The role of age of information in enhancing short-term energy forecasting

Short-term consumption and generation forecasts are expected to play an important role in future energy systems, be for system management or for informing trading platforms. Such forecasts are based on millions of measurements from thousands of Smart Meters (SMs) usually collected at a central location through wireless communication technology. However, this process is prone to delays, that hinder the performance of forecasting methods increasingly based on Machine Learning (ML). In this paper we investigate a new prioritization strategy, based on Age of Information (AoI) optimal processing of measurements at the aggregation node and demonstrate that adopting the AoI-optimal approach improves short-term forecasting. The considered workflow includes SMs sending measurements through a wireless network, an aggregation node where the proposed AoI and conventional policies are compared, followed by a forecasting module realized with 5 different techniques. Simulations with real-world data show AoI-optimal processing doubles 5-minute forecasting accuracy and improves up to 45-minute predictions by over 10%. AoI-optimal processing, unlike traditional approaches, ensures that the system has the most up-to-date measurements available for forecasting. Our findings demonstrate that the way information is processed at the aggregation node is crucial for real-time actions in smart grids, such as short-term household load forecasting.