How to Reduce Unplanned Downtime Without Expensive Infrastructure Upgrades

India met an all-time peak of 256.1 GW on 25 April 2026 without shortage, supported by roughly 65 GW of capacity added during FY 2025-26. The achievement is real. It is also the wrong place to keep pushing. Adding water at the top of the bucket is a poor strategy when nobody has measured the leak.

September 01, 2026. By News Bureau

Tags:

articles

Udit Poddar

WorkOnGrid

Infrastructure Upgrades

A distribution utility has one job. Deliver electricity to the consumer, at standard voltage and frequency, without interruption. Every other number a DISCOM is judged on sits downstream of it. Which makes it strange that unplanned downtime, the direct failure of that obligation, is the thing the sector measures worst.

Nobody is Counting Properly

Consider what a regulator can see. Of 72 DISCOMs, only 36 publish outage data of any kind. Of those 36, only 17 use SAIFI and SAIDI; the rest log interruptions by date and time without recording how many consumers were affected. Six report only planned outages. Eight have five years of history, and five publish nothing but the current day's numbers.

The definitions do not agree either. Delhi's supply code excludes planned outages and interruptions under five minutes; Haryana's includes planned outages and excludes those under three minutes. Two utilities can file the same SAIDI and mean different things. And the CEA's mandate to publish reliability indices is not statutory, compliance is voluntary, and its 2021-22 list covers only 49 of the 72 DISCOMs.

What should be measured is not complicated: SAIDI, SAIFI and CAIDI on one national definition, derived from meter-level loss and restoration timestamps rather than manual feeder logs, published monthly at feeder and DT level. None of that is a research problem.

The Reflex is to Spend

Set that against supply. India met an all-time peak of 256.1 GW on 25 April 2026 without shortage, supported by roughly 65 GW of capacity added during FY 2025-26. The achievement is real. It is also the wrong place to keep pushing. Adding water at the top of the bucket is a poor strategy when nobody has measured the leak.

The reflex is to answer with capital, and India has already committed at scale. RDSS has sanctioned INR 1.53 lakh crore for loss reduction infrastructure: substation augmentation, distribution transformer upgrades, reconductoring and feeder segregation. Against that, INR 100.9 crore was allocated to training DISCOM staff in digital systems, of which INR 23.36 crore had been spent by July 2026.

The world's wealthiest system shows where that road ends. American infrastructure law put INR 73 billion into grid modernisation between 2021 and 2026, and the sector still carries a INR 578 billion investment gap. Reliability did not improve. Between 2018 and 2024 the number of major US outages rose, with customer costs from outages reaching USD 121 billion in 2024. A far richer system, spending at record levels, cannot buy its way out of an operating problem. Neither will we.

The Instruments are Already on the Poles

India has already bought the instrumentation. As of 30 June 2026, 7.24 crore smart meters were installed across consumer, DT and feeder points, including 52.53 lakh DT meters and 2.05 lakh feeder meters. The hardware is going into the ground.

What is missing sits above the meter, and the telemetry is arriving unevenly. By October 2025, 94 percent of sanctioned feeder meters had been awarded and 64 percent were installed and communicating. At the distribution transformer level, 93 percent were awarded and 17 percent were communicating, which is about 6 percent of the 14.8 million transformers on the network. Consumer meters reached 53 million communicating by June 2026, against 196 million sanctioned.

And that is only collection. Ask what happens to the readings that do arrive and the answer is thinner. Senior DISCOM officials, quoted by Down To Earth in January, acknowledge that even advanced utilities use less than a quarter of the data their meters generate, for want of analytics teams and institutional capacity. The gap is a software and operational data layer. It is not another hardware procurement round.

Twenty Times Better, on the Same Equipment

Some utilities have already closed that gap. Figures discussed at a CEA standardisation cell meeting last October put the national distribution transformer failure rate at around 10 per cent, close to 1.3 million failures a year, with some northern states above 20 per cent. Tata Power-DDL runs below 0.5 percent and CESC Kolkata at 0.4 to 0.5 percent, attributed to remote monitoring, health indexing and inspection discipline. The causes identified were all operational: overloading, unbalanced loading, poor earthing and improper fuse coordination. A twentyfold spread between operators of identical equipment and identical standards is not a design problem. It is a question of whether anyone is watching the asset.

Three Changes, None of them Capex

Treat communication health as an operational signal, and aggregate what it reports. A meter that backfills on reconnection looks complete after the fact, so completeness cannot tell you whether supply was live. Track last-seen, expected against received reads, signal strength and concentrator telemetry, so one failed concentrator is not read as a hundred failed meters. One LV fault produces thousands of last-gasp messages; cluster them to the DT and the feeder and raise one incident with a location and a consumer count. Detection built on customer calls under-counts by design: fewer than 20 percent of affected customers report an outage at all.

Rank the failures you can predict. DT load against nameplate, phase imbalance, the frequency and duration of sustained overload, voltage excursions. Prioritise by expected consumer-minutes at risk rather than by asset age.

Verify restoration, and govern the sequence inside a workflow. First-breath signals and on-demand reads confirm supply actually returned, which is what stops nested outages and crews sent to streets already back on. Trigger, condition, action, with SLA timers set to the state's standards of performance windows and a retained audit trail, so SAIDI and SAIFI come out of timestamps rather than a year-end reconstruction. All of it has to run on the HES, MDM, CIS, SCADA and GIS a utility already operates.

What the Industry is Actually Waiting for

Which brings us to the part that is not a technology problem. None of this happens at scale because the industry lacks guidance on three things.

One definition. SAIDI, SAIFI and CAIDI computed the same way in every state, on the same thresholds and the same source data, so that two utilities filing the same number mean the same thing.

Reporting that is compulsory and deeper. Not a voluntary annual aggregate but feeder and transformer level data on a fixed cycle, with planned outages separated from unplanned.

A link between capital and uptime. Distribution investment is justified today on loss reduction. Nothing in the appraisal asks what it did to the hours a consumer actually had supply. Reward that, and the rest follows.

The national number will keep saying demand met. The consumer's number will keep saying four hours dark. Both are true, and the only reason they can coexist is that we have chosen not to measure the second one.

- Udit Poddar, Co-founder & CEO, WorkOnGrid