Mirvac Sydney Chiller Plant Optimization

Portfolio Chiller Plant Optimization across Four Mirvac Sydney Assets

Four central chilled water plants, four different control philosophies and four different data platforms. Each one modelled on its own trend data and optimized inside the building's existing BMS, with verified efficiency gains between 9.4% and 21%.

Client

Mirvac

Location

Sydney, Australia

Sector

Commercial Office

Assets

4 buildings, 13 chillers

Plant Capacity

20,656 kWr

Four plants that had nothing in common but an owner

Mirvac's Sydney assets do not share a plant design. Angel Place at 123 Pitt Street runs four PowerPax chillers of unequal size, one at 720 kWr and three at 1,700 kWr. 400 George Street runs three York centrifugals and one PowerPax. 8 Chifley Square runs two magnetic bearing York units. Building 1 at South Eveleigh runs three chillers across five staging levels.

They do not share a controls stack either. Honeywell at 400 George Street and Johnson Controls Metasys at 8 Chifley, with different data platforms again for telemetry: Switch Automation at Angel Place, Bueno at 8 Chifley and South Eveleigh.

What they did share was the usual cause of avoidable energy use in a well run building. These are high performing assets with strong NABERS ratings and competent facilities teams, and the plants were still sequenced on fixed thresholds with condenser water setpoints that did not track ambient conditions. Good buildings, conservative controls.

None of the four required new hardware. Every strategy was delivered as setpoints for the building's existing controls contractor to implement.

Asset by Asset

What was done and what was measured

Each plant was verified over its own reporting period against its own baseline. The measurement windows differ, so the kWh and carbon figures below are the savings measured within each stated window and are not annualized.

400 George Street

Sydney CBD · Premium grade office

21%Verified

The largest plant in the portfolio at 7,476 kWr, and the one with the most staging headroom. Three York centrifugals at 2,159 kWr each plus a 999 kWr PowerPax gave the optimizer a wide set of combinations to choose between, and the original fixed sequence had been using very few of them.

Chiller staging, dynamic condenser water temperature and load balancing were written into the Honeywell BMS in March 2023. The verification regression returned an R² of 0.84, the weakest fit of the three measured sites, across 123 days of measured operation. A longer window would tighten that fit, and this result should be read as the strongest in the portfolio and the least settled.

Read the full 400 George Street case study

Plant capacity: 7,476 kWr
Chillers: 4
Cooling towers: 5
Plant installed: 2011
BMS: Honeywell
Strategy live: 7 Mar 2023
Baseline period: Jan to Dec 2022
Measured: 123 days
Within: 1 Jun 23 to 9 Jan 24
Energy saved: 53,780 kWh
Carbon avoided: 43.0 tCO₂
Regression R²: 0.84

Building 1, South Eveleigh

Australian Technology Park · Commercial office

12.4%Verified, preliminary

A 2018 plant, and the newest in the portfolio. Two 2,070 kWr chillers and one 870 kWr unit across five staging levels, with telemetry delivered by API from Bueno.

Staging, condenser water temperature control and load balancing went live on 1 February 2023. The verification here is preliminary rather than a full twelve month result, and the saving is reported over the measured window only.

Plant capacity: 5,010 kWr
Chillers: 3
Staging levels: 5
Plant installed: 2018
Data platform: Bueno API
Strategy live: 1 Feb 2023
Energy saved: 14,160 kWh
Carbon avoided: 11.2 tCO₂
Verification: Preliminary

Angel Place, 123 Pitt Street

Sydney CBD · 5.5 star NABERS Energy

9.4%Verified

The most rigorously verified plant in the portfolio, and a useful demonstration that a 5.5 star NABERS asset still has efficiency left in it. Four PowerPax chillers, one at 720 kWr and three at 1,700 kWr, on three cooling towers.

Strategy was implemented in December 2022 by Environmental Automation, using data supplied by API from the Switch Automation platform. Verification used weighted average system lift as the independent variable rather than wet bulb temperature, giving a regression fit of R² 0.995 across a 265 day reporting period.

Plant capacity: 5,820 kWr
Chillers: 4
Cooling towers: 3
Plant installed: 2017
NABERS Energy: 5.5 star
Data platform: Switch Automation
Strategy live: Dec 2022
Baseline period: 298 days to Sep 2021
Reporting period: 19 Jan to 9 Oct 2023
Energy saved: 33,240 kWh
Carbon avoided: 26.6 tCO₂
Regression R²: 0.995

8 Chifley Square

Sydney CBD · 5.5 star NABERS Energy

16.4%Modelled

The smallest plant in the portfolio at 2,350 kWr. Two York YMC² magnetic bearing centrifugal chillers on three variable speed cooling towers, serving 19,289 m² of net lettable area.

Magnetic bearing chillers hold efficiency well at part load, which changes where the optimization opportunity sits. The modelled saving of 16.4% has not yet been through measurement and verification, and until it has, it is a forecast rather than a result.

Plant capacity: 2,350 kWr
Chillers: 2, magnetic bearing
Cooling towers: 3, variable speed
Net lettable area: 19,289 m²
NABERS Energy: 5.5 star
BMS: JCI Metasys
Data platform: Bueno
Modelled saving: 47,493 kWh
Modelled carbon: 38 tCO₂
Verification: Pending

How these numbers were produced

Every verified figure on this page comes from a regression of plant energy against measured cooling load, run over a defined reporting period and compared with the same regression fitted to a pre implementation baseline. The saving is the difference between what the baseline model predicts the plant would have consumed under the reporting period's actual conditions, and what it did consume.

That method matters because raw consumption comparisons are misleading. A building that gets hotter, busier or more densely occupied will use more energy on an optimized plant than it did on an unoptimized one in a milder year, and a naive year on year comparison would call that a failure.

The choice of independent variable matters too. At Angel Place, moving from wet bulb temperature to weighted average system lift raised the regression fit from 0.988 to 0.995, because lift captures what the chillers are actually working against rather than a proxy for it.

One method, four very different plants

Take the data the building already produces

Chiller, pump and cooling tower trends came through whichever platform the asset already ran, by API from Switch Automation at Angel Place and from Bueno at 8 Chifley and South Eveleigh. Nothing was installed to collect data that was already being collected.

Model the plant, not the building type

A digital twin was built for each plant from its own operating history. Four PowerPax chillers of unequal size and two magnetic bearing York units present completely different staging problems, and a strategy generated for one would be wrong for the other.

Bound the strategy with site knowledge

Facilities teams supplied the constraints that do not appear in trend data. Which sequences are off limits, which plant is unreliable, which readings to distrust. Recommended setpoints stayed inside the parameters the existing BMS already permitted.

Hand it to the incumbent contractor

The strategy was issued as setpoints and sequences and coded in by the controls contractor already responsible for the building. No Exergenics hardware sits on any of these sites and no connection to the control network is maintained.

Find the headroom in your own portfolio

Exergenics optimizes chilled water plants across commercial office portfolios, hospitals, airports and universities. No capital expenditure, no hardware and no disruption to tenants.

Talk to an expert

Download case study PDF

Measured figures are drawn from the measurement and verification report issued for each asset. Savings are stated over the measurement window given for each site and are not annualized, so they should not be summed into a portfolio annual figure. The South Eveleigh result is preliminary. The 8 Chifley Square figure is a modelled forecast and has not been verified.