Dear reader,

‍
After seven years focused almost entirely on product development, deployments, and R&D, we finally decided it was time to start sharing more of what we’ve learned building and optimizing chilled water plants around the world. This newsletter is where we’ll do that. Make sure you subscribe here.

If you work in facilities, controls, HVAC engineering, sustainability, or building operations, you already know the problem: most centralized chilled water plants still run on static control logic and rule-of-thumb strategies that rarely adapt properly to real operating conditions. That’s the gap Exergenics was built to solve.

What you’ll get from this newsletter:

• Technical deep dives into chilled water plant optimization
• Real deployment stories and lessons from live sites
• Discussions around staging, load balancing, condenser water reset, and plant thermodynamics
• Perspectives on machine learning, simulation, and practical HVAC engineering
• Analysis of where building optimization is actually heading beyond the AI hype cycle

• Verified results from projects across commercial offices, healthcare, airports, universities, hotels, casinos, stadiums, and district cooling systems.

Thank you again for joining us and happy reading!

Where it all started

Back in 2016, Iain Stewart was finishing his Master of Engineering at the University of Melbourne and had become fixated on a problem: centralized chilled water plants burn enormous amounts of energy, and the control strategies running them are almost always static rules of thumb set at design. They don’t adapt to real-world conditions. That gap between how plants are designed to run and how they actually run became the basis of a research paper, and then a desktop proof of concept. Version 0.1 of Exergenics.

In late 2019, the company was formally founded after going through EnergyLab and the Melbourne Accelerator Program. Iain’s identical twin brother Tim Stewart came on board as co-founder and COO in 2020. In 2021, the team launched its commercial product and filed an international patent protecting the core two-stage optimization engine.

What we actually do

We optimize centralized chilled water plants. Chillers, pumps, cooling towers, and the controls that run them. Our platform leverages historical telemetry data from your Building Management System (BMS); temperature, pressure, energy, flow rates, pump and fan speeds & weather data. A summer and shoulder season is typically enough to train our models; we don’t need years of data to get started. We use that operational dataset to build machine-learning models for every major component of the plant.

Below is a real example. This is a 3D response surface for a chiller, showing how its efficiency (kW/ton) changes across different part-load ratios and lift conditions. Each black dot is a 15-minute data point from the BMS. The colored surface is our machine learning model, trained on that data. If you want to spin it yourself, you can here.

‍

Once all the component models are calibrated, we stitch them together into a full plant simulation; what we call a Large Physics Model. It works like a “what if” engine. Ask it a question: what happens if I have two chillers running to meet a 5000-ton cooling load, three cooling towers at 75% speed, and dedicated condenser water pumps at 85% speed, with a 70°F wet-bulb temperature and a chilled water leaving temperature of 44°F? It gives you an answer. Individual equipment and total plant energy consumption, chiller and plant kW/ton efficiency, condenser water temperatures and flows, all of it.

But one question at a time is not where the value is. The value is in looping through millions, sometimes hundreds of millions, of operating scenarios and finding the combination of set points that delivers the best outcome.That’s the optimization engine, and it runs in two stages.

Stage one: load balancing. For any given weather condition, thermal load, and equipment combination, what’s the ideal balance of pump speed, fan speed, and chiller part-load that gets total plant energy to its lowest? We call this the sweet spot. The optimizer maps it across every plausible operating condition the plant could face (even if historically the plant has never operated there before).

Stage two: staging. Now we layer in time. As load rises through the day and drops at night, when should machines come on and off? This isn’t just about efficiency anymore. We’re also looking at runtime per cycle, short cycling, runtime balancing between high-load and low-load pairs. It’s a Pareto optimization: trade-offs between energy and mechanical health, weighted to the site’s priorities.

Here’s what makes this different: all of the heavy computation happens offline, ahead of time, in our cloud. By the time we sit down with your team, we’ve already run the simulations, identified the optimal strategies, and stress-tested them against your plant’s real operating data. There’s no agent deployed to your site that needs to “learn on the job” by experimenting with your equipment. The learning is done before we touch a single set point.

The deliverable is a functional description: a set of simple, readable control algorithms that get programmed into the existing BMS by the incumbent contractor. No new hardware. No gateway. No black box. One to two days of retrocommissioning on any modern BMS, and it’s done.

We made this architectural decision early and deliberately, based on extensive feedback from operators and technicians. Your team can examine the new logic. Your BMS partners can interrogate it. Your mechanical contractors can constrain the model based on their expert site knowledge before anything goes live.

If someone asks “why is the condenser water set point doing that?”, there’s an equation with clear variables; not a neural net making decisions you can’t trace. Every recommendation is transparent and fully interpretable. We value collaboration over competition.

Where we are today

Exergenics grew out of the University of Melbourne, and that research relationship runs deep. From the original research paper to the Melbourne Accelerator Program, to our graduate program with the Faculty of Engineering & IT, the university has been a partner since day one. What started as a research pilot in 2021 is now deployed and controlling the central plant across the entire Parkville campus. Our partnership with CSIRO followed a similar trajectory: early joint research through AIRAH iHub led to a full rollout across all CSIRO research labs nationwide.

From those research foundations, we’ve expanded across the board: commercial offices, airports, healthcare, stadiums, shopping centres, hotels, casinos, data centres, and district cooling systems. If it has a centralized chilled water plant, we can help.

Every number we publish is measured and verified against a baseline model at 3 and 12 months post-deployment. No hand-waving. This rigor has been recognized through industry awards from Engineers Australia, the Proptech Association, and AIRAH & CIBSE - global leaders pushing the boundaries of building efficiency

What we will cover in this newsletter

This newsletter will alternate between technical deep dives and stories from the field. Here’s what’s planned to come in future months:

• The condenser water temperature problem
• Why “wet bulb plus four” is such a hard rule of thumb to beat, what common optimization strategies look like, and how dynamic condenser temperature reset consistently pulls 5–15% out of a plant.
• Why air-gapped?
• The trade-offs between live cloud control, bolt-on controllers, and our functional-description approach. Why we went this way & why it is scaling so quickly.
• Deployment stories
• Real-world walkthroughs of specific changes made, their improvements, and what we learn from every site we touch.
• The power of simulation
• Beyond energy optimization, the Large Physics Model can support lifecycle asset planning and “what-if” analysis: simulating equipment upgrades, plant reconfigurations, or shifts in building usage to forecast energy and carbon impacts, derisking capital decisions before they are made.
• The physics
• Chilled water thermodynamics, optimization theory, model predictive control, and where machine learning fits into practical HVAC engineering.
• Market shifts
  The AI hype cycle in building tech, the rise of data-driven controls, and what we’re actually seeing change across different building types.

That’s Exergenics. A team of engineers and data scientists spread across Melbourne, Los Angeles & New York City, scaling fast and focused on becoming the global standard for chilled water plant optimization. We’re not only pleased to actively reduce energy consumption for our clients, we have the added benefit of contributing to a cleaner environment for the whole community. We’ve spent years building the technology. Now we want to share what we’ve learned along the way.

Got questions, topic ideas, or want to talk shop? Reply to this email. We read every one.

Thanks for reading,

Iain Stewart

Cofounder & CEO

‍