How alouach-07/wind-turbine-scada-forecasting-uq-tinyml works
This is a wind-turbine SCADA research pipeline: turbine histories are audited and converted into leakage-aware chronological windows, forecasting models produce next-step power and uncertainty, and analyses assess model behavior. A separate deployment branch distills two hybrid teachers into compact INT8 students and provides embedded inference artifacts. The README documents the broad workflow; sampled source confirms the TinyML inference path and distillation stages. Notebook internals and some unsampled model implementations are not inspected, so those links rely on README statements.
Main entry points: teachers Forecast models · train_students.py Student training · 01_eda_and_data_quality.ipynb Data audit · 02_forecasting_and_uq.ipynb Window preparation
Components 15
SCADA preparation
- Data audit
notebooks/01_eda_and_data_quality.ipynb - Window preparation
notebooks/02_forecasting_and_uq.ipynb
Forecasting and analysis
- Forecast models
models/teachers - Uncertainty evaluation
notebooks/02_forecasting_and_uq.ipynb - Forecast results
results/forecasting - Explainability
notebooks/03_explainability.ipynb - XAI results
results/xai
TinyML distillation
- Teacher targets
models/tinyml/scripts/export_teacher_targets.py - Student training
models/tinyml/scripts/train_students.py - GRU student artifacts
models/tinyml/gru_fastkan_uq_tinyml - XGBoost student artifacts
models/tinyml/xgboost_fastkan_uq_tinyml
Embedded inference
- GRU TinyML inference
models/tinyml/gru_fastkan_uq_tinyml/tinyml_inference.cc - XGBoost TinyML inference
models/tinyml/xgboost_fastkan_uq_tinyml/tinyml_inference.cc
Other components
- Kelmarsh SCADA
- Turbine operator
Connections 15
- Kelmarsh SCADA → Data audit: audited in
README.md - Data audit → Window preparation: precedes
README.md - Window preparation → Forecast models: feeds histories
README.md - Forecast models → Uncertainty evaluation: predicts uncertainty
README.md - Forecast models → Forecast results: benchmarked in
README.md - Forecast models → Explainability: analyzed by
README.md - Explainability → XAI results: exports rankings
README.md - Forecast models → Teacher targets: provides teacher outputs
README.md - Teacher targets → Student training: supplies targets
models/tinyml/scripts/train_students.py - Student training → GRU student artifacts: exports student
models/tinyml/scripts/train_students.py - Student training → XGBoost student artifacts: exports student
models/tinyml/scripts/train_students.py - GRU student artifacts → GRU TinyML inference: loads model data
models/tinyml/gru_fastkan_uq_tinyml/tinyml_inference.cc - XGBoost student artifacts → XGBoost TinyML inference: loads model data
models/tinyml/xgboost_fastkan_uq_tinyml/tinyml_inference.cc - Turbine operator → GRU TinyML inference: supplies window
- Turbine operator → XGBoost TinyML inference: supplies window