The Idle Third of the Grid & the burn nobody talks about

P2 sheet daily capacity and energy

The fleet’s present capacity — what PGCB itself declares available on the day, after derating and outages — rose from to . Generation rose too, but more slowly, so the share of available plant that produces nothing on an average day fell to a low of in 2022 and has climbed back to .

P2 sheet by PGCB area

Every station sits under one of nine PGCB areas, restated on every report. Two questions answer most of what a region is for: how much of the country’s electricity does it generate, and how much of the plant it holds is doing nothing while it does so.

They come apart badly. Barisal produces more than an eighth of national output from the least idle fleet in the country outside Mymensingh. Dhaka produces the most in absolute terms and idles more than a third of everything it holds. Khulna is the extreme: of its listed stations sit out an average day, and it turns the capacity it does have into less electricity than any other area.

Generation log hourly, 520,776 area-hours

The hourly log gives each region’s contribution hour by hour. Pooled over seven years the national curve troughs at and peaks at — the familiar evening peak, driven by lighting and cooling load rather than industry.

The regions do not share one shape. Every area but one peaks between 20:00 and 21:00; Rangpur alone peaks at midday, on a fleet small enough that one plant’s schedule sets the curve.

P2 remarks and Forecast idle-day reasons

Each idle station-day carries PGCB’s own remark, folded into a family. Weighted by the capacity sitting idle, one cause leads every year of the window.

Peak hour the measure the policy brief uses

The reason changed, not the amount

The brief’s sharpest observation is not that capacity sits idle but that the stated reason has moved. Plant held back because nobody wanted the electricity has almost vanished; plant held back for no recorded reason has taken its place. The reports bear this out.

The brief’s specific claims, against the reports

Five of the brief’s factual claims can be checked directly against the daily reports. Four hold. One does not.

station_master 203 stations

Idleness at the fleet level is a handful of large plants held out for years at a time, not a general slackness.

Capacity charges what availability costs when nothing is dispatched

A two-part power purchase agreement pays a plant twice: an energy charge for each unit delivered, and a capacity charge for being available — owed in full on a day the plant sends out nothing. Idle capacity is therefore not merely wasted. It is the one state that costs the full fixed price and returns nothing at all.

Whether the charge is actually owed depends on availability, not on dispatch, and that distinction does the work here. A plant that stood ready while BPDB failed to deliver its gas was available, and is paid. A plant broken down was not available, and its sponsor forfeits the payment. PGCB’s own remark against each idle day decides which case it is.

Why it stood still, and what that cost

Fuel shortage is not only the largest cause of idleness, it is the most expensive one: capacity contracted, built, and paid for, then left without gas to burn.

Method 15 plants, 14 of them payment records

Every figure above is capacity-days — measured directly from the daily reports — multiplied by a taka rate per MW-day. The MW-days are solid. The rate is the weak link, and it is worth being exact about how weak.

Fifteen plant pages on energytransitionbd.org state a capacity charge precisely enough to convert. Fourteen of them state what BPDB actually paid over a known window — not an estimate — which is the strongest evidence available. Those rates cluster tightly, and Anwara 300 MW, the one plant whose page states a forward annual charge, lands inside the cluster derived entirely from the others’ payment histories. Two independent readings agreeing is the reason to trust the middle of this range at all.

What it cannot do is speak for the plants it does not cover. The evidence is entirely small oil- and gas-fired plants that reached commercial operation between 2008 and 2012. Nothing here is evidence about a modern coal unit, an LNG combined cycle, or a solar farm, whose capacity charges differ in both level and structure. Rates for those are the fleet median wearing a disguise, which is why the table below reports what rests on real evidence and what does not.

BPDB daily reports fuel mix, cost and load-shedding

The sections above showed that half the fleet stands still at the peak hour. The question now widens: what does the state write down about price, fuel mix, load-shedding and renewables? The PGCB and BPDB daily generation reports record three things against every station, every day — how much capacity it has, what it delivered at the peak hour, and why it delivered nothing. Every figure in this section comes from there, and from nowhere else.

Fuel cost

৳6.06per unit, FY26 — it was ৳2.48 in FY20

Gas share

41%it was 68% in FY19

Imports

16.7%of grid electricity — was 9.3%

Solar + wind

1.36%up from 0.06% in eight years

Idle at peak

14,501MW, out of 28,034

Load shedding

3,312MW at the day peak, 9 August 2026

How much of ৳26 is fuel, and how much is not

The brief carries a calculation for FY2023-24: 12% of all electricity came from liquid fuel — 11.80 billion units, at ৳307.88 billion, an average cost of ৳26 a unit. The BPDB daily reports give their own independent figures for the same year:

In FY2023-24 liquid fuel delivered 11.72 billion units — almost exactly the brief’s 11.80. But the fuel for those units cost ৳177.6 billion, or ৳15.16 a unit. The gap against the brief’s ৳307.88 billion is ৳130.3 billion — meaning that of the ৳26, ৳15.16 is fuel and the remaining ৳11.11 is not. What the brief calls capacity charges and rent-seeking sits in that second part. The daily reports do not name it, but they size it: 42% of the price of liquid-fuel electricity.

Load-shedding came back, and came back bigger

Peak-hour load-shedding was first recorded separately in the daily reports on 1 January 2025. The twenty months since look like this — occasional through 2025, then near-daily through the summer of 2026, and larger than in any earlier month.

Each column is that month’s maximum load-shedding, in megawatts. In June 2026 there was peak-hour load-shedding on 29 of 30 days, and on 28 days in July. The largest single day was 9 August 2026 — 3,312 MW at the day peak and 2,722 MW at the evening peak. The highest figure in all of 2025 was 1,491 MW, on 26 April: more than double in a year.

The renewable figure, set beside the neighbours

The brief carries a table for 2023-25 of how many billion units each country gets from renewables a year. Against Bangladesh it says “negligible”. BPDB’s own daily reports give the number: in FY2024-25, solar, wind and hydro together produced 2.31 billion units, or 2.27% of all grid electricity. Counting solar and wind alone, 1.35 billion.

The bars share one scale, which is why Bangladesh’s is barely visible — that is the point. The Renewable Energy Policy of 2008 set a target of 10% by 2021. What the daily reports recorded in that fiscal year: 0.18% from solar and wind, 0.99% including hydro. The brief says “in practice it never reached even 1 percent”. The reports agree.

Comparative context outside the PGCB reports

Everything up to here has been an internal account: the state’s own paper on the state’s own fleet. But a number only means something with a comparison beside it. What did countries like ours do over the same fifteen years? The comparison below is of six: Vietnam, India, Pakistan, Sri Lanka, the Philippines and Bangladesh. All import-dependent, all under subsidy pressure, all with a loss-making state utility. One thing should be settled at the outset — it is not that Bangladesh did not try. In 2001, 35% of the country had electricity; by 2022, 99.4%, the steepest climb on this list. In solar home systems Bangladesh built the largest off-grid programme in the world: 6,037,642 units by December 2022, 263.5 MWp. The difference is not effort but method — which procurement instrument the state chose for buying electricity.

That distinction is what connects this section to the measurements above. Under a capacity payment an idle plant still earns, and there is no revenue mechanism a solar developer can enter, because there is no fuel pass-through to monetise and the payment does not follow output. Under an output payment an idle plant earns nothing and cheap generation wins on merit. The idle-capacity numbers in sections 01 to 06 are what the first arrangement predicts.

Solar · 2018

0.24Bangladesh, GW — Vietnam was at 0.10

Solar · 2024

0.85Bangladesh — Vietnam 18.67

Wind · 2024

0.06GW, with 710 km of coastline

Solar + wind

1.5%of electricity, 2025 — neighbours 12–22%

Reserve margin

66.1%December 2024 — India 26%, Vietnam 28%

Per capita

586kWh, 2024 — the lowest on this list

In 2015 these six countries were in effectively the same place — not one of them had reached a gigawatt of solar. Ten years on, this is where the solar and wind share of electricity stands, and the instrument each country used to get there:

The last column is the real column. Everyone else started an instrument — a published price, an auction, net metering, a deadline. Bangladesh has been announcing targets since 2008. A target does not instruct anyone to build a power station.

Year by year: who decided what, and when

Bangladesh on the left, the rest of the region on the right. The five years set in bold are the ones that turned.

Procurement design the mechanism behind the gap

In 2018 Bangladesh and Vietnam had effectively the same installed solar capacity. Two years later the gap was forty-nine times. A gap that large, that fast, is not explained by technology — both countries were buying the same Chinese panels at the same price. Nor by sunlight: Bangladesh receives 4–5 kWh/m²/day, more than northern Vietnam. The explanation sits in one place — the instrument the state chose for buying electricity, and which technologies that instrument lets in.

What each country actually did — and where it tripped

None of them is a model to copy wholesale. Each instrument arrived with a trap attached, and the traps are the most useful part for Bangladesh.

The three objections that always come up

All three are reasonable, and all three are partly true. Which part is worth separating out.

What the instrument decides: which currency the bill is paid in

Everything so far has been about renewables. But the procurement decision settles something else that is discussed less — how much of the electricity bill leaves the country. The countries that have taken this question separately have made one thing explicit: “energy security” is not one thing but three distinct axes, and each needs its own measure, its own date and its own instrument. Slogans deliver none of them.

Of the three axes, the one most within Bangladesh’s own control is the currency axis — and the reason is uncomfortable. Importing fuel is unremarkable in this region; all six countries in section 08 are import-dependent. What is unusual is that Bangladesh pays dollars for machines that do not run. Sri Lanka’s dollars ran out paying for fuel it actually burned. A large part of Bangladesh’s goes against capacity standing still — the megawatts counted in section 05. It is the only kind of import dependence that can be reduced without delivering one unit less electricity. The other two axes take years to move; this one takes a contract renegotiation.

The old instrument is closed, the new one is not settled

On 28 November 2024 the Special Provisions Act of 2010 was repealed. The instrument that produced this gap over fifteen years is now shut — the first of the brief’s thirteen demands is effectively met. But the contracts written under that Act were not voided; capacity charges run to the end of each PPA. The bill from the old instrument is still arriving.

And the new instrument? The Renewable Energy Policy 2025 targets 20% by 2030. Reaching it from January 2026 means installing about 760 MW a year — against an average of 59 MW a year over the past fourteen. Roughly thirteen times the pace. Yet the policy still carries no published price, no auction calendar and no deadline. The 2008 policy also had a target, 10%, and it also stayed a target. This is exactly where the Philippines trap sits.

Where this comparison meets the brief’s own chapters:

• 8.3 / 9.3Tariff setting — Vietnam’s FiT and India’s auctions are both, at bottom, methods of setting a tariff. This connects directly to the brief’s best-evidenced chapter.
• 8.7 / 9.7Investment strategy — investment did not stay away because risk was high; it went precisely where the capacity charge was guaranteed. A question of turning the incentive around.
• 8.9 / 9.9Legal strategy — the emptiest area in the research (three papers), and the source of the entire gap. The brief is ahead of the literature here, and this comparison supports it.
• 8.10 / 9.10Sustainable transition — twelve papers in the corpus, almost all on technical feasibility. Nothing on procurement design — which is where the Vietnam–India difference actually lies.

Consumers Association of Bangladesh 9 May 2026, para 7

The brief closes on five asks. They are policy positions, not findings, and nothing in the daily reports can confirm or refute them — they are set out here because the numbers above are the evidence they are argued from.

Provenance who published each figure

Every measured number on this page comes from a report published by a Bangladesh government body. Nothing is modelled, estimated or supplied by a third party. The daily reports were read through the open archive at energy.datahubbd.org, which mirrors what PGCB and BPDB publish, and each figure is reproducible from it.

Where the daily reports and an outside source disagree, this page reports the daily reports and says so. The method — coverage, the two sheet templates, the reconciliation against the hourly log, and what is still open — is kept as a separate note.