Real time Prognosis of Battery powered vehicles. A Machine Learning Approach

      Аннотация: In the past decade a lot of effort was put in improving the battery and related technology by researchers across the world for successful commercialization of battery powered electric vehicles. The battery management system (BMS) in an electric vehicle is an integral part of the battery technology for accurate prognostics and catering of battery. In this work we propose and implement a proactive battery management system (PBMS) which estimates the present state of charge and peak power capability of Li ion battery using an Extended Kalman Filter based Mix estimation and communicates it via wireless IEEE 805.15.4 Zigbee interface. The PBMS also learns the profile of discharge of the battery using a dual time frame gradient descent algorithm and predicts the remaining time of operation of the battery pack. Based on the learned profile of discharge the PBMS estimates the energy required for a fixed future duration and discharges the required energy using a set point controller. A case study of the upward slope road profile is made with the PBMS implemented in a laboratory model of electric car.

Ключевые слова: Machine Learning, energy management, battery powered electric vehicles