Short-term forecasting of peak electrical demand in Iceland: a data-driven approach for enhanced grid management

The increasing complexity of electricity distribution networks and variability in power demand, driven by environmental and societal factors, pose significant challenges for distribution system operators (DSOs) like RARIK. Short-term (24 hours) forecasting of peak electricity consumption is critical for ensuring grid stability and avoiding overloads. This work addresses the prediction of extreme demand events, where surges could compromise network reliability. The project supports a broader initiative leveraging Battery Energy Storage Systems (BESS) and flexibility solutions to mitigate local grid constraints. Our approach combines time series forecasting with hyper-local indicators and external factors, including weather conditions, social events, and industrial activity, to provide actionable insights for grid operators. Using historical load data from RARIK’s network, we modelled demand surges with predictors including temperature, wind speed, precipitation, and calendar events. A hierarchical structure, from individual end-user meters to aggregated substations, captured dependencies across grid levels. Advanced forecasting techniques and robust cross-validation ensured accurate predictions of peak demand under various scenarios. The model identifies periods when consumption risks exceeding critical thresholds, enabling RARIK to manage flexibility resources proactively, reduce operational risks, defer grid upgrades, and ensure reliability during extreme events. This data-driven solution offers DSOs enhanced efficiency and resilience for sustainable energy management.