A robust chaos-inspired artificial intelligence model for dealing with nonlinear dynamics in wind speed forecasting

Abstract

The global impacts of climate change have become increasingly pronounced in recent years due to the rise in greenhouse gas emissions from fossil fuels. This trend threatens water resources, ecological balance, and could lead to desertification and drought. To address these challenges, reducing fossil fuel consumption and embracing renewable energy sources is crucial. Among these, wind energy stands out as a clean and renewable source garnering more attention each day. However, the variable and unpredictable nature of wind speed presents a challenge to integrating wind energy into the electricity grid. Accurate wind speed forecasting is essential to overcome these obstacles and optimize wind energy usage. This study focuses on developing a robust wind speed forecasting model capable of handling non-linear dynamics to minimize losses and improve wind energy efficiency. Wind speed data from the Bandırma meteorological station in the Marmara region of Turkey, known for its wind energy potential, was decomposed into intrinsic mode functions (IMFs) using robust empirical mode decomposition (REMD). The extracted IMFs were then fed into a long short-term memory (LSTM) architecture whose parameters were estimated using the African vultures optimization (AVO) algorithm based on tent chaotic mapping. This approach aimed to build a highly accurate wind speed forecasting model. The performance of the proposed optimization algorithm in improving the model parameters was compared with that of the chaotic particle swarm optimization (CPSO) algorithm. Finally, the study highlights the potential of utilizing advanced optimization techniques and deep learning models to improve wind speed forecasting, ultimately contributing to more efficient and sustainable wind energy generation. This robust hybrid model represents a significant step forward in wind energy research and its practical applications. © 2024 Barış et al.