Estimation of real traffic radiated emissions from electric vehicles in terms of the driving profile using neural networks

  1. Ahmed Mohamed Wefky El Hadi Elezzazy
Dirixida por:
  1. Felipe Espinosa Zapata Director

Universidade de defensa: Universidad de Alcalá

Ano de defensa: 2013

  1. Alfredo Gardel Vicente Presidente/a
  2. Ignacio Bravo Muñoz Secretario/a
  3. Abdelbaset Awawdeh Awawdeh Vogal
  4. Omar M. Ramahi Vogal
  5. Roberto Iglesias Rodríguez Vogal

Tipo: Tese

Teseo: 373765 DIALNET lock_openTESEO editor


The increment of the use of electric vehicles leads to a worry about measuring its principal source of environmental pollution: electromagnetic emissions. Given the complexity of directly measuring vehicular radiated emissions in real traffic, the main contribution of this PhD thesis is to propose an indirect solution to estimate such type of vehicular emissions. Relating the on-road vehicular radiated emissions with the driving profile is a complicated task. This is because it is not possible to directly measure the vehicular radiated interferences in real traffic due to potential interferences from another electromagnetic wave sources. This thesis presents a microscopic artificial intelligence model based on neural networks to estimate real traffic radiated emissions of electric vehicles in terms of the driving dynamics. Instantaneous values of measured speed and calculated acceleration have been used to characterize the driving profile. Experimental electromagnetic interference tests have been carried out with a Vectrix electric motorcycle as well as Twizy electric cars in semi-anechoic chambers. Both the motorcycle and the car have been subjected to different urban and interurban driving profiles. Time Domain measurement methodology of electromagnetic radiated emissions has been adopted in this work to save the overall measurement time. The relationship between the magnetic radiated emissions of the Twizy and the corresponding speed has been very noticeable. Maximum magnetic field levels have been observed during high speed cruising in extra-urban driving and acceleration in urban environments. A comparative study of the prediction performance between various static and dynamic neural models has been introduced. The Multilayer Perceptron feedforward neural network trained with Extreme Learning Machines has achieved the best estimation results of magnetic radiated disturbances as function of instantaneous speed and acceleration. In this way, on-road magnetic radiated interferences from an electric vehicle equipped with a Global Positioning System can be estimated. This research line will allow quantify the pollutant electromagnetic emissions of electric vehicles and study new policies to preserve the environment.