The State Of Charge estimating methods for rechargeable Lead-acid batteries

State of Charge (SOC) is a key element for battery energy assessment, performing the stored energy. An accurate estimation of the SOC is fundamental for the safe and reliable operation of photovoltaic systems. To this end, the scientific literature covers a broad range of methodologies with extensive accuracy and complexity. However, the accuracy of the SOC is highly dependent on the methodology adopted. This paper investigates four methods of estimating the SOC for lead-acid battery manufacturers. For this purpose, four methods were selected and then used in practice, including the Modified Coulomb Counting (MCC) method, the Neural Network (NN), and two other machine learning based techniques, namely the Support Vector Machines (SVM) and the Nearest Neighbours Algorithm (KNN), respectively. An experimental test is considered through a detailed analysis based on a statistical evaluation covering the real cycle of charge and discharge modes. It means that the NN algorithm has given more accuracy. The algorithm showed high prediction accuracy, the majority of predictions had a relative error close to zero, reaching a maximum error value of about 0.1%. Simulation results were performed in Matlab software. All of these results confirm the accuracy and efficiency of the chosen methodology.