Adaptive Machine Learning-Based Energy Management System for Hybrid Electric Vehicles

2024-01-5108

12/02/2024

Event
Automotive Technical Papers
Authors Abstract
Content
This paper examines the effectiveness of optimizing energy management in hybrid electric vehicles by integrating adaptive machine learning algorithms with the energy management electronic control unit (ECU). Existing traditional rule-based energy management and control strategies of power distribution between internal combustion and battery struggle to adapt to dynamic driving conditions, such as rapid acceleration, frequent stop-and-go traffic, and varying terrain. These scenarios often result in sub-optimal energy utilization and performance, as the fixed rules struggle to account for the immediate demands and inefficiencies that arise in such conditions. In conditions like that, rapid acceleration demands a sudden increase in power, which can lead to inefficient fuel consumption if not managed properly, while frequent stop-and-go traffic conditions can cause the battery to drain and lead to increased fuel consumption. Varying terrain can also lead to improper power distribution, with steep inclines demanding more power from the internal combustion engine than normal terrain. In contrast, the proposed system utilizes real-time feedback from the energy management ECU, based on that it decides the power distribution between the battery and internal combustion engine in the subsequent cycle, ensuring optimal performance across a wide range of driving conditions by efficiently distributing the power between internal combustion engine and battery.
In this paper, the approach toward the development of a machine learning algorithm for energy management of hybrid electric vehicles will be shown in MATLAB. The procedure involves data collection from the energy management ECU, including parameters such as battery state of charge (SOC), vehicle speed, engine load, torque, fuel consumption rate, and temperature. This data is then used to extract relevant patterns and relationships. Subsequent machine learning algorithms are selected based on their suitability for the data characteristics and problem requirements and trained using MATLAB. After training, the model will receive real-time feedback from the energy management ECU based on collected parameters. Using this feedback the model decides the power distribution between the internal combustion engine and battery. Results, including metrics like mean absolute error (MAE) and root mean square error (RMSE) regarding fuel consumption and overall fuel consumption rates, will be plotted to validate the system’s performance before and after implementing the machine learning model. These results will be plotted on performance graphs to illustrate the improvement achieved.
In conclusion, this article showcases a comprehensive methodology utilizing MATLAB for the development and validation of an adaptive machine learning-based energy management system for hybrid electric vehicles.
Meta TagsDetails
DOI
https://doi.org/10.4271/2024-01-5108
Pages
12
Citation
Bhargav, M., "Adaptive Machine Learning-Based Energy Management System for Hybrid Electric Vehicles," SAE Technical Paper 2024-01-5108, 2024, https://doi.org/10.4271/2024-01-5108.
Additional Details
Publisher
Published
Dec 02
Product Code
2024-01-5108
Content Type
Technical Paper
Language
English