English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Due to the proliferation of renewable energy and its intrinsic intermittencyand stochasticity, current power systems face severe operational challenges.Data-driven decision-making algorithms from reinforcement learning (RL) offer asolution towards efficiently operating a clean energy system. Although RLalgorithms achieve promising performance compared to model-based controlmodels, there has been limited investigation of RL robustness insafety-critical physical systems. In this work, we first show that severalcompetition-winning, state-of-the-art RL agents proposed for power systemcontrol are vulnerable to adversarial attacks. Specifically, we use anadversary Markov Decision Process to learn an attack policy, and demonstratethe potency of our attack by successfully attacking multiple winning agentsfrom the Learning To Run a Power Network (L2RPN) challenge, under bothwhite-box and black-box attack settings. We then propose to use adversarialtraining to increase the robustness of RL agent against attacks and avoidinfeasible operational decisions. To the best of our knowledge, our work is thefirst to highlight the fragility of grid control RL algorithms, and contributean effective defense scheme towards improving their robustness and security.

Improving Robustness of Reinforcement Learning for Power System Control with Adversarial Training