Dataset: Screening-based siting and metocean-aware O&M routing for floating offshore wind farms
Description
This dataset and Python code support a screening-to-operations framework for floating offshore wind (FOW) pre-feasibility assessment across 20 Spanish coastal areas. The workflow integrates constrained geospatial screening, spatial out-of-fold machine-learning validation, candidate ranking, and metocean-aware O&M routing. ExtraTrees and LightGBM are used as surrogate models for capacity-factor and relative fatigue-demand screening indicators. The results highlight an energy–maintainability trade-off between energy/wake-aware and O&M-friendly layouts and show that high predictive accuracy does not necessarily guarantee stable Top-N candidate rankings. Weather-dependent accessibility is identified as a major driver of serviceability, decreasing substantially from calm to stormy conditions. The repository provides the datasets and Python implementation required to reproduce the spatial screening, machine-learning evaluation, ranking analysis, layout assessment, and maintenance-routing experiments reported in the study.
Files
Steps to reproduce
================================================================================ STEPS TO REPRODUCE Dataset / Framework: Offshore Wind Farm O&M Optimization Pipeline Target Repository: Mendeley Data ================================================================================ PREREQUISITES ------------- Before running the pipeline, ensure the Python environment 'turbinas2' has been created and verified using the provided 'environment.yml' or 'setup_environment.sh' script. DATASET STRUCTURE ----------------- The dataset is organized into 20 separate directories corresponding to 20 target zones. Each zone directory contains the following standardized structure: Zone_XX/ ├── files/ # Contains input parameters, datasets, and output destination ├── Phase1_M_ver_Final.py # Phase 1: Preprocessing & Data Extraction ├── Phase2_M_ver_Final.py # Phase 2: Intermediate Model Processing └── Phase3_M_Final_Python.py# Phase 3: Final Optimization & Monte Carlo Simulation REPRODUCTION STEPS ------------------ Step 1: Activate the Python Environment ---------------------------------------- Open a terminal / shell session and activate the execution environment: conda activate turbinas2 Step 2: Navigate to a Specific Zone Directory --------------------------------------------- Change your working directory to one of the 20 zone folders (e.g., Zone_01): cd path/to/Zone_01 Step 3: Execute the Multi-Phase Pipeline Sequentially ------------------------------------------------------ Run the three execution scripts in strict sequential order. Each phase depends on the intermediate outputs generated in the 'files/' directory by the previous step. 1. Execute Phase 1 (Preprocessing & Data Extraction): python Phase1_M_ver_Final.py 2. Execute Phase 2 (Intermediate Model Processing): python Phase2_M_ver_Final.py 3. Execute Phase 3 (Final Optimization & Monte Carlo Simulation): python Phase3_M_Final_Python.py Step 4: Verify Generated Outputs -------------------------------- Upon completion of Phase 3, check the contents inside the local 'files/' folder: - Exported numerical results (.csv / .xlsx) - Generated diagnostic plots - Interactive geospatial visualization maps (.html) Repeat Steps 2 to 4 for any of the remaining 20 zone folders to reproduce the full experimental dataset across all geographic regions.
Institutions
- CUNEF UniversidadMadrid, Madrid