Day-ahead Prediction of Wind Power Based on Conditional Generative Adversarial Network

Jinyong Dong, Jian Chen, Qiuwei Wu, Bo Pan, Gang Liu

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

    Abstract

    Short-term forecasting of wind power plays a critical role in day-ahead dispatching and economic operation of power systems. In this paper, a wind power prediction model based on Convolutional Neural Networks (CNN) and Conditional Generative Adversarial Network (CGAN) is proposed. With weather factor labels helping, the model is proposed to solve the problem of day-ahead prediction. In this model, K-means is applied to divide the historical wind farm data into several parts based on weather factors. At each time, the power of similar weather time can be found through grey relational analysis (GRA), which can be used as a label together with weather information. With the guidance of conditional labels, the generative model can generate samples more purposefully, and the discriminator can identify more accurately. By this way, the generator and the discriminator form a game so that CGAN model can improve the accuracy of wind power forecasting. Finally, three examples of Datang Hongxing Wind farm are given to test the validity of the proposed algorithm.

    Original languageEnglish
    Title of host publicationProceedings of 2021 IEEE Sustainable Power and Energy Conference
    PublisherIEEE
    Publication date2021
    Pages73-79
    ISBN (Electronic)9781665414395
    DOIs
    Publication statusPublished - 2021
    Event2021 IEEE Sustainable Power and Energy Conference - Nanjing, China
    Duration: 23 Dec 202125 Dec 2021

    Conference

    Conference2021 IEEE Sustainable Power and Energy Conference
    Country/TerritoryChina
    CityNanjing
    Period23/12/202125/12/2021

    Keywords

    • Conditional generative adversarial network
    • Renewable energy
    • Short-term forecasting
    • Wind power forecasting

    Fingerprint

    Dive into the research topics of 'Day-ahead Prediction of Wind Power Based on Conditional Generative Adversarial Network'. Together they form a unique fingerprint.

    Cite this