What is it about?

This study proposes a new framework, Safe-AutoSAC, that enhances integrated energy system (IES) scheduling using safe deep reinforcement learning (Safe DRL) combined with automated machine learning (AutoML). The model leverages a multi-channel Informer-based forecasting module to accurately predict uncertain loads and renewable generation. It then uses a safety-constrained reinforcement learning policy to make scheduling decisions that maximize efficiency while respecting system safety. The novelty lies in the automatic tuning of the learning process and in jointly optimizing IES operations with flexible electric vehicle (EV) demand response, enabling secure, adaptive, and cost-effective energy management under real-world uncertainties.

Featured Image

Why is it important?

As energy systems become increasingly complex due to the integration of renewables and electric vehicles, conventional scheduling approaches often fail to address real-time uncertainty and operational safety. Safe-AutoSAC tackles these challenges by integrating learning-based control with safety assurance and automatic adaptability. This is particularly critical for ensuring reliable and economical operation in future smart grids. The method’s ability to generalize across varying conditions without manual tuning makes it a scalable and practical solution for modern IES scheduling.

Perspectives

From my perspective, Safe-AutoSAC is a promising step toward fully autonomous and resilient energy management. By combining the learning power of DRL with the self-optimization of AutoML and integrating safety into decision-making, this framework offers a balanced solution that is both intelligent and trustworthy. I see great potential for applying this approach to city-wide energy systems, microgrids, and other real-time control scenarios, especially where safety and flexibility are non-negotiable.

Professor/Clarivate Highly Cited Researcher/Associate Editor of IEEE TSG/TII/TSTE Yang Li
Northeast Electric Power University

Read the Original

This page is a summary of: Safe-AutoSAC: AutoML-enhanced safe deep reinforcement learning for integrated energy system scheduling with multi-channel informer forecasting and electric vehicle demand response, Applied Energy, December 2025, Elsevier,
DOI: 10.1016/j.apenergy.2025.126468.
You can read the full text:

Read

Contributors

The following have contributed to this page