Ultra-short-term PV power forecasting based on LSTM with PeepHoles connections

Bo, Yaolong; Tang, Jiacheng; Yu, Miao; Wei, Wei

doi:10.1109/ispec48194.2019.8975140

Cited by 2 publications

(1 citation statement)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The ultra-short-term data records are also more useful than long-term data [2]. Reference [3] proposes a long short-term memory (LSTM) prediction model, where peepholes connection structure is applied to improve the ability of extracting the dynamic behaviours. Authors in [4] applied persistence method to one-hour-ahead PV power forecasting, regardless of the numerical weather prediction information.…”

Section: Introductionmentioning

confidence: 99%

Ultra-short-term PV prediction based on LSTM with a multi-head attention mechanism

Zhang

Wang

et al. 2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

With the increasing penetration and installation of renewable energy such as photovoltaic (PV) generation, ultra-short-term PV output prediction is necessary to guarantee the stability of the power system. However, the traditional methods cannot capture the important features of PV power data, and prediction accuracy cannot be guaranteed. Focusing on these problems, this paper presents a novel PV prediction method based on LSTM with a multi-head attention mechanism. The mechanism can make the matrix operation run in parallel. At the same time, the model assigns weight to different features of the input data to improve the prediction. To show the effectiveness, this paper constructs four scale types for PV power prediction. Case studies show that the multi-head attention mechanism can improve the prediction performance, and the prediction accuracy decreases following the increase of time scale.

show abstract

Section: Introductionmentioning

confidence: 99%