Emulation-Based Optimal Control of Chiller Plants

Abstract

Chiller plants of HVAC systems usually operate in part-load conditions during most time of a year. Energy efficiency (i.e. energy input ratio, EIR) of chiller plant components such as chillers, pumps, and cooling towers are the functions of part-load ratio (PLR) and other variables. However, different chiller plant components have different functions of EIR to PLR and don’t reach their respective peaks at the same part-load ratio. Based on this fact, there are possibilities that overall efficiency of a chiller plant can be improved to and maintained at the maximum via optimal control of operating numbers and set-points of various plant components at various part-load ratios. An emulation-based optimal control strategy for chiller plants is introduced in the paper. The main idea of this control strategy is to set up a virtual chiller plant as a mirror of a real system. The virtual system is composed of mathematical models that are obtained through theoretical derivation, numerical calculation or off-line test. These models can emulate energy performances of the physical components in the real system. An optimization algorithm is first run on the virtual system to search for an optimal combination of the operating number and set-points of various components to achieve the highest overall efficiency of a chiller plant. An objective function of the optimization algorithm is the overall efficiency of a chiller plant, having real-time cooling load and meteorological data as inputs, and energy and mass balance and component capacities and restrictions as constraint conditions. Once an optimal combination is identified, it will be used to control the real system operation. A mockup system has been installed and operated in a 50,000 m2 office building in Shanghai, China in order to test and verify the optimal control strategy. Preliminary testing results show that the annual overall energy efficiency of the chiller plant (with constant speed centrifugal chillers) is from 0.7 kW/Ton to 0.75 kW/Ton, about 25% less than that of the same plant controlled by normal strategy.