Eesti Energia has launched an international competition for data scientists to create a more accurate prediction model for the consumption and production of small producers. The aim of the machine-learning based solution is to reduce clients’ costs and encourage the production of green energy.

Every day, electricity suppliers must predict and purchase electricity from the exchange based on hourly prices for the consumption of their clients as well as to cover the production. If not enough electricity has been purchased, the electricity supplier will have to buy more from the balance market at a higher price than the exchange price. If there is leftover electricity, it must be sold on the balance market with a price much cheaper than the exchange price. Imbalance and balance energy cost are caused by the difference between predictions and reality. The competition will be held in Kaggle, an international community of data scientists and machine learning professionals.

Ilmar Käär, head of the business and information technology at Eesti Energia, says that a more accurate prediction model would help solve electricity production and consumption imbalances and reduce the costs.

‘There are more than 20,000 electricity producers in Elektrilevi’s network alone, where the majority is made up of solar parks. As with micro and small producers, the client’s consumption and production must be predicted at the same time, this is more complicated. At the same time, its accuracy has never been as important: a small mistake in the prediction means very high costs for the electricity supplier, as there are a lot of producers today,’ Käär said.

Solving the energy imbalance and the rising costs resulting from it would also benefit customers. ‘The prediction errors are priced into the customers’ margin as a balance cost,’ Käär explained.

In addition, excessive imbalance can bring along higher operating costs, a potential network instability and inefficient use of energy resources. ‘That is why we are involving capable data scientists from all over the world,’ Käär added. He believes that solving this problem efficiently can make the network connections of micro and small producers easier.

Teams of up to five people can participate in the international competition. The registration is open until 24 January. Prediction models can be submitted until 31 January. The models are evaluated based on MAE (the mean absolute error between the predicted return and the observed target).

The submission of works will be followed by a two-month analysis and evaluation period. The best solutions of the competition will be announced by the end of April. The six best solutions will receive a financial award, with the first place winning $15,000.

Kaggle is the largest international community focused on artificial intelligence and machine learning. It is also the organiser of international competitions related to both artificial intelligence and machine learning. Larger competitions have financial awards, which causes fierce competition all over the world.

Eesti Energia is an international energy company whose home markets are the Baltic States, Finland, and Poland. The group is engaged in both producing and selling energy as well as in offering useful and convenient energy solutions. The goal of the group is to achieve carbon neutrality in electricity production by 2035 and in the entire group’s production by 2045.