Do you understand the beauty of AI?

It´s going to get better. As it gets better, more cleantech companies like dLab will incorporate it into their system. As they do, their system will operate more effectively, and we will move even faster towards the sustainable future we aim for. This is what dLab believes!

dLab also believe in supporting the sustainable society of tomorrow by offering an intelligent platform for the energy sector.

By providing visual analytics and enabling data-driven decision-making, the possibilities are opened for a modernized way of working and a more resilient power grid.

How does dLab use AI to gain further insight to their data?

Earlier this year, we conducted a study where we with the help of so-called unsupervised machine learning tried to find clusters of patterns in a large set of our recorded data. The results were very promising, and to further develop our solution dLab is now launching an AI initiative, beginning with the development of a number of proof-of-concepts to test hypotheses that our engineers have put forth.

Our platform analyzes thousands of recordings each week, and a large number of these are categorized as a “registration” (a “registration” in dLab-terms is the lowest level of severity). These registrations in themselves currently does not provide as much useful insights to the customer as an actual disturbance does, but there may be patterns to detect from them, nonetheless.

Also, although we take much pride in how much detailed information our algorithms provide in case of a disturbance (feeder(s) affected, type of incident and key values such as maximum rms-values to mention a few), we believe these recordings might hide even more interesting tidbits, such as what the actual cause of the fault was.

For example, we have seen indications that certain component failures in the grid manifest as specific patterns in the waveforms of the measured voltages. Currently, our solution correctly identifies and categorizes these incidents as either overcurrents, earth faults, etc., but with machine learning, we could potentially do better. These indications that we have observed could be proven (or where applicable, and perhaps just as important, disproven) and even more correlations between waveform patterns and fault causes could be found. We just will not know until we embark on this endeavour.

“It is our conviction that this is a field worth investigating more”, says dLab’s CTO Fredrik Akke. “Of course, we must not abandon the more traditional approach of advanced signal processing, but by complementing our well-tried and powerful algorithms with Machine Learning and statistical analysis, I believe truly interesting results can be achieved”.

The additional information and knowledge the answers to these questions can provide would further enable our customers to take another step towards becoming a truly modern grid operator where they don’t just react to situations but are able to plan ahead and work more proactively.