Variable Refrigerant Flow System Research Articles

Evaluation of Multi-Functional Variable Refrigerant Flow System With Thermal Energy Storage and Photovoltaic-Based Distributed System for Net-Zero Energy Home Design

Abstract Efficient heating and cooling systems and renewable energy sources are crucial for effectively designing net-zero energy homes (NZEHs). The study proposes using a multi-functional variable refrigerant flow system with hydraulic heat recovery (MFVRF-H2R) to reduce heating, ventilating, and air-conditioning (HVAC) and hot water energy usage, offering a practical approach to enable NZEH solutions. Photovoltaic (PV)-based on-site power generation is utilized to achieve zero energy performance in residential buildings. A building energy simulation study is conducted to assess the effectiveness of the combined systems in various climate conditions. To develop the simulation model, the US National Institute of Standards and Technology (NIST)’s net-zero energy residential test facility is used as the benchmark for NZEH baseline models. The MFVRF-H2R system is incorporated into the NZEH baseline to propose a more-energy-efficient design with heat recovery technology. eQUEST and post-processing calculations are used to simulate NZEH performance, comparing whole-building energy end-use and PV capacity for the baseline and alternative models with MFVRF-H2R. Results demonstrate that the proposed variable refrigerant flow (VRF)-based NZEH design can provide potential energy savings of up to 32% for cooling energy under various climate zones. Moreover, the NZEH design with the proposed MFVRF-H2R can achieve up to a 90% reduction in domestic hot water usage compared to an NZEH design without VRF heat recovery technology. The study suggests that the MFVRF-H2R system can provide practical and realistic solutions for making HVAC energy-efficient by minimizing thermal waste and reusing it for other thermal parts of the building, such as hot water applications. Consequently, this study highlights the effectiveness of the MFVRF-H2R system in designing NZEHs while considering heat recovery and renewable energy technologies.

Super‐heating degree responsive control strategy of variable refrigerant flow air conditioning system for energy saving and stable operation

AbstractThe control strategy is important for variable refrigerant flow (VRF) air conditioning system to realize energy saving and stable operation. In this work, the super‐heating degree (SHD) responsive control strategy was developed to dynamically adjust the respective SHD target value (ΔTe,tar) of indoor units so as to adapt to the variable cooling load. The capacity code (Ecode) is introduced into the control model to quantify the reference target value of SHD. To responsively adapt to variable cooling load, the correction target values of SHD based on discharge temperature and deviation of the outlet air temperature are introduced in the SHD model for operation protection of the compressor and the consistency of refrigerant distribution. The enthalpy difference analogy method is developed to estimate the performance of the VRF system including cooling capacity, electric power, and energy efficiency ratio (EER). The performance was tested under the conditions of different ΔTe,tar. Experimental results demonstrate that the ΔTe,tar is an important factor to determine the operation frequency, which affects cooling capacity, power, EER, and operation status of the system. The cooling capacity and electric power at the ΔTe,tar of 9°C was about 55.2% and 63.5% lower than those at the ΔTe,tar of 1°C, and the corresponding EER was improved by 22.6%. To further verify the SHD control strategy, the VRF system was tested by the responsive control model under the condition that Ecode and ΔTe,tar were constantly automatically adjusted and changed with the actual indoor ambient temperature. Experimental results demonstrate that the VRF system can achieve a good and uniform cooling effect and realize energy saving and stable operation according to the responsive super‐heating degree control.

Operation characteristics based on a novel performance model based on capacity utilization rate of a variable refrigerant flow air conditioning system

In the context of a large market share, enhancing the actual performance of variable refrigerant flow (VRF) air-conditioning systems is of considerable importance. The operating states of indoor units (IUs) significantly influence the energy efficiency of a VRF system with multiple indoor terminals. Therefore, it is challenging to address accurate and concise representation of the operational characteristics of IUs for analyzing their impact on system performance. This study explored a novel performance index called capacity utilization rate (CUR) to illustrate the effect of IUs on system performance and developed a performance model based on this index. Simulation and experimental results demonstrated that energy efficiency and evaporating temperature decrease with a reduction in CUR under specific ambient temperatures and part-load rates. The proposed performance model was applied to real buildings in field tests to examine the operational characteristics of the VRF system. This research offers valuable insights for system design, performance simulation, and performance evaluation of VRF systems.

Power consumption prediction of variable refrigerant flow system through data-physics hybrid approach: An online prediction test in office building

Variable refrigerant flow (VRF) system contains numerous sensors and has the advance for fast response, which is suitable for building demand response (DR) management. Fast and accurate power consumption prediction of VRF system is essential for DR. As traditional prediction methods, white-box models are difficult to build on operational data, while black-box models cannot make interpretable predictions. Neither of them can meet the requirements of power consumption prediction for VRF system under demand response. Therefore, a grey box model for power consumption of VRF system is proposed in this study, which has the advantage of data-driven and interpretability.The proposed model consists of four sub-models, which predict the building thermal load, compressor frequency, high pressure state and low pressure state of VRF, respectively. These predictions are finally used as inputs to the power prediction model. The proposed model is verified on both offline test and online test. The results show that the model is capable of predicting the power consumption accurately under high time resolution. During the online test, the MAE, CV-RMSE, and R2 of the model are 1296.41 W, 24.65% and 0.90, respectively. The proposed model can be used as the evaluation tool of DR management for decision making.

Modelling variable refrigerant flow system for control purpose

Variable Refrigerant Flow (VRF) systems are gradually gaining popularity in small and medium-sized commercial and residential buildings owing to their high part-load performance, flexible control, and ease of installation and maintenance. Developing models of VRF systems to predict their performance are important for model-based control, fault diagnostic and detection. There are VRF models published in existing literatures, however those models were developed and validated in different datasets. As a result, the model accuracy cannot be directly compared. To fill this gap, this paper presents a comprehensive review of the existing VRF models, and summarizes the input/output parameters and mathematical formulas of 16 VRF models from literature (referred to as physics-based model). Next, we validate and compare the model accuracy of existing models using the same dataset. Additionally, we develop data-driven models using the state-of-art machine learning algorithms, and compare the model accuracy between existing physics-based models with data-driven models. We find the model proposed by Hu et al. in 2019, which regresses the VRF cooling capacity and COP as a linear combination of indoor and outdoor temperatures times a cubed polynomial function of compressor frequency, is the most accurate physics-based model, with a prediction error of 22.19% in the training dataset and 22.44% in the validation dataset. XGBoost is the most accurate data-driven model, with a prediction error of 19.29% in the training dataset and 22.02% in the validation dataset. The data-driven model is more accurate while the physics-based model is more generalizable. The findings of this study can help researchers to select the proper VRF model for building energy prediction, model-based optimization, and fault diagnostic and detection.

Annual Effect of the VRF Control Algorithm in Response to the TOU Rate Plan

Many countries adopt a time-of-use (TOU) rate system, in which electricity rates vary by season and time of day, to reduce power usage during peak power consumption hours. South Korea offers a TOU rate plan that depends on the electricity usage of a building and its contracted power; in this plan, the electricity rate reaches up to 300% depending on the time of day. Hence, electrically powered variable refrigerant flow (VRF) systems are increasingly being installed in small- and medium-sized buildings requiring individual cooling and heating operations. This study aims to develop a new control algorithm to reduce electricity consumption and electricity rates for cooling and heating by VRF systems in university buildings adopting the TOU rate plan and apply it to actual buildings to verify the reduction effect. The proposed control algorithm primarily consists of a module that controls the refrigerant evaporation temperature (cooling) and high pressure (heating) according to the indoor heat load and a module that controls the indoor set temperature based on the hourly electricity rate. The developed algorithm was installed in the controller of a VRF system installed in an actual university building and the annual effect was verified using the method proposed by the International Performance Measurement and Verification Protocol. As a result, power consumption was reduced by 17.8% for heating and 4.0% for cooling due to the application of the control algorithm, and the electricity rates reduced by 19.2% and 7.3%, respectively.

The impact of the wall insulation material and variable refrigerant flow system on building energy consumption and cost

The use of VRF system and insulation in building envelope are both considered as promising option to reduce energy consumption of a building. This study analysis energy saving potential of VRF system and building insulation using modelling and simulation of a typical office building. The model office building is simulated with the weather of Hyderabad, Sindh, Pakistan using EnergyPlus building simulation software. The simulation cases include conventional and VRF air conditioning systems with and without insulation to evaluate and compare the annual cooling and energy savings and payback period. Results showed that by replacing conventional air conditioner with VRF AC electrical power can be reduced by 42-45%. It is also noted that Cellulose, expanded polystyrene, extruded polystyrene and polyurethane insulations can save around 49.5, 51.4, 51.6, and 54.54% of electricity, respectively. In the case VRF air conditioner used with Cellulose, expanded polystyrene, extruded polystyrene and polyurethane insulation may reduce electricity consumption by 66.5, 67.4, 67.5 and 68.9 %, respectively. The payback period varies from 7 to 15 months. However, cellulose with VRF air conditioner has the least payback period of around 7 months. The longest payback period of around 15 months was noted for the un-insulated office building with VRF air conditioner. Despite having longer payback period, the combination of polyurethane insulation with VRF air conditioning system is the most efficient combination.

Novel approach to typical air-conditioning behavior pattern extraction based on large-scale VRF system online monitoring data

The diversity of occupant air-conditioning (AC) behavior in buildings leads to significant discrepancies in building energy consumption. Thus, establishing models that reflect diverse realistic occupant AC behavior is crucial for the building performance simulations (BPS). Owing to the lack of large-scale samples, it is difficult to extract typical AC behavior patterns from existing studies. With advances in IT methods for data collection, large-scale on-site air-conditioning (AC) operation data of variable refrigerant flow (VRF) systems have become available. Therefore, based on the Big Data of 1200 indoor units (IDUs) of VRF systems, a novel method was proposed herein to establish realistic AC behavior models and extract typical patterns. The overall framework included three steps: data collection and preprocessing, large-scale occupant AC behavior model establishment, and typical behavior pattern extraction and evaluation. First, data preprocessing approaches were implemented to address data issues. Subsequently, based on Big Data, Bayesian parameter calibration was proposed to construct large-scale occupant AC behavior models. Afterwards, using the large-scale calibrated AC behavior models, the cooling demands of the population were simulated, and k-means clustering analysis was adopted to categorize the AC behavior into five groups representing different cooling demand levels and a typical model reflecting the group feature was selected. In the model validation, a hypothesis-based test was applied to confirm the validity of extracted typical patterns in reflecting the real distribution of district cooling demands. In the results, this study identified two occupancy patterns, i.e. “Daily occupied mode” (DOM) and “Night occupied mode” (NOM), and based on two occupancy types, five typical AC behavior patterns were extracted separately. From the typical patterns, it could be noticed that 90% of occupants had relatively low cooling demand less than 40 kWh/m2 during the cooling season. Moreover, comparing to the performance of fixed AC schedule model in district cooling demand evaluation, the extracted typical AC behavior patterns improved the accuracy up to 30.4% when calculating the bias between average modelled cooling demand and measured data. Additionally, the typical AC behavior models put forward in this study can be further integrated in BPS tools, which is promising for the simulation of cooling demand at building or district scale, and realizing AC technology evaluation in future engineering projects.

Advanced Method of Variable Refrigerant Flow (VRF) System Design to Forecast on Site Operation—Part 3: Optimal Solutions to Minimize Sizes

Outdoor air conditioning systems (ACS) are used as autonomic systems as well as in combined outdoor and indoor ACS of the variable refrigerant flow (VRF) type, with variable speed compressors (VSC) as their advanced version. Methods for determining the optimal value of refrigeration capacity and providing the maximum rate of the summarized annual refrigeration energy generation increment, according to its needs at minimum compressor sizes and rational values, are applied to reveal the reserves for reducing the designed (installed) refrigeration capacity, thus enabling us to practically achieve maximum annual refrigeration energy generation as the primary criterion at the second stage of the general design methodology previously developed by the authors. The principle of sharing the total thermal load on the ACS between the ranges of changeable loads for outdoor air precooling, and a relatively stable load range for further processing air are used as its basis. According to this principle, the changeable thermal load range is chosen as the object for energy saving by recuperating the excessive refrigeration generated at lowered loading in order to compensate for the increased loads, thereby matching actual duties at a reduced designed refrigeration capacity. The method allows us to determine the corresponding level of regulated loads (LRL) of SRC and the load range of compressor operation to minimize sizes.

Advanced Method of Variable Refrigerant Flow (VRF) Systems Designing to Forecast Onsite Operation—Part 2: Phenomenological Simulation to Recoup Refrigeration Energy

This paper focuses on the application of speed-regulated compressors (SRCs) to cover changeable heat loads with high efficiency in conventional air conditioning systems (ACS) as well as in the more advanced variable refrigerant flow (VRF)-type outdoor and indoor ACS. In reality, an SRC is an oversized compressor, although it can operate efficiently at part loads. The higher the level of regulated loads (LRL) of the SRC, the more the compressor is oversized. It is preferable to reduce the size of the SRC by covering the peak loads and recouping the excessive refrigeration energy reserved at decreased actual loads within the range of regulated loads. Therefore, the range of changeable loads is chosen as the object to be narrowed by using the reserved refrigeration capacity. Thus, the general fundamental approach of dividing the overall heat load range of the ACS into the ranges with changeable and unchangeable loads, as previously developed by the authors, is applied for the range of primary changeable loads. Due to this innovative step, the principle of two-stage outdoor air conditioning according to changeable and unchangeable loads, also proposed by the authors, has been extended over the range of primary changeable loads to reduce the level of refrigeration capacity regulation and SRC size. To realize this, part of the changeable load range is offset by the reserved refrigeration capacity, leading to a reduction in the changeable load range and the SRC size by approximately 20% for temperate climatic conditions.

Advanced Method of Variable Refrigerant Flow (VRF) Systems Designing to Forecast On-Site Operation—Part 1: General Approaches and Criteria

All the energetic management and controlling strategies in ambient air conditioning systems (ACS) are aimed to match design load to current needs. This might be achieved by determining a rational value of design thermal load without overestimation that can minimize its deviation from the actual values. The application of variable refrigerant flow (VRF) systems with speed-regulated compressors (SRC) is considered as the most advanced trend in building air conditioning due to the ability of SRCs to cover changeable heat loads without lowering their efficiency. The level of load regulation by SRC is evaluated as the ratio of the load range, regulated by SCR, to the overall design load range. With this, the range of actual changeable loads is usually supposed to be covered by SRC entirely while keeping the rest, unregulated, and load range unchangeable. However, to confirm this, the rest load range behind the regulated one should be investigated to estimate the efficiency of SRC operation. Therefore, the approach to dividing the overall thermal load range of ambient air conditioning into the ranges of changeable and unchangeable loads to compare with those covered by SRC is used. From this approach, the method of rational designing and shearing a design refrigeration capacity in response to current loading, based on the principle of two-stage ambient air conditioning, has been widened on the VRF systems to estimate the efficiency of SCR application. This was realized by imposing the load ranges regulated by SRC onto the ranges of changeable and unchangeable loads within the overall range of actual loading. The proposed innovative criteria and indicators for rational shearing the load ranges to match current duties and load level evaluation can reveal the reserves for improving the efficiency of SRC compressor operation and the ACS of VRF type as a whole.

Building energy efficiency improvement using multi-objective optimization for heating and cooling VRF thermostat setpoints

The variable refrigerant flow system is one of the best heating, ventilation, and air conditioning systems (HVAC) thanks to its ability to provide thermal comfort inside buildings. But, at the same time, these systems are considered one of the most energy-consuming systems in the building sector. Thus, it is crucial to well size the system according to the building’s cooling and heating needs and the indoor temperature fluctuations. Although many researchers have studied the optimization of the building energy performance considering heating or cooling needs, using air handling units, radiant floor heating, and direct expansion valves, few studies have considered the use of multi-objective optimization using only the thermostat setpoints of VRF systems for both cooling and heating needs. Thus, the main aim of this study is to conduct a sensitivity analysis and a multi-objective optimization strategy for a residential building containing a variable refrigerant flow system, to evaluate the effect of the building performance on energy consumption and improve the building energy efficiency. The numerical model was based on the EnergyPlus, jEPlus, and jEPlus+EA simulation engines. The approach used in this paper has allowed us to reach significant quantitative energy saving by varying the cooling and heating setpoints and scheduling scenarios. It should be stressed that this approach could be applied to several HVAC systems to reduce energy-building consumption.

Experimental research on the performance and parameters sensitivity analysis of variable refrigerant flow system with common faults imposed in heating mode

Due to improper installation, operation, and maintenance, many common faults inevitably occur in the air conditioning systems. However, the same fault could demonstrate different characteristics under different load rates and fault levels; different faults may present similar fault phenomena. This paper reveals the quantitative impact of faults on system performance of VRF system with common faults imposed in heating mode through fault experiments. The system parameters sensitivity study has been carried out to understand the internal impact mechanism related to system parameters. This provides a reference for selecting characteristic parameters and fault decoupling variables and developing interpretable, professional, and reliable fault detection and diagnosis models. The results show that the outdoor unit fouling fault can cause the heating capacity to drop by 73.2%, and the indoor unit fouling can cause the system power to drop by 65.5%. The coefficient of performance drop is as high as 48.6%. The sensitivity study result shows that the system load rate is an influential parameter; it needs to be considered when characteristic parameters and fault decoupling variables are selected.

Evaluation of the performance of water-cooled VRF system for heat exchanger geometry under part-load operation

In this study, static numerical simulation was used to evaluate the performance of a water-cooled variable refrigerant flow (VRF) system for geometry of plate-type outdoor heat exchanger. In general, heat exchanger is designed based on full-load operating condition. Since most of operation for VRF system is part-load operation, part-load operating characteristics should be considered to optimize entire system. The geometry characteristics are divided into chevron angle and vertical length. Numerical models of cooling and heating modes were developed for the VRF system, system performance was represented by integrated energy efficiency ratio (IEER) for the cooling system and coefficient of performance (COP) for the heating system by AHRI standard 1230. Regardless of operating mode, 20° of chevron angle among 20°, 35° and 45° is the optimal angle of outdoor heat exchanger. The energy efficiency was proportional to the vertical length of the plate heat exchanger and inversely proportional to the chevron angle. In the cooling mode, the energy efficiency ratio (EER) calculation converged at a shorter vertical length of the heat exchanger with a decrease in operating loads. The optimal vertical length is 1.7 m with only considering full-load, and 1.2 m with considering part-load operation. In the heating mode, the tendency of energy efficiency was similar to that of the cooling mode. The optimal vertical length is 1 m which is 0.2 m shorter than the optimal length in cooling mode.

Investigation of VRF system cooling operation and performance in residential buildings based on large-scale dataset

With the increase in the energy consumption of air-conditioning (AC) systems in Chinese residential buildings, the realization of energy savings in AC systems has attracted increasing attention. The variable refrigerant flow (VRF) system is a common AC system for residential buildings in China. In most previous studies on VRF systems, onsite measurements or surveys were conducted to collect operational data. These traditional methods may face various data issues, such as limited sample sizes and invalid data, making them unable to capture the spatial and temporal performance features of VRF systems in residential buildings on a large scale. To fill this gap, with advances in data storage and transmission technology, Big Data methods have been widely used for data collection. In the present study, researchers adopted 16,985 sets of VRF system operation data from China as the database and conducted data analysis for both the spatial and temporal dimensions. Several key indicators were proposed from the two perspectives (spatial and temporal), including the part-space index (PSI), load ratio (LR), use duration (UD), and cooling energy consumption. The main findings were as follows: (1) The “part-time part-space” operation mode of residential VRF systems can be analyzed according to the statistical results of the UD and PSI. (2) An LR of <30% is the main operating condition for VRF systems in residential buildings. (3) Extracted typical LR patterns can reflect different user behavior. The statistical results obtained in this study provide a basis for VRF engineering projects.

CFD Simulation of Thermal Plume of VRF Air Conditioners for Cooling Season in Semi-Closed Spaces for Erbil Climate Conditions

The variable refrigerant flow (VRF) system is extensively utilized for the purpose of space cooling and heating in residential and commercial buildings. The VRF outdoor units are frequently placed on balconies of multi-storey buildings, and the aluminum louvers conceal them to keep the aesthetic architecture appearance. This type of installation will cause the thermal plume influence of the dissipated heat discharged from condensing units to cause high suction temperature, high energy consumption, and low coefficient of performance of the system. The numerical results showed that the discharged heat from condensing units will lead to rising thermal plume influence, causing greater suction temperatures for VRF outdoor units on top floors, even overriding the operating temperature range. It also shows that removing the inward louver in front of the discharge duct on each floor is an efficient method to deteriorate the thermal plume influence for VRF air conditioners and enhance the thermal efficiency of the entire system. For the proposed design, the average suction temperatures could be reduced by 8%, 15%, and 19% for VRF outdoor units for floors 10th, 11th, &amp; 12th, respectively. This work aims to provide guidelines for the placement of VRF air conditioners in buildings in the designing stage.

Optimal insulation’s thickness of pipes in Variable Refrigerant Flow (VRF) system—An-Najah Child Institute as a case study

Buildings use a significant amount of the world’s total primary energy, particularly in heating and cooling systems. Energy consumption of air conditioning systems can be lowered by installing energy-efficient systems that take into account piping thermal insulation. The objective of this research is to calculate the optimal thickness of insulation for piping networks with variable refrigerant flow and to investigate the environmental and economic impacts of using the optimal insulation thickness. The case study of this research is An-Najah Child Institute. The research is based on the current cost over the life cycle of the insulation, which is assumed to be 15 years, to determine the optimal thickness of the common insulation types (rock wool, extruded polystyrene, and flexible foam) for different pipe diameters and different refrigerant flow rates. The research results show that the optimum insulation thicknesses for pipe diameters from 12.7 to 50.8 mm using Rockwool are 32 to 44 mm, 28 to 38 mm for extruded polystyrene, and 11to 13 mm for flexible foam insulation. The reduction in energy losses from the optimal Rockwool insulation thickness in the case study piping system ranges from 78.5% to 81.6% compared to the uninsulated insulated piping system. In addition, the carbon emission saving by the optimal Rockwool insulation thickness is 1737, 1541, 1033, and 1160 kg CO2 for coal, oil, natural gas, and LPG fuel used in electricity generation, respectively. In addition, the study found that the physical parameters of the refrigerant, such as quality and flow rate, do not affect the optimal insulation thickness for all insulation types studied. And the thermal power saved corresponds to only 1.2% of the total system capacity

Improving operation management performance of air conditioners in offices using data envelopment analysis

Heating, ventilation, and air conditioning (HVAC) systems account for the bulk of a building's energy use, which means that potentially significant energy savings can be made by efficiently modifying how those systems are operated. Although several studies have demonstrated that optimizing the operation of HVAC equipment in combination with sensor readings can reduce potentially wasteful processes, such efforts are limited to buildings that have or can incorporate additional sensors. As an alternative approach, this paper aims to identify potentially wasted energy and provide suggested improvements for buildings that do not have installed sensors, which is important because such suggestions can extend the scope of energy reduction measures. More specifically, we hypothesized that comparing differences in user operations would make it possible to identify potentially wasteful operating times. Accordingly, we began our study by identifying target days and days for improvement using data envelopment analysis and then detected operational differences by comparing their operating times. Our new method was then applied to variable refrigerant flow systems in an office building as a case study. The obtained results showed that the potential for energy reduction was about 7.5%. These results, which helped determine the appropriate times for automatically shutting down each indoor unit, proved effective in implementing energy reduction measures. The novelty of this paper is that it clarified the energy waste that can be detected based on operating times, which is expected to lead to the development of additional options for energy reduction measures, thereby allowing building owners to implement more effective actions based on the sizes and conditions of their buildings.

Research on the semantic web representation for building operation with Variable Refrigerant Flow systems

Building metadata is regarded as the signpost in organizing massive building data. The application of building metadata simplifies the creation of digital representations and provides portable data analytics. Typical metadata standards such as Brick and Haystack are used to describe the data of the building system. However, neither Haystack nor Brick has provided definitions of the Variable Refrigerant Flow system so far. With the growing market for the Variable Refrigerant Flow, it is needed to have semantic descriptions of Variable Refrigerant Flow systems in metadata standards for data analysis and management. Addressing this challenge, this paper developed and verified the semantic descriptions of the Variable Refrigerant Flow system in Brick Schema. With the semantic representation, a building data management framework including a backend and a frontend has also been developed. Then, the model and the service framework are tested in an office building. Results indicated using the semantic model and the service framework reduces manpower and data analysis time. This paper provides a possible solution for the expression of the Variable Refrigerant Flow systems in the building semantic web. The works in this paper will support the semantic web in automation strategies for building management and scalable building operation.

A robust VRF fault diagnosis method based on ensemble BiLSTM with attention mechanism: Considering uncertainties and generalization

Developing effective fault detection and diagnosis (FDD) model is of great significance to the energy efficiency and comfort of variable refrigerant flow (VRF) system. In recent FDD studies, the consideration of incomplete information and uncertainty, as well as the generalization of model and the interpretability of fault action mechanism have become a great concern. This paper accordingly proposes a fault diagnosis strategy based on attention-BiLSTM with ensemble feature sets. Three methods including Correlation Analysis, Information Entropy, and Gini Impurity are used for feature selection and ensemble feature sets combination. The proposed model can achieve interpretability of fault diagnosis and quantification of diagnosis uncertainties through attention mechanism and ensemble feature sets. A total of four typical faults under different working conditions are used to verify the accuracy and generalization of the proposed method. The results show that the ensemble attention-BiLSTM model has a fault diagnosis accuracy rate of 98.3% if the first two most likely outputs are considered, and when some sensors fail, the accuracy rate can still remain above 90%. In addition, a generalization verification idea is proposed, that is, the model is trained under specific operating conditions and tested on datasets with different operating conditions or fault levels. The results show that the average generalization of the optimized model reaches 86%.