#1 Fix the energy market
Re-architect how energy is priced, procured, and coordinated across the grid.
Required outcomes — What “good” looks like
What “good” looks like
A successful electricity system cannot be judged only by whether it produces enough energy at the lowest theoretical cost.
The modern grid is a cyber-physical, socio-technical and economic system.
It combines:
- physical infrastructure and electrical flows;
- software, communications and automated control;
- markets, prices, contracts and investment mechanisms;
- institutions, operators and regulators;
- millions of households, businesses and distributed assets;
- human decisions, preferences, behaviours and perceptions of fairness.
Any proposed market design must therefore be evaluated across all of these domains simultaneously.
A mechanism that is economically elegant but physically infeasible is not a valid solution. A mechanism that is technically operable but socially unacceptable will not secure the participation on which it depends. A mechanism that produces desirable short-term dispatch outcomes but cannot recover the costs of the system will not support investment or remain viable over time.
The objective is not simply to optimise one variable. It is to design a system that remains feasible, trusted, investable and operable while delivering the outcomes society requires.
The system has changed
Historically, electricity demand was treated largely as passive.
Consumers used electricity when they wished, generators were instructed to follow demand, and most operational decisions were taken by a relatively small number of central actors.
That model is changing.
As the system becomes more dependent on variable renewable generation, distributed energy resources, electric vehicles, heat pumps, batteries and flexible demand, consumers are increasingly expected to play an active role in balancing the system.
People may be asked to:
- move consumption between time periods;
- allow devices to charge or discharge automatically;
- accept temporary limits on flexible loads;
- provide services through batteries, vehicles or heating systems;
- respond to prices, incentives or service-level offers;
- permit third parties to control or schedule devices on their behalf.
Demand is therefore no longer simply passive load. It is becoming an active system resource.
This moves electricity-system design into the fields of behavioural economics, behavioural science and human-centred system design.
Traditional assumptions that people are homogeneous, perfectly informed and consistently rational are not sufficient. People have different needs, constraints, capabilities, levels of knowledge, attitudes to risk and willingness to participate. Some will actively optimise their energy use. Others will require automation. Some will value savings; others will value convenience, predictability, privacy or control.
A market or control architecture that depends on human participation must represent these realities explicitly.
The system must represent its real characteristics
Any candidate solution must begin with an accurate representation of the system it is trying to coordinate.
That means representing:
- where electricity is produced and consumed;
- when energy and capacity are available;
- the physical limits of networks and assets;
- the variability and uncertainty of supply and demand;
- the controllability or uncontrollability of different resources;
- the operational state of devices and infrastructure;
- the behaviour and preferences of participants;
- the incentives created by prices, contracts and rules;
- the distribution of costs, benefits, risks and reliability.
A market design should not depend on assumptions that are no longer true.
In particular, it should not assume that:
- demand is passive;
- all consumers respond to price in the same way;
- all participants have perfect information;
- people continuously make economically rational decisions;
- electricity can always be transported from where it is produced to where it is required;
- import and export can be treated using identical one-directional logic;
- financial transactions determine the physical path of electricity;
- a nationally balanced system is necessarily locally feasible.
The mechanism must reflect the actual physical, behavioural and institutional characteristics of the grid.
A hierarchy of objectives
Before evaluating any candidate solution, the objective of the system must be stated clearly.
Different designs may pursue different combinations of objectives, including:
- maintaining secure and feasible system operation;
- supplying essential electricity reliably;
- minimising total system cost;
- reducing emissions;
- maximising the productive use of available renewable energy;
- reducing network congestion and avoidable infrastructure expenditure;
- encouraging investment in valuable generation, storage and network capacity;
- distributing costs and access fairly;
- protecting vulnerable consumers;
- preserving consumer choice and autonomy;
- reducing administrative and operational complexity;
- maintaining public legitimacy and support for electrification.
These objectives may conflict.
For example, the least-cost dispatch outcome in a single interval may not produce an acceptable long-term distribution of access. Maximising short-term network utilisation may reduce operational headroom. Strong price signals may improve economic efficiency while undermining predictability or participation. A highly centralised control system may be technically efficient but unacceptable to consumers.
Candidate solutions must therefore state:
- what they are trying to optimise;
- which constraints they treat as inviolable;
- which outcomes they are willing to trade off;
- how those trade-offs are governed;
- who bears the consequences when objectives conflict.
The evaluation framework
Candidate solutions will be assessed against a common set of tests covering:
- physical and engineering feasibility;
- operational and cyber-physical performance;
- human participation and behavioural realism;
- fairness and distributional outcomes;
- economic and market-design properties;
- investment and revenue sufficiency;
- governance, transparency and accountability;
- resilience and failure management.
Each test should be classified as one of the following.
Hard constraint
A condition that must always be satisfied.
Failure means that the proposed mechanism is physically infeasible, unsafe, financially inconsistent or otherwise incapable of operating as claimed.
Examples include thermal network limits, power balance, equipment operating constraints and accounting consistency.
Conditional hard constraint
A condition that is mandatory within a defined operating mode, time horizon or service commitment.
For example, a contracted reliability level may be hard for essential services but probabilistic for flexible demand. Revenue adequacy may be assessed over an investment period rather than in every individual interval.
Soft constraint
A desirable outcome that should be optimised, measured or improved, but may involve trade-offs.
Examples include convenience, simplicity, perceived fairness and the minimisation of consumer intervention.
Normative criterion
A criterion whose interpretation depends partly on social values, political choices or public consent.
Fairness is the clearest example. It can be measured in several ways, but the choice of fairness principle cannot be derived from physics alone.
The purpose of classification is not to imply that soft or normative criteria are unimportant. A system may satisfy every engineering constraint and still fail because people reject it, do not trust it or cannot participate in it.
1. Physical and engineering tests
Physics defines the feasible operating region.
Financial incentives cannot cause electricity to flow through a constrained line, create unavailable generation or override the physical behaviour of the network.
1.1 Power-balance test
Question: Does the mechanism maintain balance between electricity injections, withdrawals, storage actions and losses?
Pass condition: At every relevant operating interval, accepted schedules are balanced or supported by explicitly procured balancing capability.
Classification: Hard constraint.
A candidate fails if it clears transactions that cannot collectively be supplied.
1.2 Network-feasibility test
Question: Are accepted imports, exports and transfers feasible within the physical network?
The assessment must consider, where relevant:
- thermal line and transformer limits;
- voltage limits;
- power-flow relationships;
- network losses;
- topology;
- phase imbalance;
- protection constraints;
- reverse power flows;
- local hosting capacity;
- meshed and radial network behaviour.
Pass condition: No accepted allocation causes an unmitigated violation of network operating limits.
Classification: Hard constraint.
National energy balance alone is insufficient. A system can have adequate generation overall while lacking the ability to transport energy to the location where it is required.
1.3 Voltage and current consistency test
Question: Does the mechanism represent the fact that power flows arise from electrical states rather than contractual intent?
Pass condition: Allocation and control decisions are compatible with the voltage, current and impedance relationships governing the relevant network model.
Classification: Hard constraint where explicitly modelled; conditional hard constraint where state estimation is incomplete.
A market cannot assume that bilateral trades determine the physical path of electricity.
1.4 Bidirectional-operation test
Question: Can the design manage assets that may both consume and export electricity?
This includes:
- batteries;
- electric vehicles;
- solar-plus-storage systems;
- flexible industrial processes;
- aggregations whose net position may change direction.
Pass condition: Import and export constraints, prices, permissions, fairness states and control actions are represented separately where necessary.
Classification: Hard constraint.
A one-directional model of passive consumption is not sufficient for a distributed grid.
1.5 Asset-operability test
Question: Does the mechanism respect the operating characteristics of physical assets?
Relevant characteristics may include:
- maximum and minimum power;
- ramp rates;
- minimum run times;
- start-up and shut-down constraints;
- state of charge;
- charging and discharging efficiency;
- rebound effects;
- thermal comfort limits;
- availability;
- communications delay;
- response duration;
- recovery time.
Pass condition: No service is scheduled outside the feasible operating envelope of the participating asset.
Classification: Hard constraint.
1.6 Reliability test
Question: Does the design deliver the level of reliability associated with each service commitment?
A credible design must distinguish between:
- essential and non-flexible demand;
- flexible but time-constrained demand;
- deferrable demand;
- interruptible services;
- reliability products chosen by consumers;
- emergency scarcity arrangements.
Pass condition: Measured reliability is consistent with the promised service level, with breaches explicitly identified and remedied.
Classification: Conditional hard constraint.
Reliability need not be identical for every service, but it must be explicit, measurable and contractually coherent.
1.7 Scarcity-feasibility test
Question: What happens when there is genuinely insufficient energy or network capacity to satisfy all requests?
Pass condition: The design contains an explicit scarcity procedure that:
- preserves essential services;
- respects physical feasibility;
- allocates remaining capacity using declared rules;
- does not rely solely on willingness or ability to pay;
- records unmet service;
- prevents repeated disadvantage where possible.
Classification: Hard constraint during scarcity.
A solution that works only when sufficient capacity exists has not solved the allocation problem.
2. Cyber-physical and operational tests
The electricity system is operated through sensors, communications, state estimation, software and automated decisions.
2.1 Observability test
Question: Does the mechanism have sufficient visibility of the system state to make the decisions it claims to make?
Pass condition: The required inputs are measured, estimated or bounded with known confidence.
These may include:
- network loading;
- voltage;
- available generation;
- device status;
- flexibility availability;
- state of charge;
- communications health;
- forecast uncertainty.
Classification: Conditional hard constraint.
Where complete observability is unavailable, the mechanism must operate conservatively and represent uncertainty rather than treating estimates as facts.
2.2 Controllability test
Question: Can the system actually implement the allocations or schedules it produces?
Pass condition: Every accepted action has a viable control path, responsible actor, communications mechanism and fallback process.
Classification: Hard constraint.
An economically optimal schedule that cannot be enacted is not operationally valid.
2.3 State-consistency test
Question: Does the mechanism maintain a continuously updated representation of commitments, network conditions and resource availability?
Pass condition: New decisions account for previous allocations, changing conditions and outstanding obligations.
Classification: Hard constraint for stateful allocation mechanisms.
The system must not repeatedly allocate the same headroom, flexibility or stored energy.
2.4 Timing test
Question: Does the speed of sensing, clearing, communications and response match the physical process being controlled?
Pass condition: The total decision and actuation delay is compatible with the timescale of the relevant constraint.
Classification: Hard constraint.
A mechanism suitable for day-ahead scheduling may not be sufficient for voltage control, frequency response or rapidly changing local congestion.
2.5 Automation test
Question: Can routine participation occur without requiring constant consumer attention?
Pass condition: Consumers can specify preferences, boundaries and permissions in advance, with compliant devices acting automatically.
Classification: Soft constraint for optional services; conditional hard constraint where large-scale participation is necessary for system operation.
Automation should reduce cognitive burden, but should not remove meaningful control or obscure material consequences.
2.6 Graceful-degradation test
Question: Does the system remain safe when software, forecasts, communications or market outputs fail?
Pass condition: The design defines transitions between normal, constrained, scarcity and fail-safe operation, with deterministic protections taking precedence when necessary.
Classification: Hard constraint.
3. Human participation and behavioural tests
Because people are increasingly part of the balancing process, behavioural performance is a system requirement rather than a communications afterthought.
3.1 Participation-realism test
Question: Does the design rely on a plausible level and form of consumer participation?
Pass condition: Required participation rates, response frequencies and behavioural assumptions are supported by evidence or explicitly tested.
Classification: Conditional hard constraint.
A mechanism fails if system security depends on consumers behaving in ways they are unlikely or unable to sustain.
3.2 Cognitive-burden test
Question: How much effort, knowledge and attention must a participant provide?
Pass condition: Routine participation can occur without requiring consumers to continuously interpret complex prices, network conditions or market rules.
Classification: Soft constraint, becoming conditional hard where participation is operationally necessary.
3.3 Comprehension test
Question: Can participants understand the service, its risks, its benefits and the consequences of their choices?
Pass condition: The offer can be explained in clear language, and users can accurately identify:
- what they are agreeing to;
- what may be controlled;
- what they will pay or receive;
- what reliability they can expect;
- how to change or withdraw their preferences.
Classification: Soft or regulatory hard constraint, depending on consumer-protection requirements.
3.4 Choice test
Question: Does the design provide meaningful choice rather than nominal consent?
Pass condition: Participants can select among understandable service options, including appropriate non-participation or low-flexibility options.
Classification: Normative criterion; conditional hard constraint for consumer legitimacy.
The purpose of choice is to allow people to determine what is valuable to them rather than requiring the system designer to infer a universal utility function.
3.5 Automation-with-agency test
Question: Does automation preserve consumer agency?
Pass condition: Consumers can set boundaries, override non-essential actions, inspect material decisions and understand who is controlling their devices.
Classification: Soft constraint, with hard safeguards around essential services and safety.
3.6 Trust test
Question: Is the mechanism likely to earn and retain participant trust?
Trust depends on:
- predictable treatment;
- understandable rules;
- accurate billing;
- visible benefits;
- protection against exploitation;
- secure handling of data;
- reliable fulfilment of commitments;
- credible dispute resolution.
Pass condition: Trust is measured through user research, complaints, retention, opt-out behaviour and demonstrated adherence to stated rules.
Classification: Soft and empirical criterion.
A system that depends on voluntary flexibility but repeatedly surprises or disadvantages participants will eventually lose the resource it depends upon.
3.7 Education and capability test
Question: Does the design provide the knowledge required for meaningful participation without assuming expert users?
Pass condition: Information and support are proportionate to the complexity and risk of the service.
Classification: Soft constraint.
Education should support participation, but it should not be used to excuse an unnecessarily complex design.
3.8 Accessibility and inclusion test
Question: Can people with different incomes, technologies, housing arrangements, abilities and levels of digital access participate or receive equivalent protection?
Pass condition: The mechanism does not make access to affordable or reliable electricity contingent on owning expensive equipment, being digitally sophisticated or having control over the property.
Classification: Normative criterion and conditional hard constraint for essential-service protection.
4. Fairness tests
Fairness is not a single mathematical property.
It may refer to equal treatment, cost causation, equal opportunity, protection of basic needs, reward for contribution, proportionality, procedural justice or fairness over time.
Candidate solutions must therefore declare which fairness principle they implement.
4.1 Essential-services test
Question: Are essential energy services protected from market-based rationing and discretionary interruption?
Pass condition: Basic and safety-critical services are served before flexible or lower-priority uses, subject only to unavoidable emergency conditions.
Classification: Hard social constraint.
4.2 Ability-to-pay test
Question: During scarcity, does access depend primarily on wealth?
Pass condition: The mechanism prevents higher-income participants from purchasing all scarce capacity where doing so would deprive others of essential or contracted services.
Classification: Normative hard constraint during scarcity.
4.3 Cost-causation test
Question: Are costs recovered from those whose actions create or increase those costs?
Pass condition: Charges have a demonstrable relationship to the costs imposed, while recognising protections for essential demand and vulnerable users.
Classification: Soft-to-conditional-hard economic criterion.
4.4 Contribution-reward test
Question: Are participants rewarded according to the value they provide?
Pass condition: Compensation reflects measurable contribution to system outcomes rather than arbitrary category membership or historical privilege.
Classification: Soft economic and fairness criterion.
4.5 Fairness-over-time test
Question: Does the mechanism prevent the same participants from being repeatedly curtailed, delayed or disadvantaged?
Pass condition: Long-run service outcomes remain within a declared fairness bound, accounting for service level, need and previous treatment.
Classification: Conditional hard constraint within declared fairness guarantees.
4.6 Procedural-fairness test
Question: Are the rules transparent, consistent, contestable and applied as stated?
Pass condition: Participants can understand the basis of important decisions, challenge errors and obtain correction.
Classification: Normative hard constraint for legitimacy.
4.7 Distributional-impact test
Question: Who gains and who loses under the proposed design?
Pass condition: The solution reports impacts across relevant groups, including:
- income;
- housing tenure;
- technology ownership;
- geography;
- vulnerability;
- consumption pattern;
- ability to provide flexibility.
Classification: Mandatory assessment criterion.
A candidate should not be described as efficient without identifying how its costs and benefits are distributed.
5. Economic and market-design tests
The economic architecture must produce useful allocations, credible incentives and viable financial flows.
5.1 Economic-efficiency test
Question: Does the mechanism make productive use of available energy, flexibility and network capacity while avoiding unnecessary costs?
Pass condition: No feasible alternative delivers the stated objective at lower total cost without worsening another protected outcome.
Classification: Optimisation objective rather than absolute hard constraint.
Efficiency must be assessed subject to physical, reliability and fairness constraints.
5.2 Allocative-efficiency test
Question: Are scarce resources allocated to the uses that deliver the greatest declared value, subject to essential-service and fairness protections?
Pass condition: The mechanism performs at least as well as relevant benchmarks under a stated welfare or service objective.
Classification: Soft optimisation criterion.
5.3 Individual-rationality test
Question: Is participation preferable to the participant’s defined outside option?
For suppliers, this may require that an accepted transaction does not knowingly force operation below avoidable cost.
For consumers, it may require that the service delivers a benefit relative to non-participation after accounting for inconvenience, risk and loss of control.
Pass condition: Participants are not systematically made worse off by a transaction they voluntarily accept.
Classification: Conditional hard constraint.
5.4 Incentive-compatibility test
Question: Do participants benefit from providing truthful information and behaving in ways aligned with the intended system outcome?
Pass condition: Strategic misrepresentation does not produce a durable advantage, or the mechanism contains effective mitigation.
Classification: Conditional hard constraint.
Tests should consider:
- false availability;
- inflated costs;
- manipulated baselines;
- strategic withholding;
- gaming of congestion;
- repeated cancellation;
- misreporting flexibility;
- exploitation of forecast errors.
5.5 Budget-balance test
Question: Are total payments and receipts internally consistent?
Pass condition: The mechanism does not require an unexplained or structurally persistent external subsidy unless that subsidy is an explicit policy choice.
Classification: Hard accounting constraint over the stated settlement period.
5.6 Revenue-adequacy test
Question: Can the system recover sufficient revenue to fund the assets and services required to deliver the promised outcome?
Pass condition: Expected revenues are sufficient to cover justified costs and support required investment over the relevant horizon.
Classification: Conditional hard constraint over the investment horizon.
Revenue adequacy is not necessarily required in every interval. It must be evaluated across the period over which costs are incurred and recovered.
5.7 Investment-coherence test
Question: Do expected revenues encourage investment in assets that deliver genuine system value?
Pass condition: Valuable capacity, location, flexibility, controllability and reliability are rewarded, while assets that cannot contribute are not paid as though they can.
Classification: Soft-to-conditional-hard long-term criterion.
5.8 Price-signal coherence test
Question: Do the prices or control signals communicate a clear and operationally meaningful objective?
Pass condition: Participants and devices are not exposed to contradictory incentives without an explicit method for resolving them.
Classification: Soft constraint, becoming hard where conflicting signals can cause constraint violations.
6. Governance and legitimacy tests
6.1 Transparency test
Question: Can the operation and purpose of the mechanism be explained and audited?
Pass condition: Rules, objectives, constraints and settlement methods are documented, and material decisions can be reconstructed.
Classification: Conditional hard constraint for public or regulated systems.
Transparency does not require exposing sensitive personal or security information.
6.2 Accountability test
Question: Is responsibility clear when the system produces an error or harmful outcome?
Pass condition: There is an identifiable party responsible for operation, correction, compensation and escalation.
Classification: Hard governance constraint.
6.3 Contestability test
Question: Can decisions, charges and classifications be challenged?
Pass condition: Participants have access to evidence, review and correction procedures.
Classification: Normative hard constraint.
6.4 Adaptability test
Question: Can the mechanism evolve as technology, behaviour and system conditions change?
Pass condition: Rules and models can be updated without destabilising existing rights, obligations or system operation.
Classification: Soft long-term criterion.
6.5 Complexity test
Question: Does the design introduce more institutional, contractual or computational complexity than the value it creates?
Pass condition: Each additional layer has a clear function that cannot be delivered more simply.
Classification: Soft criterion.
7. Resilience and security tests
7.1 Cybersecurity test
Question: Can malicious actors manipulate measurements, bids, control signals or device behaviour?
Pass condition: The design includes authentication, authorisation, secure communications, anomaly detection and recovery procedures proportionate to the risk.
Classification: Hard constraint.
7.2 Privacy test
Question: Does the mechanism collect only the data required for its stated purpose?
Pass condition: Personal data are minimised, protected and not reused without a lawful and understandable basis.
Classification: Hard legal constraint and normative criterion.
7.3 Common-mode-failure test
Question: Could a shared algorithm, price signal, software update or communications failure cause a large number of devices to respond simultaneously in a harmful way?
Pass condition: The design includes diversity, staggering, rate limits, supervisory controls or other protections.
Classification: Hard constraint.
7.4 Recovery test
Question: Can the system restore normal operation following failure?
Pass condition: Recovery priorities, authority, communications and state reconciliation are defined and tested.
Classification: Hard constraint.
How candidate solutions will be assessed
Each candidate solution should be evaluated using a structured test record.
| Field | Required content |
|---|---|
| Test | The criterion being assessed |
| Domain | Physics, cyber-physical, behavioural, fairness, economics, governance or resilience |
| Classification | Hard, conditional hard, soft or normative |
| Claimed outcome | What the candidate says it achieves |
| Evidence | Model, proof, simulation, trial, user research, operational data or legal analysis |
| Metric | The measurable indicator used |
| Threshold | The value or condition required to pass |
| Operating conditions | The circumstances under which the result holds |
| Failure mode | What happens if the condition is not met |
| Trade-off | Which other objectives are affected |
| Result | Pass, conditional pass, fail or insufficient evidence |
| Confidence | High, medium or low |
A candidate should not receive a simple overall pass merely because it performs well on average.
The assessment should distinguish between:
- Pass: the criterion is satisfied under the stated operating conditions;
- Conditional pass: the criterion is satisfied only under identified assumptions or modes;
- Fail: the criterion is violated;
- Insufficient evidence: the claim has not yet been demonstrated.
Hard-constraint failure should normally disqualify a candidate in the operating conditions where the failure occurs.
Soft-constraint results should be presented as trade-offs rather than hidden within an aggregate score.
Normative criteria should expose the value judgement being applied. For example, a proposal should not simply state that it is “fair”; it should state whether fairness means equal allocation, need-based priority, cost causation, contribution-based reward, equal opportunity, protection of essential services or fairness over time.
The central test
The central question for every candidate solution is:
Can it coordinate the physical electricity system, produce economically coherent outcomes and secure the sustained participation of the people on whom the system increasingly depends?
A credible solution must satisfy all three.
It must be:
- physically feasible;
- operationally implementable;
- economically viable;
- behaviourally realistic;
- fair and publicly legitimate;
- resilient when forecasts, communications, markets or participants fail.
That is what “good” looks like.