Varo Project
Smart Industry and IoT
Client company:VARO Energy B.V.
Project description
The Varo Project focuses on a network of truck charging stations that supply energy to customers throughout the week. Our mission was to support them in forecasting how much energy is needed in order to purchase approximately the exact quantity of electricity and avoid paying fines.
Context
Varo Energy purchases energy on an hourly basis, adding extra complexity. The shorter purchasing period exposes demand to unexpected fluctuations, which increase the risk of shortage leading to customer dissatisfaction or high extra electricity prices. This makes it challenging to maintain a stable budget at first glance.
Results
- Research showed that features such Weather(temperature) doesn't impact significantly energy consumption in our case.
- Daily prediction algorithms could achieve relatively good performances theoretically
- The project revealed that the dataset lacked sufficient patterns for a reliable algorithm to be modelled.
- Hourly fluctuation context and the short timeframe made it difficult of creating a reliable predictive model.
- A more practical solution is to combine self logic applied with current context based on the Monkey Trader’s (1-week lag) results. This approach has proven during the testing phase that it could consistently outperform the Monkey Trader. At the same time, further research and investigations should focus on identifying and investigating on new featured data that will feed the model and improve its predictive capabilities.
About the project group
Composed of members with a shared ICT and Business background, we worked collaboratively on the VARO Project over four exciting months. The workload was divided equally, with a room for individual initiative. Indeed, all members could propose solutions alongside group efforts, which were then evaluated and voted on. The winning solution was adopted, showcasing both collaboration and innovation.
Our presentations were structured as a clear storyline, building on progress from each sprint to showcase our achievements effectively.
