Institute of Policy Dynamics
Statistics
Energy / Public Data / PGCB & BPDB

Bangladesh's Electricity Supply and Load-Shedding, 2015-2026

Eleven years of hourly grid readings, cross-checked against daily generation reports and mapped to districts and power plants. Explore supply, load-shedding, forecasts, demand growth, production cost, fuel use, and data quality.

August 2026

Bangladesh's Electricity Supply and Load-Shedding, 2015-2026 illustration

Last updated: 27 Sept 2026

Right now

The most recent hourly reading from PGCB. Supply plus load-shedding is exactly the published “demand”.

There is 387 MW of load-shedding right now. At this hour over the last 90 readings it has typically been 1,578 MW, most days between 523 and 2,250 MW.

The electricity not supplied so far today (387 MWh) would have run a day's power for about 0.3 million people. (At 560 kWh per person per year.)

15,485 MW
Supply
387 MW
Load-shedding
Today's highest: 387 MW
15,872 MW
Demand (published)
387 MWh
Electricity not supplied
Hours with load-shedding: 1

Supply and load-shedding, hour by hour

Supply and load-shedding, hour by hour05,00010K15K20K20 Sept21 Sept22 Sept24 Sept25 Sept27 Sept
SupplyLoad-shedding

Why is there load-shedding?

The NLDC daily report prints a reason beside every power station. These are the stations that generated nothing at all at the evening peak, grouped by the reason given.

Idle capacity, by stated reason

The shortfall, as BPDB itself attributes it

At the latest evening peak, 7,274 MW of generating capacity produced nothing at all , 4,209 MW of it sitting idle for want of gas alone. Source: NLDC daily report, 22 Sept.

When gas falls, the lights go out

Gas delivered to power stations against demand left unserved, shown as monthly medians. The hot-month comparison holds temperature nearly constant.

Gas supplied

Demand unserved

16.2 million units
Least gas
867 MMCFD, 33.6°C, 121 days
2.6 million units
Middle
945 MMCFD, 34°C, 119 days
0.7 million units
Most gas
1,020 MMCFD, 33.3°C, 121 days

What BPDB said the day before

BPDB's next-day load-shedding forecast is set against what actually happened. The forecast became more realistic in 2026, but zero forecasts still miss often and by more.

325 / 328 days forecast at zero
2025
170 zero forecasts followed by load-shedding, 132 MW average
172 / 223 days forecast at zero
2026
98 zero forecasts followed by load-shedding, 455 MW average
2,846 MW
Worst missed zero forecast
16 Jun 2026

Forecast and outcome, side by side

Both figures describe the evening peak at the substation end.

Is the burden shared evenly?

Zones are not shed equally on the same day. How much of a zone's own demand gets cut, and how much shortfall falls on each resident , worked out from the authorities' own table.

Share of its own demand that was cut

Shortfall per person, at the evening peak

Over the last 90 days, 7.3% of national demand was cut on average. But Mymensingh lost 20.8% , 2.8× the national average , while Chattogram lost only 3.2%.

Electricity on the map

The darker the shade, the more load-shedding in that zone. Dots are power stations and grid substations, sized by capacity or load.

A district takes the colour of the grid zone it sits in , no district-level breakdown is published inside a zone. Plant and substation positions come from matching names against OpenStreetMap; some are only accurate to the nearest town.

How now compares with previous years

Load-shedding swings with the season , heavier in summer, lighter in winter. So "versus a year ago" only means something against the same point in the calendar. Each line below is one year, laid over the same dates.

Electricity not supplied, by day of year (7-day average)

Over the last 30 days an average of 31,131 MWh a day went unsupplied. At the same point in 2025 it was 2,989 , about 10.4× more now.
31,131 MWh
2026
23.8 hours
2,989 MWh
2025
15.4 hours
15,037 MWh
2024
21.5 hours
9,518 MWh
2023
20.4 hours
9,445 MWh
2022
17.6 hours

How fast demand has grown

Evening-peak demand laid over the same days of earlier years. This is published demand: electricity delivered plus electricity cut.

+96.3%
Growth over the decade
2016-2026
7%
Average annual growth
17,403 MW
This year's busy days
2026

Evening-peak demand, recent years side by side

Demand on each year's median day

Summer demand: delivered and cut

From 2025 to 2026, published demand rose 7.6%, while delivered electricity rose 2.2%.

Demand here equals supply plus load-shedding. Consumption abandoned because power was unavailable is not counted, so true demand may be higher.

What a day of electricity costs to make

Daily generation cost and cost per unit, compared with the same point a year earlier, show both how much prices changed and why.

276.9 crore Tk
Latest day's generation cost
22 Sept 2026
7.74 Tk/unit
Cost per unit
+1.29 Tk
Change per unit in a year
2025-09

Cost per unit against the same days of earlier years

Why the unit cost rose

Zone by zone

Demand and load-shedding in each zone at the evening peak, as published in the NLDC daily report.

ZoneDemand (published)Load-sheddingShare of all load-shedding
Mymensingh1,51831467%
Dhaka6,0928417.9%
Sylhet755306.4%
Rangpur1,066265.5%
Cumilla1,37991.9%
Rajshahi1,67061.3%
Chattogram1,47200%
Khulna1,90000%
Barishal52100%

Load-shedding trend by zone, last 365 days

What the electricity is made from

The fuel split of each day's generation, in gigawatt-hours.

September: total generation +4.2%, Gas -8.9%, oil +36.2%.

Generation by fuel, average per day

A daily average keeps short and incomplete months comparable.

What each fuel delivers, and what each fuel costs

Oil supplies 9.5% of electricity and takes 27.5% of the bill.

An ordinary day, half hour by half hour

Generation by fuel over the latest 30-day window. The top band is load-shedding: electricity requested but not delivered.

3,289 MW
Peak oil generation
1,910 MW
Oil's daily swing
81.8%
2,379 MW
Peak load-shedding
00:00

Daily generation by fuel

Fuel cost per unit of electricity

Over time

PGCB's archive holds rows back to 2015, but before 2022 the load-shedding column reads "0" in almost every hour. That period is marked "not reported" rather than drawn as zero.

Average shortfall

How solid are these figures?

This report shows the authorities' own numbers. But it also matters how those numbers are built, and where two official sources fail to agree with each other.

The published "demand" is an arithmetic identity, not a measurement

In PGCB's hourly table, demand equals supply plus load-shedding exactly in 99.4% of rows. BPDB's daily report does the same with energy: demand = generated + not supplied, exact in 100% of days. So "demand" is never measured independently , it is built from the load-shedding figure. The two cannot be used to check each other, and real demand may well be higher.

Two different pages, one set of numbers

BPDB's "area-wise demand" page and the NLDC daily report's evening-peak table agree exactly on 594 of the 600 days compared. It is not a second independent source , it is the same figure printed elsewhere.

How much each year actually holds

PGCB's table has rows going back to 2015, but the demand and supply columns are almost entirely empty, and before 2022 the load-shedding column reads "0" nearly every hour. The table below shows what is really there.

YearRowsLoad-shedding above zero%
20156,074220.4%
20168,753360.4%
20178,08450.1%
20188,689110.1%
20198,671120.1%
20208,76860.1%
20218,74830%
20228,7452,27726%
20238,7374,96556.8%
20248,7735,02157.2%
20258,7533,32638%
20266,4534,15664.4%

National installed capacity is about 29,000 MW. Load-shedding figures larger than that are treated as typing errors and set aside , 3 in total.

Sources: PGCB / NLDC hourly demand-supply-loadshed, BPDB daily generation archive (NLDC PDF reports), BPDB area-wise demand, PGCB daily workbook (NLDC reports, digitised), OpenStreetMap (plant & substation locations, district boundaries)