(342p) Comparative Study of Methods for Optimal Scheduling of Centralized Chilled Water Plants Under Forecast Uncertainty | AIChE

(342p) Comparative Study of Methods for Optimal Scheduling of Centralized Chilled Water Plants Under Forecast Uncertainty

Authors 

Campos, G. - Presenter, University of California, Davis
Liu, Y., University of California, Davis
El-Farra, N., University of California, Davis
Palazoglu, A., University of California, Davis
HVAC (Heating, Ventilation, and Air Conditioning) systems are a major source of energy consumption, being responsible for roughly 14% of primary energy utilization in the US in 2018 (EIA, 2019). A specific subsystem of great importance is the centralized chiller plant, which generates chilled water used for air conditioning. These systems are widely used for large facilities with high thermal load density such as university campuses and present higher efficiency and reliability compared to decentralized units (ASHRAE, 2016). The operation optimization of these systems relies heavily on forecasts of demand, prices and temperatures, which are in turn dictated by highly random factors such as human behavior and weather conditions. The impact of forecasting is further aggravated when the frequency of the closed-loop recalculation is relaxed, e.g. to match/accommodate operator shifts.

Approaches for handling uncertainty in scheduling problems may be classified as either reactive, in which a corrective action is taken only after the realization of the uncertain event, or preventive, in which the uncertainty is modeled explicitly and actions are taken before its realization (Li and Ierapetritou, 2008). Preventive approaches typically make use of one of three techniques: (i) chance-constrained optimization, which minimizes the probability of violating certain constraints; (ii) stagewise stochastic programming, which optimizes the expectation with respect to the uncertain parameters and allows for recourse decisions; and (iii) robust optimization, which only requires uncertainty bounds and typically results in a conservative worst-case solution. As outlined by Mesbah (2016), recent literature on optimal scheduling and control of central cooling plants under forecast uncertainty has mainly focused on a chance constrained approach (Ma et al, 2014, Oldewurtel et al, 2014).

In the present work we perform a comparative study of the different approaches for optimal scheduling of central chiller plants under forecast uncertainty: stagewise stochastic programming, chance-constrained programming, and robust optimization. We propose model formulations for each strategy and demonstrate the advantages and drawbacks of each method. Furthermore, we analyze how the different strategies can complement each other and show how the choice of parameters affect the final equipment sequencing result. The study is performed with real data from the central chilled water plant from the University of California, Davis, which comprises a battery of large-scale centrifugal chillers, a large-scale TES (Thermal Energy Storage) tank and participates in the Day-Ahead electricity market from CAISO (California Independent System Operator).

References

ASHRAE (American Society of Heating, Refrigeration, & Air-Conditioning Engineers). (2016). ASHRAE HVAC Systems & Equipment Handbook.

EIA (Energy Information Administration). (2019) Annual Energy Outlook 2019.

Li, Z., Ierapetritou, M. G. (2008). Process scheduling under uncertainty: Review and challenges. Computers & Chemical Engineering 32, 4–5, pp. 715-727.

Ma, Y., Matuško, J., Borrelli, F. (2014). Stochastic model predictive control for building HVAC systems: Complexity and conservatism. IEEE Transactions on Control Systems Technology 23 (1), 101-116.

Mesbah, A. (2016). Stochastic model predictive control: An overview and perspectives for future research. IEEE Control Systems Magazine 36 (6), 30-44.

Oldewurtel, F., Jones, C. N., Parisio, A., Morari, M. (2014). Stochastic model predictive control for building climate control. IEEE Trans. Contr. Syst. Technol., vol. 22, pp. 1198–1205.