Electrical Storage, Building Mass Storage, Economic Optimization, Model Predictive Control, Embedded Batteries
Since building heating, ventilation, and air conditioning (HVAC) systems are significant consumers of primary energy, considerable efforts are being made to improve energy efficiency and decrease energy costs in these applications. Notably, substantial opportunities in the area of HVAC control exist for decreasing energy costs by shifting loads from peak periods to off-peak periods in the presence of time-varying utility prices. This load shifting is also beneficial for power companies since it results in a more constant total load allowing them to operate more efficiently. Economic model predictive control (MPC) has been shown to significantly decrease the energy costs of commercial HVAC systems via load shifting. Typically, thermal energy storage (TES) is used for this purpose by running HVAC equipment at higher rates during periods of low power prices to charge TES and at lower rates during periods of higher prices while discharging TES to meet building demand loads. However, with batteries becoming less expensive to manufacture, electrical energy storage in batteries is becoming a viable option for load shifting. Batteries can be used for both load shifting to decrease costs and revenue generation if the incentives on the electricity market are appropriate. In this work, embedded battery applications are considered. In embedded battery applications, the batteries are directly packaged with airside equipment such as air handler units (AHUs), roof-top units (RTUs), and variable refrigerant flow systems (VRFs). In this arrangement, the batteries are accessible only to the local unit and not to other units. In this paper, we propose a hierarchical control system framework for the economic optimization of distributed embedded battery units. The architecture considers both building mass storage as well as the electrical energy storage of the battery units. A high-level problem performs an economic optimization over the entire system using aggregate models. The low-level layer is broken into subsystems, each optimizing its local decisions with higher fidelity models. Advantages of this framework include no iterative communication required between subsystems, decreased computational complexity in the high-level problem allowing for real-time online implementation, and management of total demand across the entire system to reduce peak demand charges. We conclude with a simulation study demonstrating the benefits of the proposed control architecture.