Using machine learning to optimise central heating

Central heating is creating waves in New Zealand.

It’s by no means the most common or popular method of space heating in residences around the country – that distinction, according to the 2018 census, goes to heat pumps and split systems – but it’s absolutely one of the fastest-growing methods of heating a home.

Of course, there’s a reason some methods are more popular than others, and a lot of it comes down to convenience. Heat pumps and split systems are cheaper and easy to install, while central heating systems need to be tailored to a specific floor plan and are often difficult to retrofit.

Meanwhile, there’s also a significant engineering component – involving a complex process of designing the system, assigning the right heat source coupled to the right delivery system, and other crucial technical details. This process alone can take days or even weeks to deliver, depending on the engineer’s workload and the size of the project.

This process not only sheds light on technical details necessary for central heating, but also determines the estimated finished capital and running costs for the systems – making it one integral to the entire project. Subsequently, it’s common for both specifiers and clients to be anxious to get it done quickly, as delays in the engineering process can grind an entire project to a halt.

But what if there was a way to circumvent this time-consuming process? That’s the question that Auckland-based central heating specialist Waterware recently sought to answer – and eventually did, with their HeatPro Rapid service.

The service seeks to bypass the engineering leg of the central heating project by creating software that makes the calculations in a fraction of the time. The service works out which components best fits the system, as well as the quantities needed to have the most effective solution.

Darren Yearsley, Waterware’s technical director, says the system was built upon the company’s existing database of jobs which listed each project’s cost per component.

“Every design that we do creates this matrix, which will work out the cost of the underfloor heating, the hot water system, the piping network – all the separate components – on a cost per square metre basis,” says Darren.

“We built this database over a period of six months, during which we processed about 800 quotes so it was a true wealth of information.”

Darren says the solution was initially used only for error checking – to compare the prices of Waterware’s designs to industry standards and averages. But then it dawned on them that it could also be used as a completely separate, invaluable tool for specifiers and clients.

“We realised, now that we’ve got this machine learning algorithm that keeps accumulating data in the background, we can use that to do cost estimates rather than just designs,” says Darren.

“And now, rather than doing a full design upfront, which takes three and a half to four hours of engineering time, you can do a quick cost estimation, which takes two or three minutes.”

The service can also include other costs as part of the total, including installation, and lists merchant, installer and retail prices available. It can also provide an estimated running cost, and details each component of the central heating system and its capabilities.

“It’s really revolutionised our business, because it prevents the scenario of spending several engineering hours on a design which can be declined if the client deems it too expensive, which is ultimately a waste of engineering time,” Darren says. “Now, at the beginning of the process, clients will get this cost estimation which is within 2% of what the quote will be.

“And the speed of the system means they don’t need to wait two weeks to find out how much it will cost, either. The speed is something that is truly valuable in this industry.”

Learn more about Waterware and its HeatPro Rapid service.