Distributed optimization of energy consumption for electric vehicles with Enervalis

Sales of electric vehicles are on the rise. This poses a number of challenges for aging grids all across Europe by introducing new peak loads on already stressed grids. The problem is exacerbated by an increasing contribution of renewable energy sources. At the same time, EVs can also form part of the solution as mobile loads which offer flexibility to the grid by offering demand side management. In this context, it is necessary however to balance a number of (often competing) constraints foremost of which is meeting user specifications. These additional constraints include maximal self-consumption of renewable generation (e.g. solar), improving grid quality, taking advantage of market signals directly or by providing flexibility of consumption etc.

The optimization problem is tractable for a single vehicle but becomes more convoluted when applied to many (hundreds or even thousands of) vehicles, each with its own set of constraints. Furthermore, it is problematic both from a logistics (providing communication framework) and privacy (sharing user data) perspective to do completely centralized control. A number of techniques have been proposed to solve for this problem in a general setting ranging from classical optimization algorithms to meta-heuristics, distributed constrained optimization and multi agent reinforcement learning.


For more information, contact Dirk Fahland.