Optimizing Power Generation Efficiency with ML-Driven Generation Output Forecasts
BKOAI developed thermodynamics-informed machine-learning models to forecast power-generation output and support proactive dispatch and efficiency decisions
Goals
- Forecast generation output using thermodynamics-informed machine learning models.
- Enable proactive efficiency optimization across generating units.
- Consolidate fragmented operating data from compressors, turbines, and related equipment.
- Give operators timely visibility into generation output forecast and plant-performance indicators.
Challenges
- Data Fragmentation: Operating data across generating units is scattered and inconsistently formatted, hindering analysis.
- Limited Domain Alignment: Off-the-shelf machine-learning models may not capture thermodynamic behavior unique to power-generation cycles.
- Delayed Decision-Making: Operators often lack timely insight into generation-output forecasts, leading to reactive rather than proactive decisions.
Solutions
- Template-Driven Contextualization: We built simulation-driven ML models grounded in thermodynamics and historical time series data to capture the nuanced operation of each unit.
- Seamless Integration: These models were embedded into industrial platforms like Seeq and AVEVA PI, allowing for continuous prediction and efficiency monitoring.
Results
- Generation Output Forecasts: Produced forecasts that supported proactive operating adjustments.
- Custom Model Design: Each unit model reflects real-world operating behavior and plant-specific thermodynamics.
- Real-Time Visibility: Operators gained timely visibility into generation output forecasts and plant-performance indicators.
.svg){kind=small}
.svg){kind=small}
.svg){kind=small}
