|
Abstract
|
In electric vehicles, energy storage systems (ESSs) are essential for managing power fluctuations and ensuring operational safety. Supercapacitors (SCs) have recently emerged as promising ESS candidates due to their high-power density, rapid charge/discharge capabilities, and low internal losses. Integrating SCs with batteries or fuel cells in hybrid configurations can leverage the strengths of each technology while mitigating their individual weaknesses. This paper presents a novel estimation technique for supercapacitors in electric vehicles. The method involves wavelet decomposition and denoising, followed by importing low-frequency signals into a back-propagation neural network for one-step prediction to determine the state-of-charge (SOC) of the SC. The proposed method is tested with a Maxwell supercapacitor model under various charge/discharge current profiles and temperature conditions, comparing the results with conventional techniques. The artificial neural networks (ANNs) with wavelet preprocessed input demonstrate significantly improved SOC estimation accuracy across different discharge profiles.
|