Energy & forecasting
Power Trading AI System
Spot-price forecasting and trading strategies for the Japanese and Australian power markets. The repository is private, so screenshots will have to do.
2024Active
AI/ML
Solar Power Prediction System
AI-Powered Solar Energy Forecasting
Time-series models that forecast solar output for an energy-storage demonstration site, scheduled with Airflow. The sun is punctual; the clouds are the problem. The repository is private, so the dashboards here are redrawn on simulated data.
- Role
- Senior Engineer - ML Developer
- Period
- 2023.04 – 2024.12
- Status
- Completed
- Theme
- Energy & forecasting
Overview
Built a comprehensive solar power prediction system for Tainan FamilyMart Pingfeng store energy storage demonstration site. The system combines weather data, historical generation patterns, and machine learning models to provide accurate solar power forecasts.
The problem
- Weather dependency in predictions
- Seasonal variation handling
- Real-time data processing
- Model accuracy optimization
What I did
- Integrated multiple weather data sources
- Implemented seasonal adjustment algorithms
- Built robust data pipeline with Airflow
- Applied ensemble modeling techniques
Results
- Achieved 92% accuracy in daily predictions
- Reduced energy storage costs by 25%
- Successfully deployed to demonstration site
- Enhanced AI showcase for the facility
Features
- Real-time solar power prediction
- Weather data integration
- Automated data pipeline
- Interactive dashboard
- Performance monitoring