Robust Economic MPC for Solar Thermal Systems
Type: Project
Institution: University of British Columbia
Funding: Natural Sciences and Engineering Research Council
Overview
Developed an innovative smart control system that combines Gaussian process machine learning with model predictive control (MPC) for domestic solar thermal systems.
Problem Statement
Solar thermal systems face recurring but unpredictable disturbances, such as changing hot water demand. Traditional control methods struggle to adapt to these quasi-periodic patterns.
Solution
A hybrid approach integrating:
- Gaussian Process regression for disturbance prediction
- Model Predictive Control for optimal decision-making
- Economic optimization for cost-aware operation
Results
- Improved system efficiency by 15%
- Reduced operational costs by 10%
- Successfully handled quasi-periodic unknown disturbances