Forecasting Wind Speed with Autoregressive and Long-Short Term Memory Neural Network Models
Open Access
- 31 December 2025
- journal article
- Published by Izmir Akademi Dernegi in Journal of AI
- Vol. 9 (1) , 98-121
Abstract
The growing need for sustainable and efficient energy systems has intensified interest in accurate renewable energy forecasting. In particular, wind speed forecasting is vital for the reliable integration of wind energy into power systems. This study investigates and compares the performance of three different approaches for wind speed prediction: Autoregressive (AR), Long Short-Term Memory (LSTM) neural networks, and a hybrid AR–LSTM model. Real wind speed data collected from İzmir, Turkey, were used in the experiments. The AR model, a linear statistical method, was evaluated alongside the LSTM model, a deep learning method capable of capturing long-term temporal dependencies. A hybrid model was also developed to benefit from the strengths of both. Additionally, noise reduction techniques such as Moving Average and Gaussian Filtering were applied to enhance data quality and model accuracy. The results demonstrated that the LSTM model achieved the lowest RMSE value (0.084), outperforming both the AR and hybrid models. This suggests that LSTM-based models are more suitable for capturing complex and nonlinear patterns in wind speed data. The findings contribute to the development of intelligent forecasting systems for efficient renewable energy management.Keywords
This publication has 26 references indexed in Scilit:
- A novel wind model downscaling with statistical regression and forecast for the cleaner energyJournal of Cleaner Production, 2023
- Day-Ahead Forecast of Photovoltaic Power Based on a Novel Stacking Ensemble MethodIEEE Access, 2023
- A Review of multilayer extreme learning machine neural networksArtificial Intelligence Review, 2023
- Machine learning forecasts of Scandinavian numerical weather prediction wind model residuals with control theory for wind energyEnergy Reports, 2022
- Deep Learning: A Comprehensive Overview on Techniques, Taxonomy, Applications and Research DirectionsSN Computer Science, 2021
- Applying a Moving Average Filter to Reduce the Frequency Measurement Error for the Case Applying by the Zero Crossing Technique2021
- Music recommender system using restricted Boltzmann machine with implicit feedback2021
- Efficient Deep Learning Approaches for Health Informatics2019
- Noise elimination based on moving average by Guassian distribution weighting method2016
- Deep and Shallow Architecture of Multilayer Neural NetworksIEEE Transactions on Neural Networks and Learning Systems, 2015