High Performance Computing for Energy Efficient Buildings Imran Akhtar ∗ Interdisciplinary Center for Applied Mathematics, MC0519 Virginia Tech, Blacksburg, VA24061, USA akhtar@vt.edu Jeff Borggaard † Interdisciplinary Center for Applied Mathematics, MC0519 Virginia Tech, Blacksburg, VA24061, USA jborggaard@vt.edu John A. Burns ‡ Interdisciplinary Center for Applied Mathematics, MC0519 Virginia Tech, Blacksburg, VA24061, USA jaburns@vt.edu ABSTRACT Commercial buildings are the largest single consumer of en- ergy in the United States. Energy eﬃcient buildings will have a signiﬁcant impact on overall energy consumption and greenhouse gas emissions. Buildings being multi-scale, multi-physics, highly uncertain dynamic systems, its energy eﬃciency is directly linked with the design and control of various systems in buildings. Achieving substantial levels of energy savings over the life-time of a building require not only the state-of-the-art hardware technology but also a thorough computational framework which includes math- ematical algorithms, computational science methodologies and computer tools targeted for rapid analysis, optimiza- tion and control. Direct application of high ﬁdelity simula- tion models to problems of optimal design and control is not feasible. Thus, reduced-order models are often developed for an eﬃcient design and control. In this study, we present the application of high performance computing tools to perform high ﬁdelity ﬂow simulations in a typical room which serves as a basic unit in a building. Using a large data set of the ﬂow and temperature ﬁeld distributed among various proces- sors, we compute optimal basis functions in parallel. These basis functions are used in developing reduced-order models of complex systems for control and optimization purposes. Keywords High performance computing, reduced-order modeling, proper orthogonal decomposition 1. INTRODUCTION With ever increasing power and speed of computational tech- nology, high performance computing (HPC) is becoming ∗ Research Faculty † Professor ‡ Hatcher Professor in the Department of Mathematics Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, re- quires prior specific permission and/or a fee. FIT ’10, 21-DEC-2010, Islamabad, Pakistan Copyright c 2010 ACM 978-1-4503-0342-2/10/12 ... $10.00 more important in science and engineering applications. In this paper we apply HPC to the problem of simulating and design of high performance energy eﬃcient buildings. Com- mercial buildings contribute a major portion of the energy consumption and greenhouse gas emissions in the U.S. and worldwide. Buildings systems are a network of heteroge- neous components that interact nonlinearly with wide range of perturbations introduced at all levels. Thus, whole build- ing simulation is a signiﬁcant computational challenge. Eﬃ- cient implementation of design, control and optimization re- quire multiple simulations which often overwhelms the avail- able resources. These complicated and complex systems are often simpliﬁed by ﬁrst developing a reduced-order model and then applying optimal design and control strategies for eﬃcient performance. Building systems in the commercial sector are not typically designed to conserve energy. Moreover, there is no single computational tool that integrates multi-level systems and produces an optimum solution. In order to ensure guaran- teed levels of energy savings over the life-time of a building, the proposed computational framework will include math- ematical algorithms, computational science methodologies and computer tools targeted for rapid analysis, optimiza- tion and control. The main complexity in solving and an- alyzing these physics-based models arise due to inﬁnite de- grees of freedom or states in the system, since most of them are governed by partial-diﬀerential equations. The modeling must accommodate all levels of ﬁdelity ranging from physics based continuum models to reduced order models. Existing methods and computer tools are typically designed to run on single processor platforms and do not take advantage of the natural parallelism that occurs in building systems. As such, existing methods must make gross model simpliﬁca- tions that prohibit the high performance designs that can be achieved with high ﬁdelity models. The new algorithms for high ﬁdelity modeling will use high performance parallel computing for simulation, parallel optimization and sensitiv- ity analysis. These tools will enable the rapid and accurate evaluation of energy ﬂows and provide a methodology to im- prove eﬃciency through integrated optimization-based tools [6, 7]. The ultimate goal is the development of high performance computing (HPC) tools for high performance energy eﬃ- cient buildings to meet the challenges in energy sector. In the United States, the potential payoﬀ of this eﬀort is huge.