Finding Wasted Energy With Building Meter Data

Finding Wasted Energy With Building Meter Data

A building energy audit that produces results starts with meter data, not a walkthrough. Read consumption every 15 minutes, per room, then lay the booking schedule over it. Electricity drawn during hours nobody booked is your waste, measured in kWh and rupiah. That number is what you fix, and what you re-measure afterward.

Everything else in an audit — the walkthrough, the equipment list, the lighting survey — is context for reading that number correctly. Without the number, an audit is a set of opinions with a cover page.

Why can't the electricity bill show you the waste?

Because the bill is one number, for one month, from one meter at the building's incoming supply. It tells you that you spent. It cannot tell you where, or when.

Take a set of multipurpose halls rented out by the event — a campus auditorium, a community hall, a hotel function room. Air conditioning and lighting are switched on by a staff member before the event and switched off by whoever remembers. Ask the manager what one wedding costs in electricity and you get a shrug. Ask what the halls draw on a Tuesday with no bookings and you get a guess.

The manager's instinct is usually right: the AC gets left on. But an instinct cannot be put in a budget, cannot be added to a rental rate, and cannot be defended when the finance director asks why the bill went up 18%. A number can do all three.

That is the whole gap this article is about. Not "does the building waste energy" — almost every building does. The question is how much, in which room, during which hours, and what it costs at your tariff.

What do you measure, and at what resolution?

Two things, at two levels of detail.

Per space, not per building. One digital power meter on the supply to each rented hall or each floor. A building-level meter averages your problem rooms together with your well-behaved ones until the problem disappears into the mean.

Interval data, not monthly totals. An interval reading is simply a consumption snapshot taken on a fixed clock — every 15 minutes, or every minute if the meter and network allow. A month of 15-minute readings is 2,880 data points per meter. A monthly bill is one.

From those two, three numbers fall out:

The second row is the finding. The first row is what makes the finding sellable to the owner, because it turns the audit into pricing work rather than a lecture about the environment.

One more note on resolution. Fifteen minutes is enough to see a hall left on overnight. One minute is what you want if you are chasing compressor cycling or a motor that restarts every few minutes. Start at 15 and go finer only on the meter that looks wrong.

What does the data actually reveal?

The same handful of findings show up in building after building. Here is what to look for, in the order they usually cost the most:

The last row is worth its own sentence. When the sum of your sub-meters — the extra meters installed downstream of the utility's meter — falls 8% short of the building's incoming meter month after month, you have either a circuit nobody mapped or a meter reading low. Both are findings. One is a wiring drawing problem; the other is a billing problem, and if you rebill tenants from that meter, it is also a legal one. More on that below.

Here is the data path that produces those rows:

Nothing exotic. Meters speak Modbus over a two-wire RS-485 loop — the plain serial cabling that building automation has used for thirty years — a small local server collects and stores the readings, and the booking calendar supplies the other half of the answer. Ninety percent of the value comes from joining those two datasets, which is a database query, not a control system.

If you want the deeper version of the metering-to-invoice path, I wrote it up in Metering to Invoice. For the case where the meter is an old mechanical dial with no output at all, there is Reading Legacy Utility Meters With ESP32-CAM.

Which fix pays first?

Schedule first. Then setpoints — the target temperatures you ask the air conditioning to hold. Equipment last. In that order, every time.

Scheduling is the cheapest lever and the biggest one. A US Department of Energy study by Pacific Northwest National Laboratory (PNNL-25985, 2017) modelled fourteen commercial building types and found that fully applying building controls — mostly correcting schedules, setpoints and operating sequences, not replacing hardware — could cut national commercial building energy use by 29%. The measures that carried the most weight were the boring ones: turn things off when the space is unoccupied, stop running equipment outside its needed hours.

Setpoints come second. Moving a hall from 20°C to 24°C is free and nobody files a complaint, but you only know it is safe to do after you have interval data showing the room was overcooled.

Equipment last. Replacing a chiller — the central plant that makes the cold water your air handlers blow over — or relamping to LED is real money with a real payback, and it should be argued from a measured baseline: the before-number you can point back to a year later. Otherwise you buy efficient equipment and keep running it on the same broken schedule.

One design rule sits above all three: the manual override has to survive. Events run late. The committee wants the lights on for teardown. A schedule that cannot be overridden from a switch in the room gets bypassed within a month — someone wedges the contactor, the electrically driven switch sitting behind the schedule, permanently closed, and your automation is now a decoration. Put an Auto/Manual selector on each zone at the panel, let staff use it freely, and log every manual operation with its start time and duration. The log is what keeps the report honest without turning the operator into a suspect.

What does Indonesian regulation require?

Two separate regimes touch this work, and they are easy to confuse.

Energy management, if the building is large enough. Permen ESDM No. 8 of 2025 on Energy Management was set on 7 March 2025 and promulgated on 13 March 2025, replacing Permen ESDM 14/2012 (JDIH BPK record). Building operators consuming more than 500 TOE per year fall under a mandatory energy management obligation. TOE — tonnes of oil equivalent — is the standard unit regulators use to compare fuels; 500 TOE works out to roughly 5.8 GWh of electricity a year, about 485,000 kWh a month. That catches large malls, hospitals, integrated campuses and hotel complexes. The obligations are a certified energy manager, a written conservation programme, periodic energy audits and an annual performance report.

Below the threshold, none of that is compulsory. The method in this article does not change — you just run it because it pays, not because you were told to.

Legal metrology, if you bill anyone from the meter. Any meter used as the basis for a charge is a trade instrument under Indonesia's legal metrology law (UU 2/1981), which means it must carry a valid tera mark — the state verification stamp applied by the local metrology office. Permendag 24/2024 sets the re-verification intervals: electronic kWh meters every 10 years, water meters of DN 50 mm and below every 5 years, diaphragm gas meters every 10 years. A sub-meter you installed yourself and never had verified is not a paperwork gap; using it to charge a tenant is an offence under that law.

Pass-through, if you resell electricity to tenants. Permen ESDM 31/2015 lets a building manager distribute electricity to tenants without an electricity business licence on one condition: no margin. What you bill the tenants in aggregate cannot exceed what you paid PLN, per-unit charges must be separated from common-area costs, and the tariff must follow the PLN class appropriate to the tenant's use. Taking a margin makes you an electricity seller, which requires a licence and a governor-approved tariff — a path that takes months and that almost nobody chooses.

The practical consequence is a split most operators miss: the audit meter and the billing meter are two different jobs. An audit meter can be any decent three-phase meter you can read over Modbus; nobody is charged from it, so nothing legal attaches. A billing meter needs a valid tera mark and a re-verification date in your asset register. Buy accordingly, and keep the verification dates in the same system that reads the meters — it is one column, and it is the column that gets you through an inspection.

How do you run this without instrumenting the whole building?

You do not need a hundred meters to start. You need one month of honest data from the few points that matter.

1. Pick 3–5 points, not 50. The biggest rented spaces, the chiller supply, and the common-area lighting board. That is usually 60–75% of the bill.
2. Get one month of interval data. Permanent meters if the budget is there; clamp-on loggers — meters that clip around an existing cable and need no rewiring — rented for a month if it is not. Both give you the same curve.
3. Get the booking schedule for the same month. This is the half that audits skip, and it is the half that makes the data mean something. A spreadsheet of start and end times per room is enough.
4. Compute the split. kWh inside booked hours versus kWh outside them, per room, per day. Publish it as a single ratio — "41% of the halls' electricity was drawn during hours with no booking" is a sentence that ends arguments.
5. Convert to rupiah at the actual tariff, then rank. For the July– September 2026 adjustment period, PLN's B-2 low-voltage business tariff sits at Rp1,445/kWh and the B-3 medium-voltage tariff at Rp1,122/kWh, both unchanged from the previous quarter (CNBC Indonesia, 7 August 2026). Rank findings by rupiah per month, never by percentage — a 40% saving on a small load loses to a 5% saving on the chiller.
6. Fix the schedule, then re-measure the same way. Same meters, same interval, same overlay, 30 days later.

Step six is the one that gets dropped, and dropping it is what makes energy projects unfalsifiable. If you cannot show the before curve and the after curve on the same axes, you did not run an audit. You wrote a proposal.

What you hand over at the end

Four things, and none of them is a slide deck:

• A baseline. kWh per room per month, split into booked and unbooked hours, for a stated month.
• A ranked finding list. Each line with a rupiah figure per month at the building's own tariff, and the intervention that removes it.
• A control change that actually shipped. Schedules loaded, zones wired to contactors, Auto/Manual selectors in place, manual operations logged.
• A re-measure date. On the calendar, with the same method written down so a different person can repeat it.

If you also want per-room enforcement — cutting supply on non-payment, not just measuring it — the device class for that is covered in Room-Level Energy Control With Smart Breakers.

We are good at costing hardware down to the last rupiah and terrible at costing the hours our buildings run empty. The hall that cooled itself all Sunday did not do anything unusual. It did exactly what it was told, by a schedule nobody had looked at in four years — and the only reason anyone found out is that somebody finally put a meter on it and read the numbers by the hour.