Home > Research > Publications & Outputs > Approximate Policy Iteration (API) with neural ...

Electronic data

  • 2019frimpongmscbyresearch

    Final published version, 784 KB, PDF document

    Available under license: CC BY-NC-ND: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

Text available via DOI:

View graph of relations

Approximate Policy Iteration (API) with neural networks for the generalized single node energy storage problem

Research output: ThesisMaster's Thesis

Published

Standard

Approximate Policy Iteration (API) with neural networks for the generalized single node energy storage problem. / Frimpong, Richlove.
Lancaster University, 2019. 87 p.

Research output: ThesisMaster's Thesis

Harvard

APA

Vancouver

Frimpong R. Approximate Policy Iteration (API) with neural networks for the generalized single node energy storage problem. Lancaster University, 2019. 87 p. doi: 10.17635/lancaster/thesis/605

Author

Bibtex

@mastersthesis{84720ef62e824d2ea9e78daf243ad74a,
title = "Approximate Policy Iteration (API) with neural networks for the generalized single node energy storage problem",
abstract = "Energy storage problems are hard sequential decision-making problems often modelled as markov decision processes. Exact solution using dynamic programming quickly becomes implausible with large state spaces hence approximate dynamic programming using policy iteration (API) is often employed in such cases. API does not always work, one reason being that the approximation architectures used are often linear for computational tractability reasons. We propose a mathematical model which allows easier implementation of non-linear approximations with API. We use neural networks along with monte-carlo simulation to predict the future values for the generated states during the improvement step of the API algorithm. Our initial experiments suggest that the proposed method provides good results which can be further improved with more fine tuning of the neural network parameters.",
author = "Richlove Frimpong",
year = "2019",
doi = "10.17635/lancaster/thesis/605",
language = "English",
publisher = "Lancaster University",
school = "Lancaster University",

}

RIS

TY - GEN

T1 - Approximate Policy Iteration (API) with neural networks for the generalized single node energy storage problem

AU - Frimpong, Richlove

PY - 2019

Y1 - 2019

N2 - Energy storage problems are hard sequential decision-making problems often modelled as markov decision processes. Exact solution using dynamic programming quickly becomes implausible with large state spaces hence approximate dynamic programming using policy iteration (API) is often employed in such cases. API does not always work, one reason being that the approximation architectures used are often linear for computational tractability reasons. We propose a mathematical model which allows easier implementation of non-linear approximations with API. We use neural networks along with monte-carlo simulation to predict the future values for the generated states during the improvement step of the API algorithm. Our initial experiments suggest that the proposed method provides good results which can be further improved with more fine tuning of the neural network parameters.

AB - Energy storage problems are hard sequential decision-making problems often modelled as markov decision processes. Exact solution using dynamic programming quickly becomes implausible with large state spaces hence approximate dynamic programming using policy iteration (API) is often employed in such cases. API does not always work, one reason being that the approximation architectures used are often linear for computational tractability reasons. We propose a mathematical model which allows easier implementation of non-linear approximations with API. We use neural networks along with monte-carlo simulation to predict the future values for the generated states during the improvement step of the API algorithm. Our initial experiments suggest that the proposed method provides good results which can be further improved with more fine tuning of the neural network parameters.

U2 - 10.17635/lancaster/thesis/605

DO - 10.17635/lancaster/thesis/605

M3 - Master's Thesis

PB - Lancaster University

ER -