Model free based DRL control strategies have achieved positive effects on the HVAC system optimal control. However, developing deep reinforcement learning (DRL) control strategies for different ...building HVAC systems is time-consuming and laborious. To address this issue, this study proposes a transfer learning and deep reinforcement learning (TL-DRL) integrated framework to achieve the DRL control strategy transfer in the building HVAC system level. Deep Q-learning (DQN) is first pre-trained in the source building until it converges to an optimal strategy. Then, the well pre-trained DQN parameters of the first few layers are transferred to the target DQN. Finally, the target DQN parameters of the last few layers are fine-tuned in the target building. An EnergyPlus-Python co-simulation testbed is developed to investigate the cross temporal-spatial transferability of DQN control strategy in the building HVAC system level. Results indicate that the proposed TL-DRL framework can effectively improve the training efficiency of control strategy by about 13.28% when transferring the first two layers compared to that of the DRL baseline models trained from scratch, while simultaneously maintaining energy consumption and indoor air temperature in an acceptable range. The proposed TL-DRL framework provides a preliminary direction for the scalability of intelligent HVAC control strategies.
•TL-DRL integrated framework for cross temporal-spatial transfer of HVAC control.•TL-DRL model performance is sensitive to the number of deep neural network layers.•TL-DRL model can improve training efficiency of control strategy by about 13.28%.•Control strategy transfer cross the same climate zone shows better performance.
In this paper, a comprehensive review of the artificial neural network (ANN) based model predictive control (MPC) system design is carried out followed by a case study in which ANN models of a ...residential house located in Ontario, Canada are developed and calibrated with the data measured from site. A new algorithm called best network after multiple iterations (BNMI) is introduced to help in determining the appropriate ANN architecture. The prediction performance of the developed models using BNMI algorithm was significantly better (between 6% and 59% better goodness of fit for various models) when compared to a previous study carried out by the authors which used the default single iteration ANN training algorithm of MATLAB®. The ANN models were further used to design the supervisory MPC for the residential HVAC system. The MPC generated the dynamic temperature set-point profiles of the zone air and buffer tank water which resulted in the operating cost reduction of the equipment without violating the thermal comfort constraints. When compared to the fixed set-point (FSP), MPC was able to save operating cost between 6% and 73% depending on the season.
Interval prediction is a promising method that can reveal the uncertainty of building load and has been shown to effectively manage building energy systems. Previous studies focused on improving the ...performance of three types of generic single models (quantile regression, bootstrap, and lower upper bound estimation models). Multi-model ensemble methods have the advantages of reducing overfitting risk, improving stability. However, in the field of building cooling load interval prediction, there is a research gap that the performance of the multi-model ensemble method has not been conducted extensively. To compensate for this gap, firstly, referring to the information entropy in information theory, we proposed deviation entropy, which is a new concept that can be used to calculate the weight of a generic single model. Based on this, a static multi-model ensemble interval prediction method was developed. Then, we introduced a sliding time window to further upgrade the static method into a dynamic method. Finally, we used actual data to evaluate the performance of the proposed dynamic multi-model ensemble interval prediction method. The study results show that the multi-model ensemble method outperforms traditional generic single model, and the dynamic method can improve the prediction reliability by 8.5% with only 0.76% loss of prediction accuracy.
•Reviewing limitations of previous data-mining based FDD methods on HVAC systems.•Knowledge-based and model-based systems have limitations in complicated FDD process.•In FDD in large-scale HVAC ...systems, driven methods are superior to other methods.•Hybrid methods have greatpotential in finding naïve faults in FDD process.•Developed hybrid methods can overcome delays, and get less false and missed alarms.
Abnormal operation of HVAC systems can result in an increase in energy usage as well as poor indoor air quality, thermal discomfort, and low productivity. Building automated systems (BAS) collects a massive amount of data related to the operation of each component of HVAC systems. Although BAS has been implemented in many buildings over the past decade, the collected data have not been analyzed thoroughly. Some studies have relied on data-mining methods to predict, detect, and diagnose faults in HVAC systems. This paper critically reviews the existing literature and identifies the research gaps in data-driven data mining fault detection and diagnosis (FDD) methods studies on HVAC systems. In this review, data-driven based FDD methods are classified into three classes, namely supervised, unsupervised, and hybrid-learning methods. The hybrid approaches are introduced as the preferred methods among the existing approaches to be used in online FDD processes. Furthermore, some components of HVAC systems and their potential faults are discussed in detail. The outcome of this review shows that data-driven based approaches are more promising for the FDD process of large-scale HVAC systems than model-based and knowledge-based ones. Moreover, an optimal approach could involve both supervised and unsupervised learning (hybrid methods).
Predictive control offers significant advantages in nonlinear control, high thermal inertia, and dynamic control. This article uses a Systematic Reviews and Meta-Analyses methodology to review 245 ...studies on predictive control in HVAC systems over the past 12 years, focusing on Model Predictive Control (MPC) and Model-Free Predictive Control (MFPC). In cooling systems, MPC is widely applied to energy efficiency management, continuous operation and maintenance, and overall system optimization in multi-zone residences. Its advantage is its ability to respond to system dynamics and precisely control key components such as cooling towers, condensers, evaporators, and pumps. Research focuses on simplifying models, reducing computational complexity, and enhancing real-time performance. In contrast, MFPC saves energy in equipment components and overall operation through intelligent valves, agent control programs, and other methods. Research focuses on developing new reinforcement learning algorithms to improve control efficiency and reliability. MPC research in heating systems focuses on hydraulic and thermal balance in central heating systems and expands to managing renewable energy hybrid systems. The research aims to dynamically adjust to meet user thermal comfort requirements while reducing energy consumption and improving efficiency. Key technologies include modeling techniques, distributed MPC, cross-regional integrated control, and efficient renewable energy integration strategies. MFPC precisely controls heating system water supply temperature, heat pump energy efficiency, and heating terminals through model-free algorithms like deep reinforcement learning and multivariable extremum seeking control. In integrated HVAC systems, MPC research focuses on managing multi-energy systems through hierarchical decomposition and multi-layer strategies, seamless renewable energy integration and optimization, and developing multi-objective optimization and decision support tools. MFPC research includes automatic grading strategies for integrated controllers, online optimization balancing methods, multi-agent methods, and developing intelligent model-free adaptive control strategies. However, MFPC integration in practical applications still needs strengthening. This review guides researchers in selecting the best predictive control mode for various HVAC system applications.
Display omitted
•Model-free and Model-based have their own characteristics and applications.•The accuracy of the model-free predictive control must be further improved.•The application of the MPC Method has been studied extensively.•The nonlinear method still needs to be explored and improved.•Researchers should select the most appropriate predictive control mode.
•Energy flexibility in demand response building is experimentally investigated.•Passive building thermal mass and active energy storage systems are coupled.•Pre-cooling and temperature reset are ...considered to assess energy flexibility.•Short-term (0.5 h) and intermediate-term (2 h) demand response are achieved.
Heating, ventilation, and air conditioning (HVAC) systems, combined with the internal thermal mass of buildings, have been deemed to be promising means of providing demand response (DR) resources, particularly for buildings with active energy storage systems. DR resources, such as peak-load reduction potential, can provide grid-responsive support resulting in a high degree of grid involvement and high flexible electricity demand. In the DR field, the potential of HVAC load flexibility has been considered in buildings. In the future smart buildings, it is important to take advantage of demand-side resources to achieve real-time energy supply–demand balance sustainably. In this context, DR potential and characteristics of buildings play a pivotal role in DR programs. However, few studies have investigated the internal thermal mass’s heat release and DR characteristics of buildings. Thus, a systematic experiment is conducted to study the DR potential and characteristics of internal thermal mass and active storage systems. The DR resources include the passive cooling storage from furniture, building envelope and an active water storage tank. Two DR control strategies, including pre-cooling and temperature resetting, are analyzed in this study. The experimental results show that the strategies are effective for short-term (0.5 h) and intermediate-term (2 h) DR programs. For a long-term DR program, active energy storage technology such as a water storage tank is required to satisfy the occupant's comfort requirements. Hence, we conclude that passive thermal mass and active storage systems should be simultaneously considered in practical DR programs for better DR implementation.
•Entropy-based causality learning framework is proposed for building fault diagnosis.•The proposed synchronicity concept is employed to measure building symptom correlations.•Causal learning aims to ...determine Bayesian network structure.•Modelica test cases demonstrate the framework’s effectiveness.
Faults, such as malfunctioning sensors, equipment, and control systems, significantly affect a building’s performance. Automatic fault detection and diagnosis (AFDD) tools have shown great potential in improving building performances, including both energy efficiency and indoor environment quality. Since modern buildings have integrated systems where multiple subsystems and equipment are coupled, many faults in a building are cross-level faults, i.e., faults occurring in one component that trigger operational abnormalities in other subsystems. Compared with non-cross-level faults, it is more challenging to isolate the root cause of a cross-level faults due to the system coupling effects. Bayesian networks (BNs) have been studied for the root cause isolation for building faults. While promising, existing BN-based diagnosis methods highly rely on expert domain knowledge, which is time-consuming and labor expensive, especially for cross-level faults. To address this challenge, we propose an entropy-based causality learning framework, termed Eigen-Entropy Causal Learning (EECL), to learn BN structures. The proposed method is data-driven without the use of expert domain knowledge; it utilizes causal inference to determine the causal mechanisms between faults status and symptoms to construct the BN model. To demonstrate the effectiveness of the proposed framework, three fault test cases are used for evaluation in this study. Experimental results show that the BN constructed by the proposed framework is able to conduct building cross-level faults diagnosis with a comparable isolation accuracy to those by domain knowledge while maintaining less complexed BN structure.
Effective fault diagnosis of compressors in heating, ventilation, and air conditioning (HVAC) systems is critical to ensure service reliability and boost energy efficiency. HVAC compressors are ...distributed in different areas and work under heterogeneous conditions, which poses emerging challenges to their data-driven modeling. Most existing methods assume multiple category-balanced source domains for model training. Although domain adaptation and generalization methods have emerged to address the data distribution discrepancies in cross-domain fault diagnosis, limited source domains and imbalanced fault categories across domains still constrain the real-world applicability of data-driven models in HVAC compressor fault diagnosis under unseen working conditions. Therefore, this paper studies a significant fault diagnosis problem named single imbalanced domain generalization (SIDG) and proposes a corresponding network (SIDGNet) for intelligent HVAC compressor fault diagnosis. Specifically, a rare fault diagnosis module combined with focal loss is introduced to tackle the class imbalance problem. To achieve better diagnostic boundaries and resist unknown data distribution discrepancies, joint supervised contrastive learning and adversarial learning with specialized data augmentation are introduced as auxiliary modules to improve the robustness and generalizability of SIDGNet. An improved uncertainty-based dynamic weighting mechanism is developed to intelligently balance the weights of module-specific losses during training, which ensures an efficient and stable optimization process. Extensive SIDG fault diagnosis experiments conducted on HVAC compressors demonstrate the superiority of SIDGNet over existing models in SIDG scenarios.
In electric buses, the heating, ventilation and air conditioning (HVAC) system is responsible for up to 50% of the energy consumption, thereby significantly affecting the efficiency. It is therefore ...necessary to identify improved thermal settings for the bus cabin to minimize the energy consumption, while guaranteeing good thermal comfort. To achieve this goal, this paper presents the results of climatic measurements in an electric bus in Berlin, Germany. These measurements were performed for outer temperatures between 5.3 °C and 7.8 °C and four cabin temperature settings. During the measurements, several climatic parameters and the energy consumption were measured, whereas the thermal comfort (TC) was evaluated via 71 passengers’ surveys. The results show that the climatic conditions in the bus vary greatly depending on the position (up to 3 K difference in mean air temperature) and height (up to 8 K/m temperature-to-height ratio). Additionally, the surveys show that the mean value of the thermal comfort parameter TC is minimized to a value of 0.15 (corresponding to “comfortable” thermal perception) for a set temperature of 21 °C, whereas the thermal conditions are perceived as acceptable even with heating off.
•Analysis of thermal comfort, energy and climatic measurements in an electric bus.•Thermal conditions strongly depend on position and height of temperature sensors.•71 passengers' surveys were collected.•Results show that a set temperature of 21 °C equals “comfortable”.•76.7% of the passengers are comfortable even if the heating is turned off.
•A review taxonomy for human-in-the-loop HVAC operations has been proposed.•Human-in-the-loop HVAC operations have been reviewed using the proposed taxonomy.•Methods for occupancy and comfort ...characterization were systematically reviewed.•Methods for integration of human dynamics in the control of HAVC were reviewed.•Presented quantitative and qualitative performance assessment on different methods.
Heating, ventilation, and air-conditioning (HVAC) systems account for almost half of the energy consumption in buildings. By benefiting from advancements in information and communication technology, human-in-the-loop HVAC operations have drawn considerable attention in the last decade with the aim of curtailing unnecessary energy use and providing user-specific comfort zones with reduced user dedication. Future progress in thie field calls for an in-depth understanding of the current state and challenges of the human-in-the-loop HVAC systems. Therefore, using a structured literature review approach, we have investigated this field according to two parameters of human dynamics that drive user-centric operations of HVAC systems, namely, occupancy and comfort. In this review and assessment study, by proposing a five-tier hierarchical taxonomy, we have classified the studies based on their contributions to occupancy- and comfort-driven human-in-the-loop HVAC operations (e.g., occupancy detection or comfort profiling) and have presented categorization for techniques and their quantitative performance assessment. In doing so, we have accounted for the context of the studies as they relate to developments in residential and office buildings given that distinct circumstances in each context (e.g., accessibility to thermostats) have resulted in different methodologies, especially in adopting the sensing techniques and HVAC operations. Moreover, we have distinguished simulations from field evaluations to assess the actual viability and challenges in achieving desirable results in practice. Lastly, the Hype cycle model was utilized to qualitatively evaluate the developments of different technologies for human-in-the-loop HVAC operations from a research perspective.