Before You Claim Savings: Baseline and Load Profile

Part 3 of 10 in Cobler's Energy Management: The Economics of Saving Energy course. New here? See the course page.

Say a contractor tells us that a new chiller schedule will cut the office tower's electricity use by 10%. The first thing to ask is, 10% of what? And we should ask our own numbers the same thing. In Part 2 we valued the tower's three measures at about RM158,065 a year. Those values are only as good as the kWh estimates behind them.

What we need is an energy baseline. It's a record of how much energy the building uses, when it uses it and why, before anything changes. Every saving in the rest of this course is measured against a baseline, so we have to build it before the project, not after.

So let's build one for the tower, our made-up 20-storey building on the MV (medium voltage) non-domestic General tariff. We'll use four kinds of evidence:

• twelve monthly bills
• half-hourly meter data
• the building's operating schedules
• the things that push consumption up and down, such as weather. Along the way we'll see where the three measures come from, and give each one a range instead of a single number.

Twelve months of bills

The simplest baseline is a year of electricity bills. A single month isn't enough, because consumption changes through the year. Hot months need more cooling, and months with public holidays or fewer working days need less. So twelve months in a row include every season and every holiday once.

Here are the tower's bills for October 2025 to September 2026. The numbers are illustrative.

The year's total is 6,000,000 kWh. The busiest month, May, used 535,000 kWh, which is 65,000 kWh (about 14%) more than December's 470,000 kWh. Keep that swing in mind. If a project saves 2 or 3% of the building's electricity, the normal month-to-month swing is several times larger than the saving.

The tower's baseline year: consumption rises in the hottest months and falls in months with cooler weather, holidays or fewer days.

Bills and meter data

But a bill shows only one number a month. To see what happens inside a month, we need interval data, the energy recorded in every half hour. The maximum demand (MD) on the bill is the highest of these half-hourly readings in the month. Maximum Demand: Why One Bad Half-Hour Costs So Much, in the bill course, explains it. The data can come from the TNB (Tenaga Nasional Berhad) meter or from the building's own main meter. How electricity meters work explains how meters record it.

Before we trust interval data, though, we add it up and compare it with the bills. The two cover the same energy, so they should agree closely. For the tower, the half-hourly readings for the same twelve months add up to 5,944,000 kWh. That's 56,000 kWh (about 1%) less than the bills.

A gap like this has a cause, and we should find it before anyone uses the data. In the tower's case, a gateway outage in March 2026 lost three weekdays of readings. An average weekday uses about 18,600 kWh, so three missing days explain about 56,000 kWh. The team marks those days as missing and fills them from similar weekdays, and then the two sources agree.

Two other differences are normal:

• Billing periods. A TNB bill runs from one meter reading date to the next, which may not match the calendar month. So compare totals over the same dates, or over the whole year.
• Sub-meters. The meters on the chiller plant, risers and tenant floors won't add up exactly to the main meter. That's because of cable losses, meter accuracy and loads that no sub-meter covers. Why your sub-meter will never match TNB's meter explains the usual reasons. Treat the TNB meter as the reference for the bill, and use sub-meters to see where the energy goes.

The load profile

With the data checked, we can plot a day of half-hourly demand, and we call that plot a load profile. The area under it is the day's energy in kWh, and its highest point is the demand in kW. Power vs Energy: What You Actually Pay For, in the bill course, explains the difference. For a baseline, we read the profile for what the building is doing at each hour.

The tower's MD day: about 430 kW at night, a 1,600 kW peak at 2:30pm, and cooling still running after 6pm.

Four features stand out:

• The base load. Between midnight and 6am the tower still draws about 430 kW. That's the equipment that runs all the time: the server room, lifts on standby, car park lighting and fans, security systems and pumps. A base load is normal. But a base load that creeps upwards month after month usually means equipment has been left on or is wearing out.
• The morning rise. Demand climbs from about 6:30am as the cooling plant, the air handling units (AHUs) and the lights start. It reaches about 1,400 kW by late morning.
• The afternoon peak. Demand is highest in the hottest part of the afternoon. On this day it reached 1,600 kW between 2:30pm and 3:00pm, which set the month's MD. Five half hours, from 1:30pm to 4:00pm, were above 1,500 kW.
• The evening tail. After office hours demand falls, but slowly. At 7:00pm the tower still draws 750 kW.

We can sum up the whole month in one number, the load factor, which is the average demand divided by the peak demand. For September, the average is 500,000 kWh spread over about 730 hours. So the load factor is 500,000 ÷ (1,600 × 730) ≈ 0.43. A low load factor means the building uses its peak capacity only for short periods.

The weekend matters too. On a Sunday the tower uses about 10,300 kWh, close to the base load all day. On a Saturday it uses about 12,400 kWh, because the plant runs in the morning for the few tenants who work then. An average weekday, for comparison, uses about 18,600 kWh.

Schedules and the three measures

So the load profile shows us what happens. To learn why, we compare it with how the building is meant to run. For the tower, the facility team checks three sources:

• the office hours in the tenancy agreements (8am to 6pm, Monday to Friday)
• the BMS (building management system) schedules
• the sub-meters on the chiller plant.

This comparison turns up three findings:

• The cooling plant runs two hours too long. The BMS log shows every AHU stopped by 6:00pm on normal weekdays. But the chiller plant is on an old schedule that keeps it running until 8:00pm. With no AHUs taking chilled water, those two hours cool almost nothing. The plant's sub-meter shows it drawing about 220 kW on average from 6pm to 8pm.
• The condenser water pumps run at one speed. Each of the two duty pumps draws a steady 27.0 kW whenever it runs, on hot days and mild days alike. A pump with a variable speed drive (VSD) slows its motor when less flow is needed, so it could use less power at part load.
• A small amount of load sets the peak. The five half hours above 1,500 kW go over it by 10, 70, 100, 80 and 40 kW. If the BMS could hold off some non-urgent loads for those periods, the MD would fall. Candidates include the car park ventilation fans and the water transfer pumps that fill the roof tanks. A small, temporary rise in the chilled water temperature would help too.

And these findings are where the three measures from Part 2 come from. We call them M1, M2 and M3:

• M1: stop the chiller plant at 6pm, with an after-hours request button for tenants who work late.
• M2: fit VSDs to the two condenser water pumps.
• M3: BMS demand limiting that holds off selected loads in the top half hours.

For more on reading load profiles, see energy consumption analysis. For how sub-meters help find problems like these, see energy sub-metering.

Drivers

Now let's go back to the twelve bills we started with, because they didn't change at random. A few factors, which we often call drivers, explain most of the movement. For an office building, the main ones are:

• Working days. A weekday uses about 18,600 kWh and an average weekend day about 11,400 kWh. So each extra working day in a month adds about 7,000 kWh. A month with a long public holiday uses less for this reason alone.
• Weather. Hotter weather means more cooling, and one common way to measure how hot a month was is cooling degree days (CDD). For each day, we take the day's average outdoor temperature and subtract a base temperature, here 24 °C. So a day averaging 28.5 °C gives 4.5 CDD, and a day at or below 24 °C gives zero. Adding up the days gives the month's CDD, and April 2026 had 135 CDD. The tower's twelve months, after allowing for working days, give a rough slope of about 600 kWh for each extra CDD. Twelve points aren't many, so this slope is an estimate, not a precise law.
• Occupancy. A new tenant, an empty floor or a new server room changes consumption for reasons that have nothing to do with any savings project. So the facility team should record these changes with their dates.

So why does this matter? Say next April is hotter than this April. Then the tower will use more energy even if a savings project works perfectly. Without the drivers, the extra cooling would hide the saving, or a cooler month would make a failed project look successful. Part 4 uses these drivers to adjust the baseline before comparing it with the months after the project.

Savings as a range

Each of the three findings gives a saving estimate, and each estimate depends on numbers that vary. So it's more honest to give a low, central and high value, and to say what each one assumes.

For the tower, the team uses 22 working days a month, or 264 a year. None of the three measures changes Saturday morning running, so we don't count it in any saving.

• M1. The plant's average draw from 6pm to 8pm was 220 kW over the year, but the monthly averages ranged from 180 to 260 kW. Some tenants will ask for cooling after 6pm, and the team allows for that by taking 20 kW off. So the central estimate is 200 kW for 2 hours on each working day: 2 × 200 × 264 = 105,600 kWh a year.
• M2. The minimum water flow the chillers need limits what VSDs on the condenser water pumps can do, so the saving is modest. The team estimates each pump will average 19.0 kW instead of 27.0 kW. That's during the 11 hours from 7am to 6pm when both run. Together, that's 16 kW less: 16 × 11 × 264 = 46,464 kWh a year. The range allows 12 to 20 kW.
• M3. Holding off loads moves energy to other half hours rather than removing it, so the team counts no kWh saving. What it can lower is the MD. The half-hour data suggests about 100 kW off each month's MD, with a range of 60 to 120 kW.

Together, the central kWh savings are 105,600 + 46,464 = 152,064 kWh a year. That's about 2.5% of the tower's 6,000,000 kWh, or about 12,700 kWh a month. And it's much smaller than the 65,000 kWh swing between May and December. So we have to measure a saving against a baseline that allows for the drivers, because we can't read it off the bill.

The range also protects the business case. When Part 9 tests the project, it will use the low and high values to see whether the decision still holds.

Proving the saving

So now we have a checked baseline, the drivers that move it, and three measures with a range for each. But once the measures are in, someone from finance will look at the bill and ask whether the savings really happened. The bill alone can't answer, because weather, working days and tenants change it too. So how do we prove a saving that the bill can't show? We'll write the plan that does it for the tower in How to Prove Energy Savings: A Simple M&V Plan.

Check your understanding

1. The tower's half-hourly data for the year adds up to 5,944,000 kWh, but the twelve bills add up to 6,000,000 kWh. What should you do before using the data?

2. The chiller plant draws about 220 kW from 6pm to 8pm on working days. The team allows 20 kW for tenants who ask for cooling after 6pm. With 264 working days, what is the central yearly saving from stopping the plant at 6pm?

Recap

• A baseline records how much energy a building uses, when and why, before a change.
• Start with twelve months of bills, then check them against half-hourly data and explain any gap.
• Read the load profile for the base load, the rise, the peak and the evening tail. Compare it with the building's schedules to find waste.
• Identify the drivers, mainly working days, weather (measured as cooling degree days) and occupancy, so later months can be compared fairly.
• Give each saving as a low, central and high estimate.
• For the tower, the three measures are stopping the plant at 6pm, VSDs on the condenser water pumps, and demand limiting.

This is Part 3 of 10 in Cobler's Energy Management: The Economics of Saving Energy course. Previous: From Kilowatt-Hours to Ringgit and Tonnes of CO₂. Next: How to Prove Energy Savings: A Simple M&V Plan.

Cobler builds CobiNeural, a platform that shows a facility team its building's energy, water and indoor air data as live numbers across the whole site. To see how your building performs, talk to us.