Application of Neural Networks in a Sodium-Nickel Chloride Battery Management System

The main objective of this research is to employ artificial neural networks in battery management systems (BMS), considering that one of the main applications of artificial neural networks is to operate as a function approximator, mapping the functional relationship between the variables of a system, from set known samples. In this context, this work approaches a method to predict the state of charge of batteries using techniques of artificial neural networks through a real database and models of the charge curve of real sodium—nickel chloride batteries. The behavior of the BMS is analyzed, using the models found in the output curves, in which the methodology used is implemented in the MATLAB® software to obtain the load curves. In this case, the proposed method uses a Multilayer Perceptron artificial neural network, which is a Feedforward architecture with backpropagation training algorithm. Finally, it was verified that the performance of the algorithm is closely linked with the adjustments in the number of layers. With the proper adjustments, the results expressed an excellent ability to indicate the battery charge status, as well as the analysis of the stipulated errors. Besides that, it is important to highlight that the methodology can be applied to other types of batteries, such as lithium battery.