Yellow Robot

‘Machine learning’ is a phrase that you’ll sometimes hear used in reference to the software behind ‘smart’ devices.

So what is machine learning, and what role can it play in a home energy management system like carbonTRACK?

What is machine learning?

Machine learning is the process by which software makes decisions based on its own observations and insights; it is one important component in the broader field of artificial intelligence. Systems that use machine learning are more sophisticated than conventional programming, in which the software simply follows a set of inbuilt, static rules with no ability to review or improve its performance.
Software that is capable of machine learning can make conjectures, test different approaches and assess their outcomes without the direct intervention of a human intellect. Using these inferences, the system can make up new provisional rules for itself to follow – all while continuously monitoring whether these rules are helping it to achieve its prescribed goals. A machine learning system will get progressively better at calling the shots over time – even when conditions are not cast in pure black & white.

The origins of machine learning: Spam filtering

The first widespread application of machine learning was for spam / junk email filters. In the decade after the use of email became widespread and commonplace, spammers finding ways to send junk to inboxes around the world. Email service providers wanted to provide their users with a better experience by eliminating spam.
The first simplistic, static filters made their decisions based on the appearance or absence of spammy keywords in the email. These filters had a low success rate compared to modern ones, sometimes failing to block actual spam while other times erroneously marking bona fide messages as junk. Pre-programmed spam-filtering algorithms could simply not keep up with the spammers’ growing arsenal of tricks.
Rather than engaging in an arms race that would require larger and larger teams of programmers to keep ahead of spammers, the anti-spammers developed filters that could ‘think for themselves’, making decisions based on a range of feedback they received and data they acquired as they went along. These highly sophisticated filters could update themselves to keep up with new spamming tactics.

Machine learning in use today

Machine learning – as a type of artificial intelligence – may sound futuristic, but you might be surprised by how common it is. In fact, odds are that it benefits you in some way in your day-to-day life. Here are just a few examples:

  1. Google’s search engine uses machine learning to deliver better search results
  2. Facebook uses machine learning to customise your news feed based on your past ‘likes’ and interactions
  3. Netflix uses machine learning to customise its recommendations to its users

How machine learning makes carbonTRACK better

An energy management that incorporates machine learning will make decisions that deliver more benefits your home than a system that does not. carbonTRACK’s machine learning functionality is some of the most sophisticated in the home energy management space.
At any given moment, the carbonTRACK platform is working away, trying to decide on the most cost-effective way to use energy in your home without compromising comfort. The more devices in the home, the more variables there are to make decisions about. Add in factors like a solar PV system, a battery bank, time of use electricity billing and electricity spot market trading, and the decision making process can become staggeringly complex for an ordinary human with limited (or even a lot of) time on their hands.

Here are some examples of how machine learning could come into play with the carbonTRACK EMS.

Example 1: Water heating


  1. carbonTRACK notices that people in your home tend to take showers in the morning, and so pre-heats your water tank with off-peak electricity to ensure you have hot water available from the cheapest source.
  2. For most of the rest of the day, carbonTRACK leaves the heating element off unless their is excess energy from your solar panels available to do the job – but only if that solar energy wouldn’t be better used for a load of washing or a dishwasher cycle.
  3. carbonTRACK knows that you usually come home around 5pm, and that occasionally you also have an evening shower; as such, it will ensure that there is a ‘buffer’ of hot water available just in case, but may not heat the tank up to 100% capacity.
  4. If you switch from morning showers to evening showers on a regular basis, carbonTRACK will catch onto the change and within a day or two will be heating the water in your tank in the afternoons and leaving the safety buffer in the mornings – always from the most cost-effective source available.

Example 2: Air conditioning


  1. You set your home’s thermostat to 18C for the summer, and set schedules with carbonTRACK to turn on your AC unit to make sure that the temperature is perfect while you are home, using the most cost-effective energy source available to run the unit.
  2. When you’re at work or away from the house, you use carbonTRACK to schedule your AC to turn off – saving you electricity, time and worry about leaving it on all day.
  3. When you’re away over night and forget to change the temperature settings on your AC – in seconds you can use the carbonTRACK app to turn off your AC from anywhere.
  4. One day, a friend calls and asks to stay at the last minute. They’ll be arriving at the house a few hours before you get home and you want the house to be nice and cool for them when they get there. With carbonTRACK, you switch on the AC 20 mins before they arrive and the house is welcomes your guest with the perfect temperature.
Want to know more about carbonTRACK? Here's how it works.

Similar Posts

Join our newsletter