PES22 ID13 SOC estimation of Lithium-ion battery: Simulation and Comparative ... скачать в хорошем качестве

PES22 ID13 SOC estimation of Lithium-ion battery: Simulation and Comparative ...

Title: SOC estimation of Lithium-ion battery: Simulation and Comparative study of machine learning-based modelling methods Authors: Saad EL FALLAH, Jaouad KHARBACH, Hicham BEN SASSI, Abdellah REZZOUK, Mohammed OUAZZANI JAMIL Absract: Lithium-ion batteries State of Charge (SOC) estimation is essential for developing electric vehicles. The SOC is one of the most important parameters that must be determined to protect the battery, which is a task conducted by the Battery Management System (BMS). Even though the SOC is an unmeasured parameter; its value can be derived based on other measurable battery quantities, such as discharge current, the voltage across the battery terminals, and temperature. This information are usually fed to state observers or estimation algorithms, which yield a real-time value of the SOC as output. Accordingly, in this paper, different machine learning-based models for SOC estimation are investigated. This includes Ridge, Lasso, ElasticNet, LinearSVR, GradientBoostingRegressor, DecisionTreeRegressor, KNeighborsRegressor and RandomForestRegressor. To evaluate the performance of each several model charge/discharge current profiles were employed to train and improve the robustness of the model. Afterwards, a detailed and critical comparative study between the models mentioned above is performed to identify the best performing model in terms of SOC estimation. In this regard, performance indicators such as the Mean Absolute Error (MAE), the Root Mean Square Error (RMSE), and the Coefficient of Determination (COD) are adopted. The obtained results concluded that the RandomForestRegressor is the best in SOC estimation.