Estimating State of Charge for xEV batteries using 1D Convolutional Neural Networks and Transfer Learning
share
A state of charge estimator is an essential component of battery management systems used in Electric Vehicles. In the recent years deep learning algorithms have fared well as accurate and reliable state of charge estimators, owing to their high accuracy of estimation under noisy conditions and the relative ease with which they can process large amount of data. However, deep learning algorithms are extremely task specific and need to be retrained when the battery data distribution changes. This paper proposes a novel state of charge estimation algorithm consisting of one dimensional convolutional neural networks and also introduces a transfer learning framework for improving generalization across different battery data distributions. The proposed method fares well in terms of estimation accuracy, learning speed and generalization capability.
READ FULL TEXT
