Unplugg is a product from Whitesmith that provides Machine Learning tools to help you make energy products smarter. This project stemmed from the idea of providing ways to improve energy efficiency in the smart home domain. From that thought, many models were created, namely standby consumption detection, disaggregation, and forecasting. Although this post will focus on the energy use case, many of these models are actually applicable in other use cases, which accounts for their versatility and utility in the current market as you can read in our blog post How To Love Machine Learning.
The only insight most people have into their electricity consumption is the monthly bill. This tells how much electricity was spent and how much it will take to cover the expenditure: that is about all the information normal consumers expect. Which does not stop them from wondering:
- But what about my consumption pattern?
- Do I spend more than other people with the same schedule?
- Has my consumption changed drastically over my typical usage?
- Which electronic devices impact the most in my electricity consumption?
- And how can I go greener and save the most electricity?
Analysing your client’s power consumption will answer all these questions and much more: it will provide a clearer look at what is behind your client’s power bill.
###ROUTINES AND HABITS - CONSUMPTION PATTERNS
A consumer’s everyday habits reflect on the patterns of power consumption of his home, along the day, along the week. The time he showers, cooks breakfast, has lunch or dinner, turns on the tv set or the computer, or if he is absent, have a direct impact on his power profile. The work schedule, that is, if he works during the day/night or on weekends or even if he works from home, also influences this profile.
There may be other people living with your client, including children. This increases the number of activities related to use of electricity inside the house, each with its own patterns in terms of frequency and time interval, specific to the individual responsible for that activity.
Naturally, a number of consumers may have similar patterns and therefore common strategies may be prepared, with the best options for pricing and for reducing your client’s overall power bill.
The consumer builds up these patterns with past usage. But what about what is to come?
###A LOOK INTO THE FUTURE - PREDICTION
The future is not as unpredictable as you might think.
The use of power meters, some of which fall into the category of “smart meters”, for general and equipment specific power consumption measurements, allows building a history of your client’s power consumption and related expenditure. From this history, one is able to identify trends, periodicities, and irregularities.
The power supplier may then use this information so as to optimize the service they offer, and:
- predict what is happening in the near future in order to anticipate problems,
- plan in advance the distribution network or
- help your client find the best power plan for the times to come.
However, as prediction is based on historic data, changes in behaviour have to be taken into account in order to maintain the quality of the prediction models. This takes us to the next topic.
###ROUTINES ARE BORING / CHANGE POINTS
If your client’s power fingerprint changes he will probably switch consumer group and require new prediction models.
In some cases, your client’s grown-up kids have moved away, your client changed the central heating or opted for natural gas in some piece of equipment.
In this case, other (updated) solutions must now be provided to account for this change of direction.
This requires the detection of what is known as “change points”. These are time points in which a change in the power consumption profiles really arises from some actual modification in the situation, pointing to new adapted schemes.
Not all changes in behaviour are considered change-points, though.
It is noticeable that the household consumption pattern differs from weekdays to weekends. These are repeated every week and do not result from a change of habits.
Also, the needs for heating, lighting, etc. evolve gradually with the seasons, and a good description of patterns should, naturally, encompass this evolution.
Until now we have only been talking about aggregate consumption analysis, that is, whole house power consumption. This does provide an interesting insight into the interaction between household appliances and their users. However, it only provides a description and an incomplete one…
This is where disaggregation of consumption comes into play.
###GOING DEEPER - DISAGGREGATION
Disaggregation opens the way for a more in-depth analysis, identifying which appliances are being used and when, as well as their consumption patterns.
This new information allows for the development of automated systems, making your client’s house smarter and improving its power efficiency. A simple example consists of eliminating standby power consumption, thus reducing costs.
Is your client constantly using hot water? He may notice that the boiler keeps working to maintain the water temperature, even if he is not going to take a shower or wash dishes in the next few hours.
The identification of periods in which the boiler (or any other appliance) can be disconnected and the development of automated on/off schedules can also be implemented in your client’s smart home.
This is just a small part of the information we can obtain from electricity consumption alone, and Unplugg is working hard to deepen your knowledge in this and various other areas.