Eesti Energia launched an international competition among data scientists at the beginning of winter to find a more accurate solution for forecasting the consumption and production of micro-producers. The machine-learning solution aims to reduce costs for customers and promote green energy production.

According to Martin Laid, Head of E-lab, the innovation division of Eesti Energia, the challenge set by the energy company has become one of the most popular machine learning and artificial intelligence competitions ever in Kaggle, the international community of data scientists and machine learning specialists. "In a few months, more than 2,500 IT and data science experts have contributed to the challenge. In total, the participating teams have proposed nearly 45,000 different solutions to the problem," Laid said.

The international competition is open for teams of up to five members and the registration is open until January 24. Models for forecasting can be submitted until January 31. The models proposed by data scientists are evaluated on the basis of the mean absolute error, or MAE, between the estimated revenue and the objective pursued.

According to Laid, a more accurate forecasting model, which is being sought on the competition, would help to solve the problem of imbalances in electricity generation and consumption and reduce the costs arising from them.

Every day, electricity suppliers have to forecast and buy electricity from the power exchange for the next day, with hourly accuracy, to cover their customers' consumption and generation. If not enough electricity is purchased for the customers, the electricity supplier has to buy it on the balancing market at a higher price than the exchange price. If there is surplus electricity, it has to be sold on the balancing electricity market at a price significantly below the exchange price. The size of the difference between the forecast and the reality is the source of the imbalance and the cost of balancing electricity.

"Elektrilevi's network alone has more than 20,000 electricity producers, of which the majority are solar parks. Since both customer consumption and production have to be forecast at the same time for micro and small producers, forecasting is more complex. At the same time, its accuracy is much more important than in the past: a small error in forecasting means very high costs for the electricity supplier, as there are already a large number of producers today," Laid explained.

Customers would also benefit from resolving the problem of energy imbalances and the resulting rising costs, as the forecasting error percentage is priced into customers' margins as the cost of balancing electricity. In addition, excessive imbalances can lead to higher operating costs, potential network instability, and inefficient use of energy resources.

After the submission of the solutions by the participants, a two-month period of analysis and evaluation will follow, and the best solutions will be announced at the end of April. The six best solutions will also receive a cash prize, the prize for the first place is $15,000.

Kaggle is the largest international community focused on artificial intelligence and machine learning, as well as the organiser of international machine learning and artificial intelligence competitions. Cash prizes can be won on the major competitions, leading to intense competition from 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 energy production and sales, as well as providing customers with useful and convenient energy solutions. The group aims to achieve carbon neutrality in electricity production by 2035 and in the group’s production as a whole by 2045.