Energy Advice from a Machine Learning perspective

Vaidas Armonas, our Head of Data Science took part in RigaComm 2022 to share some insights how data science can help in these uncertain times.

“The average price for today (2022-10-06) in Nordpool day-ahead market is
214,68 EUR / MWh.
Read that sentence one more time.”
– Vaidas Armonas, Head of Data Science, Eliq

The sky-high electricity prices are something that we are all facing right now, and energy efficiency is on everyone’s mind. Even people who rarely, if ever, second guess their electricity usage patterns are now trying to understand how they could improve their electricity consumption. I took part in RigaComm 2022 last week to give a talk during their Machine Learning Conference, to explain how data science can help in these uncertain times. The conference covered topics such as ‘how AI process automation can help simplify organisation workflows’ and ‘how intelligent machines can take the information we have and use it to generate the information we need’, which brings us to Eliq, and why RigaComm invited me in the first place.

Experts at making Insights from Energy Data

At Eliq, we are energy data experts. We have worked in the energy space for the last ten years. During those years, the company has evolved into a Home Energy Transition platform that can deliver insights and energy advice as well as energy management services by integrating energy consumption data with external data sources such as weather and market price data.

The answer is in the Data

Let us take a more detailed look at what data we use to produce insights.

  • * Energy consumption data – primarily smart meter data – meaning at least hourly data points, but we support all the way up to monthly energy consumption numbers.
  • * Weather data – mostly temperature data, but we use cloudiness, precipitation, and other weather parameters for our services where we see they add value.
  • * Home profile – this unique data source helps us get more information from energy consumers about their homes and provide a more tailored service with otherwise limited information.
  • * Energy advice – advice collected from energy experts and adapted to the specific markets and households, complete with savings estimates for each piece of advice.

Combining the above four sources via our algorithms in different ways enables us to produce our insights.

What Insights?

Let’s start with a story. Last week, I read a thank you letter from a family who forgot to turn their hot water tap off in the basement. That’s cute, but is this related to energy insights? Yes, in a big way! Our Anomaly Alerts feature is designed to spot unusual energy consumption and alert unsuspecting consumers of strange behaviour. And don’t worry,  they’ve turned the tap off now.

Anomaly Alerts help save energy in these rare situations, but what If you wanted to also save energy in a proactive way? Well, the first thing to figure out is what is consuming energy in the first place, which takes us to energy usage categories. 

Energy Usage Categories is a feature that estimates energy consumption for categories like Always On, Fridges & Freezers, Heating, Cooling, EV, etc. Giving an idea of what consumed energy without expensive hardware upgrades.

So, now you now what is consuming the most, great. But how do your consumption patterns compare to others living in similar circumstances? The Similar Homes feature – answers this question. Are you top of the class? Or maybe there is something you can do to reduce energy consumption? It is designed to give you a pat on the back or inspiration to act!

On the note of what you can do to improve, the Energy Advisor feature will happily provide some tips based on your particular location and your specific situation and energy usage. It is designed to take the guesswork out of the decision-making process – Similar Homes says you can improve your energy consumption and Energy Usage Categories indicate that the Always On category is out of proportion? The Energy Advisor will have a recommendation or two.

This all sounds simple enough. And it is – on the outside at least. We work very hard every day to ensure that doing the right thing is simple for energy consumers by giving them relevant and actionable insights.

Ok, Insights. How?

Knowing what data we use to produce certain insights is interesting, but for the data scientists among you, it is even more interesting to hear how these insights are produced. So let us briefly touch on this topic.

Similar Homes answer the question – how do your consumption patterns compare to others living in similar circumstances? This already suggests that we need to select houses with similar characteristics. There are several ways to do this – from business rules where we define every little detail of this grouping to the Nearest Neighbours algorithm, where we allow some variance in similar groups – our approach is in the middle of these two extremes. We define a few strict rules – for house type and heating system – but allow flexibility in other parameters.

Energy and Temperature Relationship

Energy Usage Categories is a much more involved feature and we will definitely share more about it in the future, but in general, we have three different approaches to do it, which allows us to extract a maximum number of insights from the data we have for any given location.

Firstly, we do the statistical disaggregation – what would an average household with given characteristics use in that country? This is a start, but we can improve on it immensely.

So, if we have at least daily data, we can incorporate machine learning approaches to estimating how much energy the location in question consumes for heating and cooling.

To get even deeper insights, we need to have sub-hourly energy consumption data. With that, we can use signal processing approaches to estimate other categories, such as fridge & freezer, always on or electric vehicle charging.

Energy and Temperature RelationshipInsights => Advice

To estimate energy usage and provide actionable insights to enable the home energy transition, we need to combine all our data sources with those of the outside world, and apply just the right amount of data science spice. Then listen carefully to what energy consumers have to say.

From a data scientist’s perspective, that’s more than enough to get you out of bed in the morning.

We haven’t talked about energy advice at all. Or perhaps we have been all along? That’s because giving advice is relatively easy once you know all there is to know about the person (or home in this case) to whom you will provide the advice. By joining our insights with data on numerous locations with energy expertise we can provide tailored and personal advice.