State-of-Health Estimation in 12 V Lead-Acid Batteries

Reliable State of Health (SoH) estimation remains a critical challenge in battery management systems due to the complex interplay of electrochemical, thermal, and operational factors, particularly when low-cost and real-time solutions are required. To address the challenge, this paper presents a simple yet effective low-cost system for estimating the State of Health (SoH) of 12 V lead-acid batteries by integrating data-driven learning with fundamental electrochemical and electrical principles. The proposed system combines a basic estimation model based on Peukert’s law, which originates from charge conservation and rate-dependent electrochemical kinetics, with a correction stage performed on a cloud server using a machine learning (ML) model. An ESP32 microcontroller equipped with sensors for current, voltage, and temperature is used to acquire real-time battery data, reflecting underlying charge transport, internal resistance evolution, and thermal effects. The local physics-informed model provides an initial SoH estimate, which is subsequently refined using a residual ML model hosted on a cloud platform to capture nonlinear degradation mechanisms. The hybrid edge–cloud framework achieved a 35% improvement in accuracy compared to the Peukert-only approach. The estimated SoH trends are consistent with expected battery behavior governed by ionic diffusion, Ohmic losses, and thermodynamic aging under varying cycling conditions, depth of discharge, and temperature. Owing to its low cost and strong physical grounding, the proposed system is suitable for practical deployment in small UPS and renewable energy storage applications.