Data-Driven Scenarios Generation for Wind Power Profiles Using Implicit Maximum Likelihood Estimation

Wenlong Liao; Birgitte Bak-Jensen; Jayakrishnan Radhakrishna Pillai; Ruijin   Zhu; Like  Song

Data-Driven Scenarios Generation for Wind Power Profiles Using Implicit Maximum Likelihood Estimation

Wenlong Liao, Birgitte Bak-Jensen, Jayakrishnan Radhakrishna Pillai, Ruijin Zhu, Like Song

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

11 Downloads (Pure)

Abstract

The scenario generation of wind power profiles is of great significance for the economic operation and stability analysis of the distribution network. In this paper, a novel generative network is proposed to model wind power profiles based on implicit maximum likelihood estimation (IMLE). Firstly, the fake sample closest to each real sample is found to calculate the loss function used for updating weights. After training the model, the new wind power profiles are generated by feeding some Gaussian noises to the generator of the IMLE model. Compared with explicit density models, the IMLE model does not need to artificially assume the probability distribution of wind power profiles. The simulation results show that the proposed approach not only fits the probability distribution of wind power profiles well, but also accurately captures the shape, temporal correlation,and fluctuation of wind power profiles.

Original language	English
Title of host publication	Proceedings of 12th International Conference on Applied Energy (ICAE2020)
Number of pages	5
Publisher	Elsevier
Publication date	2020
Article number	0428
Publication status	Published - 2020
Event	International Conference on Applied Energy 2020 - Bangkok, Thailand Duration: 1 Dec 2020 → 10 Dec 2020 https://applied-energy.org/icae2020_cfp

Conference

Conference	International Conference on Applied Energy 2020
Country	Thailand
City	Bangkok
Period	01/12/2020 → 10/12/2020
Internet address	https://applied-energy.org/icae2020_cfp

Keywords

wind power, scenario generation, deep learning, generative network

Fingerprint Dive into the research topics of 'Data-Driven Scenarios Generation for Wind Power Profiles Using Implicit Maximum Likelihood Estimation'. Together they form a unique fingerprint.

Cite this

@inproceedings{38166603a5594af3aaf00fa762ba1747,

title = "Data-Driven Scenarios Generation for Wind Power Profiles Using Implicit Maximum Likelihood Estimation",

abstract = "The scenario generation of wind power profiles is of great significance for the economic operation and stability analysis of the distribution network. In this paper, a novel generative network is proposed to model wind power profiles based on implicit maximum likelihood estimation (IMLE). Firstly, the fake sample closest to each real sample is found to calculate the loss function used for updating weights. After training the model, the new wind power profiles are generated by feeding some Gaussian noises to the generator of the IMLE model. Compared with explicit density models, the IMLE model does not need to artificially assume the probability distribution of wind power profiles. The simulation results show that the proposed approach not only fits the probability distribution of wind power profiles well, but also accurately captures the shape, temporal correlation,and fluctuation of wind power profiles.",

keywords = "wind power, scenario generation, deep learning, generative network",

author = "Wenlong Liao and Birgitte Bak-Jensen and Pillai, {Jayakrishnan Radhakrishna} and Ruijin Zhu and Like Song",

year = "2020",

language = "English",

booktitle = "Proceedings of 12th International Conference on Applied Energy (ICAE2020)",

publisher = "Elsevier",

address = "United Kingdom",

note = "International Conference on Applied Energy 2020 ; Conference date: 01-12-2020 Through 10-12-2020",

url = "https://applied-energy.org/icae2020_cfp",

}

Data-Driven Scenarios Generation for Wind Power Profiles Using Implicit Maximum Likelihood Estimation. / Liao, Wenlong ; Bak-Jensen, Birgitte ; Pillai, Jayakrishnan Radhakrishna; Zhu, Ruijin ; Song , Like .

Proceedings of 12th International Conference on Applied Energy (ICAE2020) . Elsevier, 2020. 0428.

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - Data-Driven Scenarios Generation for Wind Power Profiles Using Implicit Maximum Likelihood Estimation

AU - Liao, Wenlong

AU - Bak-Jensen, Birgitte

AU - Pillai, Jayakrishnan Radhakrishna

AU - Zhu, Ruijin

AU - Song , Like

PY - 2020

Y1 - 2020

N2 - The scenario generation of wind power profiles is of great significance for the economic operation and stability analysis of the distribution network. In this paper, a novel generative network is proposed to model wind power profiles based on implicit maximum likelihood estimation (IMLE). Firstly, the fake sample closest to each real sample is found to calculate the loss function used for updating weights. After training the model, the new wind power profiles are generated by feeding some Gaussian noises to the generator of the IMLE model. Compared with explicit density models, the IMLE model does not need to artificially assume the probability distribution of wind power profiles. The simulation results show that the proposed approach not only fits the probability distribution of wind power profiles well, but also accurately captures the shape, temporal correlation,and fluctuation of wind power profiles.

AB - The scenario generation of wind power profiles is of great significance for the economic operation and stability analysis of the distribution network. In this paper, a novel generative network is proposed to model wind power profiles based on implicit maximum likelihood estimation (IMLE). Firstly, the fake sample closest to each real sample is found to calculate the loss function used for updating weights. After training the model, the new wind power profiles are generated by feeding some Gaussian noises to the generator of the IMLE model. Compared with explicit density models, the IMLE model does not need to artificially assume the probability distribution of wind power profiles. The simulation results show that the proposed approach not only fits the probability distribution of wind power profiles well, but also accurately captures the shape, temporal correlation,and fluctuation of wind power profiles.

KW - wind power, scenario generation, deep learning, generative network

M3 - Article in proceeding

BT - Proceedings of 12th International Conference on Applied Energy (ICAE2020)

PB - Elsevier

T2 - International Conference on Applied Energy 2020

Y2 - 1 December 2020 through 10 December 2020

ER -