Parametric Identification of a Deep-Cycle Gel Battery Using the Recursive Least Squares (RLS)

Accurate parameter identification of lead-acid battery models is crucial for developing reliable Battery Management Systems (BMS). This study investigates the identification of electrical parameters in a first-order RC equivalent circuit model of a 12 V, 55 Ah deep cycle gel battery (AVEL AL12V55AH(10HR)) using the Recursive Least Squares (RLS) algorithm. The identification approach is formulated in a linear regression structure derived from the discrete-time voltage-current relationship of the battery model. The RLS algorithm recursively updates the parameter estimates in real time by minimizing the squared prediction error between the measured and estimated terminal voltages. To generate the required excitation for system identification, a series of current step signals is injected via an IGBT-based switching mechanism. Terminal voltage and current responses are recorded across a wide range of the battery’s State of Charge (SOC). Experimental validation is performed using a dSPACE1104-based testbench. The results show that the proposed RLS-based method accurately captures the battery's dynamic behavior and provides consistent parameter estimates under a fixed sampling period and a constant forgetting factor.