Edge AI system for EV trip analytics : on-device WLTP class 3B processing and battery temperature prediction

Abstract

The thesis presents a novel edge-native architecture of real-time analysis of driving performance of an electric vehicle (EV) based on a low entry ESP32-S3 microcontroller. The device calculates the important parameters like battery temperature, periods of coasting, battery current efficiency, energy consumption (Wh/km) and regenerative generating energy collection by executing the conventional Worldwide Harmonised Light-duty Test Procedure (WLTP) Class 3B drive cycle. On-device deployment of an Edge Impulse trained TinyML regression model that predicts the battery pack temperature 2 minutes before being depleted is also an important contribution. All the commutated measures are sent over MQTT to a cloud-based Node-RED dashboard to be displayed easily. The framework shows high effectiveness, which has its inference latencies of less than 2 ms and low memory footprint is demonstrated, showing that on-device intelligence is an alternative option to standard cloud-based analytics to enhance EV safety and driving economy.