DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT
The article examines the application of digital twins as a tool for modernizing building energy management aimed at electricity consumption forecasting and improving energy efficiency. Based on a review of contemporary approaches, the study analyzes how the integration of BIM/semantic models, IoT da...
Збережено в:
| Дата: | 2026 |
|---|---|
| Автор: | |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
General Energy Institute of the National Academy of Sciences of Ukraine
2026
|
| Теми: | |
| Онлайн доступ: | https://systemre.org/index.php/journal/article/view/962 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | System Research in Energy |
| Завантажити файл: |  |
Репозитарії
System Research in Energy| _version_ | 1871104454775799808 |
|---|---|
| author | Shpak, Denys |
| author_facet | Shpak, Denys |
| author_institution_txt_mv | [
{
"author": "Denys Shpak",
"institution": null
}
] |
| author_sort | Shpak, Denys |
| baseUrl_str | https://systemre.org/index.php/journal/oai |
| collection | OJS |
| datestamp_date | 2026-07-18T12:57:51Z |
| description | The article examines the application of digital twins as a tool for modernizing building energy management aimed at electricity consumption forecasting and improving energy efficiency. Based on a review of contemporary approaches, the study analyzes how the integration of BIM/semantic models, IoT data, and operational systems with machine learning methods enables the implementation of a closed-loop control framework. Methods for electricity consumption forecasting are systematized according to time horizons and their suitability for operational tasks. Typical scenarios for improving energy efficiency through the use of digital twins are presented. A methodological framework for the implementation of digital twins in buildings is proposed, taking into account measurement and verification, model drift, interoperability, and cybersecurity. The impact of digital twins on the quality of managerial decision-making is identified, particularly in enhancing energy consumption controllability and enabling a transition to proactive control of building engineering systems based on predictive models and scenario analysis, as well as in creating conditions for the development of more flexible and sustainable building energy systems. The practical implementation cycle of digital twins in building energy management is generalized. Key limitations of digital twin deployment in buildings are outlined. The dependence of digital twin effectiveness on data consistency and the correctness of energy management problem formulation is demonstrated. |
| doi_str_mv | 10.15407/srenergy2026.02.100 |
| first_indexed | 2026-05-30T01:00:11Z |
| format | Article |
| fulltext |
© Shpak D., 2026
Це стаття відкритого доступу за ліцензією CC0 1.0 Universal
https://creativecommons.org/publicdomain/zero/1.0
100 ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86)
https://doi.org/10.15407/srenergy2026.02.100
UDC 004.94.621.311.697(477)
Denys Shpak, https://orcid.org/0009-0000-7698-1982
General Energy Institute of NAS of Ukraine, 172, Antonovycha St., Kyiv, 03150, Ukraine
e-mail: den.shpak.98@gmail.com
___________________________________________________________________________________________________
DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR
ELECTRICITY CONSUMPTION FORECASTING AND ENERGY
EFFICIENCY IMPROVEMENT
Abstract. The article examines the application of digital twins as a tool for modernizing building energy
management aimed at electricity consumption forecasting and improving energy efficiency. Based on a review
of contemporary approaches, the study analyzes how the integration of BIM/semantic models, IoT data, and
operational systems with machine learning methods enables the implementation of a closed-loop control
framework. Methods for electricity consumption forecasting are systematized according to time horizons and
their suitability for operational tasks. Typical scenarios for improving energy efficiency through the use of
digital twins are presented. A methodological framework for the implementation of digital twins in buildings is
proposed, taking into account measurement and verification, model drift, interoperability, and cybersecurity.
The impact of digital twins on the quality of managerial decision-making is identified, particularly in enhancing
energy consumption controllability and enabling a transition to proactive control of building engineering
systems based on predictive models and scenario analysis, as well as in creating conditions for the development
of more flexible and sustainable building energy systems. The practical implementation cycle of digital twins
in building energy management is generalized. Key limitations of digital twin deployment in buildings are
outlined. The dependence of digital twin effectiveness on data consistency and the correctness of energy
management problem formulation is demonstrated.
Keywords: digital twin, building, electricity consumption forecasting, energy efficiency, IoT, BMS/EMS,
machine learning, optimization.
1. Introduction
Building energy consumption is shaped by the complex interaction of engineering systems, external
conditions, and user behavior. The main consumers of electricity include HVAC systems, lighting, elevator
equipment, and IT infrastructure, while weather conditions, operational schedules, and patterns of space usage
determine temporal variability of energy demand. Under such conditions, the application of static control strategies
proves insufficiently effective, as it does not account for process dynamics and limits the potential for energy
savings without compromising comfort. In this context, digital technologies that enable the integration of data,
models, and decision-making algorithms into a unified information environment are becoming increasingly
relevant [1].
According to international energy agencies, buildings account for approximately 30–40 % of global final
energy consumption and represent a significant source of greenhouse gas emissions. Therefore, improving building
energy efficiency is one of the key directions in the development of sustainable energy systems. Modern
approaches to building management increasingly rely on digital technologies such as the Internet of Things (IoT),
energy management systems, big data analytics, and artificial intelligence [2]. One of the most promising tools for
the digitalization of building operation in this context is digital twin (DT) technology, which involves the creation
of a dynamic digital model of a physical asset synchronized with the real environment through data streams from
sensors, meters, and building management systems. Such a model not only reflects the current state of the asset
mailto:den.shpak.98@gmail.com
ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86) 101
but also enables forecasting, scenario analysis, and optimization of system operation [3], integrating geometric,
semantic, and physical models with analytical algorithms and decision-support tools, which distinguishes it from
traditional BIM models or BMS/SCADA systems [5].
One of the key functions of a digital twin in energy management is electricity consumption forecasting,
which enables advance planning of equipment operation, optimization of HVAC schedules, reduction of peak
loads and energy costs, and implementation of demand response mechanisms. In addition, predictive models can
be used for the integration of local renewable energy sources, optimization of energy storage operation, and early
detection of anomalies in engineering systems. Contemporary studies employ various approaches to building
energy consumption forecasting, including statistical methods, machine learning algorithms, and deep neural
networks. Their integration with digital twins forms the basis for adaptive energy management, in which control
decisions for building systems are made based on forecasts, data analysis, and optimization algorithms [4].
The aim of this study is to analyze current approaches to the application of digital twin technology for
electricity consumption forecasting and improving building energy efficiency, as well as to develop
methodological provisions for their implementation in the operation of building engineering systems.
To achieve this aim, the following main tasks are addressed: (1) systematization of the concept and
architecture of a building digital twin; (2) analysis of electricity consumption forecasting methods and their
suitability for different time horizons; (3) investigation of the potential of digital twins for improving energy
efficiency; (4) development of methodological provisions for implementing a digital twin based on a closed-loop
energy management approach.
2. Method and materials
The study is conducted as a review-based analysis of contemporary approaches to the application of digital
twins in building energy management systems. The research methodology is based on the systematization of
findings from scientific publications addressing digital twin architectures, energy consumption forecasting
methods, and optimization algorithms for building engineering systems. The digital twin is considered as an
integration platform for data acquisition, modeling, forecasting, and optimization of building operation modes.
Within such a system, real-time operational data from engineering systems are continuously used to update the
digital model, while the results of forecasting and optimization are applied to support control decisions, integrating
models, data, and control algorithms into a unified information system [5, 6].
The analysis of the digital twin is carried out with consideration of its key functional components, which
form a sequential logic of data processing and decision-making. The first component is the building representation,
which includes the use of Building Information Modeling (BIM), semantic descriptions of engineering systems,
and technical attributes of equipment. This enables the creation of a structured digital representation of the asset
and the establishment of relationships between spaces, zones, and engineering systems. On this basis, operational
data from multiple sources are integrated, including smart electricity meters, BMS/SCADA systems, IoT sensors,
and external information services providing weather data or occupancy schedules. Their integration ensures a
comprehensive view of building operation and forms the foundation for developing predictive models [7].
Subsequently, the digital twin incorporates electricity consumption forecasting models, including statistical
methods, machine learning algorithms, deep learning models, and hybrid approaches that combine data-driven
techniques with physical building models. The selection of a specific model type depends on data availability,
forecasting horizons, and system complexity [8]. The forecasting results are then used by control and optimization
algorithms to determine efficient operating modes of engineering systems. These include rule-based approaches,
mathematical optimization methods, Model Predictive Control (MPC), and reinforcement learning, enabling
adaptive building operation under changing conditions [9].
The final stage involves model verification and refinement procedures, including the evaluation of
forecasting accuracy, analysis of the energy impact of control actions, and periodic model retraining in response
102 ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86)
to changes in building operation. In modern energy management systems, these processes are often implemented
within the Measurement and Verification (M&V) framework, which allows for objective assessment of the
effectiveness of energy efficiency measures [10].
3. Literature review
The concept of a building digital twin and its distinction from related approaches. A digital twin is
defined as a dynamic digital representation of a physical object that reflects its structure, parameters, and current
state, and maintains continuous synchronization with the real environment through data streams from sensors and
information systems. In the context of buildings, such a model integrates various types of data and modeling
approaches, forming a comprehensive digital environment for monitoring, forecasting, and optimizing the
operation of engineering systems. Unlike traditional building information models, which are primarily used during
the design and construction phases, a digital twin is mainly focused on the operational phase of the building
lifecycle and involves continuous interaction with real-time operational data [11].
The main difference between a digital twin and building automation systems such as BMS or SCADA lies
in the presence of an integrated digital model of the object and enhanced analytical capabilities. While such
systems provide monitoring of equipment parameters and perform operational control functions, they typically do
not support advanced forecasting or scenario-based simulation. In contrast, a digital twin combines geometric,
physical, and behavioral models with data analysis and optimization algorithms, enabling a more comprehensive
approach to building management. An effective digital twin should ensure the integration of different types of
information, including geometric models of the building, characteristics of engineering systems, streams of
operational data, and analytical algorithms. Such integration forms the foundation for next-generation energy
management systems capable of adapting equipment operation modes to changing building conditions [7].
Levels and data of a building digital twin. A multi-level representation of a building allows it to be
considered as an information system in which different types of models and data are integrated to reflect the
structure, functioning, and operational characteristics of the object. At the basic level, a geometric representation
of the building is formed, describing its spatial structure, the arrangement of rooms, floors, and functional zones,
and typically based on BIM models or other digital architectural descriptions.
The semantic and physical descriptions of a building constitute key layers of the digital twin, complementing
its geometric representation. The semantic layer includes information about equipment, engineering systems, and
the relationships between them; such models enable linking system components to corresponding building zones
and play an important role in data integration and interoperability between information systems [12]. The physical
modeling layer, in turn, covers thermal, ventilation, and electrical processes within the building. Depending on the
task, these models may be based on simplified energy approaches or detailed simulations that allow evaluation of
thermal inertia, heat losses through the building envelope, and interactions between engineering systems. This
structure is further complemented by the behavioral layer, which accounts for building usage patterns, occupancy
schedules, and user behavior ‒ factors that significantly influence energy consumption and improve the accuracy
of predictive models [13].
The integration of operational data from various sources ensures the functioning of all these layers of the
digital twin and forms the basis for its analytical capabilities. The main data sources include smart electricity
meters, which generate detailed time series of consumption at both the building level and the level of individual
systems or zones [14], as well as sensor data from engineering systems, including temperatures, airflow rates,
equipment status, and operating parameters. Contextual data also play an important role, particularly weather
conditions, calendar factors, and occupancy indicators, all of which directly affect thermal loads and operating
modes. The coordinated use of these data sources enables the creation of a comprehensive digital representation
of the building and supports the development of accurate energy consumption forecasting models [15].
ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86) 103
The final layers of a digital twin are the control and economic levels. The control level encompasses
automation system logic, optimization algorithms, and equipment control strategies, while the economic level is
associated with tariff analysis, operational costs, and the evaluation of energy efficiency measures.
Considering the described components, a building digital twin can be appropriately viewed as a multi-
layered information system in which functional levels are organized as a sequence of data processing stages. At
the initial stage, data are collected from sensors, smart meters, and building automation systems. These data are
then transmitted to the integration layer, where they undergo cleaning, synchronization, and preparation for further
analysis.
The next stage involves the formation of the modeling layer, where physical and statistical models are
combined with machine learning algorithms. On this basis, energy consumption forecasting and the analysis of
possible building operation scenarios are performed. The obtained results are used to generate control decisions
and optimize the operating modes of engineering systems.
The final element of the architecture consists of monitoring and diagnostic modules that enable the
assessment of deviations between predicted and actual values, as well as user interaction interfaces, including
analytical dashboards, visualization systems, and software APIs for integration with BMS or EMS [16].
Forecast horizons and their role in energy management systems. Electricity consumption forecasting
enables the assessment of the future state of a building’s energy system and supports the formation of optimal
control decisions. In energy management systems, load forecasting is used for planning the operation modes of
engineering systems, managing peak loads, optimizing equipment schedules, and integrating renewable energy
sources [17].
Building electricity consumption forecasting is commonly classified according to the time horizon of the
prediction. Short-term forecasts, covering intervals from several minutes to a few hours, are used for real-time
control of engineering systems, particularly HVAC systems, as well as for detecting anomalies in equipment
operation. Forecasts with horizons ranging from several hours to up to two days are applied to optimize equipment
schedules, implement Demand Response mechanisms, and prepare for potential peak loads in the energy system
[18].
Medium-term forecasts, covering periods from several days to weeks, are used for analyzing building
operation modes and evaluating different scenarios of engineering system usage. In contrast, long-term forecasts,
which may span months or entire seasons, are applied for energy planning, assessing the effectiveness of
retrofitting measures, and forming energy consumption budgets [20].
Classes of electricity consumption forecasting methods. Models for forecasting building electricity
consumption can be broadly classified into statistical methods, machine learning algorithms, and deep learning
models. Each of these approaches has its own advantages and limitations, which determine its area of application.
Statistical methods are a traditional tool for time series forecasting of energy consumption. These include
autoregressive integrated moving average (ARIMA) models. The main advantages of statistical models are their
relative simplicity and interpretability; however, they have a limited ability to capture complex nonlinear
relationships between system parameters [22].
Machine learning algorithms demonstrate better performance in energy consumption forecasting tasks when
large volumes of data are available. Commonly used models include Random Forest, Gradient Boosting, XGBoost,
and LightGBM, as well as Support Vector Regression. These algorithms perform well with tabular data and allow
the incorporation of a wide range of factors, including weather parameters, time-related indicators, and building
operation characteristics.
In recent years, deep learning models have become increasingly widespread due to their ability to effectively
capture complex temporal dependencies. These include recurrent neural networks, particularly LSTM and GRU
architectures, as well as convolutional neural networks and transformer-based models for time series analysis.
104 ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86)
Such approaches demonstrate high forecasting accuracy in multi-step energy consumption prediction tasks,
although they require large training datasets and significant computational resources [23].
A separate group consists of physical and hybrid models. Physical approaches are based on thermal or
energy models of a building, taking into account the properties of the building envelope, thermal inertia, and
characteristics of engineering systems. Hybrid models combine physical principles with machine learning
algorithms, which makes it possible to improve forecasting accuracy while maintaining interpretability of the
results [1].
The classes of electricity consumption forecasting methods, along with their characteristics and limitations,
are summarized in Table 1.
Table 1. Classes of Electricity Consumption Forecasting Methods
Class of Methods Typical Models Advantages Limitations
Statistical ARIMA, SARIMA Simplicity, interpretability
Limited ability to model
nonlinear relationships
Machine Learning Random Forest, XGBoost, SVR
High accuracy, ability to handle
multiple features
Data requirements, lower
interpretability
Deep Learning
LSTM, GRU, CNN,
Transformer
Modeling complex temporal
dependencies
High data and computational
requirements
Physical Building energy models
Interpretability, physical
consistency
Complexity of calibration
Hybrid Physical + Machine Learning
Balance between accuracy and
interpretability
Implementation complexity
Input features for forecasting models. The accuracy of electricity consumption forecasting largely
depends on the set of input parameters used for model training. Forecasting models typically rely on a combination
of temporal, weather-related, and operational characteristics of the building.
The integration of temporal features, weather factors, and operational parameters enables the formation of
multidimensional feature sets, which provide a more accurate representation of building energy behavior and
improve the performance of predictive models. Temporal features include hour of the day, day of the week,
seasonal indicators, and other calendar-related parameters that reflect cyclical patterns of building usage. Weather
factors ‒ such as outdoor air temperature, relative humidity, wind speed, and solar radiation ‒ have a significant
impact on energy consumption levels. To account for the thermal inertia of buildings, lagged values of these
parameters are often incorporated into the models [17].
Occupancy indicators, such as building usage levels or user activity data obtained from access control
systems or wireless networks, are also important inputs. Additional features may include operational parameters
of engineering systems, such as temperature setpoints, equipment operating modes, or temperatures in specific
building zones [23].
The main groups of input features used in electricity consumption forecasting models are summarized in
Table 2.
Forecast accuracy evaluation metrics. The evaluation of forecasting model accuracy makes it possible to
determine the suitability of a model for practical application in building control systems. In electricity consumption
forecasting tasks, the most commonly used metrics are Mean Absolute Error (MAE) and Root Mean Squared Error
(RMSE), which quantify the deviation between predicted and actual values in physical units [24].
ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86) 105
Table 2. Types of Input Features for Electricity Consumption Forecasting Models
Feature Group Examples Characteristics
Temporal
Hour of the day, day of the week,
seasonal indicators
Reflect periodicity and cyclic patterns of consumption; enable
modeling of daily and seasonal trends
Weather
Temperature, humidity, wind, solar
radiation
Determine building thermal load; have a nonlinear impact on
energy consumption
Operational
Operating modes, temperature
setpoints, system parameters
Directly related to the control of engineering systems; reflect
current operating conditions
Behavioral
Occupancy, access data, Wi-Fi
activity
Account for user influence; may be noisy and incomplete
Lagged Previous values of parameters Allow capturing building inertia and dependence on past states
To compare models across different data scales, relative metrics are used, particularly Mean Absolute
Percentage Error (MAPE) and its symmetric modification (sMAPE). In the case of probabilistic forecasting
models, specialized metrics such as pinball loss or Continuous Ranked Probability Score (CRPS) are applied,
allowing the evaluation of the accuracy of predicted value distributions [25].
In energy management systems, particular attention is paid to the accuracy of peak load forecasting, as peak
values often determine electricity costs. Therefore, many studies complement general metrics with additional
indicators that assess the accuracy of peak load predictions.
4. Application of a digital twin for improving the energy efficiency of buildings
Unlike traditional automation systems, which operate based on fixed rules or static schedules, a digital twin
‒ through the integration of operational data, forecasting models, and optimization algorithms ‒ creates new
opportunities for improving energy efficiency and enabling adaptive control of engineering systems based on
energy consumption forecasts and real-time building conditions. The integration of forecasting models with digital
twins allows for the optimization of building electrical systems, reduction of peak loads, integration of renewable
energy sources and storage systems, as well as diagnostics and predictive maintenance. Owing to its ability to
simulate different operational scenarios, a digital twin can be used not only for real-time control but also for
evaluating the effectiveness of energy efficiency measures and engineering system upgrades [26, 4].
The main applications of digital twins in building energy efficiency are summarized in Table 3.
Table 3. Main Application Areas of Digital Twins
Application Area Description Typical Outcomes
HVAC Operation Optimization Control of temperature setpoints and ventilation
based on load forecasts, occupancy, and building
thermal inertia
Reduction in HVAC energy
consumption
Peak shaving and load shifting Forecasting peak loads and smoothing them by
adjusting equipment operation schedules
Reduction in peak demand and
electricity costs
Demand Response Adaptation of consumption to energy market
signals or DR events
Increased building flexibility
Integration of Storage and Renewables Optimization of PV systems and battery
operation based on generation and load forecasts
Increased self-consumption and reduced
energy purchases
Fault Detection and Diagnosis (FDD) Comparison of predicted and actual system
behavior
Early detection of faults and equipment
degradation
Scenario Analysis for Retrofits Evaluation of energy efficiency measures
through simulation
Forecast of energy savings and
assessment of payback period
One of the most common applications of a digital twin is the optimization of HVAC system operation. By
using load forecasts and information about the building’s thermal inertia, the system can implement strategies such
106 ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86)
as pre-cooling or pre-heating during periods of lower electricity prices. This approach reduces energy consumption
while maintaining comfortable conditions for building occupants [2].
Since, in many tariff structures, a significant portion of electricity costs is associated with peak demand,
peak load management is an important direction for improving energy efficiency. Through load forecasting
capabilities, a digital twin enables the implementation of peak shaving and load shifting strategies, which involve
adjusting equipment operation schedules and redistributing loads over time.
Digital twins are also widely used for integrating renewable energy sources and energy storage systems.
The combination of solar generation forecasts with building load forecasts allows for the optimization of battery
charging and discharging schedules, reducing electricity consumption from the grid during peak hours [5, 9].
In addition to energy optimization tasks, a digital twin can be used for fault detection and predictive
maintenance of equipment. In such systems, an expected behavior of engineering systems is established, after
which deviations between predicted and actual parameter values are analyzed. Significant deviations may indicate
sensor faults, incorrect setpoints, or equipment degradation, enabling timely maintenance actions [6].
Another promising application is the use of digital twins for scenario-based analysis of retrofits. Through
simulations in a digital environment, it is possible to evaluate the impact of various energy efficiency measures,
such as HVAC system upgrades, lighting replacement, building envelope insulation, or the installation of
renewable energy systems. This approach allows for the assessment of economic efficiency of investments prior
to their actual implementation [1, 5].
5. Results
The implementation of a digital twin in building energy management systems is a continuous process that
integrates operational data collection, modeling, forecasting, optimization of engineering system operation, and
evaluation of control results. In modern research and practical applications, this approach is described as a closed-
loop control process, where the results of analysis and forecasting are used to update models and improve control
strategies [14].
The methodological framework for digital twin implementation proposed in this study is based on a
sequence of interconnected stages that form a closed-loop energy management cycle for a building.
The initial stage involves defining energy management objectives and key performance indicators (KPIs).
These may include specific energy consumption, peak demand, electricity costs, thermal comfort parameters, or
greenhouse gas emissions [10]. At this stage, system constraints are also defined, such as acceptable indoor
temperature ranges or equipment operating limits.
The next step is the inventory of available data and building information systems. In practice, this includes
assessing the availability of smart meters, sub-metering systems, sensors of engineering systems, and building
automation systems (BMS or SCADA). An important aspect is also the evaluation of historical data availability
and the possibility of integrating external data sources, such as weather services or calendar data [14].
This is followed by data preparation and processing, which includes time series synchronization, handling
of missing values, anomaly detection, and validation of sensor performance. Data quality is a critical factor for
developing reliable forecasting models and ensuring the effective operation of the digital twin.
Based on the prepared data, a digital model of the building is developed, forming the core of the digital
twin. At this stage, the building zoning structure is defined, engineering systems are described, and relationships
between model elements and data flows are established. Basic physical constraints are also incorporated, such as
thermal properties of the building envelope or maximum equipment capacity.
The next stage involves training and calibration of electricity consumption forecasting models. This includes
defining the forecasting horizon, selecting input features, and training models using historical data. The resulting
models are evaluated using standard forecasting accuracy metrics and then integrated into the digital twin.
ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86) 107
After model development, the digital twin is integrated with building control systems such as BMS or EMS.
At the initial stages, the digital twin may operate in an advisory mode, where it provides recommendations for
optimal equipment operation while the final decision is made by a human operator. Subsequently, a transition to
semi-automated or fully automated control is possible, provided that established safety constraints are satisfied.
The obtained forecasts are used to optimize the operation modes of engineering systems, in particular for
determining optimal temperature setpoints, equipment schedules, or strategies for building participation in
Demand Response programs. The implementation of such decisions makes it possible to reduce energy
consumption and optimize electricity costs [7].
After the implementation of control actions, measurement and verification of the energy effect are carried
out. At this stage, actual energy consumption values are compared with baseline values, taking into account
external factors such as weather conditions or changes in building usage patterns. In addition to energy indicators,
occupant comfort parameters are also monitored.
The final stage of the cycle involves model correction and continuous system learning. During building
operation, changes may occur in space usage patterns, equipment upgrades, or seasonal operating conditions.
Therefore, forecasting models and control algorithms should be periodically updated and retrained using new data
[3, 4].
Based on these stages, a digital twin operation cycle is formed, in which data, models, and control decisions
continuously interact. Such an approach enables adaptive energy management of buildings and provides a
foundation for scaling digital twin technology to building portfolios or urban energy systems.
A generalized structural scheme of the closed-loop operation of a building energy management system is
shown in Fig. 1.
Figure 1. Closed-loop energy management cycle of a building
108 ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86)
6. Conclusions
This paper examined the application of digital twin technology for electricity consumption forecasting and
improving building energy efficiency. The analysis demonstrates that the integration of digital models, data
acquisition systems, forecasting algorithms, and optimization methods creates new opportunities for efficient
energy management in buildings. A digital twin acts as an integration platform that connects the physical
infrastructure of a building with its digital model, enabling analysis of the current system state, forecasting of
future loads, and generation of optimal control decisions. This enables the transition from traditional static control
of engineering systems to adaptive management that accounts for changing operating conditions, weather factors,
and occupant behavior.
The study systematized the main approaches to electricity consumption forecasting, including statistical
methods, machine learning algorithms, deep neural networks, physics-based models, and hybrid approaches. It
was shown that the choice of a specific method depends on data availability, system complexity, and the required
forecasting horizon. Hybrid and physics-informed models are particularly promising, as they combine the
interpretability of physical models with the accuracy of modern machine learning techniques.
The architecture of a digital twin for building energy management systems was also analyzed. A multi-layer
structure was proposed, including data acquisition, integration and processing, modeling, forecasting, control
optimization, as well as monitoring and diagnostics layers. Such an architecture ensures a complete operational
cycle of the digital twin ‒ from data collection to decision-making for optimal system operation.
The key result of this work is the development of a methodological framework for digital twin
implementation, presented as a closed-loop cycle. This framework enables the systematization and standardization
of the building energy management digitalization process by proposing nine clearly defined and interconnected
stages ‒ from KPI formulation to continuous model retraining. Its practical value lies in serving as a universal
roadmap for engineers and energy managers, helping to minimize technical risks associated with integrating
heterogeneous systems and ensuring transparency in decision-making throughout the building lifecycle.
Future work involves transitioning from theoretical analysis to practical validation of the proposed
framework using a pilot building. This includes data inventory and preparation, development and calibration of
hybrid forecasting models based on real operational data, and integration of these models into a decision-support
system prototype to evaluate the actual effectiveness of optimization algorithms under dynamically changing
external conditions.
The conducted analysis showed that the use of a digital twin can provide significant energy and economic
benefits through the optimization of HVAC system operation, reduction of peak loads, implementation of demand
response mechanisms, and integration of local renewable energy sources and storage systems. Additional
advantages include early detection of equipment faults, predictive maintenance, and scenario-based analysis of
energy efficiency upgrades.
At the same time, the implementation of digital twins is associated with a number of technical and
organizational challenges. The main issues include data quality and availability, the complexity of integrating
heterogeneous information systems, the need for model calibration for different types of buildings, as well as
cybersecurity and data protection concerns. In addition, the economic effectiveness of implementation largely
depends on the proper definition of energy management objectives and performance indicators.
It has been established that the use of a digital twin can provide significant energy benefits through HVAC
system optimization, peak load reduction, and the implementation of Demand Response strategies.
References
1. Zhao, H.-X., & Magoulès, F. (2012). A review on the prediction of building energy consumption. Renewable and
Sustainable Energy Reviews, 16(6), 3586–3592. https://doi.org/10.1016/j.rser.2012.02.049
2. de las Morenas, J., Belmonte, L. M., & Morales, R. (2025). Designing an AI-driven digital twin architecture for building
energy prediction. Journal of Building Engineering, 113, 113966. https://doi.org/10.1016/j.jobe.2025.113966
ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86) 109
3. Mariano-Hernández, D., Hernández-Callejo, L., Zorita-Lamadrid, A., Duque-Pérez, O., & Santos García, F. (2021). A
review of strategies for building energy management system: Model predictive control, demand side management,
optimization, and fault detect & diagnosis. Journal of Building Engineering, 33, 101692.
https://doi.org/10.1016/j.jobe.2020.101692
4. Sghiri, A., Gallab, M., Merzouk, S., & Assoul, S. (2025). Leveraging digital twins for enhancing building energy
efficiency: A review. Buildings, 15(3), 498. Retrieved March 12, 2026, from https://www.mdpi.com/2075-
5309/15/3/498
5. Bourdeau, M., Zhai, X. Q., Nefzaoui, E., Guo, X., & Chatellier, P. (2019). Modeling and forecasting building energy
consumption: A review of data-driven techniques. Sustainable Cities and Society, 48, 101533.
https://doi.org/10.1016/j.scs.2019.101533
6. Valenzuela, P. E., Ebadat, A., Everitt, N., & Parisio, A. (2020). Closed-loop identification for model predictive control
of HVAC systems: From input design to controller synthesis. IEEE Transactions on Control Systems Technology, 28(5),
1681–1695. https://doi.org/10.1109/tcst.2019.2917675
7. Jain, A., Nong, D., Nghiem, T., & Mangharam, R. (2018). Digital twins for efficient modeling and control of buildings:
An integrated solution with SCADA systems. Proceedings of the 2018 Building Performance Analysis Conference and
SimBuild (pp. 799–806). IBPSA-USA / ASHRAE. Retrieved March 12, 2026, from
https://publications.ibpsa.org/proceedings/simbuild/2018/papers/simbuild2018_C110.pdf
8. Wei, Y., Zhang, X., Shi, Y., Xia, L., Pan, S., Wu, J., Han, M., & Zhao, X. (2018). A review of data-driven approaches
for prediction and classification of building energy consumption. Renewable and Sustainable Energy Reviews, 82, 1027–
1047. https://doi.org/10.1016/j.rser.2017.09.108
9. Mariano-Hernández, D., Hernández-Callejo, L., Zorita-Lamadrid, A., Duque-Pérez, O., & Santos García, F. (2021). A
review of strategies for building energy management system: Model predictive control, demand side management,
optimization, and fault detect & diagnosis. Journal of Building Engineering, 33, 101692.
https://doi.org/10.1016/j.jobe.2020.101692
10. Tanguay, D. (n.d.). International Performance Measurement and Verification Protocol (IPMVP). Organization (EVO).
Retrieved March 12, 2026, from https://evo-world.org/en/products-services-mainmenu-en/protocols/ipmvp
11. Fuller, A., Fan, Z., Day, C., & Barlow, C. (2020). Digital twin: Enabling technologies, challenges and open research.
IEEE Access: Practical Innovations, Open Solutions, 8, 108952–108971. https://doi.org/10.1109/access.2020.2998358
12. Boje, C., Guerriero, A., Kubicki, S., & Rezgui, Y. (2020). Towards a semantic Construction Digital Twin: Directions
for future research. Automation in Construction, 114, 103179. https://doi.org/10.1016/j.autcon.2020.103179
13. Jeong, D.-Y., Baek, M.-S., Lim, T.-B., Kim, Y.-W., Kim, S.-H., Lee, Y.-T., Jung, W.-S., & Lee, I.-B. (2022). Digital
twin: Technology evolution stages and implementation layers with technology elements. IEEE Access: Practical
Innovations, Open Solutions, 10, 52609–52620. https://doi.org/10.1109/access.2022.3174220
14. Yoon, S. (2023). Building digital twinning: Data, information, and models. Journal of Building Engineering, 76, 107021.
https://doi.org/10.1016/j.jobe.2023.107021
15. Mathumitha, R., Rathika, P., & Manimala, K. (2024). Intelligent deep learning techniques for energy consumption
forecasting in smart buildings: a review. Artificial Intelligence Review, 57(2). https://doi.org/10.1007/s10462-023-
10660-8
16. Ahmad, T., Zhang, D., Huang, C., Zhang, H., Dai, N., Song, Y., & Chen, H. (2021). Artificial intelligence in sustainable
energy industry: Status Quo, challenges and opportunities. Journal of Cleaner Production, 289, 125834.
https://doi.org/10.1016/j.jclepro.2021.125834
17. Deb, C., Zhang, F., Yang, J., Lee, S. E., & Shah, K. W. (2017). A review on time series forecasting techniques for
building energy consumption. Renewable and Sustainable Energy Reviews, 74, 902–924.
https://doi.org/10.1016/j.rser.2017.02.085
18. Choo, K., Galante, R. M., & Ohadi, M. M. (2014). Energy consumption analysis of a medium-size primary data center
in an academic campus. Energy and Buildings, 76, 414–421. https://doi.org/10.1016/j.enbuild.2014.02.042
19. Ahmad, T., & Chen, H. (2018). Short and medium-term forecasting of cooling and heating load demand in building
environment with data-mining based approaches. Energy and Buildings, 166, 460–476.
https://doi.org/10.1016/j.enbuild.2018.01.066
20. Hahn, H., Meyer-Nieberg, S., & Pickl, S. (2009). Electric load forecasting methods: Tools for decision making.
European Journal of Operational Research, 199(3), 902–907. https://doi.org/10.1016/j.ejor.2009.01.062
21. Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles and Practice (2nd ed.). Retrieved March 12,
2026, from https://otexts.com/fpp2/
22. Henzel, J., Wróbel, Ł., Fice, M., & Sikora, M. (2022). Energy consumption forecasting for the digital-twin model of the
building. Energies, 15(12), 4318. https://doi.org/10.3390/en15124318
23. Ahmad, M. W., Mourshed, M., & Rezgui, Y. (2017). Trees vs Neurons: Comparison between random forest and ANN
for high-resolution prediction of building energy consumption. Energy and Buildings, 147, 77–89.
https://doi.org/10.1016/j.enbuild.2017.04.038
https://www.mdpi.com/2075-5309/15/3/498
https://www.mdpi.com/2075-5309/15/3/498
110 ISSN 2786-7633. Системні дослідження в енергетиці. 2026. 2(86)
24. Willmott, C. J., & Matsuura, K. (2005). Advantages of the mean absolute error (MAE) over the root mean square error
(RMSE) in assessing average model performance. Climate Research, 30, 79–82. https://doi.org/10.3354/cr030079
25. Gneiting, T., & Raftery, A. E. (2007). Strictly proper scoring rules, prediction, and estimation. Journal of the American
Statistical Association, 102(477), 359–378. https://doi.org/10.1198/016214506000001437
26. Hosamo, H., Hosamo, M. H., Nielsen, H. K., Svennevig, P. R., & Svidt, K. (2022). Digital Twin of HVAC system
(HVACDT) for multiobjective optimization of energy consumption and thermal comfort based on BIM framework with
ANN-MOGA. Advances in Building Energy Research, 17(2), 125–171.
https://doi.org/10.1080/17512549.2022.2136240
ЦИФРОВІ ДВІЙНИКИ В ЕНЕРГОМЕНЕДЖМЕНТІ БУДІВЕЛЬ
ДЛЯ ПРОГНОЗУВАННЯ ЕЛЕКТРОСПОЖИВАННЯ ТА
ПІДВИЩЕННЯ ЕНЕРГОЕФЕКТИВНОСТІ
Денис Шпак, https://orcid.org/0009-0000-7698-1982
Інститут загальної енергетики НАН України, вул. Антоновича, 172, Київ, 03150, Україна
e-mail: den.shpak.98@gmail.com
Анотація. У статті розглянуто застосування цифрового двійника як інструмента модернізації
енергоменеджменту будівель для прогнозування споживання електроенергії та підвищення
енергоефективності. На основі огляду сучасних підходів проаналізовано, як інтеграція
BIM/семантичних моделей, даних IoT та експлуатаційних систем із методами машинного навчання
забезпечує реалізацію «замкненого циклу» керування. Систематизовано методи прогнозування
електроспоживання за часовими горизонтами та придатністю до задач експлуатації. Наведено
типові сценарії підвищення енергоефективності завдяки цифровому двійнику. Запропоновано
концептуальну рамку впровадження цифрового двійника у будівлі з урахуванням вимірювання та
верифікації, дрейфу моделей, інтероперабельності та кібербезпеки. Визначено вплив цифрового
двійника на якість управлінських рішень у покращенні керованості енергоспоживання та переході до
проактивного керування інженерними системами будівлі на основі прогнозних моделей і сценарного
аналізу та створенні передумов для розвитку більш гнучких і сталих енергетичних систем будівель.
Узагальнено практичний цикл впровадження цифрового двійника в енергоменеджменті будівель.
Окреслено ключові обмеження впровадження цифрових двійників у будівлях. Показано залежність
ефективності застосування цифрового двійника від узгодженості даних і коректності постановки
задач енергоменеджменту.
Ключові слова: цифровий двійник, будівля, прогноз електроспоживання, енергоефективність, IoT,
BMS/EMS, машинне навчання, оптимізація.
Дата першого надходження статті до журналу: 06.04.2026
Дата прийняття статті до друку після рецензування:
Дата публікації (оприлюднення): 30.05.2026
mailto:den.shpak.98@gmail.com
|
| id | systemreorg-article-962 |
| institution | System Research in Energy |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2026-07-19T01:24:21Z |
| publishDate | 2026 |
| publisher | General Energy Institute of the National Academy of Sciences of Ukraine |
| record_format | ojs |
| resource_txt_mv | systemreorg/a0/2cb74f25f5ba3f47eb353522cd536aa0.pdf |
| spelling | systemreorg-article-9622026-07-18T12:57:51Z DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT Цифрові двійники в енергоменеджменті будівель для прогнозування електроспоживання та підвищення енергоефективності Shpak, Denys digital twin, building, electricity consumption forecasting, energy efficiency, IoT, BMS/EMS, machine learning, optimization. цифровий двійник, будівля, прогноз електроспоживання, енергоефективність, IoT, BMS/EMS, машинне навчання, оптимізація. The article examines the application of digital twins as a tool for modernizing building energy management aimed at electricity consumption forecasting and improving energy efficiency. Based on a review of contemporary approaches, the study analyzes how the integration of BIM/semantic models, IoT data, and operational systems with machine learning methods enables the implementation of a closed-loop control framework. Methods for electricity consumption forecasting are systematized according to time horizons and their suitability for operational tasks. Typical scenarios for improving energy efficiency through the use of digital twins are presented. A methodological framework for the implementation of digital twins in buildings is proposed, taking into account measurement and verification, model drift, interoperability, and cybersecurity. The impact of digital twins on the quality of managerial decision-making is identified, particularly in enhancing energy consumption controllability and enabling a transition to proactive control of building engineering systems based on predictive models and scenario analysis, as well as in creating conditions for the development of more flexible and sustainable building energy systems. The practical implementation cycle of digital twins in building energy management is generalized. Key limitations of digital twin deployment in buildings are outlined. The dependence of digital twin effectiveness on data consistency and the correctness of energy management problem formulation is demonstrated. У статті розглянуто застосування цифрового двійника як інструмента модернізації енергоменеджменту будівель для прогнозування споживання електроенергії та підвищення енергоефективності. На основі огляду сучасних підходів проаналізовано, як інтеграція BIM/семантичних моделей, даних IoT та експлуатаційних систем із методами машинного навчання забезпечує реалізацію «замкненого циклу» керування. Систематизовано методи прогнозування електроспоживання за часовими горизонтами та придатністю до задач експлуатації. Наведено типові сценарії підвищення енергоефективності завдяки цифровому двійнику. Запропоновано концептуальну рамку впровадження цифрового двійника у будівлі з урахуванням вимірювання та верифікації, дрейфу моделей, інтероперабельності та кібербезпеки. Визначено вплив цифрового двійника на якість управлінських рішень у покращенні керованості енергоспоживання та переході до проактивного керування інженерними системами будівлі на основі прогнозних моделей і сценарного аналізу та створенні передумов для розвитку більш гнучких і сталих енергетичних систем будівель. Узагальнено практичний цикл впровадження цифрового двійника в енергоменеджменті будівель. Окреслено ключові обмеження впровадження цифрових двійників у будівлях. Показано залежність ефективності застосування цифрового двійника від узгодженості даних і коректності постановки задач енергоменеджменту. General Energy Institute of the National Academy of Sciences of Ukraine 2026-05-30 Article Article application/pdf https://systemre.org/index.php/journal/article/view/962 10.15407/srenergy2026.02.100 System Research in Energy; No. 2 (86) (2026): System Research in Energy; 100-110 Системні дослідження в енергетиці; № 2 (86) (2026): Системні дослідження в енергетиці; 100-110 2786-7102 2786-7633 en https://systemre.org/index.php/journal/article/view/962/847 Copyright (c) 2026 Denys Shpak https://creativecommons.org/publicdomain/zero/1.0 |
| spellingShingle | digital twin building electricity consumption forecasting energy efficiency IoT BMS/EMS machine learning optimization. Shpak, Denys DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT |
| title | DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT |
| title_alt | Цифрові двійники в енергоменеджменті будівель для прогнозування електроспоживання та підвищення енергоефективності |
| title_full | DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT |
| title_fullStr | DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT |
| title_full_unstemmed | DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT |
| title_short | DIGITAL TWINS IN BUILDING ENERGY MANAGEMENT FOR ELECTRICITY CONSUMPTION FORECASTING AND ENERGY EFFICIENCY IMPROVEMENT |
| title_sort | digital twins in building energy management for electricity consumption forecasting and energy efficiency improvement |
| topic | digital twin building electricity consumption forecasting energy efficiency IoT BMS/EMS machine learning optimization. |
| topic_facet | digital twin building electricity consumption forecasting energy efficiency IoT BMS/EMS machine learning optimization. цифровий двійник будівля прогноз електроспоживання енергоефективність IoT BMS/EMS машинне навчання оптимізація. |
| url | https://systemre.org/index.php/journal/article/view/962 |
| work_keys_str_mv | AT shpakdenys digitaltwinsinbuildingenergymanagementforelectricityconsumptionforecastingandenergyefficiencyimprovement AT shpakdenys cifrovídvíjnikivenergomenedžmentíbudívelʹdlâprognozuvannâelektrospoživannâtapídviŝennâenergoefektivností |