No abstract

Identification approaches applied to semi-physical thermal network structures, so called gray-box modeling approaches, are popular in building science for energy audits, retrofit analysis and advanced building controls, e.g. model predictive control. However conventional identification approaches applied to thermal networks fail when there are significant unmeasured heat gains that influence building responses. This paper presents a method to obtain improved gray-box building models from closed loop data having significant unmeasured disturbances. The method estimates both physical parameters of a building thermal network model and also a disturbance model that characterizes the unmeasured inputs. The performance of the algorithm is demonstrated using numerical and experimental results.

8Penetration of advanced building control techniques into the market has been slow since buildings are unique and site-9 specific controller design is costly. In addition, for medium-to large-sized commercial buildings, HVAC system 10 configurations can be very complex making centralized control infeasible. This paper presents a general multi-agent control 11 methodology that can be applied to building energy system optimization in a "plug-and-play" manner. A multi-agent 12 framework is developed to automate the controller design process and reduce the building-specific engineering efforts. To 13 support distributed decision making, two alternative consensus-based distributed optimization algorithms are adapted and 14implemented within the framework. The overall multi-agent control approach was tested in simulation with two case studies: 15 optimization of a chilled water cooling plant and optimal control of a direct-expansion (DX) air-conditioning system serving 16 a multi-zone building. In both cases, the multi-agent controller was able to find near-optimal solutions and significant energy 17 savings were achieved. 18Keyword: Multi-agent control; Building energy system optimization; Distributed optimization; HVAC component 19 coordination 20 INTRODUCTION 21More than 40% of the primary energy usage in the United States is related to energy consumption in buildings [1] and if 22 buildings are not operated properly, a significant amount of energy is wasted. The energy savings opportunities for optimal 23 building controls are becoming widely recognized leading to growing research efforts in the past few years. However, the 24 deployment of advanced controls in buildings has been progressing very slowly due to several reasons: (1) buildings are 25 unique in terms of both building construction and heating, ventilation and air-conditioning (HVAC) system configuration, 26 which makes building-specific controller design costly; (2) optimal control of complex building energy systems is difficult 27

 We show a unique approach for simulating the dynamics of indoor air and envelopes.  The coupled approach enables a fast simulation of open space buildings.  The model is particularly appropriate for control analysis of open space buildings.  The model was validated with experimental data.  A retrofit analysis w.r.t. the change of thermostat locations is shown. AbstractSmall-to-moderate sized commercial buildings commonly use rooftop units (RTUs) to provide indoor comfort. These applications are often characterized by significant spatial variations in comfort due to poor thermostat placement and poor coordination of RTUs leading to high energy and demand costs and marginal comfort. Tools are needed that can assess both energy and comfort performance for these types of applications so that the benefits of improved RTU coordination and control can be evaluated in terms of energy and comfort and so that appropriate real-time strategies can be developed. In this paper we discuss a method for generating and coupling tractable reduced order models for the building envelope and for the indoor air dynamics. Overall assessments of the coupled model are performed for a typical sit down restaurant that employs four rooftop units. Measurements are available from the site to validate the models.