notesum.ai

Published at November 27

RL for Mitigating Cascading Failures: Targeted Exploration via Sensitivity Factors

cs.LG

cs.AI

cs.SY

eess.SY

Released Date: November 27, 2024

Authors: Anmol Dwivedi¹, Ali Tajer¹, Santiago Paternain¹, Nurali Virani²

Aff.: ¹Rensselaer Polytechnic Institute; ²GE Vernova Advanced Research

Arxiv: http://arxiv.org/abs/2411.18050v1

Refer to caption

Action Space $(\|\mathcal{A}\|)$	Agent Type	Avg. ST	$\%$ Do-nothing	$\%$ Reconnect	$\%$ Removals	Avg. Action Diversity
$-$	${\sf Do\text{-}Nothing}$	$4733.96$	$100$	$-$	$-$	$-$
$\mathcal{A}_{\sf line}\;(60)$	${\sf Re\text{-}Connection}$	$4743.87$	$99.90$	$0.10$	$-$	$1.093\;(1.821\%)$
$\mathcal{A}_{\sf line}\;(119)$	milp_agent[31]	$4062.62$	$12.05$	$1.70$	$86.24$	$6.093\;(5.12\%)$
$\mathcal{A}_{\sf line}\;(119)$ $\mu_{\sf line}=0$	$\pi^{\sf rand}_{\boldsymbol{\theta}}(0)$	$5929.03$	$26.78$	$5.85$	$67.35$	$13.406\;(11.265\%)$
$\mathcal{A}_{\sf line}\;(119)$ $\mu_{\sf line}=0$	$\text{PG-RL}~{}[\pi^{\sf physics}_{\boldsymbol{\theta}}(0)$ ]	$\mathbf{6657.09}$	$1.74$	$7.66$	$90.59$	$\mathbf{17.062\;(14.337\%)}$
$\mathcal{A}_{\sf line}\;(119)$ $\mu_{\sf line}=1$	$\pi^{\sf rand}_{\boldsymbol{\theta}}(1)$	$5327.06$	$81.51$	$0.28$	$18.20$	$3.625\;(3.046\%)$
$\mathcal{A}_{\sf line}\;(119)$ $\mu_{\sf line}=1$	$\text{PG-RL}~{}[\pi^{\sf physics}_{\boldsymbol{\theta}}(1)$ ]	$\mathbf{6603.56}$	$13.93$	$7.00$	$79.06$	$\mathbf{17.156\;(14.416\%)}$
$\mathcal{A}_{\sf line}\;(119)$ $\mu_{\sf line}=1.5$	$\pi^{\sf rand}_{\boldsymbol{\theta}}(1.5)$	$4916.34$	$92.69$	$0.01$	$7.28$	$3.406\;(2.862\%)$
$\mathcal{A}_{\sf line}\;(119)$ $\mu_{\sf line}=1.5$	$\text{PG-RL}~{}[\pi^{\sf physics}_{\boldsymbol{\theta}}(1.5)$ ]	$\mathbf{6761.34}$	$46.53$	$6.12$	$47.34$	$\mathbf{15.718\;(13.208\%)}$