Abstract David Metivier

Abstract: Mean Field Control and Disorder for Efficient Mixing of Energy Loads

Introduction of renewable energies in power systems have forced electricity management to become more flexible both for the generation and consumption. Demand Response (DR) is a control strategy aiming to address this challenge by adapting and controlling in real time resources to the demand. In particular, non-critical ensemble of loads can be used as "virtual batteries'' for a few instants, by simply be turned Off. Aggregates of Thermostatic Controlled Loads (TCLs) are good candidates to act as virtual batteries. They are cycling electric devices that switch On/Off depending on some measured parameter (typically  temperature). These loads are widely used (air conditioners, fridges, pool pumps, ...) and due to their large inertia, are interesting for DR. Approached from the standpoint of Statistical Physics, an ensemble of TCLs represents a non-equilibrium system driven away from its natural steady state by DR perturbations. After introducing the model describing large aggregate of TCLs via coupled Fokker-Planck equations, we will explore the following points: i) how randomness makes the system resilient, mixing it toward a steady state; ii) how a Mean-Field Control, simple and private, can be implemented based on an observable quantity, improving ensemble resiliency even more; iii) how disorder (variability) in the TCLs ensemble affects the system.