What is Firm Flexibility Capacity?
Firm Flexibility Capacity (FFC) is the flexibility a fleet of distributed energy resources can safely commit to a market or contract, in megawatts, at a stated confidence. A virtual power plant tells you what your fleet can technically do. The FFC tells you what you can safely promise someone else it will do.
Energy Variance assesses it from a fleet's own history: battery, solar and EV telemetry in, a calibrated probabilistic forecast through, and a defensible MW figure out, with every step of the derating and the residual risk of the commitments already on the book. The terms below are the ones an assessment is written in.
Questions it answers
How much flexibility can my fleet confidently commit?
Its Firm Flexibility Capacity: the megawatts the fleet's own history says it will meet or exceed at the confidence you choose, P95 by default, in each direction, across the delivery window you name, after the commitments already sold.
What is a fleet flexibility assessment?
A flexibility capacity assessment of a DER fleet for one window and one direction: nominal capability, forecast available flexibility from P50 to P99, the Firm Flexibility Capacity, the derating between them and the residual risk of the commitments on the book, in a report you can share.
How do you forecast the flexibility of a battery fleet?
From the fleet's own telemetry. Probabilistic forecasting models, from similar-day and quantile regression to gradient boosting, conformal prediction, scenario Monte Carlo and ensembles, are fitted on the earlier history, graded on held-back days, and the best calibrated is chosen for that fleet.
How do I know a P95 forecast is honest?
By measuring it. Every model is scored on five-minute intervals it never saw, and the share the fleet actually met is the measured confidence printed beside the FFC. Coverage, pinball loss, CRPS and expected shortfall are reported by season, time of day and temperature band.
How much of my VPP's capacity is firm?
Usually far less than its nameplate. The assessment walks from nominal capability to firm capacity step by step, so you see what state of charge, availability, customer opt-outs, response shortfall, correlated failure and network constraints each take.
How much flexibility is left after my existing commitments?
The headroom: the FFC less every commitment standing in the window at its full megawatts, whether FCAS, FCESS, network support or a bilateral, as if each is dispatched. Commitments that compete for the same power or the same energy are flagged.
What is my risk of under-delivering a commitment?
The assessment gives the chance the FFC falls short on its promise, the potential penalty on each commitment's own terms and call odds, and the residual risk left after any cover you hold, naming the costliest remaining shortfall.
How does weather affect flexibility forecasting?
Heat derates inverters and drives household load, cloud cuts solar output, and bad weather sends EVs home emptier and later. The models condition on temperature, cloud and irradiance, and calibration is checked on hot days on their own.
Can I assess a window next month or next year?
Yes. Up to sixteen days ahead the forecast runs on the weather forecast; beyond that, on last year's recorded weather with variation, and the assessment says which.
Is a simple average forecast good enough?
Every assessment answers that for your fleet. Beside it sits the common forecasting algorithm, the mean of the last five like days, with how often its figure actually held, measured the same way.
What data do I need?
Historical telemetry as a CSV: timestamp, asset, power, state of charge, and available charge and discharge, at the fleet's own interval. Missing intervals, clock changes, stale values and sign errors are reported, never silently fixed.
Which markets and assets does it cover?
The NEM and WEM in Australia and the Singapore electricity market; residential and C&I batteries, rooftop and C&I solar, and EV chargers, both managed charging and vehicle-to-grid.
Does Energy Variance control my assets?
No. It reads history and never sends a command to a device. Your VPP or DERMS keeps control; Energy Variance tells you what you can safely promise.
Firm flexibility
Firm Flexibility Capacity (FFC)
The megawatts a fleet can commit at a chosen confidence, for one direction, delivery window and location. At P95 it is the level the fleet is forecast to meet or exceed 95 days in 100, held to the worst dispatch block of the window. Nothing is held back on top of it.
Flexibility assessment
What the platform produces for a dataset, a window and a direction: the FFC, its measured confidence, the derating that leads to it, and the residual risk of the commitments on the book. It names no counterparty and no price.
Flex up
More net export to the grid: a battery discharging, a load curtailing, an EV pausing its charge. Positive by convention.
Flex down
More net import from the grid: a battery charging, solar curtailing its export, an EV charging harder. A fleet has a different figure in each direction at the same instant, and the two are never netted into one.
Nominal capability
The nameplate roll-up of a fleet's inverter and charger ratings: an upper bound on what it could ever offer. The FFC is read down from it.
P50, P90, P95 and P99
Quantiles of the fleet's forecast available flexibility. P95 is the level the fleet is forecast to meet or exceed 95% of the time, and the confidence every assessment is made at. Any level from P50 to P99 can be chosen.
Delivery window
The hours and days the FFC is assessed for. A window lands on the market's settlement block, and its worst block sets the figure.
Sustained for the whole window
The FFC held block after block across the delivery window, so the energy stored in the fleet limits it as well as its power: a battery that can discharge 5 MW cannot hold it for twelve hours.
Sustained for any dispatch block
The FFC held for any single dispatch interval of the window: five minutes in the NEM and WEM, thirty in Singapore.
Where the megawatts went
The derating breakdown from nominal capability to FFC: state-of-charge uncertainty, asset availability, customer opt-out, response shortfall, correlated failure, network constraints and commitments already sold. A figure nobody can interrogate is a figure nobody trusts.
Correlation penalty
The capacity held back because assets fail together. Shown as "what it costs to not believe in luck": the gap between the FFC and what the same fleet would claim if every asset failed independently.
Headroom
What is left to commit in a window: the FFC less every commitment standing in it at its full megawatts, as if each is called.
Over-commitment
Selling more than the FFC in a window. It appears in the derating as capacity sold beyond the FFC, and every day it applies is short if called.
Confidence and calibration
Measured confidence
The share of held-back five-minute intervals on which the fleet delivered at least the model's forecast at that level. A P95 figure that held on 96.2% of them reads 96.2% (P95). It is what "confidence" means beside every FFC.
Calibration
Whether a forecast's confidence is honest. A calibrated P95 is met 95% of the time. Under-coverage is a business risk; over-coverage is megawatts left on the table.
Coverage
The share of outcomes at or above the forecast quantile: the success ratio of a probabilistic forecast, scored by season, time of day and temperature band.
Probabilistic forecast
A forecast that gives a distribution of outcomes rather than one number, so a commitment can be sized at a confidence instead of at a guess.
Pinball loss
The quantile loss used to rank forecasting models. It rewards a quantile that is both calibrated and tight.
CRPS
Continuous ranked probability score: the error of the whole forecast distribution, not of one quantile.
PIT histogram
The probability integral transform of each outcome. Flat when a forecast is calibrated; its shape tells bias from a spread that is too wide or too narrow.
Expected shortfall (CVaR)
Given the fleet falls short of its P95, how far short on average. It is the number that prices penalty exposure, and it matters more than P95 itself.
Held-back days
Models are fitted on the earlier part of a fleet's history and graded on days they never saw, every five-minute interval inside the hours of interest.
Evidence gate
A coverage miss counts against a model only when the held-back days are enough to prove it. The day is the unit of evidence, so a short test period reads "not proven", never "fine".
Common forecasting algorithm
The comparison every assessment carries: the mean of the last five like days. A point forecast, with the confidence it actually held measured the same way as ours.
Model race
Similar-day, season-conditioned quantile, gradient-boosted (including XGBoost), conformal, scenario Monte Carlo and ensemble models raced on one scorecard; the best is chosen per dataset and then tuned.
Flexibility forecasting
Forecasting what a fleet will be able to offer, in each direction and in energy, rather than what it will consume or generate. Load forecasting and solar forecasting are inputs to it, not the answer.
Quantile forecast
A forecast stated as levels the outcome will exceed with given probabilities: P50, P90, P95, P99. Quantile regression and gradient-boosted quantile models produce one directly.
Conformal prediction
A calibration method that adjusts a model's quantiles by the errors it made on days held out from its fit, so that its stated confidence holds.
Forecast horizon
How far ahead an assessment reaches: up to sixteen days on the weather forecast, and any period beyond on last year's recorded weather with variation.
DER fleets
Distributed energy resources (DER)
Small generation, storage and controllable load behind the meter: home and commercial batteries, rooftop and commercial solar, EV chargers and dispatchable process load.
Virtual power plant (VPP)
Software that aggregates and dispatches many DER as one resource. Energy Variance sits beside it: it never sends a command to a device.
Flex owner
Whoever controls a DER fleet and sells its flexibility: a VPP operator, aggregator, retailer or battery operator.
Battery energy storage (BESS)
Residential and commercial and industrial (C&I) batteries. Flex in both directions, bounded by power, state of charge, energy capacity and round-trip efficiency.
Solar PV curtailment
Rooftop and C&I solar offers flex down only: it can curtail what it exports, and it has nothing held back to add.
EV managed charging and V2G
A managed charger flexes by pausing or speeding a charge while the car is plugged in. A bidirectional, vehicle-to-grid car also exports, and its connection window decides what it is worth.
State of charge (SOC)
How full a battery is, as a percentage of its usable energy. Where a fleet starts the day sets how much it can deliver.
Network constraint
A limit at one location on the distribution network. Flexibility relieves it only from behind that element; location is part of the product.
Dynamic operating envelope
A varying export or import limit on a connection. Read from the telemetry's available charge and discharge, so the forecast learns when it bites.
Fleet telemetry
Historical per-asset data, typically every five minutes: power, state of charge, available charge and discharge. Uploaded as CSV, with gaps, stale values, clock changes and sign errors reported, never silently filled.
Controlled days
Days somebody else dispatched the fleet. They are left out whole from every fit, forecast and score, because they show the fleet under control rather than its own behaviour.
Demand response
Changing consumption on request. Curtailable C&I process load counts as flexibility; small residential loads such as hot water and pool pumps are not modelled as flexibility.
DERMS
Distributed energy resource management system: the control platform that dispatches a fleet. Energy Variance works from its history and never from its controls.
Commitments and risk
Commitment
Capacity already sold: an FCAS or FCESS service, a network support contract, an energy bilateral. It comes off the FFC at its full megawatts, whatever the odds of a call.
Standby
An availability payment: paid per MW per hour for being ready, whether or not the fleet is ever dispatched.
Dispatch probability
The chance a commitment is called on a given day. The same every day unless the commitment names a temperature, cloud or irradiance trigger.
Potential penalty
What under-delivering a commitment could cost, priced on its own contract terms and the odds it is called.
Released capacity
A commitment's megawatts returned to the FFC on chosen days. Sold FFC is spent, so a released day is short by the whole commitment if its buyer calls.
Co-delivery
A commitment delivered by the same megawatts, at the same time, as wherever the FFC is sold: delivered once, paid twice. Marked "not reducing the FFC".
Chance of FFC being short on its promise
One minus the confidence the promise actually held: 100% less the measured confidence on a held day, more where released capacity may be called.
Cover
Hedging or insurance cover offered by a third party against a shortfall. Each offer is priced against the commitments it would stand behind.
Residual risk
The loss you remain exposed to after cover, with the costliest remaining shortfall and its dispatch probability, and a verdict against your line.
Absorbable loss
The loss you are willing to carry without cover. A shortfall inside it is acceptable; one above it is covered or flagged.
Cross-market conflict
Two commitments drawing on the same megawatts or the same energy at once, found before either is called.
Flex delivery track record
How a fleet has performed when called upon, measured against the FFC stored before each event. The next phase, built on the audit trail every assessment keeps.
Markets
NEM
The National Electricity Market of eastern and southern Australia, operated by AEMO. Five-minute dispatch and five-minute settlement, in A$.
WEM
The Wholesale Electricity Market of south-west Western Australia, operated by AEMO. Five-minute dispatch and a thirty-minute trading interval, so a commitment is held to its worst five minutes.
Singapore electricity market
Operated by EMC, with energy priced at the USEP. Thirty-minute dispatch and settlement, in S$.
FCAS
Frequency control ancillary services in the NEM: contingency raise, contingency lower and regulation.
FCESS
Frequency co-optimised essential system services, the WEM's equivalent: contingency reserve raise, contingency reserve lower and regulation.
Primary and contingency reserve
Singapore's reserve classes, priced per provider group by effectiveness. Contingency reserve's ten-minute response is one five-minute telemetry can evidence.
Network support
A distribution network (DNSP, or Western Power in the WEM) paying for flexibility at a location to relieve a constraint.
Settlement block
The interval a market settles on: five minutes in the NEM, thirty in the WEM and Singapore. A delivery window must land on it.
Market potential
What a dataset's realised flexibility would have earned at observed historical prices, under a rule you set. A look back, never a forecast or a promise.
Flex Lab
Dataset Lab
Upload fleet telemetry or generate a synthetic battery, solar and EV fleet with ground truth, inspect its data quality and see its market potential.
Forecast Lab
Race every forecasting model on the hours that matter, read the scorecard and choose the model for the dataset.
Risk Lab
Assess the Firm Flexibility Capacity for a window and a direction, add commitments, penalties and cover, and compute the residual risk.
Results Lab
The FFC, its derating, the confidence breakdown and the forecast error, compared run against run and exported as a report.