At Scale Short-Term Forecasting and Anomaly Detection for GHG emissions with Digital Twins

According to the International Energy Agency (IEA), only the transportation for end-use sectors accounted for 37% of total CO
2
emissions in 2021. To respond to this concern, this paper presents a new methodology in the field of Digital Twins (DTs) that simplifies decision-making on transportation and CO
2
emissions on large geographic areas considering contextual high spatial granularity information for predictions from multiple sources, such as traffic, citizens mobility, events in correlated areas, and weather. This method is based on a hybrid system for traffic short-term forecasting and anomaly detection developed through Data Analytics (DA) and a novel Machine Learning (ML) approach for Time Series Analysis (TSA) able to scale in large scenarios while keeping the promptness of DT environments. The proposed system was developed through the knowledge extracted by a set of real multi modal transportation datasets and demonstrates better performance against conventional techniques such as ARIMA, SARIMAX and VARMA. Finally, this work was integrated in a DT solution for the optimization of transportation and its commitment with Net Zero Targets in a use case for Transport Decarbonization.