Orchid Project

Developing a modern electricity grid where information flows in both directions between consumers and producers is critical to achieving the UK’s ambitious carbon reduction targets. Human-agent collectives are an essential part of this vision, and we are seeking to develop agents (or ‘energy avatars’) that are capable of continuously monitoring, predicting and feeding back information about energy generation and consumption within the grid, in order to satisfy individuals’ preferences for cost, carbon and comfort.

To address these aims we are developing (i) coalition formation algorithms that allow multiple self-interested parties such as renewable generators to come together with consumers to create virtual power plants that can more effectively manage the intermittent nature of these energy sources and (ii) algorithms to generate effective short term predictions of demand and supply to allow the optimisation of energy use.

Learn more about Smart Grid Technologies

Smart Grid Management in 2050 vignette

It is 2050 and anthropogenic global warming is both a well established scientific fact and an observable phenomenon to billions of people worldwide. Under the UN Framework Convention on Climate Change, the 2040 Beijing Protocol has just been ratified. This subjects all countries that have not yet achieved an 80% reduction in CO2 emissions (compared to 1990 values) to economic penalties and is severely curtailing the burning of fossil fuels. Electricity in the UK is generated from an ageing fleet of nuclear power stations,from solar farms in the South of Spain and in North-African deserts, from offshore wind farms around the coast and from micro-renewable sources (solar PV, tidal and wind) deployed across the country with the majority of homes being able to export electricity back to the grid. Heating in houses and commercial buildings is increasingly electrified through the use of ground and air source heat pumps (with biogas making up the shortfall) and transport is almost completely electrified after the peak in worldwide oil production in 2030 saw oil prices soar to $500 per barrel.

Due to this increased demand for electricity and the variable nature of renewable energy sources used to generate it, there is barely enough electricity to satisfy demand at all times. The energy profligacy of the first decade of the century is a distant memory and energy is a scarce resource. There are regular rolling blackouts as the grid fails to meet demand at different periods of the day. In particular, at times when there is less wind around the UK and increased cloud cover over southern Spain, grid operators rely on ageing gas-fired power stations that incur economic penalties that are passed directly to customers, leading to rising consumer dissatisfaction. To make matters worse, electric vehicles are putting unexpected loads on the grid, whose fundamental infrastructure was designed to feed energy from a small number of generators to a large number of consumers. In particular, as consumers charge their cars at their workplace, transmission lines delivering electricity often reach their thermal limits and eventually trip transformers, leaving consumers stranded with their disabled vehicles.

All houses are now equipped with smart meters and electricity is charged in real time, with feedback about home owners’ use of rooms and activities being used to produce real-time estimates of future demand. However, consumer protection agencies are increasingly concerned about the privacy and rights of users and are protesting against the invasive use of these meters and the ways in which energy suppliers are attempting to influence the consumer behaviour by imposing excessively severe, and unannounced, critical pricing periods. In response, many consumers are simply downloading code and ’hacking’ their meters (easily done since the national rollout of smart meters was completed in 2020 and the security features incorporated in these meters are now trivial to break) in order to subvert the pricing policies of the energy suppliers.

It is this dystopian vision of a future energy scenario that we hope to mitigate through the development of HACs within ORCHID. By building systems in which energy users and producers can coordinate and cooperate we hope to avoid blackouts by appropriately prioritising and scheduling energy use and production within the network. By devolving this coordination to agents (or ’energy avatars’ ) that are capable of continuous monitoring, predicting and feeding back information about energy generation and consumption, the individual preferences of both energy consumers and energy suppliers will be better satisfied.

Flexible Autonomy

Consumers, suppliers and grid operators need to work closely with a variety of computational agents. At the consumer level, people may allow software agents to manage energy usage in their house (e.g. through deferment of loads and storing electricity) and the purchase of short-term dynamically negotiated energy contracts from the grid. At the supplier and grid operator levels, agents will be responsible for trading energy in spot markets and identifying and correcting network faults respectively. This requires both humans and agents to have the same understanding of the on-going activities and issues (possibly requiring a dialogue between them or the use of sensors) and will require novel modes of interaction so humans can describe their future activities and desires to the agents. Similarly, when a grid operation agent detects its automated fault correction mechanisms are bringing the network to its stability limits, it may devolve control back to a human operator who will have ultimate responsibility to shed loads to make sure that the system is safe. In both cases the degree of autonomy given to agents needs to be defined, refined according to the prevailing circumstances and tested to evaluate its impact on the lifestyle of the consumer and the economic and system benefits achieved.

Agile Teaming

Given the number of small energy providers there is an opportunity for them to come together to form Virtual Power Plants (VPPs) that can adjust their overall production of electricity to match current demand. Conversely, like-minded consumers may form energy cooperatives to aggregate the demand profiles of their members into a more predictable profile in order to negotiate cheaper electricity. In both cases the formation and on-going operation of such HACs is a major challenge given that interactions in the collective may be affected due to misunderstanding or mistrust and the uncertainty underlying their joint production, or demand, of electricity. This may mean that some VPPs come together for a short period only and then break apart to form other VPPs that better suit demand at other times, Given all the actors in the system are connected through the grid, the behaviour of a given VPP/cooperative is likely to affect the performance of other teams in the system. This interdependence compounds the problem of sharing profits generated by a given team among its members to ensure they stay within the VPP or cooperative.

Incentive Engineering

The scenario highlights the need for reliable data to be provided by the smart meters to predict demand and for incentives to be provided to consumers to influence demand. Consumers also need to be incentivised to inform their smart meter, and hence the grid, of their activities in order to efficiently use energy in the home and allow better prediction of demand. This means consumers should benefit from providing energy suppliers with additional information, rather than being co-opted to do so through closed protocols. This will only work well if the incentives are correctly designed, understood and valued by the consumers. However, consumers may not participate in such exchanges if they do not trust the mechanisms or wish to maintain the privacy of their energy usage. In addition, unless appropriate safeguards are in place, consumers may be enticed to misreport their usage profile in an attempt to join the cheapest cooperatives. Hence there is a clear need to reward truthful revelation and design persuasive technologies to convince consumers to relinquish at least some of their private information and to motivate them to change their behaviour.

Accountable Information Infrastructure

Information about possible changes in demand and supply at different points in the grid, as well as the thermal state of transmission lines, needs to be aggregated by grid operators to guarantee that energy delivery is secured. This requires both metering of specific points in the network and collecting consumer usage profiles. This represents a large amount of real-time data that may be corrupted, delayed or lost in transmission, incomplete due to consumers’ privacy restrictions or the target of a potential cyber-attack. Critical decisions are likely to be made on the basis of this information, so it is important that there is an auditable trail of all decisions made within the system to maintain the accountability of the system and to ensure humans can trust their agents. Computationally efficient techniques are required to search, store and make inferences over such large amounts of incomplete and possibly incorrect data and to ensure that the output of situational awareness process is robust to misrepresentations by certain actors. The information provided to users on their smart meters will need to be proven to come from trustworthy sources in order to be credible and auditable for billing purposes. This information will also need to be easily understandable and verifiable by home owners.

agents | citizen-science | applications | crowdsourcing | disaster response | smart grid | accountable information architecture |agent-based computing | agile teaming | disaster recovery |flexible autonomy | decentralised control | energy management |HACs | human-agent interaction |human computer interaction | human agent collectives | incentive engineering | machine learning |mechanism design | ORCHID |