A reinforcement learning framework for optimal operation and maintenance of power grids

Roberto Rocchetta, L. Bellani, M. Compare, E. Zio, E. Patelli (Corresponding author)

Research output: Contribution to journalArticleAcademicpeer-review

2 Citations (Scopus)

Abstract

We develop a Reinforcement Learning framework for the optimal management of the operation and maintenance of power grids equipped with prognostics and health management capabilities. Reinforcement learning exploits the information about the health state of the grid components. Optimal actions are identified maximizing the expected profit, considering the aleatory uncertainties in the environment. To extend the applicability of the proposed approach to realistic problems with large and continuous state spaces, we use Artificial Neural Networks (ANN) tools to replace the tabular representation of the state-action value function. The non-tabular Reinforcement Learning algorithm adopting an ANN ensemble is designed and tested on the scaled-down power grid case study, which includes renewable energy sources, controllable generators, maintenance delays and prognostics and health management devices. The method strengths and weaknesses are identified by comparison to the reference Bellman's optimally. Results show good approximation capability of Q-learning with ANN, and that the proposed framework outperforms expert-based solutions to grid operation and maintenance management.

Original languageEnglish
Pages (from-to)291-301
Number of pages11
JournalApplied Energy
Volume241
DOIs
Publication statusPublished - 1 May 2019
Externally publishedYes

Fingerprint

Reinforcement learning
reinforcement
learning
artificial neural network
Health
Neural networks
space use
Learning algorithms
Profitability
health

Keywords

  • Artificial neural networks
  • Operation and maintenance
  • Power grid
  • Prognostic and health management
  • Reinforcement learning
  • Uncertainty

Cite this

Rocchetta, Roberto ; Bellani, L. ; Compare, M. ; Zio, E. ; Patelli, E. / A reinforcement learning framework for optimal operation and maintenance of power grids. In: Applied Energy. 2019 ; Vol. 241. pp. 291-301.
@article{65a40e5e212248f9af52d62e918e16a5,
title = "A reinforcement learning framework for optimal operation and maintenance of power grids",
abstract = "We develop a Reinforcement Learning framework for the optimal management of the operation and maintenance of power grids equipped with prognostics and health management capabilities. Reinforcement learning exploits the information about the health state of the grid components. Optimal actions are identified maximizing the expected profit, considering the aleatory uncertainties in the environment. To extend the applicability of the proposed approach to realistic problems with large and continuous state spaces, we use Artificial Neural Networks (ANN) tools to replace the tabular representation of the state-action value function. The non-tabular Reinforcement Learning algorithm adopting an ANN ensemble is designed and tested on the scaled-down power grid case study, which includes renewable energy sources, controllable generators, maintenance delays and prognostics and health management devices. The method strengths and weaknesses are identified by comparison to the reference Bellman's optimally. Results show good approximation capability of Q-learning with ANN, and that the proposed framework outperforms expert-based solutions to grid operation and maintenance management.",
keywords = "Artificial neural networks, Operation and maintenance, Power grid, Prognostic and health management, Reinforcement learning, Uncertainty",
author = "Roberto Rocchetta and L. Bellani and M. Compare and E. Zio and E. Patelli",
year = "2019",
month = "5",
day = "1",
doi = "10.1016/j.apenergy.2019.03.027",
language = "English",
volume = "241",
pages = "291--301",
journal = "Applied Energy",
issn = "0306-2619",
publisher = "Elsevier",

}

A reinforcement learning framework for optimal operation and maintenance of power grids. / Rocchetta, Roberto; Bellani, L.; Compare, M.; Zio, E.; Patelli, E. (Corresponding author).

In: Applied Energy, Vol. 241, 01.05.2019, p. 291-301.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - A reinforcement learning framework for optimal operation and maintenance of power grids

AU - Rocchetta, Roberto

AU - Bellani, L.

AU - Compare, M.

AU - Zio, E.

AU - Patelli, E.

PY - 2019/5/1

Y1 - 2019/5/1

N2 - We develop a Reinforcement Learning framework for the optimal management of the operation and maintenance of power grids equipped with prognostics and health management capabilities. Reinforcement learning exploits the information about the health state of the grid components. Optimal actions are identified maximizing the expected profit, considering the aleatory uncertainties in the environment. To extend the applicability of the proposed approach to realistic problems with large and continuous state spaces, we use Artificial Neural Networks (ANN) tools to replace the tabular representation of the state-action value function. The non-tabular Reinforcement Learning algorithm adopting an ANN ensemble is designed and tested on the scaled-down power grid case study, which includes renewable energy sources, controllable generators, maintenance delays and prognostics and health management devices. The method strengths and weaknesses are identified by comparison to the reference Bellman's optimally. Results show good approximation capability of Q-learning with ANN, and that the proposed framework outperforms expert-based solutions to grid operation and maintenance management.

AB - We develop a Reinforcement Learning framework for the optimal management of the operation and maintenance of power grids equipped with prognostics and health management capabilities. Reinforcement learning exploits the information about the health state of the grid components. Optimal actions are identified maximizing the expected profit, considering the aleatory uncertainties in the environment. To extend the applicability of the proposed approach to realistic problems with large and continuous state spaces, we use Artificial Neural Networks (ANN) tools to replace the tabular representation of the state-action value function. The non-tabular Reinforcement Learning algorithm adopting an ANN ensemble is designed and tested on the scaled-down power grid case study, which includes renewable energy sources, controllable generators, maintenance delays and prognostics and health management devices. The method strengths and weaknesses are identified by comparison to the reference Bellman's optimally. Results show good approximation capability of Q-learning with ANN, and that the proposed framework outperforms expert-based solutions to grid operation and maintenance management.

KW - Artificial neural networks

KW - Operation and maintenance

KW - Power grid

KW - Prognostic and health management

KW - Reinforcement learning

KW - Uncertainty

UR - http://www.scopus.com/inward/record.url?scp=85062607186&partnerID=8YFLogxK

U2 - 10.1016/j.apenergy.2019.03.027

DO - 10.1016/j.apenergy.2019.03.027

M3 - Article

AN - SCOPUS:85062607186

VL - 241

SP - 291

EP - 301

JO - Applied Energy

JF - Applied Energy

SN - 0306-2619

ER -