A Neural Network-Based Robust Online SOC and SOH Estimation for Sealed Lead–Acid Batteries in Renewable Systems

Abstract

Sealed-type lead–acid batteries are most common energy storage devices used with renewable systems. Battery state of charge (SOC) and state of health (SOH) estimation is a crucial part which requires maximum possible accuracy to ensure a secure and long-lasting battery energy storage system by cutting off charging and discharging processes at right time (100–40%). In this research work, a neural network (NN)-based simplified SOC and SOH estimation technique is proposed. Proposed technique provides an accurate online SOC and SOH estimation without battery internal parameter information. This technique only requires real-time battery voltage and current information and very simple mathematical calculations to estimate SOC and SOH, which makes it easy to implement at any low-cost micro-controller unit (MCU). Training data for NN have been acquired by using Arduino mega MCU, voltage sensor circuit, and current sensor. The NN program is designed and trained by backpropagation technique in Arduino mega. The calculated weights are further used for estimation. Terminal voltage $$(V_{\mathrm{t}})$$ and open-circuit voltage $$(V_{\mathrm{oc}})$$ are measured for different charging currents $$(I_{\mathrm{ch}})$$ and discharging currents $$(I_{\mathrm{dch}})$$ . Later on, these data are used for training the NN. Experimental results are provided to prove the preeminence of proposed SOC and SOH estimation technique.