Bi-LSTM-Based Deep Stacked Sequence-to-Sequence Autoencoder for Forecasting Solar Irradiation and Wind Speed

• DOI: 10.32604/cmc.2023.038564
• Publication Date: 2023-05-04T06:48:36Z
• Authors: Neelam Mughees; Mujtaba Hussain J...; Abdullah Mughees; Anam Mughees; Krzysztof Ejsmont
• Keywords: Autoencoder; Sequence (biology)

Wind and solar energy are two popular forms of renewable used in microgrids facilitating the transition towards net-zero carbon emissions by 2050. However, they exceedingly unpredictable since rely highly on weather atmospheric conditions. In microgrids, smart management systems, such as integrated demand response programs, permanently established a step-ahead basis, which means that accurate forecasting wind speed irradiance intervals is becoming increasingly crucial to optimal operation planning microgrids. With this mind, novel “bidirectional long short-term memory network” (Bi-LSTM)-based, deep stacked, sequence-to-sequence autoencoder (S2SAE) model for predicting irradiation was developed evaluated MATLAB. To create stacked S2SAE prediction model, Bi-LSTM-based encoder decoder top one another reduce dimension input sequence, extract its features, then reconstruct it produce forecasts. Hyperparameters proposed were optimized using Bayesian optimization algorithm. Moreover, performance compared three other deep, shallow S2SAEs, i.e., LSTM-based gated recurrent unit-based model. All these models also modeled The results simulated based actual data confirmed outperformed alternatives achieving an accuracy up 99.7%, evidenced high reliability forecasting.