Hybrid deep learning for power generation forecasting in active solar trackers. Issue 23 (16th October 2020)

• Record Type: Journal Article
• Authors: Frizzo Stefenon, Stéfano; Kasburg, Christopher; Nied, Ademir; Rodrigues Klaar, Anne Carolina; Silva Ferreira, Fernanda Cristina; Waldrigues Branco, Nathielle
• Is Part Of: IET generation, transmission & distribution. Volume 14:Issue 23(2020)
• Page Start: 5667
• Page End: 5674
• Publication Date: 2020-10-16
• Subjects: wavelet transforms -- power engineering computing -- feature extraction -- photovoltaic power systems -- learning (artificial intelligence) -- load forecasting -- signal denoising -- recurrent neural nets -- wavelet neural nets -- filtering theory -- solar power stations; signal characteristic -- feature extraction -- signal noise reduction -- solar uptake -- photovoltaic panel generation -- electricity demand -- active solar trackers -- power generation forecasting -- hybrid deep learning -- dual‐axis solar trackers -- photovoltaic power forecasting -- wavelet LSTM -- wavelet energy coefficient -- signal filtering -- nonlinear data -- long–short‐term memory; Electric power production -- Periodicals; Electric power transmission -- Periodicals; Electric power distribution -- Periodicals; 621.3105
• DOI: 10.1049/iet-gtd.2020.0814
• Languages: English
• ISSNs: 1751-8687

Abstract

Abstract : To meet the growing electricity demand for consumers, it is necessary to use more efficient systems. The solar trackers stand out among the applications that can improve the efficiency of photovoltaic panel generation by increasing their solar uptake. For solar trackers to be more efficient, they can base their position update on a generation forecast and thus perform the control action only when there is greater efficiency in this update. For generation forecast, the long–short‐term memory (LSTM) can handle a large volume of non‐linear data. Furthermore, to improve the analysis, it is possible to apply signal filtering techniques. The wavelet energy coefficient is a technique used to reduce signal noise and extract features; this technique performs the filter and preserves the signal characteristic. In this study, the authors present a combination of wavelet energy coefficient and LSTM, defined as wavelet LSTM, to perform photovoltaic power forecasting in the dual‐axis solar trackers.