Evolving Controllers for Electric Vehicle Charging

Abstract

We describe an algorithm to design controllers for the charging of electric vehicles. The controller is represented as a neural network, whose weights are set by an evolutionary algorithm in order to minimize the changes in the overall electrical consumption. The presented algorithm provides de-centralized controllers that also respect the privacy of the owner of electric vehicles, i.e. the controller does not share the information about charging with any third party. The presented controllers also require only a very small amount of memory and computational resources and are thus suitable for implementation in embedded systems.