Case Study: Elektroprivreda Srbije (EPS) achieves faster, more accurate energy forecasts and increased trading profits with Microsoft Azure

Serbian leading energy provider boosts trading profits with predictive artificial intelligence

Elektroprivreda Srbije (EPS), Serbia’s state-owned electricity generator, supplier and trader, faced costly inefficiencies from manual, spreadsheet-based demand forecasting. Dispatchers spent hours entering decades of consumption and weather data, producing error-prone hourly forecasts that increased balancing costs and reduced trading performance.

Working with Microsoft and Informatika AD, EPS deployed a solution using Azure Machine Learning, Power Apps and Power BI to ingest 20 years of data and run continual ML-driven forecasts. Forecasting time fell from two hours to 15 minutes, error margins dropped to 1.7%, and the improvements translate to around €600,000 saved annually on balancing costs plus an estimated €300,000 in additional trading profit.

Elektroprivreda Srbije

Danilo Komatina

Principal Engineer