Short-Term Solar Irradiance Forecasting Model Based on Bidirectional Long Short-Term Memory Deep Learning

Alharbi, Fahad and Csala, Dénes (2021) Short-Term Solar Irradiance Forecasting Model Based on Bidirectional Long Short-Term Memory Deep Learning. In: 2021 International Conference on Electrical, Communication, and Computer Engineering (ICECCE) :. IEEE, pp. 1-6. ISBN 9781665446020

Text (Short_Term_Solar_Irradiance_Forecasting_Model_on_Bidirectional_Long_Short_Term_Memory_Deep (1))
Short_Term_Solar_Irradiance_Forecasting_Model_on_Bidirectional_Long_Short_Term_Memory_Deep_1_.pdf - Accepted Version
Available under License Creative Commons Attribution-NonCommercial.
Download (1MB)

Abstract

Recurrent neural networks (RNNs) are the most effective technology to study and analyze the future performance of solar irradiance. Bidirectional RNNs (BRNNs) provide the key benefit of manipulating the data with two different hidden layers in two opposite directions and can feed back to the same layer of output. In this approach, the output layers can simultaneously receive information from the past (backward layers) and the future (forward layers). A bidirectional long short-term memory (BI-LSTM) model was developed and employed to predict solar irradiance values for the next 169 hours based on hourly historical data (01-01-1985 to 16-09-2020) from Tabuk city. The findings specifically demonstrate that in terms of classification and considerations, the BI-LSTM model has promising performance with notable accuracy. In addition, the model is capable of coping with the selected size of sequential data. The prediction accuracy and performance of the BI-LSTM model were highly enhanced when external data such as wind speed and temperature were provided