Electricity Networks
Smart Energy Systems Han La Poutré CWI, Amsterdam Centrum Wiskunde & Informatica
& Utrecht University
Future Scenarios: Smart Energy Systems (SES)
Smart Grids
ICT Techniques for Supply/Demand Matching • Some (first) approaches and solutions exist – E.g., first market mechanisms (Powermatcher), first clustering (microgrids), smart powermanagement (factories), ...
• A variety of settings exist – Depending on areas, possible business roles, regulation, etc. – Different solutions needed (as above)
• And: – We want (to keep our) stable electricity supply!
• How to approach this? Questions in – Electrical engineering • Network stability, energy efficiency, controlability, supply/demand matching
– Economics and business science • Who does what (services, control)
– Law • Regulation
– Computer science!!
ICT Techniques for Supply/Demand Matching • Questions in Computer science: High complexity • E.g.: Efficient solutions – Can the ICT system react fast on sudden changes – Can I indeed get all my desired power when I bid? What if not?
• E.g.: Stable solutions – How do I know that using my washing machine this afternoon, with sunny weather, indeed gives low prices? – How do I know that when the washing/drying machine starts, prices remain low for the next hours? Any guarantee?
• E.g.: Robust solutions – Will prices remain low when just one large consumer suddenly may start using power? – What happens if everybody uses the same software with similar decisions?
– (Market) mechanisms that yield sufficient certainty?
ICT Techniques for Supply/Demand Matching • Questions in Computer science: High complexity • Large amounts of data and decisions (computations) – Millions of households in NL; smart meters in network – Tbytes data per day; millions of decisions per day – Aim for decentralized ICT solutions
• Efficient ICT solutions – Efficient (market/coordination) mechanisms for supply/demand matching – Fast response and computation
• Stable ICT solutions – Low uncertainty (of market prices): predictability
• Scalable ICT solutions – For hundreds to thousands to millions of actors
• Robust solutions – Heterogeneous environments: large and small actors, .. – Reliability of ICT performance
– Typical Computer Science issues • Current research
ICT Techniques for Supply/Demand Matching • Coordination and control – Remote control • At homes, by electricity companies – Electric devices (freezer, e‐vehicle, heater)
– Actors respond via market mechanisms • Dynamic pricing • Auctions (several types) • Other market types – Negotiation
– Other forms of organisation • Contracting (day/night, advanced, ...) • Clustering (smart neighbourhoods)
Non‐technical requirement for SES
Some Important Problems in SES
Smart Grids: Supply/Demand Matching
General Problem Domain • Decentralized decision making • Economic and environmental optimization • Decentralized logistics
• Local decision makers • • • •
Limited information Adaptive to dynamic environments Repeated decisions Learning from past
Areas • Application and modeling areas – Decentralized logistics • Cargo transportation
– Patient Logistics • Hospitals
– Energy markets • Decentralized supply and demand
Economic Paradigms in ICT
ICT / Agent Techniques • Software agent • Independent intelligent software module – Can interact autonomously with other agents
• For automatic actions on behalf of a party – E.g., its owner
• Multiagent system – Many independent software agents; ICT system – Coordinating together
• Conceptual use
ICT Techniques
• Coordination between agents – Market mechanisms • Auctions – Continuous, double auctions, first price, second price, ascending bid, …
• Negotiation – Bilateral, multilateral
• Dynamic pricing – Central dynamic price determination
– Efficient and robust market mechanisms • For ICT solutions: bidding agents • Scalable; heterogenous; low uncertainty; robust
– Computable solutions • Strategies of agents • Computation of “winners” (allocations) in auctions
– Other requirements • E.g. as before
• Important field in computer science
• Adaptive, learning agents – Learning from the past • Data, actions, situations • Forecasting
• Computational intelligence techniques – E.g. neural networks, genetic algorithms, hybrid heuristics
Power Generators • Power generators, plannable – Large power plants • Gas, oil, coal, nuclear, ..
– Small (green) power generators • Biomass, batteries (e‐vehicles), μCHPs (μWKKs), warehouse CHPs, ..
– Partly “plannable”: wind parks (weather forecasts), solar cell fields,.. • Delivery or storage
Reduction of Uncertainty for Power Generators • Power generators generate (as in large scale power supply) – Default power • Amount of power determined one day ahead for a specific time slot (half hour)
– Balancing power • Additional mount of power determined “just” before a spefic time slot (half hour)
• When to deliver how much power, for which price?
Reduction of Uncertainty for Power Generators • Actors (agents) submit bids to generate power at a certain price – Default power: 1 day ahead • How much power to generate, in the folowing day at a specific time slot (half hour)
– Reserve power • Maximal additional amount of power to be delivered “just” before a spefic time slot (half hour) (“optional power”)
– 2 independent auctions per time slot • Default power and reserve power • 24 hours ahead
Reduction of Uncertainty for Power Generators • Actors (agents) submit 2 bids to generate power at a certain price – 2 independent auctions per time slot • Default power and reserve power
– Uncertainty about combined allocations • Problematic for planning • Problematic for revenue
– Often default and reserve power generation are related – How to combine these?
Reduction of Uncertainty for Power Generators • Actors (agents) submit 2 bids to generate power at a certain price – How to combine these? – Ratio between default and reserve power • For power generators • For markets (by e.g. network operators) • Chosen, fixed • Combined bid
Power Consumers • Several similar features; e.g. – Power consumption/storage devices • E‐vehicle, freezer, washing machine running at night; or combination
– Prescheduled power consumption • Couple of hours in advance
– Possible additional power consumption, if requested • • • •
Freezer: extra freezing E‐vehicle: extra charging Turn on washing machine When access power is available (uncertain green power supply)
– How to deal with • Uncertainty of reserving power for consumption at specific prices • Unexpected deviations in power availability • Combine both: interconnected consumption
Combined market mechanism • Combined market mechanism with ratio • Advantages – Precise: generators submit only one non‐linear bid – More plannable: generators choose a ratio between default and reserve power – More certain revenues
• More information: Nicolas Höning – Poster session
Comprehensibility of Agent Systems in SES
Comprehensibility of Agent Systems in SES
Some Important Problems in SES
Computer Science Research on SES
Conclusion
• Emerging field: Computer Science research on SES – Last couple of years • After EE and economics
– International results • Conferences, projects, networks
– Organisations • • • • •
NWO/STW SES program EU FP7 Agentschap.nl EIT ICT Labs institute EIT InnoEnergy institute
• Required for SES
• Smart Energy Systems – Novel ICT required – Requirements from • EE, society, regulations, economy, business
– Several basic solutions exist – More advanced solutions underway • Computer science research – Scalability, stability, heterogenous, ..
– Contact:
[email protected]