Chiller Plant Operation Optimization with Input and Model Uncertainties

Given high energy costs of chiller plants, plant operation optimization is important. Optimized results may not satisfy actual demand if ignoring input uncertainties. Without considering model uncertainties (e.g., parameters), more energy may be consumed than what is expected. Therefore, optimization with input and model uncertainties is studied. However, models with multiple parameters in the presence of uncertainties are complicated and the number of constraints increases exponentially to satisfy possible conditions caused by uncertainties. Additionally, with continuous and discrete decision variables (e.g., the number of active chillers), the problem is a mixed-integer problem. In this paper, to overcome above difficulties, uncertainties are considered for key parameters selected based on sensitivity analysis. For bound constraints, min/max values of variables with uncertainties are used to reduce the number of constraints. To obtain near-optimal solutions efficiently, a decomposition and coordination method is used. Chillers and cooling towers with the same uncertain temperature are solved together to avoid involving additional coupling constraints for temperature. Results show near-optimal solutions, short CPU times, and scalability of our method. Our method achieves significant energy savings as compared with the baseline, and has higher probability of satisfying cooling demand as compared with the method without considering uncertainties.