Litcius/Paper detail

Deep Learning Based State of Charge Prediction for EV Battery Packs

Alagar Karthick, Anandan Shailaja, S. Kaliappan, Revathi Rajaraman

20256 citationsDOI

Abstract

Powerful, secure, and long-lasting electric vehicle (EV) batteries rely on precise State of Charge (SoC) assessment. The authors of this study introduce a hybrid model that can accurately predict SoC performance in dynamic environments by combining deep learning methods for the extraction of features with a Support Vector Machine (SVM). A deep neural network is employed in the proposed method to identify intricate nonlinear correlations between the state of charge (SoC) of a battery and several observable variables. For more accurate and robust predictions than standalone approaches, the deep learning model's high-dimensional features are fed into an SVM regressor. Using standard EV battery datasets, simulations, and experiments demonstrate that the hybrid model achieves significantly better results than deep learning and classical machine learning models in terms of root mean square error and mean absolute error. Our findings provide more evidence that next-gen battery management systems could benefit from integrating deep learning with support vector machines to provide SoC estimates in real-time.

Topics & Concepts

Battery (electricity)Deep learningArtificial intelligenceSupport vector machineState of chargeArtificial neural networkComputer scienceMean squared errorMachine learningState (computer science)ObservableNonlinear systemEngineeringCharge (physics)Pattern recognition (psychology)Feature extractionDeep neural networksAdvanced Battery Technologies ResearchElectric Vehicles and InfrastructureElectric and Hybrid Vehicle Technologies
Deep Learning Based State of Charge Prediction for EV Battery Packs | Litcius