A unified real-time framework for multi-horizon motion-response forecasting of floating offshore wind turbine platforms under rapidly varying conditions

This code repository is built for the full-process simulation and analysis of wind turbines (focusing on floating wind turbines), encompassing four core functional modules. It covers the complete workflow from typhoon environment modeling and multi-condition parallel simulation to deep learning-based response prediction (with uncertainty analysis) and independent pitch control strategy research, effectively supporting studies on turbine performance, loads, and control strategies under complex wind conditions.Summary of Core ModulesParallel Condition Processing Module: Automates the simulation workflow for multi-load cases, supporting the reading of Excel configuration tables, batch generation of input files, parallel launch of OpenFAST simulations, and output organization, which significantly improves the efficiency of multi-condition simulation.Typhoon Field Modeling Module: Constructs a 3D typhoon wind field model, capable of calculating pressure distribution, gradient wind/friction wind and other parameters, and outputting wind speed/pressure data across different spatial dimensions to provide typhoon environmental boundary conditions for simulations.2DCNN-MIRB Deep Learning Module: Implements the full-process application of the MIRB model based on 2D convolutional neural networks, including data preprocessing, model training, optimization and evaluation, with core support for uncertainty analysis of prediction results (e.g., confidence intervals, error distributions).Independent Pitch Control (IPC) Simulation Module: Builds an IPC simulation framework based on the IEA-15-240-RWT reference wind turbine and UMaineSemi floating platform, which is used to study the impact of different control strategies on turbine response, loads and stability.Key NotesBefore use, configure the paths and executables of external software such as OpenFAST. To ensure experimental reproducibility, it is recommended to set random seeds in model training/optimization scripts and record key parameters.