Making the Smart Grid Happen

Research areas

Temporary Supervisor

Professor Sylvie Thiebaux

Description

Everybody has heard about the smart grid: a self-aware, self-organising, self-healing network integrating conventional and renewable (e.g. wind/solar) energy generation, distributed generation, distributed storage of energy (e.g. in batteries of electric vehicles), and giving more control to households to manage their own energy use. This is a critical part of making the planet smarter and fighting climate change. That's the future. The reality is that our power systems are just starting to be equipped with the sensors, meters, and remote controlled equipment, which will eventually enable reliable, efficient, optimal automated operation --- provided we get cracking and find out how to do this!

Goals

There is a wealth of interesting research problems at the intersection of optimisation (artificial intelligence, operations research), machine learning, and control, ranging from short (a couple of years) to much longer term, which we need to solve to turn the dream into reality. So just pick one up.

  • Alarm processing: incidents yield avalanches of alarms which overwhelm control room operators; we need intelligent, efficient techniques to summarise those large logs and determine the root causes.
  • Self-healing: when faults causes loss of power to parts of the network, it must self-reconfigure to isolate the faulty areas and resupply power to the healthy ones, taking into account a range of constraints (e.g. capacity constraints) and optimisation criteria (e.g. as load balancing).
  • Better utilisation of storage: distributed generation and the ability to store power in batteries of electric vehicles enables to shift part of the peak demand to off-peak where electricity is less expensive. But scheduling storage is a complicated optimisation problem which must take into account predicted electricity pricing, predicted usage of vehicles, and generation capacity.
  • Anticipatory Planning: How do we plan for expanding the grid? How do we reconfigure the network to ensure the grid will be easily manageable tomorrow when maintenance works are performed? How do we reconfigure in anticipation of the bushfire threatening parts of the grid?
  • Disaster preparedness and management: with a major storm approaching, how we stockpile repair equipment before and deploy repair crews after the storm hit? How do we restore service to minimise the size and time of the blackout?
  • Decentralised diagnosis and control of large networks: power grids are much too large to be managed centrally. Decentralised diagnosis and control is required to ensure real-time response. However, centralised aspects will need to be integrated to ensure reliability and optimality.
  • Data-intensive management: with all these smart meters and sensors, we will have astronomic quantities of data; how do we best use that much data to forecast demand, improve control and identify infrastructure that needs maintenance?
  • Preference aggregation and mechanism design: how do we combine potentially conflicting customer preferences for demand management and how to we design incentives and pricing formula that encourage appropriate global behavior?
  • User interface and interaction: automated solutions will not be deployed until they are entirely trusted by human; therefore, interfaces are needed to enable the visualisation, understanding and control of the network operations, and to progressively hand over control from the human to the machine.

Requirements

Background in Computer Science, Mathematics, or Engineering, and entry to PhD or Computer Science Hons.

Updated:  1 June 2019/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing