Abstract. Neural Networks have been used in system control, medicine, pattern recognition and business. The backpropagation neural network (BPNN) appear to be most popular and have been widely used in many applications. BPNN is a supervised learning technique for training multilayer feedforward neural networks. The gradient or steepest descent method is used to train a BPNN by adjusting the weights . The purpose of update numerical weights is minimize error of network between target and output. In this paper, focus with BPNN modeling with data battery for training and testing. We used discharge and Urban Dynamometer Driving Schedule (UDDS) as training data and testing data, respectively Architecture of BPNN consist of input layer, hidden layer and output layer. The otherhand, using BPNN has problem to define amount of hidden neurons. In this study, we used current or voltage as input in input layer, one hidden layer with 8 neurons and one output layer. We used Levenberg-Marquardt algorithm to get fast iteration when computation. The experiment used 2200 mAh of LiFePO4 battery. Result of this research show that Mean Squared Error (MSE) value when current as input and voltage as target is 0.021135 with regresion is 0.626. Then MSE value when voltage as input and current as target 0.029925 with regresion is 0.5213. In this study relationship between voltage and current battery is nonlinear.

After decades, the battery usage has been widespread for many applications, especially in the field of Electric Vehicle (EV). The battery is a very important component in the EV. Because the battery as the primary power source replacement of the fossil fuel. Therefore, the condition of the batteries should be always in good condition. To prevent failure of the battery for battery management system (BMS) is needed. BMS is a system to regulate the use of the battery and protects the battery from the failure of the battery supply. Many factors can be monitored at BMS, one of which is a State of Charge (SOC). SOC determination is directly related to the estimated OCV (Open Circuit Voltage). The accuracy of the estimation algorithms depend on the accuracy of the model selection to describe the dynamic characteristics of the battery. This study begins with the selection of the right model (fig.1, fig.2, fig.3) for estimating OCV. Selection of appropriate model using RLS algorithm for estimate the battery terminal voltage. Parameter that reference for determining the selection of the model is the max, min, mean, RMSE, mean RMSE of the error. Later models have been used to estimate the OCV. The result based on this research shows that modeling with n = 1 is the best result to be used in model parameter estimation and OCV battery in term of the smaller max, min, mean, rmse error. This research also show us that RLS algorithm can be estimate the parameters of the batery, OCV (fig.4), and terminal voltage of the battery with an error less than 0.1%