Student: Martijn Schoot Uiterkamp
Supervisors: Marco Gerards, Johann Hurink
Programme: Applied Mathematics - University of Twente
When an electric vehicle (EV) arrives at a house, it must be charged so that it can be used during the next day. When the EV is plugged into the house network, it will create a large increase in the total electricity consumption of the house when charging is immediately done at maximum power . This leads to large peaks in the electricity consumption and therefore causes a large amount of stress on and losses in the electricity grid. Therefore, it is better to spread the charging of the EV over a larger time interval to avoid large peaks. Also, on a neighborhood level, it is wise to take into account the charging of several EV’s at the same time, so that we avoid large peaks at the neighborhood level as well. For this last issue, a neighborhood controller can be installed, which tries to flatten the electricity consumption on the neighborhood level. To do this, it sends steering signals to the houses in the neighborhood; each house can use the received steering signal to plan its electricity consumption so that the total consumption is flattened.
An example of a steering signal that has proven itself to be very effective is a target profile. It explicitly states the amount of electricity that, ideally, should be consumed by the house at each point in time. The house is given the task to match its electricity consumption to the received target profile as good as possible. Therefore, the charging of an EV should be done in such a way that the deviation of the house’s total electricity consumption (house demand plus charged amount) from the target profile is minimized.. Several algorithms have already been developed to solve this problem to optimality when the house demand at each time is known. However, this is not the case in practice. Most existing planning algorithms for the EV charging problem therefore base their planning on predictions of the house demand. Unfortunately, it appears to be very hard to accurately predict the house demand at a certain point in time. The goal of this master assignment is to design new and improve existing online algorithms for EV charging that are robust against predictions errors in the house demand and have a fixed performance guarantee.