To main content

Data Driven Reliable and Resilient Energy System Against Disasters


Disasters such as hurricanes, earthquakes, wildfires, etc. are felt most acutely at local and regional levels. These events have exposed weaknesses in how well-prepared infrastructure operators are to keep their services and provide resilient responses. Outages and service disruptions are largely due to the inability of the affected city infrastructure (i.e., power grids) to cope with random and dynamic disruptive events, translating into resilience deficiencies. A significant challenge is the lack of data availability, sharing, and analysis for emergency planning, and restoration. The special section on Data-driven Reliable and Resilient Energy System Against Disasters in TII aims to address the data-driven approaches for power system and infrastructure reliability and resilience during small-and large-scale extreme weather events or natural disasters. The specific aim is to utilize the advancements in data mining and data processing to minimize catastrophic conditions that affect the quality of critical infrastructure operations, quality of life, and economic activities. The response to our call for this special section was very positive, and we have received many submissions from all around the world in response to the Call for Papers. Each paper was assigned to and reviewed by multiple experts in the field during the review process. Thanks to the support from the IEEE Transactions on Industrial Informatics editorial team, and the dedicated work of numerous reviewers, we were able to accept eight excellent articles covering various aspects of resilient energy system against disasters. In the following, we will introduce these articles and highlight their main contributions. The majority of papers submitted to this SS focused on machine learning applications for power distribution and transmission networks resilience. Hence, half of the papers in this SS are related to machine learning-based resilience approaches using deep learning, Gaussian process, and hybrid learning. Konstantinou and Anubi, in their paper named "Resilient Cyber-Physical Energy Systems using Prior Information based on Gaussian Process," [1] focused on the resilience of cyber-physical energy systems. They formulated their problem as a constrained optimization with prior information based on Gaussian Process. This work is especially critical for Adverse cyber-physical effects (ACEs) in a specific area of the Cyber-physical energy systems (CPES) due to the impact of a hurricane in a disaster-prone region. The paper named "Deep Learning-Based Hurricane Resilient Co-planning of Transmission Lines, Battery Energy Storages and Wind Farms" [2] by Moradi-Sepahvand, Amraee, and Gougheri presents a model for expansion co-planning of power systems with renewable sources considering resiliency aspects against extreme weather events as hurricanes. The main outcome of [2] is that including battery storage and HVDC transmission lines in the planning tools can facilitate the integration of wind farms, increasing system resiliency accordingly. The other paper that used deep learning for power distribution resilience is "Resilient Operation of Distribution Grids Using Deep Reinforcement Learning" by Mohammad Mehdi Hosseini and Masood Parvania. This paper utilizes deep reinforcement learning to develop a resilience controller for real-time dispatching of distributed generation and energy storage units after sudden outages. The proposed algorithm learns the failure development pattern of uncertain high-impact events. The fourth machine learning related paper is "Hybrid Imitation Learning for Real-Time Service Restoration in Resilient Distribution Systems" [4] by Y. Zhang et. al. The paper proposes the imitation learning framework for training for a multi-agent system based on mixed-integer programming. A hybrid policy network can handle tie-line operations and reactive power dispatch simultaneously to improve restoration performance.







  • Western Norway University of Applied Sciences
  • Uppsala University
  • SINTEF Energy Research / Energisystemer
  • Florida State University



Published in

IEEE Transactions on Industrial Informatics



View this publication at Cristin