Investigating Wind Speed Forecasting for Grid-Connected Renewable Energy Systems Using LSTM and A Day Ahead Chaining

Abstract

Insufficient access and unreliable grid service are among emerging communities' most significant challenges.
Consequently, these communities rely on fossil fuel-based technologies to provide for their electrical needs, resulting
in substantial socioeconomic implications for society. Accurate and reliable wind speed forecasting is essential for
effectively using wind energy, a Renewable Energy Source (RES) with considerable developing potential. Forecasting
wind speed (FWS) is essential for the safe and effective operation of the power grid. Despite considerable study on
various processing and validation methodologies, its application to realistic forecasting remains problematic due to
the substantial influence of newly acquired data on the outcomes of the initial fragmented sub-sequences. A Long
Short-Term Memory (LSTM) neural network-based model for wind speed prediction, validated by the A Day Ahead
Chaining (ADAC) approach, has been proposed to address this problem. A FWS model using LSTM has been
developed, ADAC has been implemented for validation, and a comparison study has been conducted to identify a
superior methodology for application on bigger datasets, enhancing accuracy and forecasting efficiency. The proposed
model emphasizes real-time data characterized by abrupt transitions and dependence on the wind speed.