Enhanced Predictive Modeling for Wind Turbines based Digital Twin Optimized Multi-Scale Fusion Self Attention Generative Adversarial Network

Wind Turbine (WT) is one of the primary renewable energy sources that provide a reliable and longlasting power supply. It does not contribute to global pollution by releasing emissions of carbon dioxide. Since wind speed is unpredictable, electricity generation forecast and monitoring for wind farms provide a challenging task. As a consequence, it restricts the capacity of managerial staff to make decisions and effectively plan for energy use. This paper proposes an Enhanced Predictive Modeling for Wind Turbines based Digital Twin Optimized Multi-Scale Fusion Self Attention Generative Adversarial Network (EPM-WT-DT-MSFSAGAN). By virtually monitoring wind turbines, this technology uses a cloud-based Digital Twin's (DT) architecture enabled by the 5G-NG-RAN (radio access network) to estimate wind speed and electrical power. The model creates a fivedimensional digital twin architecture built on the digital duplicates foundation of Microsoft Azure. The input data is collected from Onshore wind farm dataset. Next, the Multi-Scale Fusion Self Attention Generative Adversarial Network (MSFSAGAN) is used to predict the wind speed and power. Finally, Humboldt Squid Optimization Algorithm (HSOA) is employed to optimize the weight parameter of MSFSAGAN. The proposed method is implemented in Python. The proposed method achieves 25.68%, 22.54% and 31.24% higher accuracy, 28.89%, 18.26% and 23.49% lower mean square error compared to the existing methods: machine learning-based digital twin for predictive modeling in wind turbines (ML-DT-PM-WT), a digital twin of wind farms using physics-informed deep learning (DT-WF-PI-DL), and a deep learning-based predictive digital twin for offshore wind farms (DL-DT-OWT).