Towards enabling grid-interactive technologies for HVAC systems in commercial buildings

Recent reports in climate science continue to incentivize policymakers to accelerate the integration of large shares of renewable energy. This premise accelerates the needs for new technologies to accommodate the practical challenges of incorporating variable generation. Grid-interactive efficient buildings (GEB) propose a vision where buildings autonomously interact to provide grid services and be compensated accordingly. Commercial heating, ventilation and air conditioning systems (HVAC) have been deemed a prospective resource due to their high electricity consumption and existing levels of automation. This PhD thesis has the purpose of contributing to the existing body of knowledge regarding demand flexibility from HVAC systems in commercial buildings. At a high level, this work targets three overarching challenges in this research area. 

The first challenge is related to the scalable and accurate prediction of load shed potential for commercial HVAC systems. Predicting and communicating load shed potential is a desirable property for an asset that participates in advanced grid-services through an aggregator. Given this challenge, I proposed AlphaShed a domain-informed statistical model that leverages the information the HVAC system state, in particular terminal damper positions, to predict load shed potential. The model demonstrated higher accuracy than a commonly used statistical model, while retaining model simplicity for easier adoption and targeting of existing building with the collection of a few parameters. 

The second challenge involves the need for new measurement and verification protocols for appropriate valuation and compensation of building flexibility. In particular, different factors, such as assessment boundary and baseline model, are influential in determining the discrepancy between the observed shed and real load shed. Therefore, I executed an analysis of the accuracy and bias tradeoffs between these two factors on the distortions in the observed load shed events. This was done through two complementary simulation experiments that generated a dataset of load shed events on which performance metrics were calculated to generate insights into the tradeoffs of the different performance assessment approaches. The output of this work is intended to help inform the design of future performance assessment protocols. 

The third challenge involves understanding the practical barriers to the transfer of grid-interactive technologies to the existing building population with VAV systems. To tackle this challenge, I systematically analyzed the performance of existing systems, through real building datasets, in the context of five key control loops: zone temperature, zone air, air distribution, supply air temperature, and chilled water temperature. Within the context of these control loops, the results called attention to issues, such as systematic oversizing, mismatch between design and operational conditions, the prevalence of faults, and conflicting objectives with comfort and energy efficiency. 

The contributions from this research are envisioned to help advance the development of technologies for HVAC-enabled demand flexibility in commercial buildings. More importantly, this research is a step forward in the direction of GEBs, and ultimately a green and sustainable power grid.