Relative Significance of Heat Transfer Processes to Quantify Tradeoffs Between Complexity and Accuracy of Energy Simulations With a Building Energy Use Patterns Classification

Open Access
Author:
Heidarinejad, Mohammad
Graduate Program:
Mechanical Engineering
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
January 24, 2014
Committee Members:
  • Dr Jelena Srebric, Dissertation Advisor
  • Dr Jelena Srebric, Committee Chair
  • Anil Kamalakant Kulkarni, Committee Member
  • Hosam Kadry Fathy, Committee Member
  • Seth Adam Blumsack, Committee Member
Keywords:
  • Building energy simulation
  • building energy use patterns classification
  • campus buildings
  • LEED certified buildings
  • OpenStudio
Abstract:
This dissertation develops rapid and accurate building energy simulations based on a building classification that identifies and focuses modeling efforts on most significant heat transfer processes. The building classification identifies energy use patterns and their contributing parameters for a portfolio of buildings. The dissertation hypothesis is “Building classification can provide minimal required inputs for rapid and accurate energy simulations for a large number of buildings”. The critical literature review indicated there is lack of studies to (1) Consider synoptic point of view rather than the case study approach, (2) Analyze influence of different granularities of energy use, (3) Identify key variables based on the heat transfer processes, and (4) Automate the procedure to quantify model complexity with accuracy. Therefore, three dissertation objectives are designed to test out the dissertation hypothesis: (1) Develop different classes of buildings based on their energy use patterns, (2) Develop different building energy simulation approaches for the identified classes of buildings to quantify tradeoffs between model accuracy and complexity, (3) Demonstrate building simulation approaches for case studies. Penn State’s and Harvard’s campus buildings as well as high performance LEED NC office buildings are test beds for this study to develop different classes of buildings. The campus buildings include detailed chilled water, electricity, and steam data, enabling to classify buildings into externally-load, internally-load, or mixed-load dominated. The energy use of the internally-load buildings is primarily a function of the internal loads and their schedules. Externally-load dominated buildings tend to have an energy use pattern that is a function of building construction materials and outdoor weather conditions. However, most of the commercial medium-sized office buildings have a mixed-load pattern, meaning the HVAC system and operation schedule dictate the indoor condition regardless of the contribution of internal and external loads. To deploy the methodology to another portfolio of buildings, simulated LEED NC office buildings are selected. The advantage of this approach is to isolate energy performance due to inherent building characteristics and location, rather than operational and maintenance factors that can contribute to significant variation in building energy use. A framework for detailed building energy databases with annual energy end-uses is developed to select variables and omit outliers. The results show that the high performance office buildings are internally-load dominated with existence of three different clusters of low-intensity, medium-intensity, and high-intensity energy use pattern for the reviewed office buildings. Low-intensity cluster buildings benefit from small building area, while the medium- and high-intensity clusters have a similar range of floor areas and different energy use intensities. Half of the energy use in the low-intensity buildings is associated with the internal loads, such as lighting and plug loads, indicating that there are opportunities to save energy by using lighting or plug load management systems. A comparison between the frameworks developed for the campus buildings and LEED NC office buildings indicates these two frameworks are complementary to each other. Availability of the information has yielded to two different procedures, suggesting future studies for a portfolio of buildings such as city benchmarking and disclosure ordinance should collect and disclose minimal required inputs suggested by this study with the minimum level of monthly energy consumption granularity. This dissertation developed automated methods using the OpenStudio API (Application Programing Interface) to create energy models based on the building class. ASHRAE Guideline 14 defines well-accepted criteria to measure accuracy of energy simulations; however, there is no well-accepted methodology to quantify the model complexity without the influence of the energy modeler judgment about the model complexity. This study developed a novel method using two weighting factors, including weighting factors based on (1) computational time and (2) easiness of on-site data collection, to measure complexity of the energy models. Therefore, this dissertation enables measurement of both model complexity and accuracy as well as assessment of the inherent tradeoffs between energy simulation model complexity and accuracy. The results of this methodology suggest for most of the internal load contributors such as operation schedules the on-site data collection adds more complexity to the model compared to the computational time. The third objective deployed the developed building classification and energy simulation approaches to two well-instrumented case studies. In the first case study, without the use of on-site data except the building energy consumption, the developed methods successfully predict the natural gas consumption, while the electricity consumption requires additional inputs beyond the building energy consumption. The second case study exhibited an opposite pattern of outcomes, meaning the developed methods successfully predict the electricity consumption, while the natural gas consumption requires additional inputs. For the first case study with additional site visit data, and for the second case study with using monthly building indoor temperature readings, the developed methods provide accurate predictions for both natural gas and electricity consumptions that have met the ASHRAE Guideline 14 requirements, to have CV below 15%. It is important to note this guideline provides accuracy measurement criteria for a well-calibrated model for which an energy modeler typically performs multiple site visits and reviews detailed building documentation to obtain inputs for the building model. In addition, with the exclusion of three outlier months from the analyses, the results without any additional inputs have met the accuracy requirement. The conducted energy simulations for the two case studies revealed that there are key variables such as outdoor air fraction, infiltration rate, and monthly HVAC setpoints especially for the shoulder months (April-May and October-November) that are not included in the building energy database due to the difficulty of on-site measurements. Use of these variables in the building classification and modeling can increase the accuracy of energy simulation to the required level of acceptance. Overall, this study provided specific data on tradeoffs between accuracy and model complexity that points to critical inputs for different building classes, rather than an increase in the volume and detail of model inputs as the current research and consulting practice indicates.