Energy Prediction Model Research Articles

Combining geographic information and climate data to develop urban building energy prediction models in Taichung, Taiwan

Climate change in Taiwan has extended and intensified the summer season, leading to a notable surge in energy demand for cooling systems, especially in densely populated regions. Building energy usage is directly correlated with cooling degree hours (CDHs), representing the hourly temperature differential between indoors and outdoors. This study employed high-resolution Taiwan ReAnalysis Downscaling (TReAD) data to develop an urban energy prediction model focusing on localized cooling demand in central Taiwan's urban areas. Validated against actual electricity consumption data, the model achieved an R2 value of 0.76. The study reveals that urban areas exhibit a high cooling demand during the hot season, exceeding 25,000°C-h and with an annual energy consumption of 44-64 kWh/m2. Conversely, rural areas have a lower cooling demand – that is, below 8,000°C-h, with an annual energy consumption of less than 10 kWh/m2.Considering the IPCC's RCP8.5 warming scenario, October shows a 20-40% increase in cooling demand compared to July and May. This underscores the need to address rising energy consumption especially during the early and late stages of the hot season in response to climate change.

A predictive walking energy model based on gait phase with suspended backpack

Walking with heavy loads is a common task in military affairs and daily life. Considering that the shoulder and leg muscles fatigue will be caused during walking, which will affect the walking endurance and physical health. However, the suspended backpack is found to improve the energy efficiency of walking with a load. In this study, A lightweight suspended backpack is designed and proposing a model for estimating the metabolic cost of a suspended backpack based on gait phase. In this study, four inertial measurement units (IMUs) are fixed on the thigh and shank, six flexible pressure sensors are mounted on the soles of the feet and shoulders, respectively. The gait is defined as four successive phases. For each phase, the muscle tension is solved based on the muscle moment balance theory. Based on the phase segmentation method, the ECCF index is calculated by adding the gait phase constraint and backpack data calculation into the energy prediction model, and the relatively accurate data is obtained. In addition, In order to study the effects of the suspended backpack with different parameters on the cost metabolism, gait phase and biomechanics, the subjects need to carry the same load of 16.5 kg to walk 400 m at the different speeds, respectively. A group of seven healthy subjects in the same walking condition need to conduct two experiments: suspended backpack work (SB) and ordinary backpack (OB). The experimental results show that the suspended backpack can reduce plantar pressure and shoulder pressure in the SB condition. And at the speed of 5.0 km/h, ground reaction force (GRF) and shoulder reaction force (SRF) were reduced by 11.59 % and 13.22 % in the SB condition compared to the OB condition, respectively.

Study on Influencing Factors and Prediction of Tunnel Floor Heave in Gently Inclined Thin-Layered Rock Mass

In recent years, the construction of new railway tunnels worldwide has become increasingly challenging due to larger cross-sections, deeper burial depths, higher in situ stress, and more complex geological conditions. During both construction and operation, some tunnels have encountered significant issues with floor heave. This paper begins by identifying the primary causes of deformation and instability in tunnel floor structures through an investigation and statistical analysis. It then examines floor heave across more than 20 railway lines, summarizing the types, generation mechanisms, and mechanical models associated with this issue. Additionally, extensive survey data indicate that tunnel floor heave is most likely to occur in gently inclined thin-layered rock masses. Therefore, using a tunnel passing through the plate suture zone in such a rock mass as a case study, numerical simulations, theoretical analyses, and on-site monitoring were conducted. This study systematically analyzed the influence of single and multiple factors, as well as the mechanical behavior of the support system, on tunnel floor heave in gently inclined thin-layered surrounding rock. Furthermore, several key models were proposed: a tunnel floor heave estimation and load formula based on a mechanical model, a dynamic relationship between surrounding rock support force and tunnel floor heave using the Nishihara model, a tunnel floor settlement estimation formula based on deformation statistics, and a tunnel floor heave energy prediction model utilizing the B-P neural network algorithm. These conclusions have been validated and widely applied in practical engineering, providing a robust theoretical foundation and technical support for future tunnel construction.

CRISP-DM-Based Data-Driven Approach for Building Energy Prediction Utilizing Indoor and Environmental Factors

The significant energy consumption associated with the built environment demands comprehensive energy prediction modelling. Leveraging their ability to capture intricate patterns without extensive domain knowledge, supervised data-driven approaches present a marked advantage in adaptability over traditional physical-based building energy models. This study employs various machine learning models to predict energy consumption for an office building in Berkeley, California. To enhance the accuracy of these predictions, different feature selection techniques, including principal component analysis (PCA), decision tree regression (DTR), and Pearson correlation analysis, were adopted to identify key attributes of energy consumption and address collinearity. The analyses yielded nine influential attributes: heating, ventilation, and air conditioning (HVAC) system operating parameters, indoor and outdoor environmental parameters, and occupancy. To overcome missing occupancy data in the datasets, we investigated the possibility of occupancy-based Wi-Fi prediction using different machine learning algorithms. The results of the occupancy prediction modelling indicate that Wi-Fi can be used with acceptable accuracy in predicting occupancy count, which can be leveraged to analyze occupant comfort and enhance the accuracy of building energy models. Six machine learning models were tested for energy prediction using two different datasets: one before and one after occupancy prediction. Using a 10-fold cross-validation with an 8:2 training-to-testing ratio, the Random Forest algorithm emerged superior, exhibiting the highest R2 value of 0.92 and the lowest RMSE of 3.78 when occupancy data were included. Additionally, an error propagation analysis was conducted to assess the impact of the occupancy-based Wi-Fi prediction model’s error on the energy prediction model. The results indicated that Wi-Fi-based occupancy prediction can improve the data inputs for building energy models, leading to more accurate energy consumption predictions. The findings underscore the potential of integrating the developed energy prediction models with fault detection systems, model predictive controllers, and energy load shape analysis, ultimately enhancing energy management practices.

A hybrid transfer learning to continual learning strategy for improving cross-building energy prediction in data increment scenario

For a data-driven building energy prediction model, the insufficient data available for new buildings and existing buildings with limited data pose a challenge to achieving the expected prediction performance. Transfer learning (TL) and continual learning (CL) are two widely applied effective methods for addressing the shortage of building data. The actual building projects typically exist in a scenario of data increment, wherein their energy consumption data gradually accumulates with usage. Previous studies have indicated that the TL strategy can transfer the learned data distribution patterns from source buildings to target buildings of the same type. Meanwhile, the CL strategy can extract the dynamic characteristics from continuously accumulated data. Both strategies can be employed to enhance the performance of energy consumption prediction models for buildings with insufficient data. However, existing studies on data scarcity in energy prediction primarily focus on buildings with fixed data quantities, and have not provided sufficient analysis for newly constructed buildings in a data increment scenario. Therefore, this study proposes a TL-to-CL strategy to enhance the energy prediction performance of newly constructed buildings in a data increment scenario. When the available data is limited, TL should be utilized. But as the data volume increases to be sufficient, TL should be transformed to CL at the reasonable time. Additionally, this study provides a reference for determining the transform time for the TL-to-CL strategy. The validation was conducted using two years of energy consumption data from 36 buildings in the Genome Project 2 dataset. The results demonstrate that the proposed TL-to-CL strategy (transform time = 4-week) obtain the overall better prediction performance than pure TL, pure CL and combined TL-CL without transformation between the two) strategies. The prediction improvement ratio (PIR) can be as high as 0.9 compared with the traditional LSTM.

Addressing Data Scarcity in Solar Energy Prediction with Machine Learning and Augmentation Techniques

The accurate prediction of global horizontal irradiance (GHI) is crucial for optimizing solar power generation systems, particularly in mountainous areas with complex topography and unique microclimates. These regions face significant challenges due to limited reliable data and the dynamic nature of local weather conditions, which complicate accurate GHI measurement. The scarcity of precise data impedes the development of reliable solar energy prediction models, impacting both economic and environmental outcomes. To address these data scarcity challenges in solar energy prediction, this paper focuses on various locations in Europe and Asia Minor, predominantly in mountainous regions. Advanced machine learning techniques, including random forest (RF) and extreme gradient boosting (XGBoost) regressors, are employed to effectively predict GHI. Additionally, optimizing training data distribution based on cloud opacity values and integrating synthetic data significantly enhance predictive accuracy, with R2 scores ranging from 0.91 to 0.97 across multiple locations. Furthermore, substantial reductions in root mean square error (RMSE), mean absolute error (MAE), and mean bias error (MBE) underscore the improved reliability of the predictions. Future research should refine synthetic data generation, optimize additional meteorological and environmental parameter integration, extend methodology to new regions, and test for predicting global tilted irradiance (GTI). The studies should expand training data considerations beyond cloud opacity, incorporating sky cover and sunshine duration to enhance prediction accuracy and reliability.

Comprehensive Analysis of Influencing Factors on Building Energy Performance and Strategic Insights for Sustainable Development: A Systematic Literature Review

A prerequisite for decreasing the intensification of energy in buildings is to evaluate and understand the influencing factors of building energy performance (BEP). These factors include building envelope features and outdoor climactic conditions, among others. Based on the importance of the influencing factors in the development of the building energy prediction model, various researchers are continuously employing different types of factors based on their popularity in academic literature, without a proper investigation of the most relevant factors, which, in some cases, potentially leads to poor model performance. However, this can be due to the absence of an adequate comprehensive analysis or review of all factors influencing BEP ubiquitously. Therefore, this paper conducts a holistic and comprehensive review of studies that have explored the various factors influencing energy use in residential and commercial buildings. In total, 74 research articles were systematically selected from the Scopus, ScienceDirect, and Institute of Electrical Electronics Engineers (IEEE) databases. Subsequently, by means of a systematic and bibliometric analysis, this paper comprehensively analyzed several important factors influencing BEP. The results reveals the important factors (such as windows and roofs) and engendered or shed light on the application of some energy-efficient strategies such as the utilization of a green roof and photovoltaic (PV) window, among others.

Analyzing meteorological parameters using Pearson correlation coefficient and implementing machine learning models for solar energy prediction in Kuching, Sarawak

Solar energy is one of the clean renewable energy sources that can offset the rising consumption of fossil fuels. However, the meteorological parameters, such as solar irradiance, ambient and solar module temperatures, relative humidity, etc., constantly change, and so does the solar power generation. Such variations cause instability in the power grid operation due to injecting an unpredicted amount of power. Hence, solar energy prediction models capable of learning from past weather data and predicting future energy generation are highly desired for grid operation and planning. The objective of this study is to determine the suitable meteorological parameters for the solar energy prediction model based on the Pearson correlation coefficient and to implement them in different machine learning models. It is found in this study that five meteorological parameters, namely Air temperature, cloud opacity, global tilted irradiance, relative humidity, and zenith angle, correlate highly with solar energy generation. Later, based on the correlations, four machine-learning models were implemented to predict the solar power for Kuching, Sarawak. The accuracy of the models is measured through standard matrices such as root mean square error, mean square error, mean absolute error, and R-squared value.

Analysis of variables to determine their influence on renewable energy forecasting using ensemble methods

Forecasting is of great importance in the field of renewable energies because it allows us to know the quantity of energy that can be produced, and thus, to have an efficient management of energy sources. However, determining which prediction system is more adequate is very complex, as each energy infrastructure is different. This work studies the influence of some variables when making predictions using ensemble methods for different locations. In particular, the proposal analyzes the influence of the aspects: the variation of the sampling frequency of solar panel systems, the influence of the type of neural network architecture and the number of ensemble method blocks for each model. Following comprehensive experimentation across multiple locations, our study has identified the most effective solar energy prediction model tailored to the specific conditions of each energy infrastructure. The results offer a decisive framework for selecting the optimal system for accurate and efficient energy forecasting. The key point is the use of short time intervals, which is independent of type of prediction model and of their ensemble method.

Fairness-aware data-driven-based model predictive controller: A study on thermal energy storage in a residential building

Besides accuracy, fairness has been reported as another performance criterion for data-driven building models (DDBMs). To ensure data-driven-based model predictive controllers (MPCs) provide optimal control strategies based on fair or unbiased prediction, this study proposes the concept of fairness-aware data-driven-based MPC. In the proposed MPC framework, fairness-aware DDBMs constitute the prediction component that provides predicted building states to the objective function in the optimization part. To investigate the effect of improving fairness on the control performance, the proposed MPC is implemented in a residential building heated by an electrically heated floor (EHF) system, which could be considered as a thermal energy storage (TES) system. In the case study, fairness-aware DDBMs are developed to predict energy demand and indoor air temperature. Then, the predicted values are used to formulate the objective function to optimize the hourly day-ahead set-point temperature with the aim of minimizing the heating cost or maximizing peak shifting, while maintaining thermal comfort. The numerical study results show that although considering fairness improvement methods in DDBMs decreases the overall predictive accuracy, it provides fair prediction by narrowing the accuracy difference between majority conditions and minority conditions. For instance, a fairness-aware energy prediction model increases the overall mean absolute error (MAE) from 6.12 kWh to 7.56 kWh but decreases the MAE difference between a majority condition and a minority condition from 0.82 kWh to 0.14 kWh. Although improving predictive fairness comes with a price of overall predictive accuracy, fairness-aware data-driven-based MPCs show comparable peak shifting or cost-saving ability with the traditional MPC in which fairness is not considered. This study provides a reference for stakeholders to design and implement trustworthy MPCs in buildings based on fair and accurate prediction.

A real time condition based sustainable maintenance method for milling process

The manufacturing industry must reduce energy consumption during the machining process to comply with energy demand and usage restrictions. Previous research has employed various methods to take a optimization plan in machining process. However, these studies mostly rely on static machining conditions, without paying much attention for the condition variation, which may lead to a coarse energy efficiency optimization in machining process. Hence, this study introduces a real time condition based sustainable maintenance method for machining process. Firstly, a real-time data acquisition system is employed to collect and send data to the developed tool condition prediction model using deep learning. Then, the physical energy prediction model is employed to reflect the real time energy consumption behaviors based on the real time tool condition. Subsequently, a similarity analysis is employed to decide whether a new optimization is required. And, once it is required, an optimization using NSGA-II and TOPSIS is conducted. Experiments were conducted on a milling machine and compared with traditional methods, and verified using triple bottom line. Results indicated that the proposed method can reduce the energy consumption by 2.51% and 7.01%, and it can maintain a qualified product during the machining process.

Predicting energy consumption of mosque buildings during the operation stage using deep learning approach

The energy consumption resulting from the construction sector is increasing rapidly. Different building types have different energy consumption schemes, especially during their operational stage. Among these types of buildings, mosques are considered one of the intermittent building types that suffer from huge energy wastage. That’s why there is a need to monitor their operational performance through retrofitting or changing their operational scheme to enhance their operational performance and examine the effect of different alterations on the annual energy consumption. However, using traditional methods and simulation models to examine the impact of different alterations on the annual energy consumption of the mosque is considered difficult and time-consuming. Using machine learning and deep learning models, energy prediction models facilitated energy consumption predictions for improving building energy utilization. Therefore, this paper proposes an efficient approach for predicting mosques’ energy consumption using a deep learning approach. Convolution Neural Network (CNN) deep learning model is developed to predict the annual energy consumption of mosque buildings based on various operational Scenarios. 3D Laser scanning and a detailed energy analysis model with various simulation results are used to generate the mosque’s simulated energy consumption dataset. Different operational variables, such as the operational schedule, division of the mosque into zones, utilization of dimmers, presence of cooling system and the setting point, and the power consumption of the lighting system and appliances, have been utilized to estimate the annual energy consumption for the mosque to formulate the database. Using the mosque’s simulated energy consumption dataset, the CNN model was trained and tested to obtain the best model configuration with regard to the defined performance metrics. Besides, the performance of the developed model has been compared with the Support Vector Regression (SVR) model. The developed model has achieved results of 4.5% and 0.98 for the MAPE and R2, respectively. The results revealed the developed CNN model's effectiveness for predicting energy consumption for mosque buildings. This model will help compare different operational strategies for mosque buildings and determine energy consumption and energy-efficient options based on different operational characteristics.

Development of an energy prediction model for residential buildings using Artificial Neural Network

A model has been developed in this study for predicting the energy consumption of residential building sector using Artificial Neural Network. This model was based on the Multi-Layer Perceptron architecture using feed-forward back propagation algorithm for training. The Levenberg-Marquardt function and the Gradient Descent with momentum function were used as training and learning function, respectively. The mean squared error was used to check the overall performance of the developed model. Trials were performed to finalize the number of hidden layers required for the model. Regression analysis was done between the predicted and actual data to validate the proposed ANN model. The prediction for the same dataset was also performed using the traditional trend extrapolation method, and the predicted results of both were compared with the actual energy consumption data recorded by the electricity regulatory authority of the state. It has been concluded that the accuracy of predicted data using the proposed model was very high (i.e., of 99.54%) as compared to the traditionally used TEM (i.e., 91.07%). MSE achieved for the ANN model was 0.01767% and that of TEM was 0.13115%. The outcome of this study can be used at building level to achieve energy efficiency by predicting the energy consumption and at the level of distribution companies to predict the overall energy demand by the respective sector, and take measures accordingly.

OPTIMIZATION EXPERIMENTAL STUDY OF MACHINING ENERGY CONSUMPTION OF ZIG-ZAG MILLING CUTTING PATH USING RSM.

Machining operations in CNC milling which remove the work material require power and energy to activate the machine components such as spindle motor, table and tool movement in order to withstand the high friction and load between tool and work material. Energy consumption during cutting operation is greatly influenced by the machining condition and parameters. This experimental research aims to investigate how energy responds to changes in the machining parameters such as depth of cut, spindle speed, and feed rate during face milling operation of CNC machine. The high-speed steel (HSS) tool with a 10mm diameter was used to face mill the 40mm x 40mm of Aluminum 6061. The design of experiment technique using Response Surface Methodology (RSM) is utilized to optimize the experimental work. Power usage and machining time were recorded for each machining process, which is then used to determine the machining energy consumption. The interaction between machining parameter and energy is comprehensively visualized using surface and contour plot. Additionally, the ANOVA analysis investigates the feed rate as the most influential parameter to the machining energy. Finally, the regression equation of machining energy is generated with reliability (R) value of 0.88 which can be used as an energy prediction model.

Transferability and robustness of a data-driven model built on a large number of buildings

Data-driven energy prediction models have shown a great importance in building energy management. However, these models require sufficient operational data to ensure prediction accuracy, posing great challenges for buildings with scarce data. Transfer learning has emerged as a key strategy to overcome this issue, enabling cross-building prediction even with limited data availability. While existing studies have mainly focused on a single or a few specific buildings to train models, this study aims to explore the transferability of building energy prediction models across a large number of buildings. Initially, energy consumption data retrieved from 327 buildings were selected as source domain to create a pre-trained model. Different volumes of data in the target domain were then utilized to fine-tune the pre-trained model, and the resulting accuracy distribution and accuracy improvement achieved by transfer learning were examined. The study also evaluated model robustness by conducting transfer procedures at 20 different time nodes. The occurrence of negative transfer was also monitored. The results show that transfer learning can significantly improve prediction accuracy when compared with the baseline model, with a median MAPE value improving from 18.31 % to 7.76 % when using only 7 days data. Meanwhile, transfer learning using 7 days data outperformed direct prediction using 180 days of data. However, negative transfer may occur, although at a low rate, and is not related to data volume. In addition, there is a possibility that a model with high general accuracy yield biased results at certain time nodes. This work provides valuable insights into the advantages and limitations of transfer learning in building energy prediction model, which facilitates the exploitation of existing building data source for advanced data analytics.

A Regression Framework for Energy Consumption in Smart Cities with Encoder-Decoder Recurrent Neural Networks

Currently, a smart city should ideally be environmentally friendly and sustainable, and energy management is one method to monitor sustainable use. This research project investigates the potential for a “smart city” to improve energy management by enabling the adoption of various types of intelligent technology to improve the energy sustainability of a city’s infrastructure and operational efficiency. In addition, the South Korean smart city region of Songdo serves as the inspiration for this case study. In the first module of the proposed framework, we place a strong emphasis on the data capabilities necessary to generate energy statistics for each of the numerous structures. In the second phase of the procedure, we employ the collected data to conduct a data analysis of the energy behavior within the microcities, from which we derive characteristics. In the third module, we construct baseline regressors to assess the proposed model’s varying degrees of efficacy. Finally, we present a method for building an energy prediction model using a deep learning regression model to solve the problem of 48-hour-ahead energy consumption forecasting. The recommended model is preferable to other models in terms of R2, MAE, and RMSE, according to the study’s findings.

Uncertainty-aware Energy Harvest Prediction and Management for IoT Devices

Internet of things (IoT) devices are popular in several high-impact applications such as mobile healthcare and digital agriculture. However, IoT devices have limited operating lifetime due to their small form factor. Harvesting energy from ambient sources is an effective method to supplement the battery. Energy harvesting necessitates development of energy management policies to manage the harvested energy. Designing optimal policies for energy management is challenging for two key reasons: (1) ambient energy sources are highly stochastic; therefore, energy management policies must consider the associated uncertainty; (2) energy management policies must consider future energy availability while making decisions to ensure that sufficient energy is available when there is no ambient energy. Prior approaches typically consider energy in the immediate future (e.g., 1 hour) and do not account for the uncertainty in future energy harvest. This article proposes novel machine learning and dynamic optimization-based approaches to handle the two challenges. Specifically, we first develop a novel set of features and use it in a low-power neural network architecture to predict future energy availability and uncertainty. The energy predictions and uncertainty are used in a dynamic optimization algorithm to optimally allocate the harvested energy. Experiments on solar energy data over 5 years from Golden, Colorado, show that the proposed energy prediction model achieves 3.4 J mean absolute error while having a coverage of 80%. Moreover, our energy management algorithm provides energy allocations that are within 2.5 J of an optimal Oracle with 2.65 mJ to 36.54 mJ of energy overhead.

Explainable molecular simulation and machine learning for carbon dioxide adsorption on magnesium oxide

The effects of the adsorption energy of CO2 within MgO at different temperatures were investigated by molecular dynamics simulations and experimentally verified. The adsorption mechanism of CO2 within MgO was discussed and explained qualitatively. The results indicated that the diffusive adsorption of CO2 by MgO was divided into two stages, and the ability of CO2 capture by the cubic MgO performed better than that by spherical MgO. The adsorption of CO2 by the cubic MgO was mainly physical and received the inhibited adsorption behavior at the high-temperature stage (>505 K). Herein, we established a comprehensive dataset of adsorption energies and quantitatively analyzed an adsorption energy prediction model using machine learning techniques. The results demonstrated that Decision Tree Regression (DTR) and K-nearest neighbor (KNN) algorithms offer satisfactory accuracy based on root mean square error (RMSE) and R2 evaluations. This approach enables efficient and precise prediction of adsorption energies without the need for labor-intensive molecular dynamics calculations. Furthermore, we explored the influence of various features (Crystal structure, The number of Mg, The number of CO2, Temperature, Pressure, Volume, and Bond energy) on prediction performance. Lastly, we globally evaluated the relative contributions of each feature across four sets of relatively effective algorithms. This comprehensive analysis enhances our understanding of the adsorption mechanism of magnesium oxide on carbon dioxide and provides valuable insights to guide the design of the next generation of high-performance magnesium oxide materials for carbon capture and storage.

Artificial intelligence-based metabolic energy prediction model for animal feed proportioning optimization

With the progress of science and technology, Artificial Intelligence (AI) technology has become one of the mainstream technologies in the current society, providing an important driving force for human development. Thereby, in order to improve the effect of animal feeding, using AI technology to improve animal feed has become a necessary measure. Based on this, this work designs to use Long Short-Term Memory (LSTM) technology to build an intelligent prediction model of metabolic energy, which provides a reference for animal feed proportioning design. This work also explores the comprehensive performance of the LSTM model through simulation evaluation. The model is evaluated with different nodes as the main indicators. The results show that compared with the models with 5 and 20 nodes, the model with 10 nodes has better performance, and the highest data calculation accuracy of the model is about 90%. Meanwhile, the highest fitting degree of the model designed is 98.2%, and the lowest is 96.2%. It suggests that the model designed can better predict metabolic energy. This work provides technical support for expanding the application scope of AI technology and contributes to the intelligence of animal feeding.

Development of a hybrid VRF system energy consumption prediction model based on data partitioning and swarm intelligence algorithm

Accurately forecasting energy consumption is beneficial and pivotal for effectively managing variable refrigerant flow (VRF) systems. Changes in energy consumption provide an intuitive representation of the operating condition and the impact of possible faults. Energy consumption prediction is fundamental in energy conservation tasks such as fault diagnosis. Currently, many energy consumption prediction model have been applied for HVAC system widely. However black-box energy prediction models applied to VRF systems are still relatively few and crude. This study therefore proposed a hybrid energy consumption prediction method for VRF systems based on data partitioning and swarm intelligence algorithm. The model was based on back propagation neural network (BPNN), and utilized data partitioning techniques to identify the energy consumption patterns of the VRF system. For each pattern, a BPNN sub-model was trained using operating and energy consumption data of a VRF system. Then, swarm intelligence algorithm is adopted to determine the optimal architecture of each BPNN sub-model. The results demonstrate that the proposed hybrid models can achieve better prediction performance than single models like BP Neural Network. Root mean square error and mean absolute percentage error of the proposed predictive model is 202.49 and 1.95%, which has a 31.3% and 45.5% decrease compared to single BPNN model. At the same time, it is enlightening for other researchers to study the potential of data partitioning algorithm and swarm intelligence algorithm in the field of VRF system energy consumption prediction.