Grantee Research Project Results
Final Report: Sustainability Assessment Decision Optimization
EPA Grant Number: SU839277Title: Sustainability Assessment Decision Optimization
Investigators: Chen, Victoria C.P.
Institution: The University of Texas at Arlington
EPA Project Officer: Page, Angela
Phase: I
Project Period: August 1, 2017 through July 31, 2018
Project Amount: $15,000
RFA: P3 Awards: A National Student Design Competition for Sustainability Focusing on People, Prosperity and the Planet (2017) RFA Text | Recipients Lists
Research Category: Sustainable and Healthy Communities , P3 Awards , P3 Challenge Area - Sustainable and Healthy Communities
Objective:
Architectural designers and building engineers currently possess software used to analyze various performance objectives for green building. However, sustainability assessment encompasses a wide range of objectives, including energy efficiency, waste reduction, cost, social impact, and environmental impact. Moreover, no integrated approach has yet been developed that can employ the kind of flexible cross-disciplinary analysis required for sustainability assessment. The objective of the project was to close the loop on the life-cycle of building construction and operations by developing a sustainability assessment decision optimization for green building framework that employs software tools in an integrated manner.
Summary/Accomplishments (Outputs/Outcomes):
The proposed project addressed the fundamental objective of solving real world green building based decision-making problems with an integration of systems engineering, statistics, and optimization. The following project tasks were proposed: (1) Calibrate the software tools for a selected case study. (2) Develop a design of experiment process that accommodates uncertainty and a mix of many categorical and continuous variables. (3) Build multi-response treed regression metamodels to represent the relationships between computer model inputs and performance outputs, where the multi-response structure corresponds to multiple performance outputs. (4) Develop and implement a multi-objective optimization process. (5) Formulate a decision support tool which will help the user make critical decisions regarding the design variables which affect the performance of the building.
For Phase I, the selected case study was a low-rise residential building located in Atlanta, GA with a 2500 square-foot area. We employed two software tools namely, ATHENA (www.athenasmi.org/tools/impactEstimator/) for environmental impact and eQUEST (www.doe2.com/eQuest/) for energy usage. To complete task (1), a comprehensive review was conducted to match the various input factors across the two software tools. A total of 52 input factors were studied, 19 were categorical, 19 were discrete-numerical, and 14 were continuous. To complete task (2), a prior study in a Masters thesis by Marjan Sayadi was leveraged. Sayadi studied two types of experimental designs to handle the mix of discrete and continuous factors. The first experimental design used a hybrid of a mixed orthogonal array to handle the discrete factors, a low-discrepancy sequence to handle the continuous factors, and a Latin hypercube to integrate them. The second experimental design used a categorical adjustment with a low-discrepancy sequence, to enable a version that simultaneously handles continuous and discrete factors. Sayadi's comparison of the two experimental design led to the decision for the current work to utilize the second experimental design for Phase I.
The experimental design was constructed for 192 runs that seek to comprehensively explore the input factor space. Each run of the experimental design specifies settings for the input factors for one run of the software tools. In addition, for each run, an instance of uncertain input variables was sampled or set at fixed values. The uncertain variables represented room occupancy and lighting use during the operations phase of the building. For the Phase I work, we selected three output response variables of interest to represent possible performance metrics. From eQUEST, annual source energy in million British thermal unit (Mbtu) was the response output. From ATHENA, two response outputs were considered, global warming potential (GWP) and nonrenewable energy. GWP is a reference measure and is expressed on an equivalency basis related to carbon dioxide (CO2) in kilograms (kg) of CO2 equivalent. For the selected case study, 38 of the input factors were studied, including building options for the siting orientation and the foundation, wall, roof, and window systems.
To complete task 3, treed regression models were fit to study which input factors of the software tools affect the three performance metrics, annual source energy, GWP, and nonrenewable energy. Treed regression handles a mix of discrete and continuous input factors by incorporating the discrete factors in the tree and the continuous factors in the regression. However, continuous factors can also be incorporated in the tree. Hence, the treed regression was executed in two ways, first including all input factors in the tree, and second including only categorical input factors in the tree. The siting orientation input factor was not identified as affecting these three performance metrics, and ventilation was only found to affect annual source energy. Window, foundation, and wall systems had an effect on all performance metrics, while roof systems did not affect annual source energy.
To complete task 4, the empirical Pareto optimum frontier was identified using annual source energy and GWP performance metrics for the case study. The results from the 192 runs are the outputs on these two performance metrics for 192 hypothetical buildings. Among these hypothetical buildings, annual source energy ranged from 250 to 400 MBtu, and GWP ranged from 300 to 700 kg CO2. To assess performance, these two metrics were plotted for each of the 192 buildings. The resulting plot enabled a graphical illustration of the empirical Pareto optimum frontier. Points on the empirical Pareto optimum frontier are those buildings with design options that best balance the minimization of both performance metrics. Five Pareto points out of the 192 runs were identified. These five building design options are those that architects and building engineers could consider for further analysis. An examination of the specific building options selected for the five buildings on the Pareto optimum frontier identified a variety of building options, with only three of the 38 input factors yielding the exact same options across all five buildings. Specifically, these three input factors were a footprint shape of rectangle (vs. a square), an external wall finish of concrete, and an external wall color that is light. These results illustrate the complexity of optimizing the selection of building options. There is complexity in how these options work together to achieve desired performance, and there are many different combinations of these options that can yield similarly desirable performance.
Finally, to complete task (5), the findings from the above four tasks were employed to formulate a decision support tool. The framework and process for this decision-support tool were constructed, and the complete development of this tool were proposed for Phase II of this research. Specifically, Phase II proposed to develop a web-based tool that would be freely available to the public.
Conclusions:
From this Phase I study, the proposed process was tested to demonstrate its potential for complex decision optimization problems, such as the green building design process. In particular, complexity of the problem is characterized by multiple performance objectives to represent sustainability and many input factors of various types (categorical, discrete-numerical, and continuous). Our results demonstrated the ability to identify the set of empirical Pareto optimum points, corresponding to building design input settings for the case study. The ability to identify these Pareto points will facilitate a sustainability assessment decision optimization process for the end user. The approach can be generalized to address any sustainability assessment for which a variety of software tools exist to measure performance objectives of interest.
This research supported travel expenses for students to present their research at three conferences. The first conference was the Annual Meeting of the Institute for Operations Research and the Management Sciences (INFORMS), held in October 2017 in Houston, Texas. The second conference was the 2018 USA Science & Engineering Festival, where the student team participated in the EPA-P3 Exhibition. Young attendees enjoyed building structures out of toothpicks and colored mini-marshmallows, and the accompanying adult attendees enjoyed learning about green building. The third conference was the INFORMS Conference on Business Analytics & Operations Research, held in April 2018 in Baltimore, Maryland. Finally, a journal paper is currently in preparation. A listing of the presented work is given below.
Journal Articles:
No journal articles submitted with this report: View all 7 publications for this projectSupplemental Keywords:
green building, energy conservation, sustainable building, design of experiments, multiple objectives, optimizationThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.