Toward Making Ventilation Decisions Based on Expected Outcomes: A Flexible Multi-Criteria Framework

Next-generation ventilation should explicitly aim to aggregate, compare, weight, and ultimately integrate ventilation’s numerous impacts, which include the dilution of multiple indoor pollutants, introduction of multiple outdoor pollutants, direct correlations to productivity and absenteeism, and strong influences on energy consumption, peak power demand, and their associated environmental impacts. We propose a multi-criteria decision-making framework based on (i) a comprehensive, scientific accounting of the costs and benefits of ventilation, and (ii) a utility maximization formulation to allow both expression of user preferences and adjustment for uncertainty. The utility function (UF) provides the core criterion and interface to enable more intelligent and holistic ventilation, allowing leveraging of new and future capabilities, including optimized and predictive methods that make use of models or learning, dynamic control that capitalizes on transient effects, and real-time information from new sensors or interconnected information networks and energy grids. We first formulate the UF to be as expansive as possible in scope of impacts, and then conduct a series of sensitivity analyses (SA) to help refine it and identify building parameters that impact it. For the SA data, we made use of an existing simulation-based analysis of eight off-the-shelf ventilation and related strategies that combined economizing, demand-controlled ventilation, and supply air temperature reset. Outcomes were reprocessed to a daily-average basis, and included: electricity consumption, natural gas consumption, median ventilation rates, relative symptom prevalence, relative work performance, and concentrations of carbon dioxide, total volatile organic compounds, and fine particulate matter and ozone of outdoor origin. We first ask, which outcomes are ultimately worth including in the UF? We employ: a principal component analysis of outcomes to check if any are redundant; SA of the impact of each individual outcome on total utility over a range of reasonable user preferences; and an analysis of which outcomes can themselves be meaningfully affected by ventilation. We next turn to an initial evaluation of which building parameters will need to be taken into account by a dynamic strategy that is attempting to maximize utility. We conduct a separate SA where the dependent variable is the daily average ventilation rate observed under each off-the-shelf dynamic strategy, and the independent variables are ~20 varied building parameters as well as 10 explanatory covariates. This task is a crucial first step in simplifying the challenge of generalizing optimal ventilation beyond the case-study level, in the face of tremendous variability among real building and setting characteristics.