Domain model

This document tracks the domain types currently implemented in NEM.Model. It describes code that exists now; future concepts belong here only when their domain types and invariants are implemented.

classDiagram
    class Scenario {
        ScenarioId id
        string name
        DateTimeOffset periodStart
        DateTimeOffset periodEnd
        ScenarioRegion[] regions
      CostBasis costBasis
    }
    class ScenarioRegion {
        string regionId
        ScenarioGeneratingFleet[] generatingFleets
      ScenarioStorageFleet[] storageFleets
    }
    class ScenarioInterconnector {
      string fromRegionId
      string toRegionId
      Power capacity
      TransmissionCostParameters costParameters
      uint technicalLifeYears
    }
    class TransmissionCostParameters {
      DistancePowerCost capitalCost
      AnnualDistancePowerCost fixedOperatingCost
    }
    class ScenarioGeneratingFleet {
      GenerationTechnology generationTechnology
      Power nameplateCapacity
      GenerationCostParameters costParameters
      GenerationTechnologyProfile technologyProfile
        monthlyCapacityFactors
    }
    class CostBasis {
      int year
      decimal realDiscountRate
    }
      class GenerationCostParameters {
      PowerCapacityCost capitalCost
      AnnualPowerCapacityCost fixedOperatingCost
        GenerationEnergyCost variableOperatingCost
      FuelPrice fuelPrice
    }
    class GenerationTechnologyProfile {
      HeatRate heatRate
      uint technicalLifeYears
    }
    class ScenarioStorageFleet {
      StorageTechnology storageTechnology
      Energy initialEnergyCapacity
      Power initialPowerCapacity
      StorageCostParameters costParameters
      StorageTechnologyProfile technologyProfile
    }
    class StorageCostParameters {
      PowerCapacityCost powerCapitalCost
      EnergyCapacityCost energyCapitalCost
      AnnualPowerCapacityCost fixedOperatingCost
    }
    class StorageTechnologyProfile {
      uint technicalLifeYears
      double roundTripEfficiency
    }
    class PowerSystem {
        PowerSystemId id
        ScenarioId derivedFromScenario
        Region[] regions
    }
    class Region {
        string regionId
        DemandProfile demand
        GeneratingFleet[] generatingFleets
        StorageFleet[] storageFleets
        StorageTechnologyProfile[] storageTechnologyProfiles
        RegionalResourceProfile resourceProfile
    }
    class DispatchOutcome
    class SystemDispatchOutcome
    class InterconnectorFlow
    class ReliabilityMetrics
    class SystemReliabilityAssessment
    class RegionReliabilityVerdict
    class IStoragePolicy
    class GreedyPolicy
    class GreedySurplusAndIncrementalGenerationChargingPolicy
    class DispatchContext
    class StorageDecision
    class StorageIntent
    class StorageOutcome
    class SystemDispatchRun
    class RegionalDispatchRun
    class GenerationBudgetState
    class StorageSizingService
    class StorageSizingRunResult
    class RegionalSizingResult
    class RegionalBatterySizing
    class InstalledBatteryAssessment
    class EnergyLimitedAssessment

    Scenario "1" *-- "1..*" ScenarioRegion
    Scenario "1" *-- "0..*" ScenarioInterconnector
    ScenarioInterconnector "1" *-- "1" TransmissionCostParameters
    Scenario "1" *-- "1" CostBasis
    ScenarioRegion "1" *-- "1..*" ScenarioGeneratingFleet
    ScenarioRegion "1" *-- "0..*" ScenarioStorageFleet
    ScenarioGeneratingFleet "1" *-- "1" GenerationCostParameters
    ScenarioGeneratingFleet "1" *-- "1" GenerationTechnologyProfile
    ScenarioStorageFleet "1" *-- "1" StorageCostParameters
    ScenarioStorageFleet "1" *-- "1" StorageTechnologyProfile
    Scenario --> PowerSystem : ScenarioDerivation.Derive
    PowerSystem "1" *-- "1..*" Region
    PowerSystem "1" *-- "0..*" Interconnector
    Region "1" *-- "1..*" GeneratingFleet
    Region "1" *-- "0..*" StorageFleet
    Region "1" *-- "1" DemandProfile
    Dispatcher --> PowerSystem : consumes
    Dispatcher --> SystemDispatchRun : delegates the horizon
    SystemDispatchRun --> RegionalDispatchRun : creates per region
    RegionalDispatchRun --> GenerationBudgetState : owns per fleet
    Dispatcher --> IStoragePolicy : invokes per interval
    Dispatcher --> DispatchContext : constructs
    IStoragePolicy --> StorageDecision : produces
    StorageDecision "1" *-- "0..*" StorageIntent
    GreedyPolicy ..|> IStoragePolicy
    GreedySurplusAndIncrementalGenerationChargingPolicy ..|> IStoragePolicy
    Dispatcher --> StorageIntent : executes
    StorageFleet --> StorageOutcome : operates
    Dispatcher --> StorageOutcome : reconciles
    Dispatcher "1" --> "1..*" DispatchOutcome : produces per region
    SystemDispatchOutcome "1" *-- "0..*" InterconnectorFlow : solver evidence
    DispatchOutcome "1" *-- "1" ReliabilityMetrics
    SystemDispatchOutcome --> PowerSystem : validates correspondence
    SystemDispatchOutcome "1" *-- "1..*" DispatchOutcome : aggregates regional evidence
    SystemDispatchOutcome "1" *-- "1" ReliabilityMetrics
    SystemReliabilityAssessment --> SystemDispatchOutcome : assesses
    SystemReliabilityAssessment "1" *-- "1..*" RegionReliabilityVerdict
    StorageSizingService --> PowerSystem : sizes immutable candidates
    StorageSizingService --> Dispatcher : reruns whole system
    StorageSizingService --> StorageSizingRunResult : produces
    StorageSizingRunResult "1" *-- "1..*" RegionalSizingResult
    StorageSizingRunResult "1" *-- "1..*" InstalledBatteryAssessment
    StorageSizingRunResult "1" *-- "0..*" InterconnectorFlow : final evidence
    StorageSizingRunResult "1" --> "0..1" EnergyLimitedAssessment : system evidence
    RegionalSizingResult "1" *-- "1" RegionalBatterySizing
    RegionalSizingResult --> DispatchOutcome : final evidence
    InstalledBatteryAssessment --> DispatchOutcome : installed-capacity evidence
    InstalledBatteryAssessment "1" *-- "1" RegionalBatterySizing

Ownership boundaries

Scenario is the aggregate root for scenario intent. It validates identity, NEM-time period bounds, cost basis, and distinct regional fleet plans.
ScenarioRegion owns region-specific generation and storage fleet plans. It requires distinct technologies within each collection, while the same technology may appear with different capacity in another scenario region.
CostBasis fixes the year and real discount rate applied to scenario costs. RealDiscountRate is a decimal, and generation technical life is a uint. Each ScenarioGeneratingFleet owns its GenerationCostParameters and a GenerationTechnologyProfile. Generation cost parameters contain only power capital cost, annual fixed OPEX, variable operating cost in AUD/MWh generated, and fuel price. They do not contain storage energy capital cost.
Each ScenarioStorageFleet owns initial MW and MWh plus StorageCostParameters: power capex in AUD/MW, energy capex in AUD/MWh of storage capacity, and fixed OPEX in AUD/MW/year. It also owns a StorageTechnologyProfile containing technical life and round-trip efficiency. Initial MW and MWh must either both be zero or both be positive. A zero-capacity plan carries both economic and technical assumptions for Battery capacity later introduced by StorageSizingService.
ScenarioDerivation is a pure domain service that realises a PowerSystem from scenario intent plus one aligned demand series per scenario region, optional aligned additive demand components, and optional aligned regional resource profiles. It realises positive-capacity storage plans and omits zero-capacity sizing plans. Realised StorageFleet values retain their scenario technology profiles. Region also retains the profile map for zero-capacity plans, so immutable sizing candidates use each region's assumptions without making sizing scenario-aware. Its ScenarioGeneratingFleet.ToGeneratingFleet transformation derives typed short-run marginal cost as variable operating cost plus fuel price multiplied by heat rate. The realised GeneratingFleet owns that dispatch-relevant value.
DataCentreDemand expands a validated nameplate into a flat, full-load-factor component for the scenario period.
PowerSystem is the realised grid aggregate and cites its source scenario by ScenarioId. It owns one or more Region aggregates and zero or more Interconnector values.
Interconnector is owned by PowerSystem, not by either endpoint, because it belongs to neither region alone. It holds one directed transfer capacity from FromRegionId to ToRegionId, metered at the sending end, and endpoints identified by the same bare region strings compared case-insensitively that are used everywhere else. A reciprocal path is a separate interconnector. PowerSystem requires both endpoints to be its own regions and permits at most one interconnector per exact direction. WithRegions forwards the collection, so the repeated region rebuilds performed by storage sizing cannot silently drop links. ScenarioInterconnector is the matching intent, hung off Scenario alongside CostBasis because it is cross-regional, and it carries TransmissionCostParameters and a technical life for that directed capacity. TransmissionCostParameters costs a line in AUD/km/MW capex (DistancePowerCost) and AUD/km/MW/year fixed opex (AnnualDistancePowerCost): a line's cost scales with both route length and transfer capacity. There is no variable term.
Region requires one or more generating fleets with distinct generation technologies and may own storage fleets with distinct storage technologies. Its DemandProfile owns the base demand and zero or more labelled DemandComponent flows. Components must be non-negative, uniquely named case-insensitively, and exactly aligned with base demand; total demand is their element-wise sum and is the only demand consumed by dispatch.
Dispatcher remains scenario-blind. It consumes a realised PowerSystem and dispatches each owned region, producing one DispatchOutcome per region. Each outcome includes its ReliabilityMetrics. A per-region RegionalDispatchRun owns mutable execution state, including generation budgets and storage levels. It orders generating fleets by short-run marginal cost, then by technology to break ties deterministically (GenerationMeritOrder); conventional Hydro sorts into this order like any other technology and is not excluded from it. Unlike every other technology it carries a monthly energy budget (GenerationBudgetState) rather than a fuel cost, so 90% of that budget (the "paced" pool) is metered by HydroReservationState: a causal threshold controller that caps Hydro's request each interval at max(0, residualDemand - T), where T is solved by bisection each interval so that, applied over a trailing 336-interval window of past residual-demand observations, it would have spent exactly the budget affordable per interval over the intervals left in the month. This self-calibrates to whatever the demand distribution looks like (a sort-based "dispatch last" rule tried first and stranded ~93% of the budget instead of rationing it - see NEM-076). The remaining 10% (the "reserve" pool) is held back entirely from normal merit-order dispatch and spent only by RegionalDispatchRun.DispatchHydroFallback, a true last-resort backstop that runs after that region's own storage, against whatever local deficit storage could not cover. Neither pool carries into the next month, so once less than three days of the month remain the unspent reserve is released into the paced pool (GenerationBudgetState.ReleaseUnspentReserve): past that point holding it no longer buys cover, it only guarantees the energy expires unused. Both the paced cap and the release are settled once per interval in BeginInterval, so local dispatch, exports, and storage charging all price Hydro against the same allowance. RegionalDispatchRun builds an immutable storage-policy context for each interval, and executes policy intent through fleet physics.
SystemDispatchRun owns the horizon. The interval is the outer loop and the region the inner loop, so every region is at the same hour at the same time and surplus in one can serve a deficit in another. Order within an interval is generation (including Hydro's paced share, in normal merit-order position), then inter-regional transfer, then storage - and, immediately after that region's own storage, that region's own Hydro reserve fallback, which is strictly local and never visible to transfer. Each region's own sequence of operations is otherwise unchanged by the inversion, so a system with no interconnectors produces results identical to dispatching each region alone.
InterRegionalTransfer is the only place the domain meets the graph. It maps regional surplus and deficit onto nodes, delegates to the pure algorithms in NEM.Model/Algorithms, and books deliveries back as imports and exports. The algorithm layer knows nothing of regions, power, or losses; the transfer layer knows nothing of how maximum flow is found. Losses are applied over the result, never inside the search, which is what keeps it a standard max-flow problem. Exports draw on curtailment first and then start dispatchable plant in merit order; pumped hydro is excluded because storage is decided after transfer. Conventional Hydro is not excluded - its incremental headroom for an export is capped to the same per-interval pace as local dispatch (see RegionalDispatchRun.IncrementalHeadroom), so an export can substitute for local demand this interval but never draws on budget paced for a future local peak. Hydro's reserve share is never exportable at all: it is reachable only from DispatchHydroFallback, after transfer has already run for the interval.
IStoragePolicy owns storage intent and fleet ordering. It receives scalar snapshots rather than mutable fleet objects and does not own state of charge, execute storage physics, or book unserved demand and curtailment.
SystemDispatchOutcome is immutable whole-system dispatch evidence. Its factory requires exactly one hourly DispatchOutcome per PowerSystem region and checks every regional demand timeline. An unlinked system rejects nonzero regional import/export boundaries; a linked system must be created from SystemDispatchRunResult solver evidence, whose interconnector losses reconcile with regional exports less imports. InterconnectorFlow stores one non-negative directed Flow series from the link's sending region and a loss series no greater than that flow. The aggregate sums common demand, residual, storage, and per-technology flow series element-wise, zero-fills technologies absent from a region, sums storage state of charge by technology, and recalculates served load and reliability from the resulting system series. It retains validated regional outcomes and directional link evidence as read-only export evidence.
SystemReliabilityAssessment is immutable whole-system target evidence. Its factory compares the aggregate USE calculated by SystemDispatchOutcome and each retained regional DispatchOutcome against one maximum USE percentage. It passes only when the system measurement and every RegionReliabilityVerdict are within that target; system USE is calculated from aggregate demand and unserved energy, never by averaging regional percentages.
StorageSizingService is a pure whole-system orchestration service. It creates immutable PowerSystem candidates, reruns dispatch, and changes Battery storage only in regions that fail the configured USE target. Pumped hydro is fixed. Existing Battery capacity is the starting lower bound and results report total Battery capacity rather than incremental capacity.
StorageSizingRunResult validates that its regional results correspond exactly to final PowerSystem regions, their generation technologies match the final system fleets, and their dispatch timelines align with the final system demand. It also requires exactly one aligned InterconnectorFlow for every final link, matching endpoint order case-insensitively, directed capacity, and non-negative flow/loss values.
PowerSystemCostCalculator calculates a PowerSystemCostBreakdown after validating scenario, realised system, and dispatch correspondence. It requires exactly one scenario year. For each generation fleet it adds annualised power capex, one year of fixed OPEX, variable OPEX on gross generated energy, and fuel cost on gross generated energy. For every final realised storage fleet, it annualises combined power and energy capex over that technology's technical life and adds one year of fixed power OPEX. This costs total final capacity, including capacity introduced by sizing, and rejects storage without matching scenario assumptions. It produces one RegionCostBreakdown per region with annualised generation, storage, and total costs and each of their levelised costs divided by that region's DispatchOutcome.DeliveredToLoad. It derives system annual costs and served energy from those regional values. System generation cost is deterministically re-aggregated by technology so the published technology contributions exactly reconcile to the published generation total, then the system components are divided once by total served energy. PowerSystemCostBreakdown and RegionCostBreakdown are value-only results, not calculation services. Transmission is annuitised from scenario directed interconnector line-distance cost assumptions and charged once at system level, so regional costs do not sum to the system total. Line distance is the great-circle distance between the two endpoint regions' weather sites (each region's RegionalResourceProfile.Location, taken from its SolarZenithSeries latitude and longitude), so an interconnector's endpoints must both carry a resource profile. This makes NEM.Model.Economics depend on NEM.Model.Weather: a scenario that is otherwise fully specified economically now throws unless every interconnector endpoint carries weather data. That dependency should go back out if regional location ever gets a home of its own (a Region.Location, or an explicit route length on the scenario interconnector). A system result states whether transmission economics were calculated; regional results state that transmission is not modelled in their cost scope, even though they disclose incoming-link loss allocation.

The transmission distance model carries two known approximations, both kept deliberately rather than fixed, because the alternative (a real per-route distance table) is more precision than the rest of the cost model supports. Route length is a proxy, and it runs long. The distance is measured between each region's solar weather site, chosen for renewable resource quality rather than for where a transmission line terminates, so corridors whose resource sites sit far from the interconnector's real endpoints are overstated against the actual NEM route:

Link	Model	Actual route
VIC1–SA1	784 km	Heywood, ~275 km
VIC1–NSW1	732 km	VNI, ~300 km
NSW1–QLD1	964 km	QNI corridor, ~500 km
TAS1–VIC1	360 km	Basslink, ~370 km
NSW1–SA1	1,020 km	EnergyConnect, ~900 km

Reciprocal links each carry the full route length (see the TransmissionCostParameters remarks): every corridor is declared as two directed interconnectors, each costed independently at the corridor's full distance, so the same kilometre of conductor is paid for twice, once per direction. Charging each of the 5 physical corridors once at its larger directed rating gives 3,954,000 km·MW against the 7,035,547 km·MW actually charged across all 10 directed links, so this convention alone accounts for roughly 1.8× of the published transmission cost. The two approximations compound: swapping a region's solar EPW file silently changes system transmission cost, because the resource site is the only source of regional location.

Exported run results cite scenario and power-system identities rather than serialising the domain object graph.

Input provenance

Demand and weather paths are CLI configuration used to locate inputs; they are not part of Scenario identity. A result records the filename, input schema version, and SHA-256 digest of the exact bytes parsed for each input. The digest is the reproducibility boundary when a configured path is later overwritten.

The upstream demand archive names remain descriptive provenance from the demand artifact. They do not replace the demand artifact digest.

Time and units

Scenario periods and model series use fixed NEM market time (UTC+10). Result generation timestamps use UTC because they describe when an artifact was created; seeing +10:00 period bounds and a +00:00 generated timestamp in the same result is intentional.

Money and cost-rate quantities use decimal; measured physical quantities use double. They meet only inside typed conversion methods, where finite, non-negative physical values are explicitly converted to decimal. Money divided by energy served to load produces EnergyPrice in AUD/MWh served. GenerationEnergyCost is AUD/MWh generated and is used for variable operating and fuel-derived costs on gross generation. FuelPrice multiplied by heat rate produces a GenerationEnergyCost. EnergyCapacityCost is one-time AUD/MWh of storage capacity and produces Money only when multiplied by storage Energy.

PowerSystemCostBreakdown retains delivered energy separately from annual generation and storage costs. Its denominator is total DispatchOutcome.DeliveredToLoad, not per-fleet generation allocation. Storage asset cost does not add charging energy: gross generation VOM and fuel already price generation used for charging and therefore include storage losses. The storage component is annualised storage asset cost divided by the same served energy denominator; it is not a standalone LCoS. These costs are modelled estimates, not audited figures; decimal prevents base-10 accumulation artefacts from appearing as model defects.

Each RegionCostBreakdown retains the equivalent annual costs and delivered energy for one region. Its three levelised costs use only that region's DispatchOutcome.DeliveredToLoad; they are not divided by total system served energy. PowerSystemCostBreakdown.Regions carries these regional values while its existing system totals remain the exact sums of the regional annual costs and delivered energy.

Flow series are interval-average MW and integrate to MWh through FlowSeries.Integrate(). The dispatch invariant is:

generation + discharge + imports + unserved
    = demand + charge + exports + curtailment

That identity is per region. Summing it across the system leaves exports exceeding imports by exactly what transmission consumed, so the system-level identity carries a losses term instead of import and export terms:

generation + discharge + unserved
    = demand + charge + curtailment + transmission losses

SystemDispatchOutcome.TransmissionLosses is exports - imports, and is cross-checked every interval against the loss the transfer solver reports directly. Nothing enters or leaves the system as a whole: every export is another region's import plus the loss incurred on the way.

DispatchOutcome.DeliveredToLoad is demand - unserved; it is the regional load served by generation, storage discharge, and imports, and is the SLCoE denominator. Storage charging is recorded only as total charge; dispatch evidence does not retain a surplus-versus-incremental-generation source split.

Per-fleet delivered and charge series are consistent bookkeeping allocations, not physical attributions. RegionalDispatchRun produces these allocations as storage operations execute; DispatchOutcome stores the supplied immutable series and enforces their invariants. Surplus charging is booked to each fleet by the amount its curtailment is reduced, following dispatch merit order (see GenerationMeritOrder; in practice this only ever touches Solar/Wind, since every other technology's curtailment is always zero). Incremental-generation charging is booked to its named source fleet. Per-fleet delivered generation is the remainder after curtailment and charge. These rules close each interval exactly without reconstructing allocations from regional totals. The resulting per-fleet identity is:

fleet generation = fleet curtailment + allocated fleet charge
  + allocated generator-supplied load

Published delivered generation uses PerFleetDelivered rather than generation minus curtailment.

DispatchOutcome.RenewableShare is calculated from that same delivered generation using explicit GenerationTechnology classification. Grid-scale renewable share is delivered Solar, Wind, and Hydro energy divided by total delivered generation. Native renewable share is delivered Solar and Wind energy divided by DemandProfile.BaseDemand energy, excluding additive demand components. Both fractions are clamped to 0 through 1 and are zero when their denominator is zero or no relevant renewable fleet exists. Sweep scalars copy these definitions from the canonical delivered-generation and base-demand artifact series before base demand is externalized.

Generator-supplied load sums to deliveredToLoad - discharge - imports + exports, while allocated fleet charge sums to regional charge. This distinction is necessary because storage discharge and imports serve load but are not current-interval generation by any generating fleet.

An unlinked PowerSystem sets imports, exports, and transmission losses to zero. Regional artifacts retain non-negative imports and exports plus annual net imported energy. Their transmission-loss series is an incoming-link accounting attribution: each directed link's loss is assigned to its receiving region, and the regional sums reconcile to the system total. It is not a physical measurement of where loss occurred along a link or a regional transmission charge. Only a system artifact publishes directional forward/reverse link series and the canonical reconciled total transmission-loss series.

A region without storage also sets charge and discharge to zero. For such a regional outcome, generation - curtailment is generation delivered to load and uses the reduced identity:

delivered generation + unserved = demand

With storage, generation - curtailment can also include generation diverted to charging. Load served is then demand - unserved, and publishing a storage run requires charge and discharge series alongside generation and curtailment to preserve the full dispatch identity.

Reliability reports both unserved energy (USE) as a percentage of demand and hours served, plus peak single-hour unserved power. USE percentage is the binding reliability measure; hours served and peak unserved are diagnostics and must not be compared directly with an energy-based reliability target.

Storage sizing

Storage sizing defaults to the NER target of 0.002% USE. A caller supplies a commercial maximum Battery power, maximum Battery energy, and pass bound. New or undersized Battery storage is first raised to 30 MW and 120 MWh. Every sized candidate preserves a minimum four-hour duration.

After the floor, growth is geometric: each iteration offers the region's current Battery power doubled, its current energy doubled, and both doubled together, each clamped to the configured maxima and to the four-hour duration floor. The reliability metrics do not set the growth factor; they choose between the resulting probes. Every candidate is dispatched across the whole system, and the search takes the one that materially reduces USE by the most, breaking ties on peak unserved power. When none improves USE, the search returns StorageNoLongerImprovesReliability; this identifies solver stagnation and does not claim whether generation timing or storage policy is the underlying cause. Explicit MW, MWh, and pass bounds also provide termination; reaching a capacity bound before a larger probe is feasible returns BatteryCapacityLimitReached, while exhausting dispatch passes returns PassLimitReached.

When a dispatch fails its USE target, EnergyLimitedAssessment also evaluates the whole PowerSystem. It sums generator availability and demand from every aligned region, applying the same resource and generation-budget rules as dispatch. When total available generation energy is below total demand energy, the run returns EnergyLimited with system-wide available energy, demand, shortfall, and the intervals where total available MW is below total demand MW. Storage is excluded because it cannot add energy. The assessment is attached to StorageSizingRunResult, rather than attributed to a region, because it is a system-level proof. A total-energy shortfall proves infeasibility even if future interconnectors permit unrestricted transfers; it does not establish network feasibility when total energy is adequate.

Growth and every refinement probe dispatch the entire linked system. The growth phase advances a failing region in deterministic ordinal region order; once every region complies, full-system probes refine each changed region's power and energy to 1 MW and 1 MWh precision while retaining only compliant candidates and their matching interconnector-flow evidence. The result is a deterministic coordinate-wise near-frontier point, not a globally minimum or cost-optimal point.

Storage

StorageFleet is an immutable storage-archetype configuration and an interval state-transition operation. A positive requested flow discharges to the grid; a negative requested flow charges from the grid. Its state of charge is always validated within zero and its configured energy capacity.

Each fleet owns its energy and power capacities; its duration is derived as MWh divided by MW. The same storage abstraction supports battery and pumped-hydro fleets with different fleet capacities. Both limits bind each interval.

Each scenario storage plan supplies a technology profile; there are no technology-name defaults in the domain. A realised fleet retains that profile. Its round-trip efficiency is applied once while charging: input MWh multiplied by efficiency becomes stored MWh. Discharge removes and delivers stored MWh one-for-one, so a charge-discharge cycle loses (1 - efficiency) of the grid energy used to charge it. Round-trip efficiency is constrained to the inclusive range from zero to one.

Dispatcher initializes each storage fleet at its StorageFleet.SeedEnergy for a dispatch run and threads the returned state of charge into the next interval. Seed energy is zero unless the scenario declared installed capacity for that fleet, in which case StorageSeedPolicy assumes an opening balance of 80% of installed capacity for PumpedHydro and 50% for every other technology (see StorageSeedPolicy, NEM-076). The seed is fixed from installed capacity at scenario load and is never recomputed as storage sizing grows a fleet, so sizing can never earn itself free energy by growing. DispatchOutcome records one interval-beginning StockSeries per storage technology. The dispatcher constructs a fresh DispatchContext after generation has been dispatched to demand and before storage operates. The context contains signed residual power, resolution, storage levels and operating headroom, and current incremental-generation headroom and short-run marginal cost for each generation fleet. Positive residual means unmet demand; negative residual means would-be-curtailed surplus.

GreedyPolicy is a stateless surplus-only policy. For a deficit it requests discharge; for a surplus it requests charging sourced only from that surplus. GreedySurplusAndIncrementalGenerationChargingPolicy is the dispatcher's default stateless policy. It has the same deficit behavior and Battery-before- PumpedHydro storage priority. In a surplus interval, it first allocates surplus, then may request incremental Coal and Gas generation to charge remaining storage headroom. In a balanced interval, it may request those incremental sources directly. It uses ascending short-run marginal cost, then generation technology, to order incremental sources. Both policies limit intent using the headroom snapshots. The fleet still clamps every request and remains the authority for power limits, energy limits, and round-trip loss.

A policy returns one StorageDecision per interval. The decision contains zero or more StorageIntent values, each targeting one storage technology with a requested MW flow and, for charging, its energy source. A fleet can receive one discharge intent, one surplus-charge intent, and one incremental-generation charge intent per generation technology. The dispatcher processes the intents in order. Each executable intent invokes that fleet's Operate transition and produces a separate StorageOutcome containing actual delivered MW and final state of charge. An intent can be skipped without an outcome when no deficit, surplus, or incremental-generation headroom remains. Outcomes are therefore per-fleet state transitions, not a collection attached to the policy decision; the dispatcher aggregates their actual flows into DispatchOutcome.

The dispatcher reconciles actual flow rather than requested flow. Remaining demand deficit becomes unserved energy, while unused surplus becomes curtailment. Partial or rejected charging is silent: it does not create unserved energy or cost. Charging is recorded as non-negative surplus-sourced and incremental-generation-sourced series whose sum is total charge. An incremental-generation intent identifies the generation technology, and accepted charging consumes that fleet's current capacity and monthly energy-budget headroom.