Sensor Drift Fault Diagnostics and Control with Temporal Adaptive Range Prediction for Electric Vehicles

Abstract

ABSTRACT: Machine learning is vital in the electric vehicle industry, but challenges remain in addressing fault diagnosis errors and range prediction inaccuracies. The "Sensor Drift Fault Diagnostics and Control with Temporal Adaptive Range Prediction" technique addresses these issues. Existing ML algorithms for fault diagnostics can cause unpredictable vehicle behavior, such as sudden acceleration or braking. The Hierarchical Iterative Proximal (HIP) Diagnostic and Control Algorithm mitigates these anomalies by accounting for sensor mounting drift. Most current range prediction techniques inaccurately estimate the battery energy consumption rate, leading to discrepancies between estimated and actual driving ranges. The Adaptive AReXo network, a novel hybrid range prediction approach, enhances range prediction in EVs by considering discharge and C-rate variability. These techniques improve the accuracy of battery energy consumption estimates and range predictions, yielding high accuracy and low RMSE in range prediction and anomaly behavior.