Figure 1. Visual look of new features in Fortum app
Figure 2. Normalized sum of monthly consumed energy in location 10. Red curve indicates the moving average.
Figure 3. Normalized sum of monthly consumed energy in location 3. Red curve indicates the moving average.
Figure 4. Electricity consumption graph from the beginning of October 2017
Figure 5. Temperature graph from the beginning of October 2017
Figure 6. Freezing cycles of refridgerators illustrated. Blue curve is original signal. Red curve is filtered with a high-pass filter.
By combining the minds of three electrical engineering and two business students, we started to ponder what the Fortum app is missing and what else could be done with more real time data. We also got ideas from our own life situations and problems we had stumbled upon.
What it does
Our solution is based on the use of real time and cumulative history data. We would revise the whole concept of Fortum’s offering for B2C customers. First of all, smart plugs could be added to the value offering. Competing firms in this area are Belkin and Vattenfall. However, Belkin is not a direct competitor as they are only selling the actual plugs and do not provide electricity. Therefore, their offering is more focused on additional features not vital to everyday life. Vattenfall on the other hand is also offering electricity, and their smart plugs enable consumers to control devices remotely and switch them on whenever wanted. One plug costs 52€. As competing over customers with price is difficult in electricity industry, this could be a future competitive advantage for Fortum - through which to differentiate and improve customer engagement. Fortum could for example give their new customers one plug to their use during the customership. The actual purpose of the plugs for the customers is to get control over their devices. They could turn them off and on remotely, instantly or by using a timer.
The second point is enriching the Fortum app: the attached Figure 1 demonstrates the new features. Color codes and faces could be added to inform the user of the situation of his/her household: green together with a happy face imply that everything is in control, orange background with a face of concern lets you know that some of your devices are on, and red together with a sad face implies that something is wrong, and you should take actions to fix the situation. This could for example save the customer from water damage if the app realizes that a fridge or freezer has broken down. Consumption patterns as in Figure 2 and 3 could be used as baseline reference for detection of anomalies and deviations from normal. ON/OFF -function would enable the user to utilize the explained smart plugs easily with one swipe. A timer function could be added through which e.g. a sauna or a coffee maker could be turned on according to the preset schedule.
Consumption profile would show more specific details about the usage of different devices, periodical averages, historical data etc. in a visual form. Utilizing machine learning algorithms for identifying recurring patterns of consumption would also bring valuable insights in addition to those obtained with expertise in the field. Also, as the prices for the next day are always known, the app could give saving/scheduling tips. For example, it could imply that washing laundry two hours later than normally would save a certain amount of money. Finally, the app could send the user push notifications such as alerts of concerning situations and the above explained consumptions tips. The app would show the core information of the consumption, and if the customer is interested to know more, additional and more specific data would be available through the app. Data visualization and personalization are key aspects of our solution.
How we built it
We utilized the data Fortum provided for the challenge through the API. Time series data were analyzed in Matlab. For example, in Figure 4 it can be seen that during Monday and Tuesday of the sample period, low amplitude wawing was likely caused by the cold weather (seen in Figure 5) in the Greater Helsinki Area.
For identifying different electric loads such as refrigerators, band pass filters were used as seen in Figure 6, where the DC/low-frequency offset has been filtered out to leave the freezing cycle apparent. Because provided data was uneven and had time gaps, rolling average filters were used as well. For sketching the new app functions, we used GIMP as a photo editor.
Challenges we ran into
During this project there were some challenges puzzling us concerning the data provided. The electricity consumption data, namely, was not completely clean as there were some latency/delay induced indexing problems at times. Also, time naturally sets its own limits and determines what and how advanced outcomes can even be created in just 48 hours. However, that was mainly just taken as a positive challenge to overcome and did not cause too much concern in our team.
Accomplishments that we're proud of
We managed to create a pithy demo and a presentation for Fortum to show and propose how their current available data could be utilized even more beneficially to serve regular household consumers.
What we learned
We learned that through the use of real time data in consumption, many things will become possible. With more time and better technology, the possibilities are endless.
What's next for team HW
By utilizing appliance specific consumption monitoring or user input during identification period, training of neural network could be done. Hopefully, Fortum will get new ideas from our solutions, and the project could move on to a more concrete stage.