Varo Charge Station Forecasting
Smart Systems
Semester programme:Smart Industry and IoT
Client company:Varo Energy
Bram van den Nieuwenhof
Petran Kuijpers
Jurre Jaspers
Giel Joosten
Nick van Heugten
Bram Laros
Project description
VARO Energy’s current problem entails that they are using a so called “Monkey trader” in order to buy their energy, the monkey trader simply buys the amount of energy that was used the previous week. They want something that will predict the energy use better then the monkey trader.
The goal of the Varo Energy project is to try and add the current algorithm made by a previous project group onto another charging station, the original algorithm was focussed on a big charging station only for trucks. The goal is to add the algorithm made for the truck station onto two smaller charging stations that are available to everyone. The hope from VARO Energy is that the same algorithm could be applied with minimal changes in order to promote standardization.
Context
VARO Energy is a leading European energy company focused on the transition to sustainable energy while maintaining reliable energy supply. The company sources, manufactures, and distributes both conventional (biofuels) and sustainable energies (biogas, e-mobility solutions). VARO is committed to reducing carbon intensity and decarbonizing its operations to meet sustainability goals. The concrete reason for this assignment is to help VARO Energy optimize its energy procurement for electric vehicle (EV) charging points.
Results
If we must make one conclusion for the entire project, we need to look back at the first meeting in week 1 and what the main goal of our project is:
“Is it possible for us (Varo Energy) to predict energy usage with one model, or do we need a different model for every charging station?”
The short answer: Yes, it is possible to predict multiple charging stations using one model.
To give more details about this answer. What we found was that the only thing that really changes here is the predictability of the charging station. To give an example, the Bodaanweg (Truck charging) was much less predictable than the Colosseumweg (Normal car and delivery van charging). This is due to the Bodaanweg having very sudden high charging spikes while the Colosseumweg has a much smoother consumption.
The only thing that changes other than predictability is what the optimal median is. We calculated these for the Bodaanweg and the colosseumweg. They were both different from eachother. However, after discussing with the stakeholder (Remco) we found out that a model in which only the median changes is still considered one model by them. It’s highly likely that the optimal median changes throughout time. So, to keep the optimal median for all charging stations these should be recalculated after a period has passed.