An Enhanced Cascaded Deep Learning Framework for Multi-Cell Voltage Forecasting and State of Charge Estimation in...

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Sustainable Energy

Background:

• Traditional Battery Management Systems (BMS) face performance limitations in electric vehicles (EVs) operating in tropical climates due to high temperatures.
• Accurate State of Charge (SoC) estimation and voltage prediction are critical for EV operational efficacy and safety, particularly under varying environmental conditions.

Purpose of the Study:

• To introduce and validate a novel two-tiered deep learning framework for enhanced BMS in EVs, specifically addressing challenges in tropical climates.
• To improve the precision of battery voltage and SoC predictions using a Long Short-Term Memory (LSTM) network architecture.

Main Methods:

• A two-stage Long Short-Term Memory (LSTM) framework was developed: LSTM-1 for individual cell voltage prediction and LSTM-2 for SoC estimation.
• The model utilized multivariate time-series data, including voltage history, vehicle speed, current, temperature, and load metrics from dynamometer testing of a 120-cell Lithium Iron Phosphate (LFP) battery pack.
• Experiments simulated urban driving conditions with varying speeds (6-40 km/h) and load conditions (0-20%) under high-temperature scenarios.

Main Results:

• The proposed deep learning BMS achieved a 15% improvement in SoC estimation accuracy compared to traditional methods under simulated real-world driving conditions.
• The framework effectively handled temperature-dependent voltage fluctuations and captured complex temporal and inter-cell dependencies.
• The system demonstrated superior performance in high-temperature and variable-load environments characteristic of tropical climates.

Conclusions:

• This research presents the first deep learning-based BMS optimization validated in tropical climates, establishing a new benchmark for EV battery management in such regions.
• The enhanced BMS framework significantly improves EV reliability and operational safety, supporting the growth of electric mobility.
• The two-tiered LSTM approach offers a robust solution for precise battery monitoring and management under challenging environmental conditions.