Digitalization – brought down to earth

I assume you have heard the word. DIGITALIZATION. It is buzzing in the media, a development so disruptive that it is called the fourth industrial revolution. The potential is huge: Machine learning, artificial intelligence, automation… “Digitalize or die”, they say. Offering threats and opportunities. Well, let me bring this fuzzy doom-day buzzword-filled topic down to earth for a little while. Let me share my specific experiences with regards to this topic during the last half a year – on my rotation in an operational part of Statkraft.

Hair pulling

Statkraft is a huge trader, both in speculative and physical products. We buy and sell electricity, certificates, guarantees and similar products in many different markets. We produce large amounts of renewable energy ourselves, aggregate energy from more than thousand different plants owned by others, and also deliver large amounts directly to large industrial companies. The values distributed are tremendous, and all of this would not be possible without efficient tools.

When your brain is suddenly all zeros and ones…

The unit in which I have been working as a trainee during the last six months, has responsibility for development of the trading and customer business in the Nordics. The unit has several full time employees developing models, optimizing tools and services and inventing new products. As part of this team I was assigned the task to develop a model that we could provide as a service for our customers. I am afraid I cannot reveal the details of the service, but it was something we had been looking at for a while and not quite figured out. A huge challenge, with a fair amount of hair pulling and also a lot of fun.

Shit in – shit out

Statkraft possesses an enormous amount of data, and with access to all of that you can potentially do a lot. However, for a model to work properly, you need to make sure the data you feed in is of great quality. “Shit it – shit out” is something a learned this half a year, and it sums it up quite well: If your data is useless, a fancy model does not really provide much value.

I tested out many different proceedings – different machine learning models, supervised and not supervised, validation techniques, anomaly detection, backtesting etc. And along the way I learned a lot about these things – things that you normally hear about, but maybe think are too complicated to be relevant for you. The thing is that many of machine learning models are actually off-the-shelf products. They are already developed and open source, waiting for you to make use of them. The hard part is to understand it, to be able to implement it, choose the data and create useful input (feature engineering). And of course make it bullet-proof for future changes and anomalies.

The amazing thing with programming is the incredible simple logic behind it. You have a problem you want to solve, but you need to be able to formulate it in a simple manner in a language that a computer may understand. It requires some brain power to do exactly that, which is also why computers will not completely take over the society. Someone needs to develop and maintain the logic behind them. As well as validate the input and output. A motivation to take on some extra credits in programming?

Summer project

During my rotation, I was lucky to have great support in not only two of my colleagues, Laxman and Vigdis (who are both super skilled at programming), but also 6 intelligent summer students. My unit was responsible for running this year’s summer project. And relevant for our time, this year’s topic was “Digitalization in the energy markets”. I was so lucky as to having the task of recruiting the 6 students and coordinating the project, together with Laxman and Vigdis. The assignment was to develop a new consumption forecasting model for one of Statkraft’s customers. We thought it might be too much of a challenge for the students, but they approached it with an admirable eagerness and focus. It was really inspiring to have them on board, discuss with them and see how both they and the model developed during the summer.

The summer students on their final presentation

In the end I was so proud of what the summer students delivered.On their final presentation for the entire company, and with our CEO and the corporate management on first row, they literally blew away our colleagues. They inspired many people, and with the tool they had developed, with great results, they also had the credibility needed to be bold and give advice for further digital development of the company.

What an amazing group of intelligent and reflected students. I learned a lot from these guys, and really hope that our paths will cross again in the future. If you want to read more about their journey and experiences this summer, I sincerely recommend reading their blog (link). They explain a lot of technical topics in a simple manner, if you want to learn more about machine learning, block chain etc.

With that I say thank you. My trainee period has come to an end – two amazing years, packed with interesting and varied tasks, units, colleagues and experiences. Now, I am taking on a permanent position as Advisor in our Corporate Strategy Development unit – something I have been very much looking forward to.

Thank you for two great years as a trainee together with these amazing people 🙂 Here from our trip to Amsterdam this fall.

All the best.


Leave a Reply