Toti, Andrea Pozzi

TL;DR

This manuscript introduces an innovative solution to confront the inherent challenges often associated with conventional predictive control strategies for constrained battery charging by employing the imitation learning paradigm to address scenarios where battery parameters are uncertain, and internal states are unobservable.

Abstract

In the realm of battery charging, several complex aspects demand meticulous attention, including thermal management, capacity degradation, and the need for rapid charging while maintaining safety and battery lifespan. By employing the imitation learning paradigm, this manuscript introduces an innovative solution to confront the inherent challenges often associated with conventional predictive control strategies for constrained battery charging. A significant contribution of this study lies in the adaptation of the Dataset Aggregation (DAGGER) algorithm to address scenarios where battery parameters are uncertain, and internal states are unobservable. Results drawn from a practical battery simulator that incorporates an electrochemical model highlight substantial improvements in battery charging performance, particularly in meeting all safety constraints and outperforming traditional strategies in computational processing.

Deep-MPC: A DAGGER-Driven Imitation Learning Strategy for Optimal Constrained Battery Charging

TL;DR

This manuscript introduces an innovative solution to confront the inherent challenges often associated with conventional predictive control strategies for constrained battery charging by employing the imitation learning paradigm to address scenarios where battery parameters are uncertain, and internal states are unobservable.

Abstract

In the realm of battery charging, several complex aspects demand meticulous attention, including thermal management, capacity degradation, and the need for rapid charging while maintaining safety and battery lifespan. By employing the imitation learning paradigm, this manuscript introduces an innovative solution to confront the inherent challenges often associated with conventional predictive control strategies for constrained battery charging. A significant contribution of this study lies in the adaptation of the Dataset Aggregation (DAGGER) algorithm to address scenarios where battery parameters are uncertain, and internal states are unobservable. Results drawn from a practical battery simulator that incorporates an electrochemical model highlight substantial improvements in battery charging performance, particularly in meeting all safety constraints and outperforming traditional strategies in computational processing.
Paper Structure (12 sections, 9 equations, 4 figures, 1 table, 1 algorithm)

This paper contains 12 sections, 9 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Fundamentals
  3. Battery model
  4. Model Predictive Control for battery charging
  5. Imitation Learning
  6. DAGGER-based Imitation Learning Approach for Optimal Constrained Battery Charging
  7. Dataset Aggregation - DAGGER algorithm
  8. Simulation Results
  9. Dataset Generation and Training Stage
  10. Case of study of battery charging applying DAGGER-based method
  11. Computational performance and statistical analysis
  12. Conclusions

Figures (4)

  • Figure 1: Battery State Comparison: a) Battery state of charge comparison between the expert agent (dash-dotted red line) and the proposed DAGGER method (crossed blue line), showcasing satisfactory reference tracking by both control methods, b) Battery voltage profiles under the control of the expert agent (dash-dotted red line) and the proposed DAGGER method (crossed blue line) to ensure constraint satisfaction, c) Core temperature profiles to assess constraint compliance, with control strategies depicted in consistent colors and styles.
  • Figure 2: Comparison of Current Profiles: This figure presents a direct comparison of the current profiles generated by the expert agent (MPC) and the DAGGER approach over time. The expert agent's current profile is represented by a dash-dotted red line, while the profile of the DAGGER method is depicted as a crossed blue line.
  • Figure 3: Online Computational Time Comparison: MPC (orange) vs. DAGGER (blue) with increasing prediction horizon.
  • Figure 4: Distribution of current errors: A statistical analysis of DAGGER-based approach with respect to the expert agent (MPC).