Optimization of a vehicle’s electric fan parameters through evolutionary algorithms aiming a higher energy efficiency
Palavras-chave:
Genetic Algorithms, Cooling System Simulation, Electric Fan ControlResumo
In current automotive projects, the presence of new technologies aimed to maintain a better energy
efficiency and, consequently, a reduction in fuel consumption is very common. Most of these technologies need a
specific calibration, in order to increase the efficiency of the component, keeping its performance within acceptable
limits. Regarding cooling system components, the electric fan driven by Pulse Width Modulation (PWM) is an
example of technology used to enhance the vehicle’s energy efficiency with less demand from the alternator and,
consequently, less engine torque. Its calibration, however, is often performed in an experimental way, and many
tests are performed until satisfying values are found for the calibrated parameters. The present work aims to use a
computational model of the cooling system to adjust the fan control parameters of a vehicle controlled by a Fuzzy
Logic based strategy. The optimization of the control parameters is performed through three different genetic
algorithms - Standard Genetic Algorithm (SGA), Differential Evolution (DE) and Particle Swarm Optimization
(PSO) - which are compared with each other and with the Nelder Mead algorithm (considered as a reference
for comparison, since it is the algorithm used in previous works). The main purpose is to show a comparison
of the coolant temperature, the percentage of the Duty Cycle and the energy spent by the electric fan, using the
computational model, with each of the optimization algorithms discussed, being the energy the main parameter
to be considered in the comparison. The results demonstrate a reduction of energy spent by the electric fan in
relation to the reference for all the studied genetic algorithms, being the PSO and DE algorithms presenting the
better results, with an energy reduction percentage of the order of 15%, followed by SGA algorithm, that presents
an energy reduction percentage of the order of 12%.
