Ai-mv-based energy savings calculation apparatus using physical data and virtual factor data and method thereof

Abstract

The present invention relates to an AI-MV-based energy saving amount calculation device and method using physical data and virtual factor data. An energy saving amount calculation device according to one embodiment of the present invention may include: a data receiving unit that receives physical data including energy production data, energy consumption data, and environmental data, and virtual factor data including user behavior data, occupancy status data, and equipment operation pattern data; a physical factor generating unit that preprocesses physical data to generate a physical factor matrix for each predetermined time interval; a virtual factor generating unit that preprocesses virtual factor data to generate a virtual factor matrix for each predetermined time interval; a model generating unit that generates an energy performance model based on a physical factor matrix, a virtual factor matrix, and actual energy usage during a reference period; a baseline generating unit that inputs a physical factor matrix and a virtual factor matrix corresponding to the time interval to be evaluated into the energy performance model to generate a dynamic energy baseline, which is the predicted energy usage for the time interval to be evaluated; and a saving amount calculation unit that calculates the energy saving amount by comparing the dynamic energy baseline with the actual energy usage corresponding to the time interval to be evaluated.

Description

Technology Field

[0001] The present invention relates to an AI-MV-based device and method for calculating energy savings using physical data and virtual factor data. More specifically, it relates to a technology that generates an energy performance model based on physical data including energy production data, energy consumption data, and environmental data, and virtual factor data including user behavior data, occupancy status data, and equipment operation pattern data, generates a dynamic energy baseline to calculate energy savings, further separates and calculates energy savings into physical-based and behavior-based energy savings, and generates an integrated net metering transaction unit utilizing verified savings. Background Technology

[0002] Various technologies are utilized in the field of energy management for homes and buildings. Smart meters and power meters measure power consumption in real time, while solar inverters efficiently convert and manage power generated by solar power systems. Additionally, environmental sensors collect environmental information, such as temperature, humidity, and illuminance, to provide useful data to energy management systems. These technologies play a crucial role in monitoring energy usage and generation, and in improving energy efficiency based on this data. Furthermore, Measurement and Verification (M&V) technologies are widely used to analyze collected data to evaluate and improve energy performance.

[0003] However, existing technologies based on Virtual Net Metering (VNM), ISO 50006, and International Performance Measurement and Verification Protocol (IPMVP) have several limitations. In particular, there is a lack of functionality to separately calculate energy savings resulting from changes in user behavior. This makes it difficult to evaluate energy saving effects more precisely and to analyze the impact of changes in user behavior on energy efficiency. Furthermore, since a structure capable of dynamically analyzing energy data by simultaneously reflecting physical and virtual factors has not been sufficiently presented, the precision and flexibility of energy management are limited.

[0004] These technical limitations act as obstacles to integrating solar power generation and verified energy savings into trading units, or to aggregating them for utilization as aggregated power resources. For example, existing technologies do not adequately support attempts to efficiently trade electricity generated from solar power systems or to collectively manage energy savings from multiple users or buildings for use as power resources. Consequently, new approaches are required to enhance the efficiency of energy management and trading, and to maximize the potential for utilizing aggregated power resources. The problem to be solved

[0005] The objective of the present invention is to provide an AI-MV-based energy saving calculation device and method using physical data and virtual factor data, which can calculate energy saving amounts using physical data and virtual factor data and utilize this for various energy management and trading purposes. means of solving the problem

[0006] In one embodiment of the present invention, an energy saving amount calculation device may be provided. The energy saving amount calculation device may include: a data receiving unit that receives physical data including energy production data, energy consumption data, and environmental data, and virtual factor data including user behavior data, occupancy status data, and equipment operation pattern data; a physical factor generating unit that preprocesses the physical data to generate a physical factor matrix for each predetermined time interval; a virtual factor generating unit that preprocesses the virtual factor data to generate a virtual factor matrix for each predetermined time interval; a model generating unit that generates an energy performance model based on the physical factor matrix, the virtual factor matrix, and actual energy usage during a reference period; a baseline generating unit that inputs the physical factor matrix and the virtual factor matrix corresponding to the time interval to be evaluated into the energy performance model to generate a dynamic energy baseline, which is the predicted energy usage for the time interval to be evaluated; and a saving amount calculation unit that calculates the energy saving amount by comparing the dynamic energy baseline with the actual energy usage corresponding to the time interval to be evaluated.

[0007] In one embodiment of the present invention, the energy saving amount calculation device includes solar power generation amount data for energy production amount data, and energy consumption amount data includes at least one of power consumption amount data, power import and export power amount data, charging and discharging amount data of an energy storage device, and smart meter or power meter measurement data, and environmental data may include at least one of external temperature data, solar radiation data, humidity data, and weather data.

[0008] In one embodiment of the present invention, user behavior data includes other heating usage pattern data, occupancy status data includes at least one of occupancy status data and occupancy personnel data, and equipment operation pattern data may include at least one of cooling operation pattern data, heating operation pattern data, hot water usage pattern data, lighting usage pattern data, and ventilation equipment operation pattern data.

[0009] In one embodiment of the present invention, preprocessing may perform at least one of outlier removal, missing value correction, normalization, and time synchronization.

[0010] In one embodiment of the present invention, the model generation unit generates a plurality of energy performance models and evaluates the model performance based on at least one of the coefficient of determination, significance probability, F-test, and cross-validation error to select a final energy performance model.

[0011] In one embodiment of the present invention, the energy saving amount calculation unit can calculate the energy saving amount by separating it into a physical-based energy saving amount and a behavior-based energy saving amount.

[0012] In one embodiment of the present invention, the reduction amount calculation unit calculates a reliability index by reflecting at least one of the model fit, prediction error rate, and data completeness of the energy performance model, and can calculate a verified reduction amount by applying the reliability index as a verification coefficient to the energy reduction amount.

[0013] In one embodiment of the present invention, a transaction unit generating unit may be further included to generate an integrated net metering transaction unit used in a Virtual Net Metering (VNM) system or an energy trading platform by combining a solar power generation amount and a verified reduction amount based on energy production data.

[0014] In one embodiment of the present invention, the transaction unit generation unit may be configured to aggregate the integrated net metering transaction units of a plurality of building energy nodes to generate aggregated power resources, and to provide the aggregated power resources to an energy trading platform or to link with a virtual power plant (VPP) operating system.

[0015] In one embodiment of the present invention, the method for calculating energy savings may include: a data receiving step for receiving physical data including energy production data, energy consumption data, and environmental data, and virtual factor data including user behavior data, occupancy status data, and equipment operation pattern data; a physical factor generation step for preprocessing physical data to generate a physical factor matrix for each predetermined time interval; a virtual factor generation step for preprocessing virtual factor data to generate a virtual factor matrix for each predetermined time interval; a model generation step for generating an energy performance model based on the physical factor matrix, virtual factor matrix, and actual energy usage during a reference period; a baseline generation step for inputting the physical factor matrix and virtual factor matrix corresponding to the time interval to be evaluated into the energy performance model to generate a dynamic energy baseline, which is the predicted energy usage for the time interval to be evaluated; and a savings calculation step for calculating energy savings by comparing the dynamic energy baseline with the actual energy usage corresponding to the time interval to be evaluated. Effects of the invention

[0016] According to one embodiment of the present invention, by receiving physical data and virtual factor data and preprocessing them to generate a physical factor matrix and a virtual factor matrix, a dynamic energy baseline reflecting both physical and virtual factors can be generated. According to one embodiment of the present invention, by comparing the dynamic energy baseline with actual energy usage to calculate energy savings, and by separating and calculating these into physical-based energy savings and behavior-based energy savings, savings based on changes in user behavior can be calculated separately. Furthermore, according to one embodiment of the present invention, by integrating the calculated energy savings and solar power generation to generate a verified savings amount reflecting a reliability index, and converting this into an integrated net metering transaction unit, it can be utilized as an energy trading platform and a collective power resource. Brief explanation of the drawing

[0017] FIG. 1 is a block diagram illustrating the configuration of an AI-MV integrated calculation system according to one embodiment of the present invention. FIG. 2 is a block diagram illustrating the configuration of an energy saving amount calculation device according to one embodiment of the present invention. FIG. 3 is an exemplary diagram illustrating a physical factor generation unit and a virtual factor generation unit according to an embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a physical factor matrix according to one embodiment of the present invention. FIG. 5 is an exemplary diagram illustrating the process of generating a verification reduction amount of an AI-MV analysis engine according to one embodiment of the present invention. FIG. 6 is an illustrative diagram for explaining a transaction unit generation unit according to an embodiment of the present invention. FIG. 7 is an illustrative diagram for explaining the function of the AI-MV integrated calculation system in the overall energy market system according to one embodiment of the present invention. FIG. 8 is a system flowchart for explaining a method for calculating energy savings according to an embodiment of the present invention. Specific details for implementing the invention

[0018] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in relation to one embodiment. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the invention is limited only by the appended claims, including all equivalents to those claimed therein, provided they are appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.

[0019] The components according to the present invention are defined by functional distinction rather than physical distinction, and can be defined by the functions each performs. Each component may be implemented as hardware or as program code and processing units that perform each function, and the functions of two or more components may be included and implemented in a single component. Therefore, it should be noted that the names assigned to the components in the following embodiments are not intended to physically distinguish each component but are assigned to imply the representative function performed by each component, and that the technical concept of the present invention is not limited by the names of the components.

[0020] FIG. 1 is a block diagram illustrating the configuration of an AI-MV integrated calculation system according to one embodiment of the present invention.

[0021] Referring to FIG. 1, an AI-MV (Artificial Intelligence-Measurement and Verification) based integrated calculation system (1) may include a building energy node (10), an energy saving amount calculation device (20), and an energy trading node (30). The AI-MV based integrated calculation system (1) may collect energy production, energy consumption, environmental information, and user behavior information generated in individual buildings or households. Based on the collected information, the AI-MV based integrated calculation system (1) may calculate a dynamic energy baseline and an energy saving amount. The AI-MV based integrated calculation system (1) may be a system that provides the calculated results in a data form that can be used for trading or settlement.

[0022] A building energy node (10) may refer to a site unit where energy data is generated. For example, a building energy node (10) may be a single-family home, a multi-family home, an office, a commercial building, a public building, an industrial building, a data center, a school, a hospital, or a small factory. However, the present invention is not limited thereto. A building energy node (10) may include a smart meter (110), a solar inverter (120), an environmental sensor (130), and an energy management system (140).

[0023] The smart meter (110) can collect metering data related to power usage. The smart meter (110) can measure power consumption, power imports, and power exports. The smart meter (110) can also collect detailed data such as the amount of power used at each metering point.

[0024] The solar inverter (120) can collect information related to power generation from the solar power generation system. The solar inverter (120) can collect solar power generation amount data over time. The solar inverter (120) can collect data including the inverter's current output and power generation status information.

[0025] The environmental sensor (130) can collect data related to the surrounding environment of the building energy node (10). The environmental sensor (130) can measure external temperature and solar radiation data. The environmental sensor (130) can also collect environmental data such as humidity and weather conditions.

[0026] The energy management system (140) can collect virtual factor data related to user behavior and facility operation, or receive it in conjunction with an external system. The energy management system (140) can collect data on the user's past energy usage history, occupancy status, and number of occupants. The energy management system (140) can collect data such as cooling operation patterns, heating operation patterns, hot water usage patterns, lighting usage patterns, ventilation facility operation patterns, and other electric heat usage patterns.

[0027] Data collected from a building energy node (10) can be divided into physical data and virtual factor data. Physical data may include energy production data, energy consumption data, and environmental data. For example, energy production data may include solar power generation data, and energy consumption data may include power consumption data, power import and export data, charging and discharging data of an energy storage device, and smart meter or power meter measurement data. Environmental data may include external temperature data, solar radiation data, humidity data, and weather data.

[0028] Virtual factor data may include user behavior data, occupancy status data, and equipment operation pattern data. For example, user behavior data may include other heating usage pattern data. Occupancy status data may include occupancy status data and occupancy count data. Equipment operation pattern data may include at least one of cooling operation pattern data, heating operation pattern data, hot water usage pattern data, lighting usage pattern data, and ventilation equipment operation pattern data.

[0029] The energy saving amount calculation device (20) can receive physical data and virtual factor data collected from the building energy node (10). The energy saving amount calculation device (20) can generate a physical factor matrix and a virtual factor matrix using the received data. The energy saving amount calculation device (20) can generate an energy performance model based on the generated matrices and calculate a dynamic energy baseline and an energy saving amount for the time interval to be evaluated.

[0030] The energy saving amount calculation device (20) may include a data receiving unit (210), a physical factor generating unit (220), a virtual factor generating unit (230), a model generating unit (240), a baseline generating unit (250), a saving amount calculation unit (260), and a transaction unit generating unit (270).

[0031] The data receiving unit (210) can receive physical data including energy production data, energy consumption data, and environmental data, and virtual factor data including user behavior data, occupancy status data, and equipment operation pattern data. The physical factor generation unit (220) can preprocess the received physical data to generate a physical factor matrix for each predetermined time interval. The virtual factor generation unit (230) can preprocess the received virtual factor data to generate a virtual factor matrix for each predetermined time interval. The model generation unit (240) can generate an energy performance model based on the physical factor matrix, virtual factor matrix, and actual energy usage during a reference period. The baseline generation unit (250) can input the physical factor matrix and virtual factor matrix corresponding to the time interval to be evaluated into the energy performance model. The baseline generation unit (250) can generate a dynamic energy baseline, which is the predicted energy usage for the time interval to be evaluated. The energy saving amount calculation unit (260) can calculate the energy saving amount by comparing the dynamic energy baseline with the actual energy usage corresponding to the time interval to be evaluated.

[0032] The energy saving amount calculation device (20) can generate a verified saving amount by reflecting a reliability index on the calculated energy saving amount. The transaction unit generation unit (270) can generate an integrated net metering transaction unit by combining the solar power generation amount and the verified saving amount as needed. In one embodiment of the present invention, the transaction unit generation unit (270) can generate an aggregated power resource by aggregating the integrated net metering transaction units of a plurality of building energy nodes (10). The transaction unit generation unit (270) can provide the generated aggregated power resource to an energy trading platform (310) or link it with a virtual power plant (VPP) operating system (330).

[0033] The energy trading node (30) may be an external linkage node that utilizes data-based results generated by the energy saving amount calculation device (20). The energy trading node (30) may include an energy trading platform (310), an energy settlement system (320), and a Virtual Power Plant (VPP) operation system (330). The energy trading node (30) may support energy transmission and reception relationships, cost settlement relationships, or market operation relationships between building energy nodes based on integrated net metering trading units or aggregated power resource data generated by the trading unit generation unit (270).

[0034] The integrated net metering transaction unit and aggregated power resource generated by the energy saving amount calculation device (20) may be in the form of data rather than power itself. The integrated net metering transaction unit and aggregated power resource may be data that quantifies, verifies, and structures energy production and saving amounts. Such data resources may be distinguishable from the actual transmission and distribution of power or physical power flow.

[0035] The AI-MV-based integrated calculation system (1) can be configured in a hierarchical structure according to function. The building energy node (10) can perform the function of collecting field data. The energy saving amount calculation device (20) can perform calculation and verification functions. The energy trading node (30) can perform trading, settlement, and operation linkage functions. Through this structure, the AI-MV-based integrated calculation system (1) can operate as an intermediate layer that converts energy data generated at the individual building level into a verified data resource that can be utilized in the market.

[0036] FIG. 2 is a block diagram illustrating the configuration of an energy saving amount calculation device according to one embodiment of the present invention.

[0037] Referring to FIG. 2, an energy saving amount calculation device (20) according to one embodiment of the present invention may be configured to include a data receiving unit (210), a physical factor generating unit (220), a virtual factor generating unit (230), a model generating unit (240), a baseline generating unit (250), a saving amount calculation unit (260), and a transaction unit generating unit (270). The energy saving amount calculation device (20) can comprehensively analyze energy-related data collected from multiple sources. The energy saving amount calculation device (20) can perform the role of precisely calculating the energy saving amount. In one embodiment of the present invention, the model generating unit (240), the baseline generating unit (250), and the saving amount calculation unit (260) may be collectively referred to as an AI-MV analysis engine.

[0038] The data receiving unit (210) can receive physical data and virtual factor data from the building energy node (10). The data receiving unit (210) can serve as an interface for exchanging data with an external device through a plurality of communication protocols. For example, the communication protocols may be Modbus, BACnet, MQTT (Message Queuing Telemetry Transport), RESTful API (Application Programming Interface), OPC-UA (Open Platform Communications Unified Architecture), SNMP (Simple Network Management Protocol), and FTP (File Transfer Protocol). However, the present invention is not limited thereto. The received data may be stored in an internal database or memory and utilized in subsequent processing steps.

[0039] The physical data received by the data receiving unit (210) may include energy production data, energy consumption data, and environmental data. In one embodiment of the present invention, the energy production data may include solar power generation data. The energy consumption data may include at least one of power consumption data, power import and export data, charging and discharging data of an energy storage device, smart meter measurement data, and power meter measurement data. The environmental data may include at least one of external temperature data, solar radiation data, humidity data, and weather data.

[0040] Virtual factor data received by the data receiving unit (210) may include user behavior data, occupancy status data, and equipment operation pattern data. In one embodiment of the present invention, user behavior data may include other heating usage pattern data. Occupancy status data may include at least one of occupancy status data and occupancy personnel data. Equipment operation pattern data may include at least one of cooling operation pattern data, heating operation pattern data, hot water usage pattern data, lighting usage pattern data, and ventilation equipment operation pattern data.

[0041] In one embodiment of the present invention, the data receiving unit (210) can process the received data by classifying it according to temporal characteristics. The data receiving unit (210) can separately receive training data for a reference period for generating an energy performance model and prediction data for an evaluation target time interval for calculating energy savings. The data receiving unit (210) can additionally receive actual energy usage data for the evaluation target time interval and provide it to the savings amount calculation unit (260). This classification of data can ensure clarity in the model training and verification process and the actual savings amount calculation process.

[0042] The physical factor generation unit (220) can generate physical factor vectors for each predetermined time interval by preprocessing physical data received from the data receiving unit (210). The physical factor generation unit (220) can remove noise or errors that may be included in the raw data. The physical factor generation unit (220) can perform the role of converting the data into a form used for analysis. The preprocessing process can serve as a basis for improving the quality of the data and increasing the accuracy of the energy performance model. The physical factor generation unit (220) can organize the preprocessed data for each time interval into a vector form. The physical factor generation unit (220) can generate a physical factor matrix by arranging the vector-form data in chronological order.

[0043] The preprocessing performed by the physical factor generation unit (220) may include a plurality of data cleaning techniques. For example, the preprocessing may perform at least one of outlier removal, missing value correction, data normalization, time synchronization, data scaling, feature engineering, and data smoothing. However, the present invention is not limited thereto. Outlier removal may be a process of filtering abnormal data caused by measurement errors, etc. Missing value correction may be a process of estimating missing data using statistical methods or interpolation methods. Data normalization may be a process of increasing the stability of model learning by converting data with different units and ranges into a constant range. Time synchronization may be a process of unifying the time standards of data collected at different intervals.

[0044]

[0045] Referring to Equation 1, the physical state at a specific time interval t can be expressed by a physical factor vector Xt. The physical factor vector Xt can represent the physical elements of the corresponding time interval by integrating them into a single structure. Here, G(t) can represent energy production at the time interval t, for example, solar power generation data. L(t) can represent energy consumption characteristic data at the time interval t, and may include, for example, power consumption data by facility, circuit, or application. P imp (t) can represent power import data in the time interval t. P exp (t) can represent power output data in the time interval t. B ess (t) may be data representing the charge amount, discharge amount, or charge / discharge state of the energy storage system (ESS) during the time interval t. In addition, T out (t), Irr(t), and Hum(t) may be environmental data representing ambient temperature, solar radiation, and humidity, respectively, during the time interval t, and W(t) may represent weather condition data during the time interval t. Such a physical factor vector X t It can comprehensively represent the physical aspects of an energy system. Meanwhile, in this specification, the prediction target for calculating energy savings may be the total actual energy consumption, and L(t) included in the physical factor vector Xt may not be the same value as the total actual energy consumption itself, but may refer to consumption characteristic data as an explanatory variable used to predict the total actual energy consumption. Through this, the energy performance model can predict the total actual energy consumption of the time interval to be evaluated by considering multiple physical factors.

[0046] The virtual factor generation unit (230) can generate virtual factor vectors for each predetermined time interval by preprocessing virtual factor data received from the data receiving unit (210). The virtual factor generation unit (230) can perform the role of converting unstructured or categorical data, such as user behavior data or equipment operation patterns, into numerical data that can be learned by an energy performance model. For example, occupancy status data in text form, such as 'occupied' or 'absent', can be converted into 1 and 0, respectively. In addition, the virtual factor generation unit (230), like the physical factor generation unit (220), can ensure the reliability of the data by performing preprocessing processes such as removing outliers and correcting missing values.

[0047] The virtual factor generation unit (230) can organize data for each time interval, which has undergone preprocessing and transformation processes, into a vector form. Subsequently, the virtual factor generation unit (230) can generate a virtual factor matrix by sequentially arranging each vector according to the flow of time. The virtual factor matrix can be used as an input to an energy performance model along with a physical factor matrix. Through this, the energy performance model can predict energy consumption by comprehensively considering not only physical environmental factors but also user behavior patterns or equipment operation habits that affect actual energy usage. In addition, the virtual factor matrix can be utilized by dividing it into a real virtual factor matrix representing the actual user behavior state and a reference virtual factor matrix representing the reference user behavior state to calculate behavior-based energy savings.

[0048]

[0049] Referring to Equation 2, user behavior and equipment operating status in a specific time interval t are the virtual factor vector Z t It can be expressed as. The above virtual argument vector Z tIt can quantify non-physical factors affecting energy usage and integrate them into a single data structure. Here, Occ b (t) may be binary data indicating occupancy status during the time interval t. Occ n (t) may be data representing the number of occupants in the time interval t. Cool(t), Heat(t), DHW(t), Light(t), and Vent(t) may be data representing the operating patterns of cooling, heating, hot water supply, lighting, and ventilation equipment, respectively, in the time interval t. These data may indicate whether the equipment is operating or its operating intensity. Misc(t) may be data representing the usage patterns of other electric heating devices in the time interval t. The notation at the top right of the formula may signify a transpose operation, and accordingly, each factor may be arranged in the form of a column vector. The above virtual factor vector Z t It can convert user behavior factors and equipment operation factors that affect energy consumption into objective data. This can improve the predictive performance of energy performance models and provide a basis for separately calculating behavior-based energy savings by utilizing the difference between the predicted values ​​obtained by applying the actual virtual factor vector and the reference virtual factor vector, respectively, under the same physical factor conditions.

[0050]

[0051] Referring to Equation 3, the Physical Factor Matrix (PFM) can be constructed by arranging physical factor vectors over the entire analysis period in chronological order. Here, the PFM represents the physical factor matrix for the analysis period. Each row of the matrix consists of physical factor vectors X1, X2, … , X for a specific time interval. TIt may correspond to T. T may represent the total number of time intervals included in the analysis period. For example, if data is collected at 1-hour intervals and analyzed for one year, T may be 8,760. Such a matrix structure can reflect the time-series characteristics of physical data over time. Furthermore, the aforementioned physical factor matrix can be directly utilized as input data for training or inference of an energy performance model. Through this, physical states for multiple time intervals can be processed in a consistent format, and the efficiency of data processing can be improved.

[0052]

[0053] Referring to Equation 4, the Virtual Factor Matrix (VFM) can be constructed by arranging virtual factor vectors over the entire analysis period in chronological order. Here, the VFM can refer to the virtual factor matrix for the analysis period. Each row of the matrix consists of virtual factor vectors Z1, Z2, … , Z for a specific time interval. T It can be composed as follows. T can represent the number of time intervals identical to the physical factor matrix. Through such a matrix structure, temporal changes in user behavior and equipment operation patterns can be represented as a single dataset. Furthermore, the virtual factor matrix can be composed of a real virtual factor matrix reflecting the actual user behavior state or a reference virtual factor matrix reflecting the reference user behavior state. Consequently, the energy performance model can receive both the physical factor matrix and the virtual factor matrix as inputs simultaneously. Through this, it can learn the complex interaction between physical and behavioral factors, and can separately calculate behavior-based energy savings by applying the real virtual factor matrix and the reference virtual factor matrix, respectively, under the same physical factor conditions.

[0054] The model generation unit (240) can generate an energy performance model based on a physical factor matrix, a virtual factor matrix, and actual energy usage corresponding to the reference period. The model generation unit (240) can derive a functional model representing the correlation between a plurality of explanatory variables included in the physical factor matrix and the virtual factor matrix and actual energy usage using machine learning or statistical analysis techniques. At this time, the model generation unit (240) may be configured to automatically optimize hyperparameters including at least one of a learning rate, the number of hidden layers and nodes, a dropout rate, and a weight regularization coefficient through grid search, random search, or Bayesian optimization techniques to maximize the prediction accuracy of the selected algorithm. Here, the actual energy usage may be a target value for prediction for calculating the amount of energy saved. On the other hand, each factor included in the physical factor matrix and the virtual factor matrix may be an explanatory variable for predicting the actual energy usage. Accordingly, the energy performance model can perform the function of predicting expected energy consumption under specific physical and hypothetical conditions, for example, the total measured energy consumption over the time interval under evaluation.

[0055] The model generation unit (240) can generate an energy performance model using multiple algorithms. According to another embodiment of the present invention, the model generation unit (240) may adopt a federated learning structure that updates a global energy performance model by combining only the learned weights after individually learning PFM and VFM data collected from multiple households on local devices for each household, without transmitting them to a central server. Through this, sensitive virtual factor data related to the energy usage habits of individual households can be protected, and at the same time, a universal dynamic baseline model reflecting various residential environments and regional characteristics can be constructed to maximize the generalization performance of the reduction amount calculation.

[0056] For example, energy performance models may be Multiple Linear Regression, Support Vector Machine, Random Forest, Gradient Boosting, Artificial Neural Network, Recurrent Neural Network (RNN), or Long Short-Term Memory (LSTM) network models. However, the present invention is not limited thereto.

[0057] The model generation unit (240) can generate multiple candidate models and evaluate the performance of each candidate model to select a model that satisfies preset criteria as the final energy performance model. The model performance evaluation can be performed based on the coefficient of determination (R-squared), significance probability (P-value), F-test, or cross-validation error. Additionally, the finally selected energy performance model can calculate predicted energy consumption that reflects the actual user behavior state when a physical factor matrix and an actual virtual factor matrix are input during the time interval to be evaluated, and can calculate predicted energy consumption that reflects the reference user behavior state when the same physical factor matrix and a reference virtual factor matrix are input. Through this, the reduction amount calculation unit (260) can separate and calculate the behavior-based energy reduction amount by utilizing the difference in predicted values ​​based on the difference in virtual factors under the same physical conditions.

[0058]

[0059] Referring to Equation 5, the energy performance model can be expressed in the form of a general function. Here can represent the predicted energy consumption at a specific time interval t. The function f can represent the energy performance model function itself, which predicts energy consumption by taking physical and virtual factors as inputs. The inputs to the above function f can be the physical factor vector Xt and the virtual factor vector Zt at the time interval t. Here, X t may include physical conditions such as external conditions, solar radiation, power import volume, and energy storage device status, and Z tmay include user behavior or equipment operation conditions such as occupancy status, the number of occupants, and the operation patterns of cooling, heating, lighting, and ventilation equipment. Accordingly, Equation 5 may be a mathematical expression indicating that energy consumption can be determined by the complex interaction of physical environmental factors and user behavior factors. Additionally, the specific form of the function f may be determined through the learning process of the model generation unit (240). Meanwhile, the function f is the same physical factor vector X t Regarding the actual virtual argument vector Z t Alternatively, a reference virtual factor vector can be optionally received to calculate different predicted energy usage, which can then be utilized for the separate calculation of behavior-based energy savings.

[0060] The baseline generation unit (250) can generate a dynamic energy baseline using the energy performance model generated by the model generation unit (240). The baseline generation unit (250) can calculate the predicted energy consumption by inputting a physical factor matrix and a virtual factor matrix corresponding to the time interval to be evaluated into the energy performance model. At this time, the virtual factor matrix may be a real virtual factor matrix reflecting the actual user behavior state or a reference virtual factor matrix reflecting the reference user behavior state. The energy performance model can predict the expected energy consumption during the corresponding time interval based on the input physical factor matrix and virtual factor matrix.

[0061] In one embodiment, the baseline generation unit (250) can input the physical factor matrix and the reference virtual factor matrix of the time interval to be evaluated into the energy performance model to calculate a reference predicted energy consumption that reflects the reference user behavior state. The reference predicted energy consumption can be used as a dynamic energy baseline. Additionally, the baseline generation unit (250) can input the physical factor matrix and the actual virtual factor matrix of the same time interval to be evaluated into the energy performance model to further calculate a predicted energy consumption that reflects the actual user behavior state. Accordingly, the reduction amount calculation unit (260) can calculate the total energy reduction amount by separating it into a physical-based energy reduction amount and a behavior-based energy reduction amount by comparing the reference predicted energy consumption, the predicted energy consumption reflecting the actual user behavior state, and the actual energy consumption.

[0062] The above dynamic energy baseline may not be a fixed reference value. That is, the dynamic energy baseline can be dynamically adjusted to reflect external environmental conditions that change over time. In addition, a reference virtual factor matrix distinct from the actual virtual factor matrix may be separately established for the separate calculation of behavior-based energy savings. Through this, the baseline generation unit (250) can generate a dynamic energy baseline in a form that allows the influence of changes in user behavior to be separated under the same physical conditions.

[0063]

[0064] Referring to Equation 6, the process of calculating the dynamic energy baseline can be expressed. Here, B(t) may represent the dynamic energy baseline value at time t. The above dynamic energy baseline value B(t) is the physical factor vector of the evaluation period in the energy performance model function f. and a reference virtual argument vector reflecting the reference user behavior state It can be obtained by inputting. That is, the dynamic energy baseline B(t) can be the predicted energy consumption when the actual physical conditions of the time interval under evaluation are reflected, but the user behavior conditions are set to a reference state.

[0065] For example, one might assume that the amount of energy saved after a household starts an energy saving campaign is being evaluated. In this case, physical factors such as the actual outside temperature and solar radiation during the campaign period ( ) can be input directly into the energy performance model.

[0066] On the other hand, virtual factors such as occupancy status and equipment usage patterns are not actual behavioral states after the campaign, but rather reference virtual factor vectors corresponding to states before the campaign or without reduction measures applied ( It can be set to ). Through this, the dynamic energy baseline B(t) can be calculated.

[0067] The above dynamic energy baseline B(t) can serve as a reference point for comparison with actual measured energy consumption. In addition, the same physical factor vector ( The actual virtual argument vector( By comparing the predicted value entered with the dynamic energy baseline B(t), the behavior-based energy savings resulting from changes in user behavior can be calculated. Furthermore, by comparing the dynamic energy baseline B(t) with the actual energy usage, it can be used to calculate the total energy savings or the physical-based energy savings.

[0068] The reduction amount calculation unit (260) can calculate the energy reduction amount by comparing the dynamic energy baseline generated by the baseline generation unit (250) with the actual energy usage corresponding to the time interval to be evaluated. Here, the dynamic energy baseline may be a predicted energy usage that reflects the actual physical conditions of the time interval to be evaluated and reflects the standard user behavior state. Additionally, the actual energy usage may be the total energy usage actually measured during the time interval to be evaluated. The reduction amount calculation unit (260) can calculate the total energy reduction amount by subtracting the actual energy usage value of the same time interval from the dynamic energy baseline value of the specific time interval. If the calculated value is positive, it can be determined that energy has been saved because the actual usage is less than the predicted usage. Conversely, if the calculated value is negative, it can be interpreted that additional energy has been used because the actual energy usage is greater than the predicted usage.

[0069] In one embodiment of the present invention, the energy saving amount calculation unit (260) can calculate the total energy saving amount by separating it into a physical-based energy saving amount and a behavior-based energy saving amount. Specifically, the energy saving amount calculation unit (260) can apply a real virtual factor matrix reflecting the actual user behavior state and a reference virtual factor matrix reflecting the reference user behavior state to the energy performance model, respectively, while maintaining the same physical factor matrix for the same evaluation target time interval. Accordingly, the energy saving amount calculation unit (260) can calculate a first predicted energy usage with the real virtual factor matrix applied and a second predicted energy usage with the reference virtual factor matrix applied.

[0070] In this case, the second predicted energy consumption may correspond to a dynamic energy baseline, and the difference between the first predicted energy consumption and the second predicted energy consumption may be calculated as a behavior-based energy saving amount attributable to changes in user behavior. Additionally, the difference between the first predicted energy consumption and the actual energy consumption may be calculated as a physical-based energy saving amount attributable to physical factors such as improvements in physical facilities, enhancement of envelope performance, or improvement in facility efficiency. Through this, the saving amount calculation unit (260) can separate and quantitatively evaluate multiple contributing factors constituting the total energy saving amount.

[0071]

[0072] Referring to Equation 7, the method for calculating energy savings can be defined. Here, S(t) may represent the total energy savings over a specific time interval t. Additionally, B(t) may represent the dynamic energy baseline over the time interval t, and E act (t) may represent the actual energy consumption measured in the time interval t. Accordingly, Equation 7 may represent the principle of calculating the total energy savings based on the difference between the predicted energy consumption, which reflects the reference user behavior state and the actual physical conditions of the time interval being evaluated, and the energy consumption actually measured in the same time interval.

[0073] In other words, the total energy savings mentioned above can be defined as the difference between the estimated energy consumption in a baseline state (corresponding to the absence of energy-saving efforts) and the actual energy consumption after the application of energy-saving efforts. Since the present invention utilizes a dynamically calculated energy baseline rather than a fixed reference value, the energy saving effect can be evaluated while reflecting changes in actual physical conditions, such as ambient temperature, solar radiation, humidity, power imports, or the status of energy storage devices, during the time interval under evaluation. For example, even in situations where heating demand increases due to a drop in ambient temperature, the corresponding physical condition can be reflected in the dynamic energy baseline; thus, it is possible to distinguish between simple seasonal changes and actual savings effects for a more accurate evaluation.

[0074] In addition, S(t) calculated by the above mathematical formula 7 may be the total energy saving amount, and subsequently, the behavior-based energy saving amount may be calculated separately by comparing the predicted values ​​to which the actual virtual factor and the reference virtual factor are applied, respectively, under the same physical conditions. Furthermore, the physical-based energy saving amount may be calculated by excluding the behavior-based energy saving amount from the above total energy saving amount, or by comparing the predicted energy usage reflecting the actual virtual factor with the actual energy usage.

[0075]

[0076] Referring to Equation 8, the calculated total energy savings can be expressed by separating them into two components. That is, the total energy savings S(t) is the physics-based energy savings S phys (t) and action-based energy savings S beh It can be expressed as a sum of (t). Here, S phys (t) may represent the amount of savings attributable to changes in physical factors, such as replacement of high-efficiency equipment, improvement of insulation performance, improvement of equipment efficiency, or improvement of control logic. In addition, S beh(t) may represent the amount of savings resulting from changes in user behavior patterns, such as the user adjusting the indoor temperature setting, turning off unnecessary lights, or operating home appliances while avoiding high energy consumption times.

[0077] Such separate calculation can be made possible through a structure that uses both physical factors and virtual factors as inputs to an energy performance model. Specifically, for the same time interval subject to evaluation, if actual virtual factors and reference virtual factors are applied while keeping physical factors the same, the difference between the two predicted values ​​can be calculated as the behavior-based energy saving amount. Additionally, physical-based energy saving amount can be calculated using the difference between the predicted energy consumption reflecting actual virtual factors and the actual energy consumption. Accordingly, the saving amount calculation unit (260) can separate and quantitatively evaluate multiple contributing factors constituting the total energy saving amount.

[0078] Through such separate calculation functions, the causes of savings can be analyzed more specifically. For example, it is possible to distinguish to what extent the savings are attributed to equipment replacement or building performance improvements, or to changes in actual user behavior. This allows for providing personalized energy-saving feedback to users or serving as a basis for designing energy-saving policies or incentives. Meanwhile, as described in this specification, energy production data, such as solar power generation, may be included in the input of the energy performance model as part of physical factors; however, it may be processed separately from physical-based energy savings to prevent duplication with the generation contribution reflected separately when calculating the integrated net metering transaction unit.

[0079] The transaction unit generation unit (270) can generate a tradable data unit based on the energy savings calculated by the savings amount calculation unit (260) or the verified savings amount that has undergone reliability verification. In one embodiment of the present invention, the transaction unit generation unit (270) can combine the verified savings amount with solar power generation amount based on energy production data. Through this, an integrated net metering transaction unit can be generated. The integrated net metering transaction unit can be a virtual energy asset utilized in a Virtual Net Metering (VNM) system or an energy trading platform.

[0080] In one embodiment of the present invention, the transaction unit generation unit (270) can perform the function of generating aggregated power resources by aggregating integrated net metering transaction units generated from individual building energy nodes. The transaction unit generation unit (270) can aggregate savings and power generation amounts generated from multiple participants to form a single resource. The generated aggregated power resource can participate in the power market and be utilized as a Demand Response resource. Additionally, the generated aggregated power resource can be linked with a Virtual Power Plant (VPP) operating system and utilized in energy services that contribute to the stability of the power grid. The data generated by the transaction unit generation unit (270) may be an information structure to support trading and settlement. This can be distinguished from existing power transmission and distribution systems that represent the physical flow of power.

[0081] FIG. 3 is an illustrative diagram for explaining a physical factor generation unit and a virtual factor generation unit according to an embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a physical factor matrix according to an embodiment of the present invention.

[0082] Referring to FIGS. 3 and 4, the energy saving amount calculation device (20) may include a configuration that processes physical data and virtual factor data, respectively, and converts them into a matrix form suitable for analysis. Specifically, the physical factor generation unit (220) may receive physical data, undergo a data preprocessing process, and then generate a physical factor matrix (PFM). Similarly, the virtual factor generation unit (230) may receive virtual factor data, perform data preprocessing, and then generate a virtual factor matrix (VFM).

[0083] The physical factor generation unit (220) can perform the role of receiving physical data from various sources and processing it into a standardized data structure. The physical data input to the physical factor generation unit (220) may include energy production data, energy consumption data, and environmental data. For example, the energy production data may be solar power generation data, wind power generation data, geothermal power generation data, fuel cell power generation data, small hydropower generation data, biomass power generation data, and energy harvesting data. However, the present invention is not limited thereto. In addition, the energy consumption data may be power consumption data, power import and export power data, charging and discharging data of an Energy Storage System (ESS), smart meter measurement data, power meter measurement data, gas usage data, and water usage data. However, the present invention is not limited thereto. The environmental data may be external temperature data, solar radiation data, humidity data, weather data, wind speed data, fine dust concentration data, and precipitation data. However, the present invention is not limited thereto. These physical data may be collected from different sensors, meters, inverters, or external data providers, and as a result, the measurement cycle, data format, occurrence of missing data, and noise levels may differ. Therefore, the physical factor generation unit (220) may perform a data preprocessing process to convert raw data with different characteristics to meet common analysis criteria.

[0084] The data preprocessing process may include multiple steps, including outlier removal, missing value correction, normalization, and time synchronization. The physical factor generation unit (220) may perform outlier detection and processing logic to remove abnormal measurements that may occur due to sensor malfunction, errors in the communication process, or temporary system failures. For example, outliers may be identified using statistical techniques based on mean and standard deviation.

[0085]

[0086] Mathematical Equation 9 can represent a criterion for determining whether a data point deviates significantly from a normal distribution. In Mathematical Equation 9, can represent a specific measurement of the i-th data item at time t. Also, can mean the statistical average value of the i-th data item over a long period. can represent the standard deviation of the same data item. Under the condition of mathematical formula 9, that is, if the absolute value of the difference between a specific data value and the mean value exceeds three times the standard deviation, the corresponding data point may be considered an outlier. For example, when external temperature data at a specific time follows a distribution with a mean of 25 degrees and a standard deviation of 2 degrees, if a value is measured that is more than 3 standard deviations (6 degrees) away from the mean, such as 32 degrees or 18 degrees, the physical factor generation unit (220) may determine this value as an outlier. The determined outlier may be removed from the system or replaced through a correction process using adjacent normal data values, thereby preventing distortion that may occur during the model learning process and improving the reliability of the analysis.

[0087] The physical factor generation unit (220) and the virtual factor generation unit (230) can perform a missing value correction function to process values ​​that are missing during the data collection process. If data at a specific point in time is not collected due to a communication interruption or a temporary shutdown of the equipment, the empty values ​​can be filled with appropriate values ​​to maintain the continuity of the model. For example, a linear interpolation method can be applied to estimate the missing values ​​using valid data located before and after the point in time when the missing values ​​occurred.

[0088]

[0089] Mathematical Equation 10 can mathematically express a linear interpolation method that corrects missing values ​​in data with temporal continuity. In Mathematical Equation 10 can represent the corrected value for the i-th data item at a specific time t where a missing value occurred. t1 may represent the nearest valid data point immediately before the missing point, and t2 may represent the nearest valid data point immediately after the missing point. class can represent the actual measured valid data at time points t1 and t2, respectively. This equation can estimate missing values ​​by calculating the value at the position corresponding to time t on the straight line connecting the two valid data points. For example, if the power consumption at 10:00 is 100 kWh and the consumption at 12:00 is 120 kWh, but the data for 11:00 is missing, the consumption at 11:00 can be corrected to 110 kWh according to Equation 10. Through this correction process, data incompleteness can be resolved, and the quality of input data for time series data analysis models can be improved.

[0090] The physical factor generation unit (220) and the virtual factor generation unit (230) can perform a normalization process to convert various data items having different units and ranges into a consistent scale. For example, temperature data may have a range of tens in Celsius (°C) units, while solar radiation data may have a range of hundreds or thousands in W / m² units. Since this difference in scale can cause problems where certain variables have an excessive influence on model learning, normalization may be required to adjust all data items to a common range between 0 and 1 or between -1 and 1.

[0091]

[0092] Equation 11 can represent the Min-Max Normalization method for scaling data. In Equation 11 can mean the data value after normalization is completed. can represent the original measurement value at time t. class and can represent the minimum and maximum values ​​within the entire period or reference period, respectively, of the corresponding t-th data item. This formula calculates the relative position that a specific data value occupies within the entire data range and can convert it into a value between 0 and 1. For example, when the minimum value of solar radiation data is 0 W / m² and the maximum value is 1000 W / m², the measured value of 750 W / m² at a specific point in time can be converted into a normalized value of 0.75 through the calculation of (750-0) / (1000-0). Through this normalization process, all input variables can be made to contribute to model training on an equal scale, which can contribute to increasing the convergence speed of the artificial intelligence model and stabilizing its performance.

[0093] In the time synchronization process, the physical factor generation unit (220) and the virtual factor generation unit (230) can align multiple data streams collected at different intervals based on a common time interval. For example, smart meter data can be collected at 15-minute intervals, solar inverter data at 5-minute intervals, and external weather data at 1-hour intervals. To integrate and analyze this data, the physical factor generation unit (220) can set a common time interval, for example, a 15-minute unit, and aggregate or rearrange all data into representative values ​​(e.g., average value, sum value, final value) of that interval. Through this process, a data matrix with the time axis aligned can be generated, which can serve as a basis to ensure that the energy performance model is learned with the physical state and user behavior at a specific point in time accurately corresponding.

[0094] The virtual factor generation unit (230) can perform the role of receiving and processing virtual factor data related to user behavior and the operation of the facility. The virtual factor data input to the virtual factor generation unit (230) may include user behavior data, occupancy status data, and facility operation pattern data. For example, user behavior data may include other heating usage pattern data, home appliance usage history data, away mode setting data, sleep mode setting data, energy saving mode participation data, electric vehicle charging schedule data, and Demand Response (DR) participation history data. However, the present invention is not limited thereto. Occupancy status data may include at least one of occupancy status data and occupancy count data. Facility operation pattern data may include cooling operation pattern data, heating operation pattern data, hot water usage pattern data, lighting usage pattern data, ventilation facility operation pattern data, HVAC operation data, and dehumidifier operation data. However, the present invention is not limited thereto. The virtual factor generation unit (230) can generate a virtual factor matrix (VFM) by performing the same preprocessing steps described above, such as outlier removal, missing value correction, normalization, and time synchronization, based on this data.

[0095] Referring to FIG. 4, a specific example of a physical factor matrix (PFM) can be illustrated in a table format. The table illustrated in FIG. 4 may represent the flow of time in the row direction and various physical data items in the column direction. For example, the table in FIG. 4 may include fields such as Time, power consumption characteristic data (E_load), solar power generation amount (E_pv), ambient temperature (Temp), solar irradiance (Solar), humidity (Humidity), load state (Load_state), and energy storage device state (ESS_state). Here, E_load, E_pv, and ESS_state may be example fields corresponding to L(t), G(t), and Bess(t) of Equation 1, respectively. Such a structure can demonstrate that physical factor vectors are constructed to comprehensively represent the physical energy state of a building for specific time intervals, and that a time series set of these physical factor vectors forms a physical factor matrix.

[0096] Each field in Fig. 4 can represent a specific state of the energy system. The Time field can express the time-series characteristics of the data by representing sequential time indices such as t1, t2, and t3. Power consumption characteristic data (E_load) can represent power consumption characteristics by facility, circuit, or application during the corresponding time interval, and solar power generation amount (E_pv) can represent the amount of power produced through the solar power generation system. Outdoor temperature (Temp), solar irradiance (Solar), and humidity (Humidity) can represent major external environmental factors that directly affect the building's energy consumption, particularly heating and cooling loads. Load state (Load_state) can be a binary state value indicating whether major electrical facilities are operating (ON) or stopped (OFF). Energy storage device state (ESS_state) can be a multi-state value indicating whether the energy storage device is charging (Charging), idle (Idle), or discharging (Discharging).

[0097] For example, the data at time t1 in FIG. 4 may indicate a state where power consumption characteristic data is 1.5 kWh, solar power generation is 0.8 kWh, the load is operating, and the energy storage device is charging. Subsequently, at time t3, it can be confirmed that both power consumption characteristic data and solar power generation have increased, the load has stopped, and the energy storage device has switched to a discharged state. A physical factor matrix constructed in this manner can systematically record dynamic changes in energy production, consumption, storage, and the external environment over time, and can function as core input data necessary for the model generation unit (240) to learn the complex correlation between these change patterns and the actual energy usage. In addition, this data structure can be utilized as basic data for the reduction amount calculation unit (260) to separate and calculate the total energy reduction amount into physical-based energy reduction amount and behavior-based energy reduction amount by distinguishing and applying actual virtual factors and reference virtual factors under the same physical conditions. Meanwhile, solar power generation data may be included in the input of the energy performance model as part of the physical factors, but it may be processed separately from the generation contribution reflected when calculating the integrated net metering unit.

[0098] In one embodiment of the present invention, the physical factor matrix (PFM) may additionally include data related to power quality. For example, qualitative aspects of energy usage may be reflected in the analysis by including data such as voltage levels, current magnitude, power factor, frequency fluctuations, and Total Harmonic Distortion (THD). This information can be utilized to model the impact of abnormal operating conditions of specific facilities or power grid instability on energy consumption, and can enable more precise energy performance analysis.

[0099] In conclusion, the physical factor generation unit (220) and the virtual factor generation unit (230) can perform the key role of collecting and refining raw data and converting it into a time-series aligned matrix structure. The physical factor matrix (PFM) and virtual factor matrix (VFM) thus generated can represent the building's physical conditions and the user's behavioral conditions as structured data by separating them. Subsequently, the model generation unit (240) combines these two matrices and uses them as input, thereby enabling the construction of an integrated energy performance model that simultaneously considers the impact of external environmental changes and user behavior changes on energy consumption. This can serve as a foundation for the inventor to implement a technical concept that goes beyond simple energy usage comparison, analyzes the fundamental drivers of energy consumption patterns, and generates a dynamic baseline based on this analysis.

[0100] FIG. 5 is an illustrative diagram for explaining the process of generating a verification reduction amount of an AI-MV analysis engine in one embodiment of the present invention.

[0101] Referring to FIG. 5, the AI-MV analysis engine can perform sequential processing steps of generating an energy performance model, generating a dynamic energy baseline, calculating energy savings, calculating a reliability index, and calculating verified savings. Here, the AI-MV analysis engine can be understood as an integrated concept including a model generation unit (240), a baseline generation unit (250), and a savings calculation unit (260). Each component can organically interact with one another to perform quantitative evaluation and verification of energy savings.

[0102] The model generation unit (240) can generate one or more energy performance models based on a physical factor matrix, a virtual factor matrix, and actual energy usage data collected during a reference period. Here, the actual energy usage may be a value to be predicted for calculating the amount of energy saved, and each factor included in the physical factor matrix and the virtual factor matrix may be an explanatory variable for predicting the actual energy usage. The energy performance model can perform the function of mathematically modeling the energy usage characteristics of a building, thereby predicting the expected energy usage under specific physical and virtual conditions. The reference period data used for model training should, in principle, include continuous time-series data of at least 12 months to sufficiently learn the seasonal variability of heating and cooling loads and the annual behavioral patterns of users, but depending on the data collection environment, it may be composed by combining data from critical periods that are representative of each season (e.g., at least 2 weeks per quarter). The model generation unit (240) can build an optimized model by considering various characteristics such as the building's use, size, equipment system, and climate conditions. In this process, the model generation unit (240) can effectively reflect the time-series characteristics of the input data or the non-linear relationship between variables.

[0103] The model generation unit (240) can generate an energy performance model using various algorithms. For example, the energy performance model may be a statistical-based regression analysis model such as multiple linear regression and non-linear regression, a machine learning-based prediction model such as Support Vector Machine (SVM), Random Forest, and Gradient Boosting, a deep learning-based time series prediction model such as Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM), a building energy simulation model such as EnergyPlus and TRNSYS, or a hybrid model combining two or more of these. However, the present invention is not limited thereto. As an example, a statistical-based regression analysis model may have the advantage of being able to interpret the influence of physical factors and virtual factors on energy consumption in the form of coefficients.

[0104] In one embodiment of the present invention, the model generation unit (240) can generate a plurality of energy performance models and then compare and analyze the performance of each model according to a predetermined evaluation criterion. The model generation unit (240) can select the model with the best evaluation result as the final energy performance model. As performance evaluation criteria, the coefficient of determination (R2), the adjusted coefficient of determination (Adjusted R2), the significance probability (p-value), the F-test, the coefficient of variation (CV), the root mean square error (RMSE), or the cross-validation error may be utilized. For example, the model generation unit (240) can calculate the above evaluation indicators for each candidate model and then determine that the higher the coefficient of determination (R2) and the adjusted coefficient of determination (Adjusted R2), and the lower the root mean square error (RMSE) and the cross-validation error, the better. Additionally, the model generation unit (240) can calculate an overall performance score by assigning preset weights to multiple evaluation indicators, and can select the candidate model with the highest overall performance score or the smallest prediction error as the final energy performance model. Through this process, the model generation unit (240) can selectively apply a customized model that more accurately reflects the energy usage characteristics, seasonal variability, occupancy patterns, photovoltaic power generation characteristics, and equipment operation characteristics of a specific building.

[0105] The baseline generation unit (250) can perform the function of generating a dynamic energy baseline based on the final energy performance model selected by the model generation unit (240). The baseline generation unit (250) can calculate the predicted energy consumption by inputting the physical factor matrix and the virtual factor matrix corresponding to the time interval to be evaluated into the final energy performance model. At this time, the virtual factor matrix may be a real virtual factor matrix reflecting the actual user behavior state or a reference virtual factor matrix reflecting the reference user behavior state. In one embodiment, the baseline generation unit (250) can calculate a reference predicted energy consumption reflecting the reference user behavior state by inputting the physical factor matrix and the reference virtual factor matrix of the time interval to be evaluated into the energy performance model, and the reference predicted energy consumption may become the dynamic energy baseline. Additionally, the baseline generation unit (250) can further calculate the predicted energy consumption reflecting the actual user behavior state by inputting the physical factor matrix and the real virtual factor matrix of the same time interval to be evaluated into the energy performance model.

[0106] The dynamic energy baseline generated in this manner may not be a fixed value based on energy usage at a specific point in the past. That is, the dynamic energy baseline can be dynamically adjusted to reflect changes in environmental conditions such as outdoor temperature, solar radiation, and humidity, energy production conditions such as solar power generation, and building operation conditions such as occupancy status, cooling and heating operation patterns, and lighting usage patterns. However, for the separate calculation of behavior-based energy savings, the dynamic energy baseline may be generated as a counterfactual baseline that reflects the actual physical conditions of the time interval being evaluated, while setting user behavior conditions to a reference state. Accordingly, the dynamic energy baseline may represent the expected energy usage when no energy saving activities are performed during the time interval being evaluated. By generating such a dynamic baseline, the baseline generation unit (250) can improve the accuracy and reliability of the energy savings calculation. For example, even when the outdoor temperature on a specific day is higher than average and cooling demand increases, the dynamic energy baseline can prevent overestimation or underestimation of the savings by reflecting these changes in physical conditions.

[0107] The energy saving amount calculation unit (260) can calculate the total energy saving amount by comparing the dynamic energy baseline, the predicted energy usage reflecting the actual user behavior state, and the actual energy usage. Additionally, the energy saving amount calculation unit (260) can calculate the behavior-based energy saving amount using the difference between the predicted values ​​to which the actual virtual factor and the reference virtual factor are applied, respectively, under the same physical factor conditions, and can calculate the physical-based energy saving amount using the difference between the predicted energy usage reflecting the actual virtual factor and the actual energy usage. Through this, the total energy saving amount can be quantitatively evaluated by separating it into the physical-based energy saving amount and the behavior-based energy saving amount.

[0108] Subsequently, the AI-MV analysis engine calculates a reliability index based on the aforementioned calculated energy savings and can calculate the verified savings using the reliability index. Accordingly, the AI-MV analysis engine can integrally perform a series of analysis processes, from the energy performance model creation stage to baseline creation, savings calculation, reliability verification, and the derivation of the final verified savings.

[0109] The reduction amount calculation unit (260) can calculate the total energy reduction amount by comparing the dynamic energy baseline generated by the baseline generation unit (250) with the actual energy usage of the time interval to be evaluated. The total energy reduction amount can be calculated by subtracting the actual energy usage value from the dynamic energy baseline value. The reduction amount calculation unit (260) can additionally perform the function of separating the calculated total energy reduction amount into a physical-based energy reduction amount caused by physical factors and a behavior-based energy reduction amount caused by virtual factors such as changes in user behavior. This separate calculation can provide a basis for clearly identifying the cause of energy reduction and quantitatively evaluating the contribution of each factor.

[0110] The reduction amount calculation unit (260) can calculate a confidence index to evaluate the reliability of the calculated energy reduction amount. The confidence index can be calculated by comprehensively reflecting at least one of the model fit of the energy performance model, the model's prediction error rate, and the completeness of the data used for model training and baseline generation. For example, the model fit can be normalized using a statistical indicator such as the coefficient of determination, the prediction error rate can be converted into an accuracy indicator using an indicator such as RMSE or Mean Absolute Percentage Error (MAPE), and the completeness of the data can be calculated based on the ratio of missing values ​​in the total data. The reduction amount calculation unit (260) can calculate the final confidence index by assigning weights to each indicator.

[0111] The reduction amount calculation unit (260) can calculate the final verified reduction amount by applying the calculated reliability index as a verification coefficient to the total energy reduction amount. The verified reduction amount can be calculated by multiplying the total energy reduction amount by the reliability index, and this may be a result that reflects uncertainty based on the performance of the model and the quality of the data. Since the verified reduction amount calculated in this way is not a reduction amount based on simple prediction, but a result in which statistical significance and data reliability have been verified, it may be suitable for use as a settlement standard in virtual net metering (VNM) transactions or energy trading platforms. Therefore, the reduction amount calculation unit (260) can provide advanced analysis functions that go beyond simply calculating the energy reduction amount to perform separation by cause, reliability evaluation, and final verification.

[0112]

[0113] Mathematical Equation 12 may represent an example of a statistical-based regression analysis model. In the above mathematical equation can mean the predicted energy consumption at time t. can be a constant term representing the basic energy usage when all other factors are 0. represents the i-th physical factor value at time t, and m can mean the total number of physical factors. represents the j-th virtual argument value at time t, and n can mean the total number of virtual arguments. and ε₀ are regression coefficients for the i-th physical factor and the j-th fictitious factor, respectively, which can represent the magnitude and direction of the influence of each factor on energy consumption. Finally, may be an error term representing a random error that is not explained by the model. The model generation unit (240) can quantitatively model the relationship of various factors affecting energy consumption through such a regression equation.

[0114]

[0115] Equation 13 represents the total energy saving S(t) calculated in the time interval t as the physics-based energy saving S phys (t) and behavior-based energy savings S beh The relationship of distribution as (t) can be expressed. Specifically, S phys (t) is the physics-based contribution factor to the total energy savings S(t) in the time interval t It can be calculated by multiplying, and S beh (t) is the behavior-based contribution factor to the total energy savings S(t) above. It can be calculated by multiplying. Here may be a coefficient representing the proportion of the total energy savings attributable to physical factors, and may be a coefficient representing the proportion of the total energy savings attributable to changes in user behavior or equipment operation patterns. For example, It can reflect the contribution of physical improvement effects, such as replacement with high-efficiency equipment, improvement of insulation performance, improvement of equipment efficiency, or improvement of control performance. In addition, It can reflect the contribution of behavioral factors such as changes in occupancy patterns, adjustment of heating and cooling setting conditions, turning off lights, or changes in equipment operation habits. Based on this relationship, the reduction amount calculation unit (260) can quantitatively separate the total energy reduction amount into reduction amounts for each cause. Furthermore, the above-mentioned physical-based contribution coefficient and behavior-based contribution factor It can be defined to represent the relative contribution of each factor to the total savings, as shown in mathematical formula 14.

[0116]

[0117] Equation 14 is the physics-based contribution factor used in Equation 13 and behavior-based contribution factor It may be a formula for calculating. In the above mathematical formula can represent the proportion of the total energy savings attributable to physical factors in a specific time interval t, and θ can represent the proportion of the total energy savings attributable to changes in user behavior or equipment operation patterns during a specific time interval t. C phys (t) may be the contribution of physical factors corresponding to the difference between the predicted energy consumption, reflecting actual virtual factors, and the measured energy consumption in a specific time interval t. This may represent the degree of contribution of changes in energy consumption on the consumption side caused by physical factors such as improved equipment efficiency, improved insulation performance, or improved control performance, and can be distinguished from the contribution of solar power generation that is separately reflected when calculating the integrated net metering transaction unit. In addition, C beh(t) may be a virtual factor contribution corresponding to the difference between the predicted energy consumption reflecting a reference virtual factor and the predicted energy consumption reflecting an actual virtual factor at a specific time interval t. This may represent the degree of contribution caused by changes in user behavior or operating patterns, such as changes in occupancy patterns, adjustments to heating and cooling settings, turning off lights, or changes in equipment operation habits. In one embodiment, and are respectively C phys (t) and C beh (t) can be calculated by normalizing it to the total contribution amount, and accordingly, the sum of the two contribution ratios can be 1. The reduction amount calculation unit (260) calculates the above contribution ratios and then multiplies each of them by the total energy reduction amount S(t) to obtain the physical-based energy reduction amount S phys (t) and behavior-based energy savings S beh (t) can be finally calculated. Through this, the separate calculation of the total energy savings defined in Equation 13 can be quantitatively implemented.

[0118]

[0119] Mathematical Equation 15 may represent an example of a formula for calculating a reliability index that quantifies the reliability of the energy saving amount calculation results. In the above formula, CI(t) may represent the reliability index at time t or at a specific evaluation interval. Q fit (t) may be a normalized value of an indicator representing the goodness of fit of the energy performance model. Qerr(t) may be a normalized value of an accuracy indicator corresponding to the model's prediction error rate. Q comp(t) may be a normalized value of an indicator representing the completeness of the data used in the modeling. w1, w2, and w3 are weights assigned to the model fit, accuracy, and data completeness indicators, respectively, and can be set so that their sum is 1. The reduction amount calculation unit (260) can generate an objective reliability indicator that comprehensively considers various aspects of the reduction amount calculation process through this equation.

[0120]

[0121] Mathematical Equation 16 may be an equation representing the final step of calculating the verified reduction amount. In the above mathematical equation, S v (t) may represent the verified savings at time t. CI(t) may be the reliability index at time t calculated through Equation 15. S(t) may be the total energy savings at time t calculated from the difference between the dynamic energy baseline and the actual usage. This equation represents the process of deriving the final savings, corrected for uncertainty, by applying the verification factor CI(t), which reflects the reliability of the model and data, to the total energy savings. For example, if the reliability index is calculated as 0.9, this may mean that 90% of the calculated total savings can be recognized as a reliable, verified savings. The verified savings calculated in this way can contribute to ensuring transparency and fairness in energy trading.

[0122] FIG. 6 is an illustrative diagram for explaining a transaction unit generation unit according to an embodiment of the present invention.

[0123] Referring to FIG. 6, the transaction unit generation unit (270) can receive verified reduction amount data and solar power generation amount data. The transaction unit generation unit (270) can perform the function of generating an integrated net metering transaction unit. The transaction unit generation unit (270) can perform the function of generating aggregated power resources. The transaction unit generation unit (270) can transmit the generated aggregated power resource data to the energy trading node (30).

[0124] The transaction unit generation unit (270) can perform the role of combining verified reduction amount data calculated from the reduction amount calculation unit (260) and solar power generation amount data received through the data receiving unit (210). The combination process may be a process of integrating the energy production-side value and the energy demand-side value into a single data unit. The internal processing function of the transaction unit generation unit (270) may be divided into an integrated net metering transaction unit generation stage and a collective power resource generation stage.

[0125] The transaction unit generation unit (270) can first generate an integrated net metering transaction unit by combining the solar power generation amount and the verified reduction amount for a specific building energy node. The integrated net metering transaction unit may be a basic data unit generated at the individual building energy node level. The integrated net metering transaction unit may provide quantitative indicators that can be directly utilized in a Virtual Net Metering (VNM) system or an energy trading platform.

[0126]

[0127] Mathematical Equation 17 is an integrated net metering transaction unit (U) for a specific building energy node (n). n It can represent a relationship for calculating (t). The integrated net metering transaction unit is the solar power generation data (G) of the corresponding building energy node at time (t). n (t)) and verification reduction amount data (S v,n It can be calculated as the sum of (t). Solar power generation data (G n(t)) can represent the physical production of energy. Verification reduction data (S v,n (t)) may refer to a hypothetical energy supply achieved through energy efficiency improvements and behavioral changes. Integrated net metering trading unit (U n (t)) can function as a data unit that integrates physical energy production and virtual energy savings to express them as a single economic value.

[0128] The transaction unit generation unit (270) can then perform the function of aggregating integrated net metering transaction units generated from multiple building energy nodes. Through the aggregation process, the transaction unit generation unit (270) can gather small energy units from individual nodes to generate aggregated power resources of a size tradable in the market. The generation of aggregated power resources can be an important step in enabling small-scale distributed resources to participate in the market.

[0129]

[0130] Mathematical Equation 18 can represent a relationship for calculating aggregated power resource data (A(t)) at time (t). The aggregated power resource data (A(t)) is the integrated net metering transaction unit (U) of each building energy node (n) with respect to the total number of building energy nodes (N) to be aggregated. n It can be calculated by summing all (t). The simple summation method may be a method of calculating the total amount by considering the contribution of each node as having equal value. The simple summation method can be used to determine the total potential of aggregated resources.

[0131] In one embodiment of the present invention, the transaction unit generation unit (270) can calculate aggregated power resources by reflecting the individual characteristics and status of each building energy node. For example, factors such as data reliability, program participation status, real-time trading possibility, contract conditions, and market rules may differ from node to node. To reflect the node-specific factors, the transaction unit generation unit (270) uses a node-specific weight (w n (t)) can be calculated. The transaction unit generation unit (270) can generate aggregated power resources in a weighted summation manner by applying the calculated weights.

[0132]

[0133] Mathematical Equation 19 is weighted aggregate power resource data (A w (t)) can represent the relationship for calculating. Weighted aggregated power resource data is the integrated net metering transaction unit (U) of each building energy node (n). n (t)) weight(w) of the corresponding node n It can be calculated by summing all values ​​multiplied by (t). The transaction unit generation unit (270) can calculate the amount of effective resources that can be practically utilized for trading or operation, rather than the simple total amount, through a weighted summation method. Weight (w n The introduction of (t) can contribute to increasing the reliability and predictability of collective power resources.

[0134]

[0135] Equation 20 is the weight (w) assigned to the n-th building energy node at time (t). n The relationship for calculating (t) can be expressed. Weights (w n (t)) is the reliability coefficient (CI n (t)), participation coefficient (P n (t)), tradability coefficient (T n (t)), contract coefficient (C n (t)), and the market operating coefficient (M nIt can be calculated as a product of (t). Each coefficient can have a value between 0 and 1, reflecting specific conditions. The transaction unit generation unit (270) can comprehensively apply multiple evaluation criteria through the coefficients.

[0136] Reliability Factor (CI) n (t)) may be an indicator reflecting the confidence level of the verified savings calculated at the corresponding node. Confidence coefficient (CI n (t)) can be linked to the reliability index calculated by the reduction amount calculation unit (260). For example, under conditions where the model fit of the energy performance model is high, the prediction error rate is low, and the data completeness is excellent, the reliability coefficient (CI) n (t)) can be set to a high value, such as 0.95. On the other hand, data loss may occur frequently due to sensor malfunctions or communication failures. In such cases, the reliability factor can be set to a low value, such as 0.4. A low reliability factor may reduce the contribution to the collective power resources.

[0137] Participation coefficient (P n (t)) may be a binary indicator representing whether the node is participating in a specific program or trading session. For example, the node may be formally registered in a virtual net program or a set resource program. The node may express its intention to participate during the corresponding time interval. If both conditions are met, the participation factor may be set to 1. If the node is not registered in the program or has suspended participation, the participation factor may be set to 0. If the participation factor is set to 0, the node may be excluded from the set resource calculation for that time.

[0138] Tradability Factor (T n(t)) may be an indicator that reflects in real time whether the integrated net-margin transaction unit of the node is technically in a tradable state. For example, if a problem occurs such as a real-time communication failure, non-transmission of metering data, verification failure, failure to meet the minimum transaction volume, or a settlement system linkage error, the tradability factor may be set to 0. If the tradability factor is set to 0, the node may be unable to perform actual transactions. The node may be controlled so that it is not included in the aggregate resource.

[0139] The contract factor (Cn(t)) represents the percentage that can be reflected in aggregation according to the contract conditions between the node and the operator. For example, if it is agreed to utilize only 80% of the verified savings under contract with a specific building energy node as an aggregation resource, the contract factor can be set to 0.8. If a contract for 100% reflection is concluded with another node, the contract factor for that node can be set to 1.0.

[0140] Market Operating Coefficient (M n (t)) may be an adjustment factor assigned according to the operating rules of the power market operator or VPP operating system. For example, during peak hours when power system stability is important, the value of resources may be recognized at 100% and the market operating factor may be set to 1.0. On the other hand, during non-peak hours when the system is stable or when conservative resource operation is required, only 90% may be recognized and the market operating factor may be set to 0.9.

[0141] The energy trading node (30) may be a component that receives aggregated power resource data from the trading unit generation unit (270). The energy trading node (30) may be a collection of systems that perform actual energy trading, settlement, and resource operation based on the received data. Details regarding the energy trading node (30) that overlap with those described in FIG. 1 may be omitted. The energy trading node (30) may include an energy trading platform (310), an energy settlement system (320), and a virtual power plant (VPP) operation system (330).

[0142] The energy trading platform (310) can perform functions such as matching trading participants, calculating trading volume, or concluding contracts based on aggregated power resource data. The energy settlement system (320) can generate cost deduction, netting processing, or final settlement data for each participating node by utilizing aggregated power resource data. The virtual power plant operation system (330) can utilize VPP operation, such as bidding, responding to dispatch instructions, or contributing to grid stabilization, by treating resources aggregated from multiple building energy nodes as a single power plant.

[0143] In one embodiment of the present invention, a transaction unit generating unit (270) or an energy trading node (30) may be linked to a distributed ledger-based system to ensure the reliability and transparency of transaction data. The transaction unit generating unit (270) may generate a transaction record corresponding to a verified reduction amount, an integrated net metering transaction unit, or an aggregated power resource. The generated transaction record or the hash value of the record may be recorded in a distributed ledger-based system. The value recorded in the distributed ledger-based system may ensure the integrity of the data. The value recorded in the distributed ledger-based system may be used to prevent duplicate settlements.

[0144] A transaction record may be a data structure containing detailed information about a transaction. For example, a transaction record may include building energy node identification information, transaction target time interval information, verified savings information, solar power generation information, integrated net metering transaction unit information, aggregated power resource information, reliability index information, transaction creation time information, unique transaction identifier, settlement status information, information on transaction participants, and other metadata necessary for settlement or verification. However, the present invention is not limited thereto.

[0145] The transaction unit generation unit (270) can store the original of the generated transaction record in an internal database. The transaction unit generation unit (270) can record only the hash value generated from the original record in the distributed ledger. The method of recording only the hash value in the distributed ledger can reduce the burden of storage capacity. The method of recording only the hash value in the distributed ledger can provide a means to verify whether the transaction history has been tampered with or forged afterward. For example, at the settlement stage, the hash value can be recalculated from the transaction record in the internal database. The calculated hash value can be compared with the hash value recorded in the distributed ledger. Through comparison, the integrity of the data can be verified.

[0146] FIG. 7 is an illustrative diagram for explaining the function of the AI-MV integrated calculation system in the overall energy market system according to one embodiment of the present invention.

[0147] Referring to FIG. 7, the AI-MV-based integrated calculation system of the present invention can be deployed and function within an ecosystem structure within the overall energy market system. FIG. 7 exemplarily illustrates the value chain of the existing energy market and a future ecosystem based on Virtual Net Metering (VNM). The functional positions occupied by the building energy node (10), energy saving amount calculation device (20), and energy trading node (30) of the present invention within the overall market structure can be explained. According to the legend shown in FIG. 7, Internal Nodes may refer to nodes that serve as the center of value exchange within the energy market ecosystem. Auxiliary Nodes may refer to nodes that support market formation or institutional operation, even though they are not direct agents of transactions. Supporting Nodes may refer to nodes that perform technical or operational support functions.

[0148] The government can serve as the overarching entity designing the policies, systems, regulations, and support structures of the entire energy market. It can also establish indirect relationships with auxiliary enablers and various market participants. For instance, the government can determine the overall direction of the market ecosystem by establishing policies for the deployment of renewable energy, mandatory energy efficiency measures, and rules for the electricity trading market. Furthermore, the government can play a role in laying the legal and institutional foundations for the introduction of new business models, such as Virtual Net Metering (VNM) schemes or Virtual Power Plants (VPPs).

[0149] The energy market ecosystem can be structured according to the value chain of manufacturing, generation, distribution, and consumption. Tier 2 enablers may correspond to the manufacturing sector. For example, Tier 2 enablers may be polysilicon companies, ingot companies, module companies, PV manufacturers, wind turbine manufacturers, battery cell manufacturers, and inverter manufacturers. However, the present invention is not limited thereto. These companies can establish the physical foundation of the energy ecosystem by performing the role of producing and supplying materials, components, and finished products that constitute solar power generation facilities.

[0150] Tier 1 enablers may fall under the generating domain. Tier 1 enablers may include generating companies, private generating companies, small-scale photovoltaics (SV), and virtual power plants (VPPs). In addition to generating companies and private generating companies responsible for large-scale centralized generation, there may be small-scale photovoltaics (SPPs) that include multiple distributed energy resources. Virtual power plants (VPPs) can integrate distributed small-scale photovoltaics (SPPs), energy storage systems (ESS), and other resources using information and communication technology (ICT). Virtual power plants (VPPs) can play a role in participating in the electricity market by controlling these integrated resources as if they were a single power plant.

[0151] Service providers may be in the transmission and distribution or distribution sectors. For example, service providers may be the Korea Power Exchange (KPX), Korea Electric Power Corporation (KEPCO), virtual net metering (VNM) operators, demand response operators, electric vehicle charging operators, energy consulting firms, and data analysis service companies. However, the present invention is not limited thereto. The Korea Power Exchange (KPX) may be responsible for the operation of the power market and grid management. Korea Electric Power Corporation (KEPCO) may be responsible for the transmission and distribution of power, supply, and final billing settlement. A virtual net metering (VNM) operator may connect power produced from distributed resources with multiple electricity consumers. A virtual net metering (VNM) operator may be an entity that operates a virtual net metering system that offsets production and consumption.

[0152] Consumers may correspond to the consuming domain, which may be associated with the building energy node (10) of the present invention. Regarding the building energy node (10), details that overlap with those described in FIG. 1 are omitted. In the context of FIG. 7, the building energy node (10) may represent the end consumers of the energy market. For example, consumers may be industrial consumers, commercial consumers, individual consumers, public sector consumers, agricultural consumers, data centers, and electric vehicle users. However, the present invention is not limited thereto. Industrial consumers may participate in the trading of Renewable Energy Certificates (RECs). Individual consumers may participate in the use of renewable energy through a virtual net metering system, etc., even in a state where they do not possess solar power facilities (PV unhold).

[0153] Regarding the energy saving amount calculation device (20), details that overlap with those described in FIGS. 1 and FIG. 2 are omitted. In the ecosystem of FIG. 7, the energy saving amount calculation device (20) may be located between the Service Providers area and the Consumers area. The energy saving amount calculation device (20) may function as a data-based trusted data layer. The energy saving amount calculation device (20) may receive physical data and virtual factor data from multiple building energy nodes (10). Based on the received data, the energy saving amount calculation device (20) may perform the role of generating energy saving amount, verified saving amount, integrated net metering transaction unit, and aggregated power resource data.

[0154] The energy saving amount calculation device (20) can provide the generated reliable data to service providers. For example, the energy saving amount calculation device (20) can provide verified saving amount data to a virtual net metering operator (VNM Operator) to support accurate net metering settlement. Additionally, the energy saving amount calculation device (20) can generate aggregated power resources by aggregating integrated net metering transaction units calculated from multiple building energy nodes (10). The energy saving amount calculation device (20) can provide the generated aggregated power resources to an energy trading platform. Alternatively, the energy saving amount calculation device (20) can link the generated aggregated power resources with a virtual power plant (VPP) operating system.

[0155] In one embodiment of the present invention, the energy saving amount calculation device (20) can support the implementation of an artificial intelligence (AI)-based virtual net metering (AI-based VNM) system. The behavior-based energy saving amount data, which is precisely calculated by the energy saving amount calculation device (20) by taking into account changes in user behavior, can enhance existing physical power generation-based net metering models. Additionally, the energy saving amount calculation device (20) can provide a data basis that enables individual energy sharing trading. For energy trading between individuals or between buildings to be reliable, an objectively verified trading unit may be required. The integrated net metering trading unit generated by the energy saving amount calculation device (20) can perform this role.

[0156] In one embodiment of the present invention, the energy saving amount calculation device (20) can also be utilized in social and policy application fields. For example, the verified saving amount calculated through the energy saving amount calculation device (20) can be converted into a monetary value. The converted value can be used as a resource to support the energy poor (E poor). In addition, the verified saving amount can be converted into an energy voucher. The converted voucher can be used for the payment of renewable energy electricity bills, etc. This can contribute to establishing a virtuous cycle structure in which energy efficiency improvement activities lead to the creation of social value.

[0157] Regarding the energy trading node (30), details that overlap with those described in FIGS. 1 and FIG. 6 are omitted. In FIG. 7, the energy trading node (30) can comprehensively refer to the entities involved in market operation and trading, which are centered around Service Providers. The energy trading node (30) can receive aggregated power resource data from the energy saving amount calculation device (20). Based on the received data, the energy trading node (30) can perform power market trading, virtual net metering settlement, virtual power plant (VPP) resource operation, etc. The verified data provided by the energy saving amount calculation device (20) can increase the operational efficiency and transaction reliability of the energy trading node (30).

[0158] The ecosystem of FIG. 7 can be supported by various auxiliary enablers. For example, auxiliary enablers may be electric vehicles (EVs), energy storage systems (ESS), energy management systems (EMS), demand response (DR), smart grid infrastructure, blockchain-based trading systems, and Internet of Things (IoT) sensors. However, the present invention is not limited thereto. Energy management systems (EMS) and demand response (DR) resources can function as supporting nodes that optimize energy usage and regulate load on the consumer side. Electric vehicles (EVs) and energy storage systems (ESS) can function as auxiliary nodes that store energy or supply power to the grid (Vehicle-to-Grid; V2G). Electric vehicles (EVs) and energy storage systems (ESS) can contribute to increasing the flexibility and stability of the power grid.

[0159] In conclusion, FIG. 7 can show that the energy saving amount calculation device (20) of the present invention can function as a data intermediary and value creator of the future energy market ecosystem, going beyond a simple data analysis device. The energy saving amount calculation device (20) can connect distributed building energy nodes (10) with energy trading nodes (30), such as a virtual power plant (VPP) or a virtual net meter (VNM). The energy saving amount calculation device (20) can generate and distribute reliable data. Through this, the energy saving amount calculation device (20) can contribute to increasing the efficiency, transparency, and participation of the entire energy ecosystem.

[0160] FIG. 8 is a system flowchart for explaining a method for calculating energy savings according to an embodiment of the present invention.

[0161] Referring to FIGS. 1 to 8, an energy saving amount calculation method according to one embodiment of the present invention can be implemented as a data processing flow performed between a building energy node (10), an energy saving amount calculation device (20), and an energy trading node (30).

[0162] In step S101, the building energy node (10) can collect physical data and virtual factor data for a reference period. Here, the physical data may include energy production data, energy consumption characteristic data, and environmental data. Additionally, the virtual factor data may include user behavior data, occupancy status data, and equipment operation pattern data.

[0163] In step S102, the building energy node (10) can transmit the collected reference period data to the energy saving amount calculation device (20). Data transmission can be performed periodically or non-periodically via a wired or wireless communication network.

[0164] In step S103, the energy saving amount calculation device (20) can generate a physical factor matrix and a virtual factor matrix for a reference period. The energy saving amount calculation device (20) can perform preprocessing processes such as removing outliers, correcting missing values, or normalization on the received physical data and virtual factor data. Subsequently, the energy saving amount calculation device (20) can structure the preprocessed data into a physical factor matrix and a virtual factor matrix, respectively.

[0165] In step S104, the energy saving amount calculation device (20) can generate an energy performance model. The energy saving amount calculation device (20) can train the energy performance model based on a physical factor matrix of a reference period, a virtual factor matrix, and actual energy usage corresponding to the reference period. The generated energy performance model may be a mathematical model representing the energy usage characteristics of a specific building.

[0166] In step S105, the building energy node (10) can collect physical data and actual virtual factor data for the time interval to be evaluated. Here, the time interval to be evaluated may refer to a specific period for which the amount of energy saved is to be calculated.

[0167] In step S106, the building energy node (10) can transmit physical data and actual virtual factor data of the time interval to be evaluated to the energy saving amount calculation device (20). The data can be collected in real-time or near-real-time and transmitted to the energy saving amount calculation device (20).

[0168] In step S107, the energy saving amount calculation device (20) can generate a physical factor matrix and an actual virtual factor matrix for the time interval to be evaluated. This step can be performed in a manner similar to step S103. Additionally, the energy saving amount calculation device (20) can set a reference virtual factor matrix based on virtual factor data of a reference period or a reference user behavior pattern corresponding to a state where the reduction measure is not applied. Accordingly, the energy saving amount calculation device (20) can distinguish and use an actual virtual factor matrix reflecting the actual user behavior state and a reference virtual factor matrix reflecting the reference user behavior state for the same time interval to be evaluated.

[0169] In step S108, the energy saving amount calculation device (20) can generate a dynamic energy baseline for the time interval to be evaluated. The energy saving amount calculation device (20) can input the physical factor matrix and the reference virtual factor matrix for the time interval to be evaluated generated in step S107 into the energy performance model generated in step S104. Accordingly, the dynamic energy baseline generated may be a predicted energy usage that reflects the actual physical conditions of the time interval to be evaluated, while setting the user behavior conditions to a reference state. Additionally, the energy saving amount calculation device (20) can input the same physical factor matrix and the actual virtual factor matrix into the energy performance model to further calculate a predicted energy usage that reflects the actual user behavior state.

[0170] In step S109, the building energy node (10) can collect actual energy usage data corresponding to the time interval to be evaluated. Here, the actual energy usage data may be total energy usage data actually measured during the time interval to be evaluated.

[0171] In step S110, the building energy node (10) can transmit the collected actual energy usage data to the energy saving amount calculation device (20). The actual energy usage data can be used to accurately calculate the energy saving amount.

[0172] In step S111, the energy saving amount calculation device (20) can calculate the energy saving amount. Specifically, the energy saving amount calculation device (20) can calculate the total energy saving amount by comparing the dynamic energy baseline generated in step S108 with the actual energy usage received in step S110. For example, the total energy saving amount S(t) can be calculated as the value obtained by subtracting the actual energy usage E_act(t) from the dynamic energy baseline B(t). Additionally, the energy saving amount calculation device (20) can calculate the behavior-based energy saving amount using the difference between the predicted energy usage reflecting the reference virtual factor matrix and the predicted energy usage reflecting the actual virtual factor matrix under the same physical factor conditions, and can calculate the physical-based energy saving amount using the difference between the predicted energy usage reflecting the actual virtual factor matrix and the actual energy usage.

[0173] In step S112, the energy saving amount calculation device (20) can calculate a reliability index. The reliability index may be an indicator that comprehensively evaluates the performance of the energy performance model and the data quality. The energy saving amount calculation device (20) can calculate the reliability index by considering the model fit, the prediction error rate, and the data completeness.

[0174] In one embodiment of the present invention, the reduction amount calculation unit (260) may calculate a reliability index by reflecting at least one of the model fit, prediction error rate, and data completeness of the energy performance model. For example, the reduction amount calculation unit (260) may determine the reliability index by combining the coefficient of determination (R-squared), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), data omission rate, outlier rate, data collection cycle compliance rate, or sensor reliability score. However, the present invention is not limited thereto.

[0175] In step S113, the energy saving amount calculation device (20) can calculate a verified saving amount. The energy saving amount calculation device (20) can apply the reliability index calculated in step S112 as a verification coefficient to the total energy saving amount calculated in step S111. Accordingly, the verified saving amount calculated may be a value that reflects uncertainty due to model performance and data quality.

[0176] In step S114, the energy saving amount calculation device (20) can generate an integrated net metering transaction unit. The energy saving amount calculation device (20) can combine energy production data, such as solar power generation amount, with verified savings amount. Here, the verified savings amount may be a value corresponding to energy savings on the consumption side, and the solar power generation amount may be reflected as a separate generation contribution. Therefore, the integrated net metering transaction unit may be a data-based asset that integrates physical energy production and virtual energy savings without duplication.

[0177] In one embodiment of the present invention, the transaction unit generation unit (270) can combine solar power generation amount and verified reduction amount based on energy production data. The transaction unit generation unit (270) can generate an integrated net metering transaction unit used in a Virtual Net Metering (VNM) system or an energy trading platform (310).

[0178] In step S115, the energy saving amount calculation device (20) can generate aggregated power resources. The energy saving amount calculation device (20) can generate aggregated power resource data by aggregating integrated net metering transaction units generated from multiple building energy nodes (10). The aggregated power resource may be a data structure for managing multiple distributed energy resources as a single virtual resource.

[0179] In step S116, the energy saving amount calculation device (20) can transmit the generated aggregated power resource data to the energy trading node (30). Through the transmission of the data, market participation of distributed energy resources can be made possible.

[0180] In step S117, the energy trading node (30) can operate aggregated power resources. The energy trading node (30) can perform processing associated with trading, operation, and settlement based on the transmitted aggregated power resource data. For example, the energy trading platform (310) can determine the amount of market participation possible, and the energy settlement system (320) can set a settlement reference value.

[0181] In one embodiment of the present invention, the transaction unit generation unit (270) can generate aggregated power resources by aggregated the integrated net metering transaction units of a plurality of building energy nodes (10). The transaction unit generation unit (270) may be configured to provide aggregated power resources to an energy trading platform (310) or to link with a virtual power plant (VPP) operating system (330).

[0182] The energy saving calculation method of the present invention as described above can be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination.

[0183] The program instructions recorded on the above-mentioned computer-readable recording medium may be those specifically designed and configured for the present invention, or they may be those known and available to those skilled in the art of computer software.

[0184] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, and flash memory.

[0185] Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware device may be configured to operate as one or more software modules to perform processing according to the present invention, and vice versa.

[0186] Although various embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. It is understood that various modifications can be made by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention. Explanation of the symbols

[0187] 1: AI-MV-based Integrated Estimation System 10: Building Energy Node 110: Smart Meter 120: Solar inverter 130: Environmental sensor 140: Energy Management System 20: Energy saving calculation device 210: Data receiver 220: Physical Factor Generation Unit 230: Virtual argument generation section 240: Model creation section 250: Baseline generation section 260: Savings Calculation Unit 270: Transaction Unit Creation Section 30: Energy trading node 310: Energy Trading Platform 320: Energy Settlement System 330: Virtual Power Plant Operating System

Claims

Claim 1 A data receiving unit that receives physical data including energy production data, energy consumption data, and environmental data, and virtual factor data including user behavior data, occupancy status data, and equipment operation pattern data; a physical factor generating unit that preprocesses the physical data to generate a physical factor matrix for each predetermined time interval; a virtual factor generating unit that preprocesses the virtual factor data to generate a virtual factor matrix for each predetermined time interval; a model generating unit that generates an energy performance model based on the physical factor matrix, the virtual factor matrix, and actual energy usage during a reference period; a baseline generating unit that inputs the physical factor matrix and the virtual factor matrix corresponding to the time interval to be evaluated into the energy performance model to generate a dynamic energy baseline, which is the predicted energy usage for the time interval to be evaluated; and a reduction amount calculation unit that calculates the energy reduction amount by comparing the dynamic energy baseline with the actual energy usage corresponding to the time interval to be evaluated.An energy saving amount calculation device comprising: a saving amount calculation unit, wherein the saving amount calculation unit calculates the energy saving amount by separating it into a physical-based energy saving amount and a behavior-based energy saving amount; wherein the saving amount calculation unit calculates a reliability index by reflecting at least one of the model fit, prediction error rate, and data completeness of the energy performance model; and calculates a verified saving amount by applying the reliability index as a verification coefficient to the energy saving amount; wherein the energy saving amount calculation device further comprises a transaction unit generation unit that generates an integrated net metering transaction unit utilized in a Virtual Net Metering (VNM) system or an energy trading platform by combining the solar power generation amount based on the energy production data and the verified saving amount; wherein the transaction unit generation unit generates an aggregated power resource by aggregating the integrated net metering transaction units of a plurality of building energy nodes, and is configured to provide the aggregated power resource to an energy trading platform or link it with a Virtual Power Plant (VPP) operating system; wherein, when generating the integrated net metering transaction unit, the transaction unit generation unit combines the generation contribution corresponding to the solar power generation amount and the saving contribution corresponding to the verified saving amount An energy saving amount calculation device characterized by processing separately, structuring the integrated net metering transaction unit to include the generation contribution and the saving contribution as quantified data items respectively, and aggregating the integrated net metering transaction units of the plurality of building energy nodes to generate an aggregated power resource in the form of data including energy production and verified savings. Claim 2 An energy saving calculation device according to claim 1, wherein the energy production data includes solar power generation data, the energy consumption data includes at least one of power consumption data, power import and export data, charging and discharging data of an energy storage device, and smart meter or power meter measurement data, and the environment data includes at least one of external temperature data, solar radiation data, humidity data, and weather data. Claim 3 An energy saving amount calculation device according to claim 1, wherein the user behavior data includes other electric heating usage pattern data, the occupancy status data includes at least one of occupancy status data and occupancy personnel data, and the equipment operation pattern data includes at least one of cooling operation pattern data, heating operation pattern data, hot water usage pattern data, lighting usage pattern data, and ventilation equipment operation pattern data. Claim 4 An energy saving amount calculation device according to claim 1, wherein the preprocessing performs at least one of removing outliers, correcting missing values, normalization, and time synchronization. Claim 5 An energy saving amount calculation device according to claim 1, wherein the model generation unit generates a plurality of energy performance models and evaluates model performance based on at least one of the coefficient of determination, significance probability, F-test, and cross-validation error to select a final energy performance model. Claim 6 delete Claim 7 delete Claim 8 delete Claim 9 delete Claim 10 A method for calculating energy savings performed in an energy savings calculation device comprises: a data receiving step of receiving physical data including energy production data, energy consumption data, and environmental data, and virtual factor data including user behavior data, occupancy status data, and equipment operation pattern data; a physical factor generation step of preprocessing the physical data to generate a physical factor matrix for each predetermined time interval; a virtual factor generation step of preprocessing the virtual factor data to generate a virtual factor matrix for each predetermined time interval; a model generation step of generating an energy performance model based on the physical factor matrix, the virtual factor matrix, and actual energy usage during a reference period; a baseline generation step of inputting the physical factor matrix and the virtual factor matrix corresponding to the time interval to be evaluated into the energy performance model to generate a dynamic energy baseline, which is the predicted energy usage for the time interval to be evaluated; and a savings calculation step of calculating an energy savings amount by comparing the dynamic energy baseline with the actual energy usage corresponding to the time interval to be evaluated.A method for calculating energy savings, comprising: a step of calculating energy savings, wherein the step of calculating energy savings separates the energy savings into physical-based energy savings and behavior-based energy savings and calculates them; after calculating the energy savings, calculates a reliability index by reflecting at least one of the model fit, prediction error rate, and data completeness of the energy performance model; and calculates a verified savings amount by applying the reliability index as a verification coefficient to the energy savings amount; wherein the method for calculating energy savings further comprises a step of generating a transaction unit, wherein, after calculating the verified savings amount, the method combines the solar power generation amount based on the energy production data with the verified savings amount to generate an integrated net metering transaction unit utilized in a Virtual Net Metering (VNM) system or an energy trading platform; wherein the step of generating a transaction unit is characterized by aggregating the integrated net metering transaction units of a plurality of building energy nodes to generate aggregated power resources, and providing the aggregated power resources to an energy trading platform or linking them with a Virtual Power Plant (VPP) operating system; wherein, when generating the integrated net metering transaction unit, the generation contribution corresponding to the solar power generation amount and the verified savings amount corresponding to A method for calculating energy savings, characterized by processing savings contributions separately, structuring the integrated net metering transaction unit to include the generation contribution and the savings contribution as quantified data items, and aggregating the integrated net metering transaction units of the plurality of building energy nodes to generate an aggregated power resource in the form of data including energy production and verified savings.

Images