On-Road Magnetic Emissions Prediction of Electric Cars in Terms of Driving Dynamics Using Neural Networks

Abstract

Abstract-This paper presents a novel artificial neural network (ANN) model estimating vehicle-level radiated magnetic emissions of an electric car as a function of the corresponding driving pattern. Real world electromagnetic interference (EMI) experiments have been realized in a semi-anechoic chamber using Renault Twizy. Time-domain electromagnetic interference (TDEMI) measurement techniques have been employed to record the radiated disturbances in the 150 kHz-30 MHz range. Interesting emissions have been found in the range 150 kHz-3.8 MHz approximately. The instantaneous vehicle speed and acceleration have been chosen to represent the vehicle operational modes. A comparative study of the prediction performance between different static and dynamic neural networks has been done. Results showed that a Multilayer Perceptron (MLP) model trained with extreme learning machines (ELM) has achieved the best prediction results. The proposed model has been used to estimate the radiated magnetic field levels of an urban trip carried out with a Think City electric car.

Extreme Learning Machines with Simple Cascades

“…See Bin (et al, 2011) for a comparison of Echo State and ELM. We also note that Wefky (et al, 2013) define cascade networks in general as we define our shallow cascade network topology, and Tissera and McDonnell (2015) use a layered auto-associative structure which could be described as a cascade network, though they do not express this in their paper.…”