Work

Learning-augmented model predictive control for energy management in hydrogen-battery hybrid trains: A comparative deep learning study

Year: 2026

Type: article

Abstract: In hybrid vehicle applications, integrating hydrogen fuel cells and batteries presents energy management challenges that require real-time power optimization to protect components, improve fuel economy, and maintain charge levels. This paper introduces a learning-augmented model predictive control (MPC) framework for hydrogen-battery hybrid trains, comparing four deep learning models, long short-term memory (LSTM), gated recurrent unit (GRU), convolutional neural network (CNN), and echo state network (ESN), for power demand prediction. Trained on data from Italy's Brescia-Iseo-Edolo railway and validated on two other lines, the system uses a zero-dimensional model combining train dynamics, fuel cell polarization, and battery energy. Starting from a baseline MPC without machine learning, the study evaluates models over the round trip. Results indicate architecture-specific trade-offs: ESN has the lowest prediction error and excels in precise power demand prediction; LSTM/GRU models support fuel economy and performance; CNN struggles across scenarios. Prediction accuracy declines with longer horizons, especially for CNN.

Source: Energy and AI

DOI: 10.1016/j.egyai.2026.100839

Authors: Yosef Berhan Jember, Massimo Santarelli, Giuseppe Santangelo, Clemente Mirko

Institution: Politecnico di Torino

Language: English

Topic: Electric and Hybrid Vehicle Technologies

Subfield: Automotive Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Affordable and clean energy, Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $3,630