Premium
A hybrid multi‐output‐predictive modelling based NSGA II approach for dimensions design optimization of battery pack module for electric vehicles
Author(s) -
Ruhatiya Chaitanya,
Gia Bao Pham N.,
Quan Tram L.,
Tho Quan T.,
Xinyu Li
Publication year - 2020
Publication title -
energy storage
Language(s) - English
Resource type - Journals
ISSN - 2578-4862
DOI - 10.1002/est2.130
Subject(s) - battery (electricity) , battery pack , decision tree , perceptron , multilayer perceptron , computer science , linear regression , regression analysis , regression , engineering , artificial neural network , artificial intelligence , machine learning , mathematics , statistics , power (physics) , physics , quantum mechanics
Battery module safety is a major concern for the commercial success of electric vehicles (EVs). Concurrently, it is also important to have a mechanically sound and ergonomically fit battery pack design. To solve this problem, a hybrid multi‐output‐predictive modelling based NSGA II approach is proposed. In this approach, the multiple predictive modelling methods (linear regression, regression with AdaBoost, decision tree regression and multi‐layer perceptron [MLP]) are applied to predict the deformation, natural frequency and mass of battery pack enclosure. By performing the comparative analysis of these methods, the decision tree regression model was selected for deformation, MLP with tanh function for frequency and MLP with ReLU function for mass. Further, these selected models were optimized using NSGA II which resulted in optimum combination of input variables for achieving the maximum deformation (0.0019 m), minimum natural frequency (91.60 Hz), and mass (12.41 kg) simultaneously.