Putting the “Smarts” into the Smart Grid: A Grand Challenge for Artificial Intelligence

The developed world has grown to its current state of monetary wealth because of the exploitation of cheap energy in the form of fossil fuels.  However, oil is becoming increasingly scarce, with peak production predicted to be reached within twenty years.  To further complicate matters, much of the oil reserves that remain lie in environmentally and politically vulnerable areas.  This makes oil subject to a higher chance of supply side shocks which drive energy prices higher.  Once we do run out of oil, we will need to transition to mostly renewable electrical energy sources such as wind, solar, and tidal energy.  Current projections already estimate the world’s electrical energy demand will increase by 76% by 2030, based on 2007 levels.  In order to accommodate the increased demand in electricity, the grids upon which the transmission of electricity depends must be adapted to the renewable infrastructure.  The new “smart” grids must both integrate widely distributed generators with varying outputs, and manage prosumers, who consume and produce electricity based on their local conditions and requirements.  Perhaps the most striking obstacle in the implementation of a new grid system is the artificial intelligence (AI) that must be developed to control energy flow, as investigated by (Ramchurn et al. 2011).—Donald Hamnett
Ramchurn, S., Vytelingum, P., Rodgers, A., Jennings, N., 2011. Putting the “smarts” into the smart grid: a grand challenge for artificial intelligence. University of Southampton, 1–9.

            Ramchurn and his colleagues organized the challenges involved with creating the smart grid’s AI into five categories:  demand-side management, electric vehicles, virtual power plants, energy prosumers, and self-healing networks.  Each of these possesses its own obstacles.  The researchers took into account the current state-of-the-art technologies to determine the AI advances that must be made to incorporate the aforementioned categories.
            In order for the grid to be safe and efficient, it must perfectly balance supply and demand.   The current grid adjusts supply to meet power demand, but in a renewable grid with varying output, demand-side management would help to reduce demand when energy supply is insufficient.  Artificial Intelligence that responds to price levels, owners’ preferences, and constraints on the grid could be used to flatten this demand curve.  For example, appliance use could be set to a timer which runs during low demand hours.  Also, both individuals on the consumer end and operators of the grid must have the ability to control and predict energy use.
            Another consideration of the smart grid is the future use of the electric vehicle (EV).  These vehicles put a large load on the grid due to the need for a rapid charge capable of a reasonable range of travel.  A system to predict individual and aggregate charging demands of EVs, and incentives to decentralize charging would need to be provided by AI to prevent tripping transformers.
            In the smart grid, virtual power plants (VPPs) are one proposed method of combining heterogeneous actors into aggregates.  AI would be necessary to model the complex interactions of a grid to form VPPs.  One of the difficulties in this regard is that on a constrained grid, individual actions affect all other parties in the grid.  A fair profit-sharing outcome would need to be reached with the most efficient VPPs possible being formed.
            The emergence of the energy prosumer with renewable energy on the smart grid requires AI that can predict prosumer profiles.  The consumption and generation prediction could then be used with price predictions to inform energy trading.  Profit maximization could be reached with such predictions, as could human-grid interactions that take into account prosumers’ preferences.
            A self-healing network requires real-time information to be shared between different nodes, which can coordinate to balance supply and demand.  The AI would need to estimate voltage and phase distribution given prosumers’ demand and supply.  Lastly, the AI would need to make predictions accurately even when faced with incomplete information.
            Switching to renewable energy is not only a matter of energy production, but also one of infrastructure and technology that are much more sophisticated than the current grid, including Artificial Intelligence.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s