Tytuł pozycji:
Learning from the COVID-19 Pandemic to Improve Critical Infrastructure Resilience using Temporal Fusion Transformers
During the COVID-19 pandemic, traditional demand prediction models drastically failed mostly due to altered consumption patterns. Accurate forecasts are essential for ensuring grid stability. This paper analyzes the performance of the Temporal Fusion Transformer (TFT) model during the COVID-19 pandemic aiming to build resilient demand prediction models. Through detailed analysis, we identify which features may contribute to improved performance during large-scale events such as pandemics. During lockdowns, consumption patterns change significantly, leading to substantial errors in existing demand prediction models. We explore the impact of features such as mobility and special day considerations (e.g., lockdown days) on enhancing model performance. We demonstrate that periodic updates on a monthly basis make the model more resilient to changes in consumption patterns during future pandemics. Moreover, we show how improvements in prediction accuracy translate to real-world benefits, such as enhanced grid stability and economic advantages, including reduced energy waste. Additionally, we discuss the implications for energy-critical infrastructure, considering disruptive scenarios like future pandemics.
1. This research has received funding from the European Union’s Horizon Europe research and innovation programme under the grant agreement No. 101073821 (SUNRISE) and 101070052 (TANGO).
2. Thematic Sessions: Regular Papers
3. Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki (2025).
