Conference Agenda

Heat pump water heaters (HPWHs) could significantly reduce energy costs and greenhouse gas emissions from water heating, the second largest energy use in residential buildings. Today, most HPWHs use electric resistance heating elements to maintain comfortable water temperatures even during large water draws. Unfortunately, heating elements significantly decrease energy efficiency, and their current and voltage requirements may necessitate costly electrical work in older homes. This paper develops and field-tests a model predictive control (MPC) system that enables a HPWH with no heating elements to maintain comfort at high efficiency. By contrast to most prior experimental studies on water heater MPC, which often use perfectly-forecasted water draws in controlled laboratory settings, this paper reports field tests from a real home with three full-time occupants. The occupants' water draws are forecasted using a machine learning model and a scalable training methodology. This paper also presents occupant feedback on thermal comfort, as well as an Internet of Things infrastructure that enables real-time data acquisition and control. In the MPC formulation, the energy savings were 11% with the same thermal comfort as the manufacturer's constant set-point control. An adjusted MPC formulation substantially improved thermal comfort while modestly increasing energy costs.

With the increasing adoption of renewable energy in the power grid, the future of building energy systems is transitioning toward a distributed, multi-source energy framework. Notably, heating and cooling needs account for approximately 30% of a building's energy consumption. To realize decarbonization, This paper proposed a residential solar-air source hybrid heat pumps water heating system. And optimize the use of energy storage and the thermal mass of building structures to significantly alleviate strain on the power grid and enhance the integration of renewable energy sources. we apply adaptive model predictive control to a multi-heat source building heat pump system with solar and air energy, in conjunction with heat pumps for heating. Additionally, an online updated modeling method, based on the simplification of the physical mechanism model and the incorporation of real-time data updates, enhances the accuracy and adaptability of the model predictive control, thereby improving optimization outcomes Our primary objective is to achieve energy efficiency and demand response through the intelligent orchestration of these heat sources and effective heat storage. Through extensive testing under typical daily climatic conditions in Beijing and real time price (RTP), our model predictive controller consistently demonstrates 17.8% energy saving and 31.9% cost savings when compared to conventional rule-based control strategies.

High performance HVAC equipment such as cold climate heat pumps is crucial to the nation’s building electrification goals. The operation of HVAC systems is dynamic, subject to ambient conditions, controls and occupants’ interactions via thermostats. Moreover, a wider adoption of heat pumps relies on customers’ satisfaction on both energy efficiencies and thermal comfort. We developed a Modelica-based modeling framework that can simulate the dynamic operation and occupant’s thermal comfort of residential HVAC equipment, including air conditioners, furnace, heat pumps, fans and thermostats, in realistic house environment. The framework leverages multiple latest development from DOE Building Energy Modeling Program including Modelica Buildings Library, Spawn of EnergyPlus, ResStock and BOPTEST, from which high-fidelity and computationally efficient component models are applied in the framework to be used for evaluating various smart thermostat algorithms. These can include, for example, basic functions such as occupant-sensing based and/or schedule-based setpoint setbacks.

We utilized the framework to conduct annual energy consumptions and thermal comfort simulations for over 250 houses across the United States and its climate zones. Our preliminary analysis shows different thermostat setback strategies may lead to energy savings range from 2% to more than 20%. In this paper, we focus on demonstrating the details of various HVAC equipment models and their operating modes such as controls of auxiliary heaters, frost and defrost operations and speed controls of double-speed and variable speed heat pumps. We believe the framework could be a useful virtual testbed for studying equipment impacts to building electrification.

Ground-source integrated heat pumps (GSIHPs) represent a promising heating, ventilation, and air-conditioning (HVAC) technology to effectively save an energy in residential buildings. The GSIHPs can significantly reduce energy consumption with efficient variable-speed components and water heating, providing four operating modes of space heating, space cooling, domestic hot water (DHW), and combined space cooling and DHW. We conducted comprehensive laboratory tests to assess the performance of a GSIHP, serving as preliminary evaluations for its anticipated operation in a residential net-zero energy building (NZEB). The laboratory tests include ASHRAE 206-2013, ASHRAE 118-2, and DOE FR 10 CFR Part 430. These tests informed the development of a performance map, considering variables such as air temperature, entering liquid temperature, flowrate, capacity, and power. The updated map was then incorporated into the predeveloped TRNSYS model. We conducted a TRNSYS simulation to explore the GSIHP performance under different climates for NZEBs. The performance improvement of the GSIHP compared to the conventional GSHP was evaluated across various climates. The ground heat exchanger (GHX) length was determined based on the building loads at each climate zone using the Kavanaugh and Rafferty (K&R) method . The size of photovoltaic (PV) panels affected by the solar radiation was determined to achieve net-zero energy. Additionally, the effect of GHX design was analyzed in terms of short-term (1 year) and long-term (10 years) energy performance. An economic analysis was conducted to ascertain optimal GHX lengths to minimize installation costs.

Building automatic fault detection and diagnosis (AFDD) technologies have shown great potential for energy savings. Literature on building AFDD research mainly focuses on traditional data available from building automated systems (BAS) or one-time measurements. In this research, we investigate the capability of acoustic emission (AE), a non-traditional data source, to support AFDD in real building heating, ventilation and air-conditioning (HVAC) systems. Experiments were conducted to generate four different AE datasets under different operational scenarios for HVAC systems, where faults were manually injected. The first dataset consists of acoustic data collected from acoustic sensors placed at two different positions (inside/outside) of the same air-cooled chiller under abnormal and normal operations; the second dataset includes acoustic data collected from two identical air-conditioner (AC) outdoor condenser units under abnormal and normal operations; the third one contains acoustic data collected from multiple air diffusers in an experimental residential home under abnormal and normal operations; and the fourth dataset is acoustic data collected under various severity levels of fault conditions occurring in a condenser unit for different time periods. Short-time Fourier Transform (STFT) is used to transform the time series to time-frequency spectrogram, and two different approaches, standard machine learning (ML) and end-to-end deep learning (DL), are used as AFDD strategies to validate the efficacy of AE for the fault detection. For the ML approach, averaged frequency at each time is derived as features fed into random forest classifier; for the DL approach, spectrograms are directly fed into multilayer perceptron. 5-fold CV is repeated 10 times to reduce randomness and avoid overfitting. Experimental results show that AFDD using acoustic data by both the ML and the DL present satisfactory detection performances. For random forest classifier, the averaged fault detection rates are 0.925, 1.0, 1.0 and 0.88 for the four datasets respectively. For multilayer perceptron model, the averaged fault detection rates are 0.97, 1.0, 1.0 and 0.879 respectively. We conclude the use of AE has great potential to support AFDD in the building systems.

The rapid electrification of residential buildings has created rising concerns about the ability of the power grid to deal with seasonal electric space conditioning peaks. Better control of residential building heating and cooling systems can help mitigate this problem. However, unlike other control areas where extensive data are often available, each residential building presents a unique system for which data is very limited and distributed sensing is rarely implemented. This complicates the development of advanced controllers. This paper combines machine learning with prior information about building physics to greatly reduce the amount of data required to learn a control-oriented thermal model of residential buildings. Using field data from an all-electric residential building in a cold climate, this work investigates different model orders and compares several machine learning techniques for disturbance predictions. The combination of a simple thermal circuit model with a support vector machine performs best among the investigated candidates. Given only four weeks of training data from comparatively low-cost sensors, this combination can predict indoor temperatures with a root-mean-square-error of 0.4 ℃ and thermal loads 2.3 kW, which is comparable to that of a higher model requiring a Kalman filter to monitor not observed states.

Heating, ventilation and air conditioning (HVAC) systems, as key components in commercial building air conditioning systems, significantly impact indoor environmental quality and building energy efficiency. This paper focuses on developing a physics-based dynamic Bayesian network (PBDBN) that integrates a dynamic scheme for detecting and diagnosing faults in building HVAC systems. Contrary to traditional data-driven BN methods that infer structure from statistical processes, our approach constructs the BN structure grounded on physical equations, with coefficients determined through data-driven processes. Moreover, unlike reference model-based BN methods that keep the building model separate from the BN structure, our approach significantly simplifies and incorporates the building model directly into the BN structure. A detailed structure construction of the proposed PBDBN is demonstrated in this paper. The model is evaluated by injecting minor sensor faults into the HVAC system. The results show that the proposed PBDBN-FDD has significant improvements compared to the existing DBN-FDD methods.