PhD Position F/M Learning and control algorithms for distributed energy resources (Campagne CORDI-S)

Contract type :
Fixed-term contract

Level of qualifications required :
Graduate degree or equivalent

Fonction :
PhD Position

Assignment

With massive and distributed renewable generation and distributed storage (batteries, EVs, electric heating/cooling systems…), there is an urgent need to develop distributed control algorithms for power grids that can scale up to real-time control and coordination of millions of individual agents (e.g. prosumers or devices). The main objective is to propose new distributed control algorithms that can manage system-wide supply-demand balance. The main challenge is in the distributed nature of resources: individual prosumers or devices have limited computation and communication capabilities, and are subject to environmental disturbances, QoS constraints, and privacy issues. Distributed control at such a large scale requires a design of simple signals for millions of devices.
The control architecture will assume a signal that is sent to the individual agents and that reflects the system needs (e.g. demand-supply mismatch). A local agent will then take decisions based on received signal and local measurements and QoS constraints. The main effort will be in the local control design that: respects the needs for the individual agent (prosumer or device), and at the same time leads to the simple aggregate model that can be then used to analyze the overall system behavior.
The approach will combine approximation methods for large populations of agents (e.g. fluid approximations or mean-field setting), optimization, and reinforcement learning.
Assignments :
Extend the approach proposed in [1] to take into account:
    heterogeneity of resources,
    predictions for weather conditions or user behavior,
    predictions for reference signals.
Develop new learning and control algorithms that do not require a known model for the individual agents.
For a better knowledge of the proposed research subject :
[1] Neil Cammardella, Ana Busic, Sean P. Meyn. Kullback-Leibler-Quadratic Optimal Control in a Stochastic Environment. IEEE Conference on Decision and Control (CDC),158-165, 2021.

Main activities

Main activities (5 maximum):
    Litterature study
    Develop new distributed control algorithms
    Implement the proposed algorithms in Python or Matlab
    Write research papers and present the results at seminar talks and conference

Skills

Technical skills and level required: master degree in computer science, applied mathematics or electrical engineering with some courses in optimization and probability, Phython or Matlab, good knowledge or reinforcement learning is optional and highly appreciated

Languages: English, French is not required

Benefits package

• Subsidized meals
• Partial reimbursement of public transport costs
• Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
• Possibility of teleworking and flexible organization of working hours
• Professional equipment available (videoconferencing, loan of computer equipment, etc.)
• Social, cultural and sports events and activities
• Access to vocational training
• Social security coverage