Testing machine learning algorithms for intelligent distributed charging management at company locations.
What is the problem? Avoiding load peaks to reduce load peaks and power price costs and meeting the requirements of the grid operator. Current technologies are limited in particular in terms of costs, flexibility, and complexity management.
What problems emerge? Simultaneous charging processes lead to high costs or high grid loads. Available solutions are expensive, unsuitable, or (economically) inefficient. In the worst case, this leads to high costs due to charging processes and the necessity of increasing the power at the grid connection point.
What is the perfect solution to the problem for the target group? A simple, self-learning charging management system with low procurement and operating costs.
What exactly is the solution offered by the K-ES Spotlight and how is the problem solved? Developing software solutions for intelligent distributed charging management for optimum utilization of the existing flexibility while taking the local topology into account, all based on self-learning algorithms.
The project is interesting to:
Providers of energy and charging management solutions, utility companies, fleet operators, operators of charging infrastructure
Developing self-learning algorithms for intelligent charging management at company locations
Developing a cloud-based software solution involving low computing power requirements and costs
Testing in a real environment (company location)
Deep reinforcement learning: Self-learning algorithms that use information about the existing charging infrastructure as well as data points of completed charging processes to learn independently i) to optimize charging processes and ii) shift charging processes