On February 24 2022, after the attack on Ukraine by the Russian Federation, the war began for everyone, each defending the country by working, doing what he and she can and helping the Armed Forces of Ukraine. Work on the economic front helps to protect Ukraine too.
Providing the country with electricity is necessary for the army, business and every citizen. Those who, unfortunately, have experienced light outages, understand even more how important it is not only for work, but for maintaining morale and life. We thank all power engineers from the generation and distribution of electricity for their excellent work, for our part we help the following:
One of the solutions in a-Gnostics, which also continues to work during the war, is to forecast electricity consumption in Ukraine. It is necessary to make an hourly forecast for market for the day ahead. The solution for forecasting energy consumption we started to develop before the launch of a new electricity market in Ukraine on July 1, 2019, and successfully implemented it for several enterprises. Machine Learning and Artificial Intelligence are used for more accurate forecasting.
Over the years of experience, it has been proved that the best mathematical models for forecasting electricity consumption are LSTM, XGBoost, Random Forest. It is necessary to get accurate facts and weather forecasts, historical facts of electricity consumption, and to get hourly facts of consumption every day to make a forecast for tomorrow.
More accurate forecasting can save companies up to USD 60K per month (the average calculation. The exact number depends on the corporation actual consumption and the current size of the imbalance).
Before the war, electricity consumption in Ukraine was 480.5 million kWh per day. The average daily accuracy of the forecasts was 97%. The deviation in each hour did not exceed 5-7%. Consumption and the features that affect consumption has changed significantly during the war.
The advantage of using Neural Networks and Machine Learning is that computer algorithms can find various factors that are not obvious or have changed significantly, calculations are performed much faster. Human can search for and understand such connections for a much longer time.
At the beginning of March 2022, the models used by a-Gnostics to forecast electricity consumption in Ukraine were adapted to the new consumption, the accuracy almost reached the pre-war level.
Important is the fact that during the war Ukraine joined ENTSO-E, the European network of electricity transmission system operators, which unites operators in 39 countries on the European continent. We thank NEC Ukrenergo and all Ukrainian power engineers for their excellent work.
Joining the grid further enhances the benefits of accurate forecasting — peak consumption hours in Ukraine occur slightly earlier than in Western Europe, electricity can be bought cheaper, and when demand falls in Ukraine, Europe begins to peak and sell more expensive. Artificial Intelligence adapts to such conditions faster, and will help power engineers in their daily work. Together we will win!