Edge-Based Short-Term Energy Demand Prediction

The electrical grid is gradually transitioning towards being an interconnected area of the smart grid, where embedded devices operate in autonomous manner without any human intervention. An important element for this transition energy demand prediction, since needs have substantially increased due to introduction new and heavy consumption sources, such as electric vehicles. Accurate especially short-term durations (i.e., minutes hours), allows operators produce substantial amount needed satisfy demand–response equilibrium avoid peak electricity load conditions that may also lead blackouts densely populated areas. However, achieve accuracy level, machine learning (ML) models require extensive training with historical measurements, which usually resource intensive (e.g., memory processing power). Hence, deriving accurate predictions demands challenging absence external factors environmental data from different regions seasons categorical values bank/bridging holidays ML model. Additionally, existing work focuses on model execution Cloud platforms, does not real-time requirements predictions. To address these challenges, article presents a method considers build profile each consumer multi-access edge computing (MEC) framework. based Temporal Fusion Transformer (TFT) model, it learn temporal dependencies gathered measurements predict satisfying accuracy. applied home management system testbed containing photovoltaic systems, meters, sensors actuators detecting temperature, humidity radiation) well storage systems additional supply source. MEC framework deployed concentrator TFT executed low requirements, ensuring security do leave location they are produced.