SOC Estimation of Li-Ion Battery Using Hybrid Artificial Neural Network and Adaptive Neuro-Fuzzy Inference System

Abstract

Several applications have used Li-Ion batteries, including automobiles, portable power sources, renewable energy-based microgrid systems, and aerospace. Overcharging or over-discharging a battery may reduce its lifespan. Hence, a precise State of Charge (SOC) evaluation is essential, giving information about how long a battery can be used safely. Since batteries are the most expensive part of most applications, they need a battery management system (BMS) to monitor the values of battery parameters. To have a proper BMS, the determination of SOC is crucial for the battery. It is critical to expect SOC accurately and quickly. Different estimation techniques have been available for SOC estimation. Blending different methods can reduce the possibility of error. This work focuses on the SOC estimation of Li-Ion batteries using data-driven methods of Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Interference System (ANFIS). The proposed methods have been trained using the data taken from a degrading battery for both charging and discharging cycles. The proposed method's results have been compared by analyzing the Root Mean Square Error value (RSME). RSME represents the standard deviation of the predicted value from the actual value. As a result, as the error decreases, the accuracy rises. The simulation validation shows that the ANN method has achieved an RSME of 0.026, and the ANFIS method has achieved 0.0209. A hybrid combination of these two methods produced more accurate and fast results. RSME for the Hybrid model is 0.0004128, which shows that the result is much better than the individual model.