Data Driven Parameter Estimation for Battery Modelling Using System Identification Technique

The battery models have become an important part for the design of battery-powered systems. The first step in the development of an accurate battery model is to build and parameterize an equivalent circuit that reflects the battery’s nonlinear behavior and dependencies on temperature, SOC, SOH, and current. These dependencies are unique to each battery’s chemistry and need to be determined using measurements performed on battery cells of exactly the same type as those of the controller. This paper gives an overview, methodology and result on data driven battery model using system identification technique. Derived by the common approaches using standardized tests, e.g. hybrid pulse-power capability (HPPC), with predetermined charge and discharge sequences, the existing parameter based battery model are in use. Equivalent linear circuit battery models of different complexity were then tested and evaluated in order to identify parameter dependencies at different state of charge levels and temperatures. This battery model generally works fine at its beginning of life, but as battery ages i.e. at Middle of life and End of life of the battery this battery parameter tends to change. In order to have accurate estimation of the battery consumption, losses and behavior of parameter over the time, battery parameters need to be continuously adjusted, corrected and estimated. Hence, battery parameter estimation is of utmost importance. This technique will be helpful to extract the battery parameter from vehicle battery data and tune it’s parameter. Simple battery models like single RC equivalent circuit model and R-equivalent circuit model produce goodness of fit at average 80 % or higher, if computational complexity is not be issue then battery model can be done for 2RC equivalent or PNGV(Partnership for New generation of Vehicles) or extended PNGV model.