VRLA Battery Monitoring Instrumentation Using DEKF for Weather Stations

Abstract

Valve-Regulated Lead-Acid (VRLA) batteries serve as an essential backup power source for BMKG Automatic Weather Stations (AWS) in remote locations, making real-time condition monitoring a critical instrumentation need. This study designs, implements, and evaluates an integrated ESP32-based instrumentation system for real-time State of Charge (SOC) monitoring of VRLA batteries, emphasizing hardware engineering, metrological calibration, embedded firmware architecture, and system-level validation. The novelty lies in the end-to-end integration of traceable sensor calibration, dual-core FreeRTOS firmware, and a model-based estimator into one validated instrumentation platform for VRLA batteries in unattended AWS operation, a combination rarely reported in prior IoT battery-monitoring studies. The system integrates a calibrated INA226 current-voltage sensor and a DHT22 temperature-humidity sensor, with a Dual Extended Kalman Filter (DEKF) based on a second-order Thevenin model used strictly as a supporting SOC estimator. Testing covered sensor calibration, functional black-box testing, SOC accuracy and convergence robustness, NVS-based restart recovery, and real-time computational load. Results show calibrated voltage deviation below 0.25% and current below 3%, steady-state voltage RMSE of 4.78-9.23 mV, SOC self-correction under 30 minutes, and continuous operation exceeding 48 hours without loop failure, providing a reliable basis for condition-based maintenance of BMKG's AWS battery network.