A building management system and a method for controlling the same
The building management system addresses challenges in HVAC modeling and real-time data integration by using advanced modeling and user feedback, ensuring efficient and adaptive HVAC operation with reduced errors and improved energy efficiency.
Patent Information
- Application Number
- PCT/TR2025/050471
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-12
- Publication Date
- 2026-11-19
AI Technical Summary
Conventional Building Management Systems (BMS) face challenges in accurately modeling thermal behavior, adapting to changing conditions, and integrating real-time data, leading to suboptimal HVAC operation, energy waste, and lack of sophisticated fault detection and diagnostic capabilities.
A building management system that utilizes real-time data and advanced modeling techniques to generate tailored recommendations for HVAC systems, incorporating user feedback and natural language interaction, and supports hybrid calibration for continuous model refinement.
Enables efficient and responsive HVAC management by minimizing errors, optimizing energy usage, and adapting to changing conditions, while providing accurate and actionable insights through integrated large language models.
Smart Images

Figure TR2025050471_19112026_PF_FP_ABST
Abstract
Description
[0001] DESCRIPTION
[0002] A BUILDING MANAGEMENT SYSTEM AND A METHOD FOR CONTROLLING THE SAME
[0003] FIELD OF THE INVENTION
[0004] The present invention relates generally to building management system and a control method for said building management system. The present invention relates more specifically to an improved data communication for the building management system.
[0005] BACKGROUND
[0006] Building Management Systems (BMS) are designed to track and coordinate various systems within a building, ensuring efficient operation and control. They are particularly beneficial for managing HVAC (heating, ventilation, and air conditioning) systems, which generally comprise multiple HVAC devices distributed across different areas of a building and the other energy consuming members disposed in the building. These systems aim to improve energy efficiency, occupant comfort, and operational effectiveness by automating and optimizing building functions.
[0007] As buildings become more complex and energy efficiency standards more stringent, conventional BMS approaches face several challenges. One issue is the difficulty in accurately modeling the thermal behavior of a building, which can vary significantly based on factors such as occupancy patterns, weather conditions, and equipment performance. This can lead to suboptimal HVAC operation and energy waste. Additionally, static control strategies may not adapt well to changing conditions or unexpected events, potentially compromising comfort or efficiency.The calibration of building models and control systems presents another challenge. Manual calibration processes can be time-consuming, error-prone, and may not capture the full complexity of building dynamics. This can result in models that drift from reality over time, leading to decreased performance and increased energy consumption. Moreover, traditional BMS often lack sophisticated fault detection and diagnostic capabilities, making it difficult to identify and address equipment issues or inefficiencies promptly. Furthermore, many existing BMS struggle to effectively incorporate real-time data from diverse sources, limiting their ability to make informed decisions and respond quickly to changing conditions. Furthermore, existing calibration methods often lack the ability to combine both user feedback and real-time sensor data. Moreover, most approaches do not leverage natural language interaction, standardized semantic tagging, or scalable multi-site coordination, which are increasingly necessary in large building portfolios.
[0008] In the prior art, among others, United States Patent no. US2012232701 discloses a method and system for optimizing building energy usage. The method comprising receiving a plurality of input values associated with a building or plurality of buildings. The method then constructs a thermal and an electrical load model based on the inputs and constructs an overall energy model, the overall energy model being based on the thermal and electrical load models. The method next generates a plurality of demand models and optimizes the demand models using complex multivariate optimization techniques, wherein optimizing is based on usage data and energy rules. Finally, the method displays recommendations based on the optimized model or generating realtime, complementary control instructions based on the optimized model, the determination based on client preferences.
[0009] SUMMARY OF THE INVENTIONThe aspect of the present invention is to provide a building management system enabling management of user requests according to at least one device within a building management system by an efficient and optimized data communication.
[0010] In accordance with the aspect of the invention, claim 1 discloses a building management system comprising at least one device disposed within a building and a data management server having: a system data module configured to obtain condition variables associated with the at least one device and the building, a modeling module configured to generate an actual model based on the at least one condition variable, a calibration module configured to generate a calibrated model by adjusting the at least one condition variable, and a recommendation module configured to generate at least one recommendation related to the at least one condition variable by comparing the calibrated model with a predefined target model. The device may be any energy consuming element in the building such as HVAC element, lighting element, cameras or any other plug load devices. This building management system enables efficient and optimized control of devices by utilizing real-time data and advanced modeling techniques to generate tailored recommendations for improved performance.
[0011] In a preferred embodiment of the present invention, the recommendation module is configured to provide a control instruction to the at least one device based on the at least one recommendation. The recommendation module is connected with an automation system configured to control and monitor at least one working parameter of at least one device. By automating the implementation of recommendations, the system minimizes the potential for errors that may occur during manual input or interpretation of suggested changes. The automated provision of control instructions based on recommendations represents an advanced approach to building management, leveraging data analysis and system intelligence to achieve efficient and responsive operations of the devices.The modeling module comprises a manual editor which allows a user inputs at least one user feedback related to at least one parameter of the digital model. The user feedback may be related with the location of at least one HVAC devices or any other constructional parameters of the actual model. The user feedback may also be related with the digital parameters of the actual model such as actual temperature or actual sensor outputs. This manual editor enables the user to make manual adjustments to the model based on observed system performance. The user feedback ensures that the model reflects any discrepancies between predicted and actual building performance, improving its accuracy. By integrating the user's inputs, the model becomes more realistic and tailored to the specific conditions of the building. As a result, the building management system can manage energy usage more efficiently and provide solutions that are better suited to user needs.
[0012] The calibration subsystem comprises a digital editor which calibrates the actual model by using at least one of an energy consumption data and an ambient sensor data. The energy consumption data helps optimize the model’s energy behavior based on the building’s actual energy usage, while the ambient sensor data takes into account environmental factors such as temperature, humidity, and light levels. This process enables the actual model to be continuously updated with real-time data, ensuring that the building management system can make more accurate decisions. As a result, necessary adjustments can be made to enhance energy efficiency and adapt to changing conditions within the building.
[0013] The building management system implements a method for managing operation of at least one device disposed within a building. This method comprises several steps executed by a data management server to optimize the performance and energy efficiency of devices. The method begins with obtaining condition variables associated with the device and the building. These condition variables may include various parameters such as building characteristics, HVAC equipment specifications,operational data, and environmental measurements. By collecting a comprehensive set of condition variables, the method establishes a foundation for accurate modeling and analysis of the HVAC system. Next, the method comprises generating an actual model based on the obtained condition variables. This actual model represents the current state and performance of the HVAC system within the building. The model incorporates the collected data to create a digital representation of the HVAC system, including its components, operational parameters, and interactions with the building environment. The method then proceeds to generate a calibrated model by adjusting at least one condition variable. This calibration process refines the actual model to improve its accuracy in representing real-world conditions. The calibration may involve iterative adjustments to align the model predictions with observed data from the building and HVAC system. Finally, the method generates at least one recommendation by comparing the calibrated model with a predefined target model. This comparison allows the system to identify areas where the HVAC performance deviates from desired standards or optimal operating conditions. Based on this analysis, the method produces recommendations for improving HVAC efficiency, reducing energy consumption, or enhancing occupant comfort.
[0014] In another aspect, the system integrates a Large Language Model (LLM) that interprets user inputs and contextual data to generate adaptive recommendations. Additionally, the invention supports hybrid calibration using both manual user edits and real-time automation data. Furthermore, it enables region-based energy optimization across multiple sites by aggregating standardized data structures.
[0015] BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The building management system of the present invention is illustrated in the attached figure, where:
[0017] FIG. 1 is a schematic diagram showing the building management system according to an embodiment of the invention.The elements illustrated in the figures are numbered as follows:
[0018] 1. The building management system
[0019] 2. Device
[0020] 10. System data module
[0021] 20. Modeling module
[0022] 30. Calibration module
[0023] 31. Manual editor
[0024] 32. Digital editor
[0025] 40. Recommendation module
[0026] 50. Automation system
[0027] 100. Data management server
[0028] DETAILED DESCRIPTION
[0029] A building management system (1) according to the present invention comprises at least one device (2) disposed within a building and a data management server (100) having a system data module (10) configured to obtain condition variables associated with the at least one device (2) and the building, a modeling module (20) configured to generate an actual model based on the at least one condition variable, a calibration module (30) configured to generate a calibrated model by adjusting the at least one condition variable and a recommendation module (40) configured to generate at least one recommendation related to the at least one condition variable by comparing the calibrated model with a predefined target model.
[0030] The system data module (10) is configured to acquire a set of condition variables associated with both the device (2) and the building. These condition variables may include building location, usage purpose of the building, floor plan of the building,equipment information of the at least one device (2), energy consumption data of the at least one device (2), predefined energy standards, ambient sensor data, energy consumption, airflow rate, compressor speed or fan speed of the at least one device (2). In some embodiments, the condition variables are not limited to the physical or operational characteristics of HVAC equipment, but may further include structured metadata collected from building management systems (BMS), programmable logic controllers (PLCs), or cloud-based loT platforms. These platforms may expose realtime data points such as temperature readings, energy consumption, device status, and configuration parameters via standard protocols (e.g., BACnet, Modbus, MQTT, REST APIs), which are integrated into the system data module for model generation and recommendation evaluation.
[0031] The modeling module (20) generates an actual model based on the condition variables obtained by the system data module (10). In some embodiments, the actual model may be a physics-based building energy model. For example, the actual model may contain three-dimensional coordinates for a plurality of objects or structures that relate to the building's energy consumption, and the actual model may estimate the building's energy consumption based on modeled physical processes relating to these objects and structures. The actual model includes physics-based equations, or other tools that can be used to predict energy consumption for a building. In some embodiments, the actual model may be an EnergyPlus™ model (EnergyPlus software is published by the United States Department of Energy). Other suitable building energy models may be used.
[0032] The calibration module (30) is responsible for generating a calibrated model by adjusting one or more of the condition variables. This calibration process helps to improve model’s accuracy in representing the real-world conditions of the building and HVAC system.The building management system (1) also comprises a recommendation module (40) configured to generate recommendations by comparing the calibrated model with a predefined target model. This comparison-based approach enables the recommendation module (40) to identify areas where the device performance deviates from desired standards or optimal operating conditions. The predefined target model serves as a benchmark for ideal system performance. In some examples, the predefined target model is based on building energy standards, such as ANSI / ASHRAE / IES standards. These standards provide guidelines for energy-efficient building design and operation, incorporating best practices and performance targets for various building systems, including HVAC elements and lighting elements. By utilizing building energy standards as a basis for the predefined target model, the recommendation module (40) can evaluate the calibrated model against industry-recognized benchmarks. This comparison allows the recommendation module (40) to identify specific areas where the building's HVAC system may be underperforming or consuming more energy than recommended by the standards. The recommendation module (40) analyzes the differences between the calibrated model and the predefined target model across various parameters, such as energy consumption, temperature control, and system efficiency. Based on these comparisons, the recommendation module (40) generates specific recommendations for improving HVAC system performance and aligning with the target model. Recommendations generated by the recommendation module (40) may include suggestions for adjusting HVAC control settings, modifying equipment operation schedules, or implementing energy-saving strategies. These recommendations are tailored to the specific characteristics and conditions of the building, as represented in the calibrated model. The comparison-based approach used by the recommendation module (40) offers several technical benefits. By leveraging established building energy standards, the system can provide recommendations grounded in industry-accepted best practices. This approach helps ensure that the suggested improvements align with recognized energy efficiency goals and performance metrics. Additionally, the use of a predefined target model based onbuilding energy standards allows for consistent evaluation across different buildings and HVAC systems. This consistency facilitates benchmarking and performance tracking over time, enabling building managers to assess progress towards energy efficiency goals and identify long-term trends in HVAC system performance.
[0033] According to another embodiment of the present invention, the recommendation module (40) is configured to provide a control instruction to the at least one device (2) based on the at least one recommendation. The recommendation module (40) is configured to communicate with the automation system (50) configured to control and monitor at least one working parameter of at least one device (2). When the recommendation module (40) generates a recommendation for improving performance or efficiency, the building management system (1) can translate this recommendation into specific control instructions. The control instruction comprises at least one of a direct command issued through a building automation protocol (e.g., BACnet, Modbus, KNX, Sedona, or Niagara), a semantic command defined using an ontology such as BRICK or Project Haystack for addressing HVAC setpoints, operational states, or scheduling parameters, an override command to temporarily adjust the operational state of a component (e.g., fan speed, damper position, valve control), a persistent reconfiguration of control logic or parameters via a supervisory control platform, a schedule adjustment instruction affecting future time-bound behavior of HVAC subsystems, a recommendation-translated command synthesized by a Large Language Model (LLM) that is automatically mapped to device-specific instructions based on its metadata and control point capabilities. These control instructions are then transmitted to the relevant devices (2) within the building by using the automation system (50). By directly sending control instructions to devices (2), the building management system (1) can quickly implement recommended changes, reducing the delay between identifying an optimization opportunity and realizing its benefits. The control instructions can be dynamically adjusted based on real-time data from the building and devices (2), allowing for responsive and adaptive management of the system. Theautomated approach allows for frequent, small adjustments to operations based on ongoing analysis and recommendations, enabling continuous optimization of system performance. Rapid and precise implementation of optimized settings can lead to improved energy efficiency by ensuring that devices (2) operate at their most efficient parameters. The building management system (1) can also record the control instructions sent to devices (2), providing valuable data for performance tracking, compliance monitoring, and future optimization efforts.
[0034] In some embodiments, the control instruction generated by the recommendation module (40) may be abstracted in a semantic layer using ontologies such as BRICK or Haystack. These semantic instructions can then be translated into device-specific control commands using a middleware or integration engine that supports standard protocols like BACnet / IP, Modbus RTU / TCP, or proprietary VRF / VRV protocols.
[0035] According to another embodiment of the present invention, the recommendation module (40) configured to utilize Large Language Model (LLM) to interpret user queries and generate natural language responses related to at least one condition variable or control instruction. The integration of an LLM in the recommendation module (40) offers several technical advantages for building management. By leveraging advanced natural language processing capabilities, the LLM enables the recommendation module (40) to generate more nuanced and context-aware suggestions for HVAC optimization. The LLM can process and analyze complex patterns in the building data, energy consumption trends, and environmental factors to produce recommendations that are tailored to the specific characteristics and requirements of the building. The use of an LLM allows the recommendation module (40) to consider a wider range of variables and their interrelationships when formulating suggestions. This comprehensive analysis can lead to more sophisticated optimization strategies that may not be immediately apparent through traditional rule-based systems. Another technical advantage of integrating an LLM is the ability to generate recommendationsin natural language format. This feature enhances the clarity and accessibility of the suggestions, making them easier for building managers and operators to understand and implement. The LLM can adapt its language and terminology to match the user's level of expertise, ensuring that recommendations are communicated effectively to both technical and non-technical stakeholders. The LLM integration also enables the recommendation module (40) to provide more detailed explanations and rationales for its suggestions. By articulating the reasoning behind each recommendation, the system helps users make more informed decisions about HVAC management and fosters trust in the automated recommendations. Furthermore, the LLM can process and respond to natural language queries from users, allowing for more intuitive interaction with the building management system. This capability enables users to ask complex questions about HVAC performance, energy efficiency, or specific building conditions, and receive detailed, contextually relevant responses. The adaptability of the LLM allows the recommendation module (40) to continuously improve its performance over time. As the model processes more data and user interactions, the recommendations can become increasingly refined and tailored to the specific needs of the building and its occupants.
[0036] The recommendation module (40) utilizes a large language model (LLM) not only for query interpretation but also for transforming optimization results into actionable insights, aligned with user-defined constraints or comfort parameters.
[0037] Additionally, the LLM can assist in interpreting and summarizing large volumes of building performance data, providing concise reports and insights that help building managers quickly identify trends, anomalies, or areas requiring attention.
[0038] According to another embodiment of the present invention, the recommendation module (40) is configured to receive a recommendation request from a user interface related to a specific condition variable and to generate the at least one recommendationbased on the specific condition variable. The recommendation request may specify a particular condition variable, such as indoor temperature, energy consumption of a specific HVAC unit, or airflow rate. Upon receiving the request, the recommendation module (40) processes the selected condition variable in the context of the current calibrated model and compares it with the predefined target model. Based on this comparison, the recommendation module (40) generates at least one targeted recommendation aimed at optimizing the selected condition variable. This allows users to interactively query the system and obtain focused insights tailored to specific operational or environmental aspects of the building. Such targeted recommendations improve system adaptability and allow for more precise control over HVAC performance.
[0039] In some embodiments, the building management system (1) supports multi-site configuration, optimization, and recommendation workflows across a portfolio of buildings, such as retail chains, bank branches, or office campuses. The data management server may aggregate condition variables and operational data across multiple locations, and allow centralized control, benchmarking, and policy enforcement. A user may initiate a region-specific analysis by querying the system via a natural language prompt, such as: “What can I do to save 20% energy across my branches in the Aegean Region?”. To process such a prompt, the recommendation module (40) may leverage a Large Language Model (LLM) integrated with a structured data interface. The LLM interprets the query, identifies the target subset of buildings (e.g., all facilities tagged with "Aegean Region" metadata), and generates a set of actionable recommendations tailored to regional characteristics such as climate conditions, occupancy patterns, HVAC equipment type, and historical energy performance. These recommendations may include bulk adjustments to setpoints, revised scheduling policies, equipment retrofit suggestions, or control strategy changes. The system may then allow the user to apply the recommended configurations selectively or globally, using a centralized command dispatch framework compatiblewith the underlying building automation systems (e.g., BACnet broadcast, MQTT publish, or REST-based orchestration). This architecture supports scalable, Al-assisted energy management for enterprise-level building portfolios, where insights and decisions are informed by aggregated data and localized constraints, and are accessible through natural language interaction.
[0040] In some embodiments, the building management system (1) manages multiple buildings across a geographical portfolio. Centralized coordination enables benchmarking, bulk recommendations, and energy savings enforcement through region-specific LLM queries.
[0041] According to another embodiment of the present invention, the recommendation module (40) is be configured to prioritize recommendations based on their potential impact on energy efficiency, cost savings, or other relevant criteria. This prioritization helps building managers focus on the most effective improvements and allocate resources efficiently. In some examples, the recommendation module (40) may also consider factors such as local climate conditions, building occupancy patterns, and specific energy efficiency goals when generating recommendations. This contextual awareness allows for more tailored and relevant suggestions that account for the unique characteristics of each building and its HVAC system.
[0042] According to another embodiment of the present invention, the recommendation module (40) is configured to display at least one recommendation on a user interface associated with the building management system. The user interface may be implemented as a web-based dashboard, a desktop application, or a mobile interface accessible to building operators or system administrators. The displayed recommendation may include textual suggestions, visual indicators, or graphical overlays on the visual model of the building, indicating optimal settings, performance issues, or suggested actions. The interface may also allow users to view additionalcontext or reasoning behind each recommendation, such as the relevant condition variables, deviations from target values, or potential energy savings.
[0043] According to another embodiment of the present invention, modeling module (20) is configured to generate a visual model of the building showing a location of the at least one device (2) and at least one condition variable and to transmit the visual model to a user interface associated with the building management system. The visual model may contain three-dimensional coordinates for a plurality of objects or structures that relate to the building's energy consumption, and the actual model may estimate the building's energy consumption based on modeled physical processes relating to these objects and structures.
[0044] According to another embodiment of the present invention, the calibration module (30) comprises a manual editor (31) configured to receive an edited visual model from the user interface, wherein the edited visual model comprises at least one modification related to the location of the at least one device (2) and / or at least one condition variable and the data management server is configured to update the condition variables stored in the system data module (10) based on the edited visual model. The manual editor (31) allows a user inputs at least one user feedback related to at least one parameter of the visual model. This manual editor (31) enables the user to make manual adjustments to the visual model based on observed system performance. The user feedback ensures that the model reflects any discrepancies between predicted and actual building performance, improving its accuracy. In certain embodiments, the system comprises a BACnet scanner or similar discovery tool that interfaces with building automation networks to identify and retrieve relevant control points, schedules, and historical data. These elements may be automatically mapped to the corresponding condition variables in the energy model. For example, a BACnet Analog Value object representing a zone temperature setpoint can be parsed and used to update the heating or cooling threshold within the model. Similarly, schedule objects (BACnetSchedule or Calendar objects) can be associated with time-based occupancy or equipment operation profiles within the simulation framework. The manual editor (31) allows users to visualize and refine this imported data through an intuitive interface, while the backend system ensures semantic and spatial alignment between scanned building control data and the actual building zones or equipment modeled.
[0045] According to another embodiment of the present invention, the data management server (100) comprises an automation system (50) configured to control and monitor at least one working parameter of at least one device (2) and the calibration module (30) comprises a digital editor (32) configured to communicate with the automation system (50) and to calibrate the actual model by using at least one working parameter monitored by the automation system (50). The data management server (100) includes an automation system (50) configured to control and monitor working parameters of the devices (2). This automation system (50) collects real-time data on various operational aspects of the devices (2). The digital editor (32) communicates with the automation system (50). This digital editor (32) utilizes the working parameters monitored by the automation system to calibrate the actual model automatically. By leveraging data from the automation system (50), the digital editor (32) can continuously refine the model based on current HVAC performance, ensuring up-to-date accuracy. The digital editing process relies on objective measurements from the devices (2), reducing the potential for human error in model calibration. Automated digital editing allows for efficient calibration of complex systems with multiple devices (2) and numerous condition variables.
[0046] The combination of manual and digital editing capabilities in the building management system (1) provides a comprehensive approach to model refinement and calibration. This dual approach allows for both expert human input and data-driven automation, resulting in a more accurate and adaptable representation of the building's HVAC system. The flexibility offered by these editing options enables the buildingmanagement system to accommodate a wide range of building types, HVAC configurations, and operational requirements.
[0047] According to another embodiment of the present invention, the calibration module (30) is configured to generate the calibrated model by minimizing an error metric between the actual model and the condition variables. This error minimization approach allows the calibration module (30) to fine-tune the model parameters, ensuring a closer alignment between the model predictions and the observed data from the building and HVAC system. The calibrated model is generated using both manual user feedback and real-time operational data, and the calibration is accepted when the statistical accuracy thresholds defined by CRMSE, MAE, and R2score are met.
[0048] According to another embodiment of the present invention, the calibration module (30) is configured to generate the calibrated model when a centered root mean square error (CRMSE) is equal to or less than 15% or alternatively, when at least one of the following conditions is met: a mean absolute error (MAE) is less than a predefined threshold, a coefficient of determination (R2score) exceeds 0.85, a normalized mean bias error (NMBE) is within ±5%, the residual mean between the calibrated model and actual model is statistically insignificant at a 95% confidence level based on a two-tailed t-test. This threshold provides a quantitative measure of the model's accuracy and serves as a benchmark for determining when the calibration process has achieved a satisfactory level of precision. The use of CRMSE as an error metric offers several advantages in the calibration process. CRMSE provides a normalized measure of the difference between predicted and observed values, allowing for meaningful comparisons across different variables and scales. By setting a specific CRMSE threshold, the calibration module (30) can ensure consistent model quality across various building and HVAC configurations. The calibration process may involve iterative adjustments to the condition variables, with the calibration module (30) repeatedly evaluating the error metric and refining the model parameters until thedesired level of accuracy is achieved. This iterative approach allows for the incorporation of new data and the continuous improvement of the model over time. In some embodiments, the calibration module (30) may evaluate the accuracy of the calibrated model by comparing it to the actual model using one or more statistical error metrics. These metrics may include, but are not limited to, centered root mean square error (CRMSE), mean absolute error (MAE), coefficient of determination (R2), and normalized mean bias error (NMBE). For example, the calibrated model may be considered sufficiently accurate when the CRMSE is equal to or less than 15%. Alternatively, acceptable calibration may be achieved if the MAE remains below a predefined application-specific threshold, or if the R2score exceeds 0.85, indicating strong correlation between predicted and observed values. Furthermore, the calibration may also be validated statistically, such that the mean of residuals (differences between predicted and observed values) is found to be statistically insignificant at a 95% confidence level using a two-tailed t-test. This multi-metric approach ensures the flexibility and robustness of the calibration process under varying operational conditions and system complexities, and aligns with recognized industry standards such as ASHRAE Guideline 14 and the International Performance Measurement and Verification Protocol (IPMVP).
[0049] In one exemplary use case, the modeling module (20) generates a thermal simulation of a commercial building using a dynamic simulation engine such as EnergyPlus. The model incorporates building geometry, occupancy profiles, HVAC schedules, internal loads, and external weather data. The actual building performance is monitored via sensors measuring zone temperature, airflow rates, and energy consumption of HVAC equipment.
[0050] The calibration module (30) compares the EnergyPlus simulation outputs (e.g., hourly zone temperatures and HVAC electricity consumption) against real-time measurements obtained from the building management system. During this process: A CRMSE of11.2% is calculated between measured and simulated zone temperatures. The R2score for hourly HVAC electricity demand is computed as 0.91. The residual mean difference for airflow rate is found to be statistically insignificant (p > 0.05), based on 168 hourly samples (7 days). Based on these results, the calibration module (30) accepts the model as sufficiently accurate. The calibrated model is then used by the recommendation module to suggest optimized HVAC operation schedules and equipment setpoints, aligned with energy conservation targets derived from ANSI / ASHRAE / IES energy standards.
[0051] According to another embodiment, the calibration module (30) may incorporate machine learning algorithms trained on historical performance data. Alternatively, the calibration logic may reside on edge computing gateways directly interfacing with local BMS networks to reduce latency.
[0052] In certain embodiments, the building management system (1) incorporates a data standardization layer designed to address inconsistencies in point naming conventions across different building automation systems (50) and integrators. Since point names (e.g., temperature sensors, damper positions, fan speeds) may vary significantly between buildings or even within the same site due to differing commissioning practices, the system includes a semantic harmonization engine to normalize and standardize such data. This engine may utilize rule-based parsing, metadata analysis, and machine learning-based inference to identify the functional meaning of each point. For example, a point named "ZT Al", "ZoneTemp-AlOl", or "Al Zone T" may all be semantically interpreted as a “Zone Air Temperature” point, and mapped to a standardized label such as zone air temperature based on a defined ontology (e.g., Project Haystack, BRICK Schema, or custom taxonomies). The system may also use spatial metadata (e.g., room tags, floorplan coordinates, equipment hierarchy) to disambiguate similar points across zones or buildings.To ensure compatibility across various BMS platforms, the building management system (1) includes a semantic standardization layer. This engine aligns point names, control schedules, and sensor types with ontologies such as Project Haystack or BRICK Schema.
[0053] A method for controlling a building management system (1) having at least one device (2) according to the present invention comprises the steps of obtaining, by a system data module (10), condition variables associated with the at least one device (2) and the building, generating, by a modeling module (20), an actual model based on the condition variables, generating, by calibration module (30), a calibrated model by adjusting at least one condition variable, generating, by a recommendation module (40), at least one recommendation related to the at least one condition variable by comparing the calibrated model with a predefined target model. Initially, a system data module (10) obtains one or more condition variables related to both the devices (2) and the building. These variables may include equipment data, energy consumption, ambient sensor readings, building layout, and predefined energy standards. Using the acquired condition variables, a modeling module (20) generates an actual model that represents the current state and behavior of the HVAC system within the specific building context. Subsequently, a calibration module (30) refines the actual model by adjusting one or more condition variables, resulting in a calibrated model. This step may involve data from real-time monitoring systems or manual input from building operators to ensure accuracy. Finally, a recommendation module (40) compares the calibrated model to a predefined target model, which may reflect energy efficiency goals or regulatory standards. Based on this comparison, the recommendation module (40) generates one or more recommendations related to the condition variables, such as operational adjustments, energy-saving strategies, or equipment reconfiguration suggestions. This method enables dynamic, data-driven management of HVAC systems for improved efficiency and performance.According to another embodiment the present invention, the method further comprises the steps of generating a visual model of the building showing a location of the at least one device (2) and at least one condition variable and transmitting the visual model to a user interface. This visual representation is created by the modeling module (20) based on the data obtained from the system data module (10). Once generated, the visual model is transmitted to a user interface, allowing building operators, facility managers, or system administrators to interactively view and analyze the current HVAC system status and related environmental variables. The visual interface improves situational awareness and supports more informed decision-making regarding system calibration and optimization.
[0054] According to another embodiment the present invention, the method further comprises the steps of receiving, by a manual editor (31), an edited visual model from the user interface, wherein the edited visual model comprises at least one modification related to the location of the at least one device and / or at least one condition variable and updating, by the data management server, the condition variables based on the edited visual model. The visual model, which may depict the layout of the building and the location of devices (2), can be modified by a user to reflect changes such as repositioning a device (2) or altering one or more condition variables e.g., updating airflow parameters, adjusting zone boundaries, or redefining sensor inputs. The edited visual model is then received by the manual editor (31), which parses the modifications and forwards the updated data to the data management server (100). This approach allows for flexible, user-driven configuration and enhances the adaptability and accuracy of the building management system.
[0055] According to another embodiment the present invention, the method further comprises the steps of controlling and monitoring, by an automation system (50), at least one working parameter of at least one device (2) and calibrating, by a digital editor (32) configured to communicate with the automation system (50), the actual model by usingat least one working parameter monitored by the automation system (50). The automation system (50) may include building management controllers, programmable logic controllers (PLCs), or sensor-actuator networks capable of retrieving real-time data such as temperature setpoints, fan speed, compressor speed, or energy consumption rates. A digital editor (32), communicatively linked to the automation system (50), is configured to retrieve this real-time operational data and utilize it for calibrating the actual model. The calibration process involves adjusting model parameters so that the actual model more accurately reflects the real-world operation of the HVAC system. By integrating live performance data into the modeling and calibration process, the system ensures continuous accuracy, adaptability, and optimal control of HVAC operations within the building environment.
[0056] According to another embodiment the present invention, the method further comprises the steps of receiving a recommendation request from a user interface related to a specific condition variable and generating the at least one recommendation based on the specific condition variable. Once the recommendation request is received, the recommendation module (40) analyzes the current state of the selected variable within the context of the calibrated model and compares it with the corresponding parameter in the predefined target model. Based on this comparison, the building management system (1) generates at least one targeted recommendation aimed at optimizing the specific condition variable.
[0057] According to another embodiment the present invention, the method further comprises the steps of transmitting a control instruction to the at least one device (2) based on the at least one recommendation. Once the system determines an optimal adjustment such as modifying a temperature setpoint, changing airflow rate, or altering compressor or fan speed, the recommendation module (40) converts this recommendation into a machine-readable control instruction. This control instruction is then transmitted to the corresponding device (2) via a communication interface integrated with the automationsystem (50). The control instruction may be executed automatically, or optionally after user approval, depending on system configuration. This step enables the building management system (1) to close the loop between data analysis and real-time action, allowing for autonomous optimization of HVAC performance in response to dynamic environmental and operational conditions.
Claims
CLAIMS1. A building management system (1) comprising: at least one device (2) disposed within a building and a data management server (100) having:a system data module (10) configured to obtain condition variables associated with the at least one device and the building,a modeling module (20) configured to generate an actual model based on the at least one condition variable,a calibration module (30) configured to generate a calibrated model by adjusting the at least one condition variable anda recommendation module (40) configured to generate at least one recommendation related to the at least one condition variable by comparing the calibrated model with a predefined target model.
2. The building management system (1) according to Claim 1, wherein the recommendation module (40) is configured to provide a control instruction to the at least one device (2) based on the at least one recommendation.
3. The building management system (1) according to any of Claim 1 or Claim 2, wherein the recommendation module (40) configured to utilize Large Language Model to interpret user queries and generate natural language responses related to at least one condition variable or control instruction.
4. The building management system (1) according to any of Claim 1 to Claim 3, wherein the recommendation module (40) is configured to receive a recommendation request from a user interface related to a specific conditionvariable and to generate the at least one recommendation based on the specific condition variable.
5. The building management system (1) according to any of Claim 1 to Claim 4, wherein the recommendation module (40) is configured to prioritize recommendations based on potential impact on energy efficiency, cost savings, or other relevant criteria of the recommendations.
6. The building management system (1) according to any of Claim 1 to Claim 5, wherein the recommendation module (40) is configured to display at least one recommendation on a user interface associated with the building management system (1).
7. The building management system (1) according to any of Claim 1 to Claim 6, wherein the condition variables comprise at least one of building location, usage purpose of the building, floor plan of the building, equipment information of the at least one device, energy consumption data of the at least one device (2), predefined energy standards, ambient sensor data, energy consumption, airflow rate, compressor speed or fan speed of the at least one device.
8. The building management system (1) according to any of Claim 1 to Claim 7, wherein modeling module (20) is configured to generate a visual model of the building showing a location of the at least one device (2) and at least one condition variable and to transmit the visual model to a user interface associated with the building management system (1).
9. The building management system (1) according to Claim 8, wherein the calibration module (30) comprises a manual editor (31) configured to receivean edited visual model from the user interface, wherein the edited visual model comprises at least one modification related to the location of the at least one device (2) and / or at least one condition variable and the data management server (100) is configured to update the condition variables stored in the system data module (10) based on the edited visual model.
10. The building management system (1) according to any of preceding claims, wherein the data management server (100) comprises an automation system (50) configured to control and monitor at least one working parameter of at least one device (2) and the calibration module (30) comprises a digital editor (32) configured to communicate with the automation system (50) and to calibrate the actual model by using at least one working parameter monitored by the automation system (50).
11. The building management system (1) according to any of preceding claims, wherein the predefined target model is based on at least one building energy standard such as ANSI / ASHRAE / IES.
12. The building management system (1) according to any of preceding claims,, wherein the calibration module (30) is configured to generate the calibrated model by minimizing an error metric between the actual model and the condition variables.
13. The building management system (1) according to Claim 12, wherein the calibration module (30) is configured to generate the calibrated model when a centered root mean square error (CRMSE) is equal to or less than 15% or alternatively, when at least one of the following conditions is met: a mean absolute error (MAE) is less than a predefined threshold, a coefficient of determination (R2score) exceeds 0.85, a normalized mean bias error (NMBE) is within ±5%, the residual mean between the calibrated model and actualmodel is statistically insignificant at a 95% confidence level based on a two- tailed t-test.
14. A method for managing operation of at least one device (2) disposed within a building by a data management server (100) comprising the steps of obtaining, by a system data module (10), condition variables associated with the at least one device (2) and the building,generating, by a modeling module (20), an actual model based on the condition variables,generating, by calibration module (30), a calibrated model by adjusting at least one condition variablegenerating, by a recommendation module (40), at least one recommendation related to the at least one condition variable by comparing the calibrated model with a predefined target model.
15. The method according to Claim 14 further comprising the steps of generating a visual model of the building showing a location of the at least one device (2) and at least one condition variable andtransmitting the visual model to a user interface.
16. The method according to any of Claim 14 or Claim 15 further comprising the steps ofreceiving, by a manual editor (31), an edited visual model from the user interface, wherein the edited visual model comprises at least one modification related to the location of the at least one device (2) and / or at least one condition variable andupdating, by the data management server (100), the condition variables based on the edited visual model.
17. The method according to any of Claim 14 to Claim 16 further comprising the steps of:controlling and monitoring, by an automation system (50), at least one working parameter of at least one device (2) andcalibrating, by a digital editor (32) configured to communicate with the automation system (50), the actual model by using at least one working parameter monitored by the automation system (50).
18. The method according to any of Claim 14 to Claim 17 further comprising the steps of:receiving a recommendation request from a user interface related to a specific condition variable andgenerating the at least one recommendation based on the specific condition variable.
19. The method according to any of Claim 14 to Claim 18 further comprising the step of:transmitting a control instruction to the at least one device (2) based on the at least one recommendation.
