Adaptive Neural Network-Based Event-Triggered SOC Observer With Application to a Stochastic Battery Model

Area of Science:

• Electrical Engineering
• Control Systems
• Artificial Intelligence

Background:

• Accurate battery state of charge (SOC) estimation is vital for safe, reliable, and efficient battery operation.
• Traditional methods often require extensive parameter extraction and can be computationally intensive.
• Existing observers may struggle with online parameter variations and computational efficiency.

Purpose of the Study:

• To propose an adaptive neural network (NN)-based event-triggered observer for accurate battery SOC estimation.
• To develop an observer that can handle unknown nonlinearities and adapt to changing battery parameters online.
• To reduce computational load through an event-triggered mechanism while ensuring system stability.

Main Methods:

• Establishment of a stochastic battery equivalent circuit model (ECM).
• Utilization of an adaptive NN to approximate unknown nonlinear system dynamics.
• Implementation of an event-triggered mechanism (ETM) for online weight updates, optimizing computational cost.
• Design of an adaptive radial basis function (RBF) NN-based observer with stability analysis using Lyapunov theory.
• Derivation of a strictly positive lower bound for interevent time to prevent Zeno behavior.

Main Results:

• The proposed adaptive NN observer accurately estimates battery SOC.
• The event-triggered mechanism effectively reduces computational cost by updating weights only when necessary.
• The observer demonstrates robustness against initial state deviations and sensor noise, as validated by experiments and simulations.
• Stability of the observer is theoretically proven, and Zeno behavior is successfully excluded.

Conclusions:

• The adaptive NN-based event-triggered observer provides an accurate and computationally efficient solution for battery SOC estimation.
• The method ensures reliable battery management by adapting to system dynamics and handling uncertainties.
• This approach offers a promising direction for enhancing battery management systems in various applications.