Portfolio loss waterfall

Net budget (P50)

The P50 energy budget after the losses it allows for in each category. In this sample it is the P50 before the losses the waterfall counts, less each category's allowance.

Weather

The change in budget caused by the weather actually measured on site. It is not an operational loss, so it is shown in grey.

Budget at actual weather

The net budget re-run at measured conditions. Only the resource changes: the model and every loss allowance keep the percentages the budget was built on, applied to the energy the measured weather would have produced. This is the fair target for the period, and the starting point for judging operations.

Allowance and result

Each category's allowance is the share of expected energy the budget assumed it would lose. The bar is allowance less actual loss, so up means better than budget.

Expected energy

The site's own yield model, such as the PVsyst model behind the P50, run on measured conditions, and used in the absolute losses view. It keeps the model's normal losses (temperature, irradiance angle, mismatch, cabling, transformers and auxiliary load) and stops before the losses this waterfall counts. Solar uses measured plane of array irradiance and module temperature, and rear-side irradiance for bifacial modules on trackers, before the inverter AC limit; IEC 61724-1 governs how those measurements are taken, but it does not validate the model itself. In this sample the solar model is the plane of array irradiance, corrected for module temperature, times a performance ratio for the losses it keeps: 0.94 for Najd's bifacial trackers, 0.93 at Rimal, 0.88 for Qamar's monofacial fixed tilt, and 0.79 to 0.88 at Kestrel, lower in winter light. At the Gulf sites the hourly beam and diffuse irradiance are calibrated month by month to the long-term averages of the Global Solar Atlas, with a Linke turbidity of 4 to 5 for the dust and haze. Wind uses measured free-stream wind speed through the warranted power curve, corrected for air density, less the modelled electrical losses to the revenue meter. Turbulence, shear and how representative the wind measurement is all affect how far that estimate can be defended, so a real assessment states the uncertainty.

How the sample is built

Every site has an hourly model of 2025: sun position, cloud and temperature at the solar sites, a wind speed series through the turbine power curve at the wind farm, the lead and lag engines' run windows at the CHP, and the charge and discharge schedule at the battery. Each fault cause has its own asset type, rate per asset per year, season, time of day, chance of clearing remotely and repair time once someone arrives, and its energy is what that asset could have made in those hours. Rare failures are rare: at each site most months have no grid outage and no HV fault, and a few voltage dips a year trip several units at once. The UK weather follows the broad pattern of 2025: the sunniest spring on record, the warmest summer, a dull October, and Storms Éowyn, Floris, Amy and Bram.

Which losses apply where

Every technology: grid and curtailment, including grid disturbance trips away from the CHP, set point and export limit, balance of plant and unexplained. Solar only: inverters and strings offline, trackers stowed or stuck, soiling, clipping, inverter thermal derating, row shading, snow, module underperformance and degradation above the P50 rate. Wind only: turbine faults and scheduled maintenance, wake, icing, high wind restarts, power curve underperformance and yaw misalignment. CHP only: engine trips, failed starts, service overruns, engines held at part load, heat-led control, gas supply pressure, cooling faults, ambient and gas quality derating, engine condition and turbocharger fouling. Battery only: PCS faults, isolated racks, BMS communication faults, module spares, a cable joint defect, chiller and auxiliary supply faults, fire detection false alarms, state of charge limits, missed EMS schedules, container thermal derating, capacity fade and cell imbalance. Categories with the same name hold different items for each technology, so compare them within a technology first. The event types map to the information categories of IEC TS 63019 for PV and IEC 61400-26-1 for wind, so the same events can be reconciled with a contract availability figure.

CHP dispatch

A CHP plant is dispatched, not driven by the weather. Its expected electricity is what the engines should have produced at their generator terminals, at rated output and the engine maker's reference conditions, in the hours they were called to run. That is the basis of the efficiency guarantee and of CHPQA power output. For the CHP the end bar is the generator meters, not export: the energy centre's own load and the private-wire supply come out of it before the export meter. The energy centre's own load and the step-up transformer losses are about 3% here. The lead engine follows the private-wire load, which never falls below about 2 MW, so it runs at full output: around the clock from September to May, and 07:00 to 23:00 from June to August, when running at night would only add heat to dump. That is about 7,900 hours a year. The lag engine is heat-led in a day window: 16 to 18 hours a day from November to February, about 8 hours in April and October, weekday middays from June to August, and not at all in May and September, when the lead engine alone makes about the heat the network takes. The private-wire customer takes power first; only the rest is spill export to the grid, and the export limitation scheme and DNO constraints act on that spill alone. Derating for high intake air temperature and gas quality is shown as a design and weather loss rather than taken out of expected. Planned services are not in expected, because the engines were not called while they ran; only the hours beyond the planned duration count as an overrun. The grey bar becomes dispatch: the plant ran more or fewer hours than budget because of heat demand or market prices, which is a commercial decision rather than a loss. In views that mix CHP with solar or wind it reads weather and dispatch.

CHP heat

Heat has its own waterfall in MWhth, measured inside the energy centre at the CHP heat meters after the dump radiators, so boiler heat and the network beyond the energy centre are outside it. Budget heat moves with degree days to give the heat budget at actual demand. Engines not running is the heat the boilers had to make because an engine was down, event by event: with one engine down, the other engine carries what it can, and the thermal store covers the first hours of an outage, so in summer this bar is close to zero. From November to February the lead engine runs around the clock and the lag engine 16 to 18 hours a day, and the boilers still top up; in May the lead engine alone makes about what the network takes. In those months heat drawn from the store is made again by the boilers and earns no credit. Recovery performance is the fall in heat recovered per MWh of gas against the commissioning baseline, corrected to the actual return temperature, less the share the network would not have taken anyway. The engine maker quotes thermal output within plus or minus 8%, so the bar shows a trend, here exchanger fouling that built up to May and cleared after the June clean, rather than a breach of the guarantee. Engines held back is the heat the engines did not make because derating and engine condition held them below full output, on the days the network needed all the heat the engines could make. The heat lost to the export limit and DNO constraints is valued with those items in the electricity view instead, so it is not counted twice. Boilers covering is the rest of the demand met by gas boilers while the CHP was available and running it would have been cheaper. When boilers were the cheaper choice it is dispatch, not a loss. Heat and electricity are never added together in MWh.

Heat dumped

Recovered heat rejected through the dump radiators because the network did not need it. It does not reduce heat delivered against demand, so it is not a bar in the heat waterfall. It burned gas for no heat sale and lowers the CHPQA Quality Index, so it is in the fuel view. It can still pay when power prices are high; the test is whether the electricity earned more than the extra gas. Dumping while the boilers are firing shows up as boilers covering. Dump radiators are often unmetered, in which case the heat is worked out from the water-side flow and temperature drop across the radiators, or by difference between the engine heat meters and heat delivered.

Fuel view

Gas burned, measured on gross calorific value, split into electricity, useful heat, dumped heat, and exhaust and other losses, beside the engine maker's full-load figures converted from net to gross calorific value. Running efficiency includes part load, ambient conditions and gas quality, so a real assessment corrects the guarantee to the actual load, ambient conditions and gas quality, and applies the contract tolerance, before calling any gap a shortfall. CHPQA uses the same gross basis: the year's measured power efficiency and useful heat efficiency, dumped heat excluded, set the Quality Index and how much of the scheme counts as Good Quality CHP.

Battery dispatch

A battery is dispatched by its trading strategy, not by the weather. Its expected discharge is what it would have exported against the optimiser's schedule had rated power and the usable energy allowed by the warranted state of health curve been available. It charges from 01:00 to 05:00 for the morning discharge and from 11:00 to 15:00 for the evening one. At the battery a fault during the charging window costs the discharge it prevents, so it is listed with that energy. A short fault costs nothing if the PCS can catch up before the discharge starts. The grey bar becomes dispatch: the battery cycled more or less than budget because of spreads and the optimiser's strategy, which is a commercial result rather than a loss. This view covers energy only; lost availability for frequency response, Balancing Mechanism or Capacity Market positions is valued separately. Metered export for a battery is energy discharged to the grid; what it imported to charge is in the round-trip view.

Round-trip view

Energy imported to charge the battery and run the site, split into energy exported, PCS and transformer losses, battery losses and auxiliary load such as chillers and controls, beside a reference split built from the guarantee and the design loss budget. Here the guarantee is measured at the point of connection with auxiliary load included; a real assessment also corrects for the change in stored energy over the period. Many warranties measure at the battery terminals and leave auxiliary load out, and the guaranteed figure is a test-cycle value at a stated C-rate and temperature that declines with age, so check the definition before comparing.

State of health

Measured energy capacity as a share of its beginning-of-life rating, from periodic capacity tests at the conditions the battery supply agreement sets, with the BMS estimate in between, against the warranted curve. Fade faster than the curve is a performance loss on cycles that use the full window, and the basis for a capacity warranty claim once a test to the warranty's procedure confirms it.

Valuing storage

A lost MWh of discharge is valued at its margin: the average discharge price less the charging cost at the guaranteed round-trip efficiency. This sample uses 2025 GB averages of £120 a MWh to discharge and £50 to charge, a £70 spread around a wholesale average of about £81. Energy a battery could not discharge is often sold later at a lower price rather than lost, so a real assessment values the shortfall against the optimiser's own counterfactual at half-hourly prices, and adds lost availability payments and non-delivery penalties. Extra import from a poor round trip is valued at the charging price, and auxiliary load at its own supply cost, which can carry charges that charging import does not.

Valuing CHP

A lost MWh of CHP electricity is valued at its margin over gas: the power price, less the gas it would have taken at the guaranteed electrical efficiency and the per-hour charge in the long-term service agreement. Private-wire power is priced by time of use, £160 a MWh from 07:00 to 23:00 and £100 at night. That is about £150 on average, about 12% below what manufacturers paid for grid power, and this sample values every MWh at that average. The heat that went with it is valued in the heat view at the boiler gas it displaced, at an 85% boiler efficiency on gross calorific value, so the full cost of an engine outage is the two together; each event shows both. The export limit and DNO constraints act only on spill export, so they are valued at the spill margin instead. Spill earns about £81 a MWh, less £99 of gas, £10 of service charge and £8.60 of Carbon Price Support, a loss of about £37 a MWh. Carbon Price Support is £3.31 a MWh of gas, and spill pays it because the CHP exemption covers only power used on site or supplied by private wire. To that is added the boiler gas for the heat these losses cost, on the days the network needed all the heat the engines could make. The value comes out positive only in the months when the network needed nearly all of it, and negative in the rest: spill export on its own does not cover the gas, so in those months these losses saved money. Manufacturers' gas ran from 4.2p a kWh in the first quarter to 3.5p in the second half, which moves this margin by about £18 a MWh, so a real assessment uses half-hourly prices. Carbon on the private-wire margin is left out of this sample; at the average UK ETS price over the year to November 2025, about £49 a tonne, it would take about £23 a MWh off that margin if the energy centre is in the scheme.

Design and weather

Wake, clipping at the inverter AC limit (a DC/AC ratio of 1.30 in this sample), inverter thermal derating, row shading, snow, icing and the wait to restart after a high wind cut-out; at the CHP, high ambient and gas quality derating; at the battery, container thermal derating. Mostly set by layout, equipment choice and climate rather than by operations, so shown in grey, though settings and upkeep can add to them.

Grid and curtailment

Network outages, grid disturbances and operator instructions, split into paid and unpaid. A voltage dip outside the grid code ride-through envelope that trips several units at once is a grid event until they reconnect automatically; a unit that does not come back on its own is a separate fault for the O&M contractor or the OEM, and a trip on a dip inside the envelope is a plant fault. On a non-firm connection offered before April 2023, ANM curtailment is the owner's risk. Unpaid curtailment at the GCC sites is either inside the PPA's yearly allowance or still waiting for a deemed energy claim.

Set point and export limit

Energy held back by the plant controller beyond what the grid actually asked for; at the CHP, engines held at part load, cautious heat-led control and the export limitation scheme; at the battery, state of charge limits and missed EMS schedules. These losses often fall between the O&M scope and the asset manager, which is why they persist.

Balance of plant

ACB trips, protection relay trips (including G99 interface protection at GB distribution-connected sites, or G59 at plant connected before 2019), transformer derating and HV faults; at the CHP, gas supply pressure and cooling and heat rejection faults, including knock trips caused by a fouled intercooler, because cleaning is the O&M contractor's; at the battery, auxiliary supply, chiller and fire detection faults, and a cable joint defect inside the EPC contractor's defects liability period.

Equipment availability

Inverters, strings and turbines offline, trackers stowed or stuck, and scheduled turbine maintenance; at the CHP, engine trips, failed starts and service overruns, each with any wait for spares inside it; at the battery, PCS faults, isolated racks, BMS communication faults and waits for battery modules.

Performance

Losses operations, O&M contracts or warranties can act on. For solar: soiling, degradation faster than the P50 assumed, and module underperformance. A module warranty is a floor on module power with a first-year allowance, so degradation faster than the P50 is the owner's until a flash test shows the modules below that floor. For wind: power curve underperformance, which stays with the owner while it is inside the warranty tolerance and no IEC 61400-12-1 test shows more, and yaw misalignment. For the CHP: engine condition, and turbocharger and intercooler fouling. For the battery: capacity fade above the warranted rate and cell imbalance.

Unexplained

Expected energy, less every counted loss, less metered export. The budget allows nothing for it. Keeping it small and honest is the real test of the data. Inside 1% of expected, with data completeness of 95% or more, it is within model and meter tolerance, so it is not ranked as the largest variance; it can come out negative in some months.

Counting each MWh once

Losses are worked out asset by asset and hour by hour as a stack. Each cause takes only the energy it accounts for beyond the causes before it, in this order: design and weather (wake, the inverter AC limit, shading, snow and icing), which sets what each asset could actually have exported; then grid and curtailment, balance of plant, equipment availability, set point and export limit, and performance. Curtailment and outages are then measured against the same potential that deemed energy and energy-based availability use. Balancing Mechanism volumes are settled against the Final Physical Notification instead, so the bar shows the energy the turbines lost and the contract position shows how much of it was paid. A set point loss is only energy a setting change could have recovered. The sample follows that order hour by hour: an inverter that fails while the plant is curtailed loses only what the rest of the plant could not make up, a tracker row held flat while its inverter is at its AC limit loses almost nothing, and no fault starts on equipment that is already out further up its power path. Whatever remains between that and metered export is unexplained. A real assessment applies the rule at the metering interval, with the same exclusion rules as the O&M contract, and keeps the interval-level reconciliation so every MWh can be traced to its cause.

Curtailment and set points

Separating what the grid asked for from what the plant controller held back only stands up when the grid instruction and the controller's behaviour are both logged, time-aligned and kept as evidence.

Actionable

The sum, site by site, of every shortfall against allowance in set point and export limit, balance of plant, equipment availability and performance. A site that beats its allowance does not offset another that misses it.

Percentages

Every percentage is a share of expected energy, so the bars add up.

Energy value view

Each MWh valued at the site's tariff, except the CHP's export limit and DNO constraints, which are valued at the spill margin. Hollin Edge Wind sits under the Renewables Obligation at 0.9 ROCs a MWh, so its MWh is worth the wind capture price, taken as 85% of the £80.63 baseload average, plus 0.9 ROCs at £75.10 each (the £67.06 buy-out price for 2025-26 and the latest £8.04 recycle value): about £136 a MWh. This is the value of the energy, not lost revenue or a recoverable sum: paid curtailment was compensated, so it cost the site little, depending on how the compensation was calculated, and merchant sites earn the price at the time rather than a flat tariff. Paid curtailment stays in so the bridge closes, and the owner chart marks it as paid. A commercial impact figure would add compensation, merchant prices, contract penalties and what can actually be recovered.

Largest shortfall

The single site and category furthest below its budget allowance for the selected view and period, ranked in the unit on screen, so in pounds in the energy value view. Degradation above the P50 rate has no allowance of its own, so all of it counts as excess when the page names the main causes. Unexplained is only ranked when it is more than 1% of expected or the data is less than 95% complete. It carries the spark in the waterfall, and its next step comes from a fixed rule for each category, not from a written opinion.

Events

Discrete losses, such as trips, faults, outages and curtailment instructions, are listed one event at a time. Each event carries its asset, first alarm, attendance, restoration, root cause, owner, contract position and evidence, and the events add up exactly to the item above them. Hours lost are the time to attend plus the repair once someone is there. An event that runs past a month end is listed in each month it touches, with the energy it lost in that month, so the monthly bars carry each loss in the month it happened. A fault that cost less than 10 kWh, such as a solar inverter trip at night, is not listed. At the battery a fault during the charging window costs the discharge it prevents, so it is listed with that energy. In a real assessment events come from the SCADA alarm log, protection relay records, work orders and the instruction logs.

Remote resets

A stop that clears on a remote reset is listed with no attendance. Only causes that need neither a part nor a visit can clear that way: a unit that did not reconnect after a grid disturbance, comms losses, controller restarts and firmware rollbacks, turbine pitch and converter fault resets, and tracker wind stow and clock resets. When the remote reset fails, the event shows the attendance and the reset on site.

Continuous losses

Soiling, wake, degradation, derating and similar losses have no single event. They are worked out hour by hour against the same model and shown by asset and by month, with the measurement behind them and what it found.

Response target

The time from the first alarm to someone on site, as the O&M or service agreement sets it. At the solar sites it is tiered by what failed: 4 hours for an inverter station, feeder or the whole site, 24 hours for a single string inverter or tracker block, 48 hours for a combiner box, and 5 days for a single string or tracker row. It is 12 hours for the wind farm, 6 hours for the battery and 8 hours for the CHP in this sample. Nobody climbs a turbine in winds above about 14 m/s, so at the wind farm some repairs wait for the wind to drop after the technician is on site. Scheduled maintenance, service overruns, spares waits and interface protection trips are not callouts, so they have no response target. Events reset remotely have no attendance.

Repeat events

Three or more events on the same asset in the period. The same breaker, inverter or engine failing again usually means the root cause was never fixed, only reset. Scheduled maintenance and planned service overruns are not counted.

Data completeness

The share of intervals with a reading from the revenue meter and the plant's own counters, weighted by expected energy: inverters and the weather station at solar sites, turbine SCADA at the wind farm, the engine controller and the gas and heat meters at the CHP, and the PCS and BMS energy counters at the battery.

Completeness is not validity

A revenue meter can report every interval and still be wrong: scaled incorrectly, wired with reversed CT polarity, or out of step in time. A real assessment checks validity separately, reconciles the revenue meter against inverter and SCADA totals, and states how confident it is in the model, alongside completeness.

Sample data. The seven sites, their names, tariffs, allowances, efficiencies, events and every number on this page are fictional, generated to show how the waterfall behaves over the full year 2025. Prices are 2025 averages from public sources, and the UK weather follows the broad pattern of 2025. Equipment models are named to make the sample realistic; the faults and figures attached to them are invented.