Machine learning-based carbon intensity estimation in smart grids with behind-the-meter distributed energy resources: A comprehensive review and future perspectives

Abstract

The rapid transition toward low-carbon energy systems has intensified the need for accurate carbon intensity (CI) estimation at high temporal resolutions. The CI estimation is intended to support carbon-aware energy management along with decarbonization strategies. Conventional approach based on average emission factors often fail to capture variations in electricity generation which are dynamic, particularly under increasing penetration of renewable energy resources (RERs) in energy mix. Recent advances in smart meter infrastructures and Behind-the-Meter (BTM) based distributed energy resources (DERs) viz., solar photovoltaic systems have created various prospects for improving the CI estimation. This review presents an all-inclusive assessment of recent advancements in spatio-temporal CI estimation using smart meter data, BTM based SPV disaggregation, machine learning (ML) based analytics and cybersecurity frameworks within in smart grids. This work critically examines preprocessing techniques, energy disaggregation methods, carbon emission flow models in line with robust forecasting approaches. Moreover, cybersecurity threats, mainly false data injection attacks (FDIAs) are discussed along with privacy preserving approaches viz., anomaly detection and encrypted mechanisms. This review further reveals that integrating CI estimation with cybersecurity measures enhances the overall resilience of carbon-aware smart grid applications. This integration also supports demand response programs (DRPs) by improvising the operational decision making in secured manner. This review also reveals the most promising ML approach among the discussed ones viz., hybrid models which are the combination of deep learning and ensemble model for precise forecasting in smart grid applications. Also, emerging artificial intelligence (AI) driven approaches are also conferred which are promising in estimating the CI aspect. Furthermore, this review identifies existing gaps and highlights emerging trends by emphasizing practical insights for researchers and policymakers to develop secure and data-driven smart grid infrastructure for sustainable energy transition to power a cleaner energy future.