Transcript Slides

Trading Agents in the Smart
Electricity Grid
Perukrishnen Vytelingum, Sarvapali D. Ramchurn,
Thomas D. Voice, Alex Rogers and Nicholas R. Jennings
University of Southampton
Background: The Wholesale Electricity Market
• Two-stage mechanism typical in the wholesale electricity
market:
– Day-ahead mechanism
– Real-time balancing mechanism
iDEaS : Intelligent Decentralised Energy-Aware Systems
2
Background
• The popular electricity market mechanisms are:
– The LMP (Locational Marginal Pricing) mechanism (e.g. in different areas in US)
– The pool-based mechanism (e.g. the National Grid in the UK and the Nord Pool in
Scandinavian countries)
• We focus on the former where pricing is based on nodal location within
the grid.
• A supplier buys electricity in the wholesale market for its consumers.
• Generators sell electricity in the wholesale market.
• Transmission lines with a capacity
• The suppliers submit their demands, generators their supply. Thereon,
the system operator computes the optimal allocation (that maximises
the system utility subject to physical constraints of the lines).
iDEaS : Intelligent Decentralised Energy-Aware Systems
3
Motivation
• Because the LMP mechanism assumes truthful behaviour of
traders, the system can be gamed.
• The most effective way to deal with this shortcoming is
regulation of the traders.
• With more and more generators in a future of microgeneration, regulation of every generator becomes less
feasible.
• A mechanism that does not make such an assumption (of
truthful behaviour) would be robust against gaming and
would not require regulation.
iDEaS : Intelligent Decentralised Energy-Aware Systems
4
Objective
• We first model the smart grid (at a micro-level) as a multiagent system with selfish, profit-motivated traders and a
system operator.
• Given these different players, we need to develop a novel
market mechanism for the wholesale electricity market (at
the day-ahead and real-time stage) that does not assume
truthful behaviours of agents.
• Our mechanism should work both at a day-ahead and realtime level.
• Finally, we need to benchmark our mechanism with the
current LMP mechanism as providing an optimal (assuming
complete and perfect information from traders).
iDEaS : Intelligent Decentralised Energy-Aware Systems
5
The Novel Electricity Market Mechanism
• Based on the Continuous Double Auction. Similar format as in
financial markets like NASDAQ.
• Combines the economics of markets and the physics of
electricity transmission networks.
• Consists of three main parts:
– The trading mechanism (day-ahead)
– The security mechanism
– The online balancing mechanism (real-time)
• Agents can use the Zero-Intelligence strategy or the novel AAEM strategy (designed for our market mechanism) to trade in
the market.
iDEaS : Intelligent Decentralised Energy-Aware Systems
The Trading Mechanism
• Multiple buyers and sellers are allowed to compete and
submit orders at any times.
• Bids and asks are queued in bid and ask orderbooks.
• Defined by the trader id, quantity they require, maximum
price the buyer is will the pay and minimum price the seller is
willing to take and the node position in the grid.
iDEaS : Intelligent Decentralised Energy-Aware Systems
The Trading Mechanism
• Market mechanism is defined by its set of protocols:
– The quote-accepting policy
– The clearing policy
– The pricing policy
iDEaS : Intelligent Decentralised Energy-Aware Systems
The Security Mechanism
• Ensures the system is secure at all times by moderating the
total demand and supply at each node of the electricity grid.
• The volume of each additional trade (i.e. a match between a
bid and an ask) is moderated for system security, i.e. none of
the transmission line overloads.
– We use DC approximation flows to calculate the resulting flows from a
potential trade and ensure they do not exceed line capacity.
iDEaS : Intelligent Decentralised Energy-Aware Systems
Pricing of Transmission Lines
• Information about the transmission lines in made
public:
– transmission cost (per unit) for each transmission line.
– Current through each line (given all the trades in the
grid).
– The capacity of the line
• Given the flows through the grid, we can price the cost
of transmission for each potential trade.
• Furthermore, given the flow from a potential trade of
+1, we can calculate the buyer’s and seller’s price for a
quantity of +1. We term this value the DLMP for buyers
and sellers.
iDEaS : Intelligent Decentralised Energy-Aware Systems
The Trading Mechanism
• Market mechanism is defined by its set of protocols:
– The quote-accepting policy
– The clearing policy
– The pricing policy
Continuous Clearing
iDEaS : Intelligent Decentralised Energy-Aware Systems
The Trading Mechanism
• Market mechanism is defined by its set of protocols:
– The quote-accepting policy
– The clearing policy
– The pricing policy
bid
Buyer’s transaction price
Transmission cost for trade
Seller’s transaction price
ask
iDEaS : Intelligent Decentralised Energy-Aware Systems
The Online Balancing Mechanism
• Real-time balancing of demand and supply using offers from
the bid and ask orderbooks:
– Buyers bought less than needed have to cover their short position by
buying from the ask orderbooks.
– Buyers that bought more than required have to cover their long
position by selling the extra power.
– Sellers that sold more than they can produce need to buy that power
from other sellers to cover their short position.
• By having to cover a long or short position, traders end up
with poorer prices at the balancing-phase where the low bids
and high asks remain.
• Incentivised to accurately predict demand and supply.
iDEaS : Intelligent Decentralised Energy-Aware Systems
Empirical Evaluation
• Within 92% to 99% of optimal (where optimal assumes
complete and perfect information)
• Evaluated for different topologies
• Optimal LMP breaks down with malicious agents
iDEaS : Intelligent Decentralised Energy-Aware Systems
Empirical Evaluation
• Increased efficiency with learning transmission lines.
• The emergent effect of the learning behaviours is a less
congested grid and more efficient allocation.
iDEaS : Intelligent Decentralised Energy-Aware Systems
Conclusion
• We designed a novel trading mechanism that combines the
economics of financial markets and the physics of the
electricity grid.
• Buyers and sellers can trade electricity in a day-ahead and a
real-time market subject to transmission line constraints.
• We empirically demonstrate a high efficiency without having
to assume that agents are truthful, risking gaming in the grid.
• We show that with a learning mechanism, the system
operator can adapt the transmission line charges for
congestion control in the system and improve system
efficiency.
iDEaS : Intelligent Decentralised Energy-Aware Systems
Questions?
17