Analysis and Development of a method for state of charge estimation of batteries based on Kalman filters

Nowadays problems related with environmental sustainability are a big social concern. Because of that, have increased the demand of renewable energy sources in substitution of fossil fuels .

The automobile industry started betting on the developement of EV's . This vehicles are environment friendly using batteries instead of fuel fossil.

The technology on batteries have evolved. So batteries, today, have better performance, are smaller and safer. However, this improvement in technology alone doesn't guarantee that the batteries meet the requirements for given application . Batteries continue to be the bottleneck in the production of EV's. In order to improve their behavior is necessary develop also batteries management system (BMS). The BMS is responsible for monitoring the battery cells keeping it operating in a voltage range and appropriate temperatures. Its application will similarly ensure the power required to meet the requirements of the EV [3].

One of the key factors in BMS is the estimation of the state of charge (SOC) [3]. There are several methods of estimation such as the Coulomb Counting, methods based battery models, methods based on neural networks, fuzzy logic and application of the Kalman filter (KF) or its variants for nonlinear systems. This last method is touted as being the one that estimates the SOC with less error [3]. This method gives information about the state estimation and also gives estimation error and can be implemented in real time. KF is a otimal filter for linear system.  Batteries have, however, a nonlinear behavior. Before the application of the filter the equations that describe their behaviour must be linearized around operating point. The two approaches most commonly used to perform this linearization are the extended Kalman filter (EKF) and unscented Kalman filter (UKF) [4]. In the case of application to EVs is still important that the estimation is made in real time once that the user need to know the available energy in is vehicle at each instant.

There are many researchs on the estimation methods and on the implementation of KF and variants, the batteries motivation of the dissertation is that it continues to be a problem with open solution. By the above, it is provided a greater degree of freedom in approach . The issue becomes even more interesting and challenging when considering the impact that an improvement in estimation method because the global EV system .