Optimisation Algorithms for an Electric Smart Grid Controller

Contact: Wajeb Saab

Background:

The major concern brought about by the increasing penetration of renewable energy sources (PV, wind), is the grid’s inability to handle these new distributed and volatile energy generators. Conventional grid control methods reach their limitations in this regard. Commelec is a recently proposed framework for controlling electrical grids in real-time [1]. In addition to addressing the aforementioned issues, it can scale up to control a grid of arbitrary size.

One main aspect of Commelec is that it employs a separation of concern. There are two types of agents in Commelec: simple-minded resource agents (RAs) and smart grid agents (GAs). RAs are specific to each type of resource (PV, Battery) and their main functionality is to map the internal state of the resource into a common device-independent language. This allows the implementation GA to be generic, independent of the topology of the grid that it controls. Moreover, the job of the GA is reduced manipulating abstract data and solving a purely mathematical optimisation function.

A nice 7-minute video presenting Commelec can be found on this page.

The optimisation function to be solved by the GA has been developed, and several solutions are in place to solve it in real-time (10-50 milliseconds). These solutions mainly employ gradient descent, set projection, and load flow analysis. However, challenges and interesting open problems are still present in several aspects of the solution. The goal of this project is to design and implement new algorithms and compare them with existing ones.

Challenges:

  • Optimisations to be performed over sets with no explicit representation
  • Optimisations to be performed over sets that might not be convex

Project Goals:

  • Understand the Commelec framework
  • Review existing solutions, implement them, and integrate them with the GA code
  • Propose and design novel solutions and improve upon existing ones

Required skills:

  • Optimisation, Linear Programming, Geometric Analysis, Set Theory
  • C/C++ skills or equivalent programming skills
  • Interest in Power Grids

Benefits :

  • Direct impact on a real-life project: Commelec will be deployed in the Swiss city of Rolle, providing real-time control for a grid of 600 households
  • Potential for a PhD thesis with the lab on the Commelec project

References:

[1] Bernstein, Andrey, et al. “A composable method for real-time control of active distribution networks with explicit power setpoints. Part I: Framework.”Electric Power Systems Research 125 (2015): 254-264.

Supervisors: Roman Rudnik, Wajeb Saab