Method and apparatus for training power demand prediction model

Abstract

According to an embodiment of the present disclosure, a method for training a power demand prediction model may comprise the steps of: acquiring total power consumption data indicating the amount of power consumed by multiple home appliances in a single household; obtaining multiple pieces of segment data by dividing the total power consumption data at predetermined time intervals; obtaining multiple pieces of sequence data corresponding to the multiple pieces of segment data by embedding each of the multiple pieces of segment data; masking target sequence data included in the multiple pieces of sequence data; and performing unsupervised learning on a machine learning model to acquire at least one piece of context data on the basis of a correlation between the target sequence data and each piece of sequence data neighboring the target sequence data among the multiple pieces of sequence data.

Description

Method and device for learning a power demand forecasting model

[0001] The present disclosure relates to a method and device for learning a power demand prediction model.

[0002] This disclosure relates to research conducted with the support of the Small and Medium Business Technology Information Promotion Agency's Startup Growth Technology Development Project (TIPS) 'Development of a Data-Based Energy Consumption Decomposition Algorithm for Building Energy Conservation' (Project Identification Number: 1425177049, Project Number 00224500, Ministry Name: Small and Medium Venture Business, Research Period: 03 / 01 / 2023 ~ 02 / 28 / 2024).

[0003]

[0004] Recently, power supply is not sufficiently meeting demand. Addressing the growing demand by increasing power generation capacity can lead to resource waste and environmental pollution. Against this backdrop, accurately predicting power demand and detecting and managing abnormal power generation are becoming increasingly important technological issues for improving power efficiency.

[0005]

[0006] In order to solve the above-described problem, the present disclosure provides a learning method and device for a machine learning model configured to predict the power demand of a household based on the results of learning the household's power consumption pattern.

[0007]

[0008] According to one embodiment of the present disclosure, a method for learning a power demand prediction model includes the steps of: obtaining total power data representing the amount of power consumed by a plurality of home appliances in a household; obtaining a plurality of segment data by dividing the total power data into predetermined time intervals; obtaining a plurality of sequence data corresponding to the plurality of segment data by embedding each of the plurality of segment data; masking target sequence data included in the plurality of sequence data; and performing unsupervised learning on a machine learning model to obtain at least one context data based on a correlation between each of the sequence data neighboring the target sequence data among the plurality of sequence data, wherein the machine learning model can be configured to output predicted power data after the arbitrary power data based on the at least one context data in response to inputting arbitrary power data.

[0009] According to one embodiment, the step of obtaining a plurality of sequence data corresponding to the plurality of segment data by embedding each of the plurality of segment data includes the step of obtaining a plurality of individual power amount data representing the amount of power consumed by each of the plurality of home appliances by decomposing any segment data included in the plurality of segment data by frequency, and the step of obtaining one sequence data based on the plurality of individual power amount data, wherein the one sequence data can be included in the plurality of sequence data.

[0010] According to one embodiment, a method for learning a power demand prediction model may be such that one sequence data is configured as discretized data by performing K-means clustering based on a plurality of individual power data.

[0011] According to one embodiment, the method may further include a step of fine-tuning a machine learning model based on at least one context data.

[0012] According to one embodiment, a method for learning a power demand prediction model may further include a step of inputting arbitrary power data into a machine learning model to obtain expected power data, and a step of decomposing the expected power data by frequency to obtain a plurality of individual expected power data corresponding to each of a plurality of home appliances.

[0013] According to one embodiment, in response to detecting an outlier in the expected power consumption data, the method may further include extracting one appliance associated with the outlier from among the plurality of appliances.

[0014] According to another embodiment of the present disclosure, a computer program recorded on a computer-readable recording medium may be provided to execute a method for learning a power demand prediction model.

[0015] According to another embodiment of the present disclosure, an apparatus for learning a power demand prediction model includes a memory, at least one processor connected to the memory and configured to execute a computer-readable program included in the memory, wherein the program is configured to perform unsupervised learning on a machine learning model to obtain total power data representing the amount of power consumed by a plurality of home appliances in a household, obtain a plurality of segment data by dividing the total power data into predetermined time intervals, obtain a plurality of sequence data corresponding to the plurality of segment data by embedding each of the plurality of segment data, and obtain at least one context data based on a correlation between each of the plurality of sequence data neighboring the target sequence data and the target sequence data, and the machine learning model may be configured to output predicted power data after the arbitrary power data based on the at least one context data in response to inputting at least one sequence data based on the arbitrary power data.

[0016]

[0017] According to some embodiments of the present disclosure, an electricity demand forecasting model optimized for a specific household may be provided.

[0018]

[0019] FIG. 1 is a schematic diagram showing the operation of a power demand prediction model according to one embodiment of the present disclosure.

[0020] FIG. 2 is a schematic diagram showing a power consumption pattern learning process of a first machine learning module according to an embodiment of the present disclosure.

[0021] FIG. 3 is a schematic diagram of a power demand prediction model learning system (300) according to one embodiment of the present disclosure.

[0022] FIG. 4 is a block diagram showing the internal configuration of an information processing system and a user terminal according to one embodiment of the present disclosure.

[0023] FIG. 5 is a flowchart of a method for learning a power demand prediction model according to one embodiment of the present disclosure.

[0024]

[0025] Hereinafter, specific details for implementing the present disclosure will be described in detail with reference to the attached drawings. However, in the following description, specific descriptions of widely known functions or configurations may be omitted if they may unnecessarily obscure the gist of the present disclosure.

[0026] In the attached drawings, identical or corresponding components are assigned the same reference numerals. Furthermore, in the description of the embodiments below, duplicate descriptions of identical or corresponding components may be omitted. However, even if a description of a component is omitted, it is not intended that such component is not included in any embodiment.

[0027] The advantages and features of the disclosed embodiments, and the methods for achieving them, will become clearer with reference to the embodiments described below, along with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below and may be implemented in various different forms. These embodiments are provided solely to ensure the completeness of the disclosure and to fully inform those skilled in the art of the scope of the invention.

[0028] The terms used in this specification will be briefly explained, followed by a detailed description of the disclosed embodiments. The terms used in this specification have been selected from widely used, current terms, taking into account the functions of the present disclosure. However, these terms may vary depending on the intentions of engineers working in the relevant field, precedents, the emergence of new technologies, etc. Furthermore, in certain cases, terms may be arbitrarily selected by the applicant, in which case their meanings will be described in detail in the description of the relevant invention. Therefore, the terms used in this disclosure should not be defined simply as names of terms, but rather based on their meanings and the overall content of the present disclosure.

[0029] In this specification, singular expressions may include plural expressions unless the context clearly indicates that they are singular. Furthermore, plural expressions may include singular expressions unless the context clearly indicates that they are plural. When a part of the specification is said to "include" a component, this does not mean that other components are excluded, but rather that other components may be included, unless otherwise specifically stated.

[0030] FIG. 1 is a schematic diagram illustrating the operation of a power demand prediction model (100) (hereinafter, referred to as a “prediction model” or a “machine learning model”) according to one embodiment of the present disclosure. As illustrated, the prediction model (100) may be configured to output predicted power data (140) subsequent to the power data (130) in response to inputting power data (130). To this end, the prediction model (100) may include one or more machine learning modules (110, 120). For example, the prediction model (100) may include a first machine learning module (110) for learning the power demand pattern of a household from which the power data has been acquired based on previously acquired power data. Additionally or alternatively, the prediction model (100) may include a second machine learning module (120) configured to output predicted power data (140) following any power data (130) from power data (130) input in real time based on the result of learning the power demand pattern of a household.

[0031] Meanwhile, the operation of the prediction model (100) may be performed on at least one processor of an information processing system (or computing device). A detailed description of the configuration of the information processing system is described below in FIG. 4.

[0032] The processor can obtain total power data (not shown) representing the amount of power consumed by multiple home appliances within a household. For example, the processor can obtain daily power data representing the amount of power consumed by multiple home appliances within a household during a day. Then, the processor can learn the household's power consumption pattern from the total power data (or daily power data) through the first machine learning module (110). For this purpose, the first machine learning module (110) can include an encoder of a transformer. The operation of the first machine learning module (110) is described in more detail below in FIG. 2 .

[0033] The processor may update the second machine learning module (120) based on data associated with the power consumption pattern obtained from the first machine learning module (110). Specifically, the processor may perform fine-tuning on the second machine learning module (120) based on data associated with the power consumption pattern obtained from the first machine learning module (110). Accordingly, the power demand prediction performance of the second machine learning module (120) for the corresponding household may be optimized.

[0034] The processor can obtain power consumption data (130) representing the amount of power consumed by multiple home appliances in a household for a certain period of time (e.g., 10 minutes, 15 minutes, 30 minutes, 1 hour, etc.). In response, the processor can obtain data associated with the power consumption pattern of the household from the first machine learning module (110). That is, the data associated with the power consumption pattern obtained from the first machine learning module (110) can be transmitted to the second machine learning module (120). Then, the processor can obtain expected power consumption data (140) subsequent to the power consumption data (130) based on the data associated with the power consumption pattern through the second machine learning module (120).

[0035] FIG. 2 is a schematic diagram showing a power consumption pattern learning process of a first machine learning module (110) according to one embodiment of the present disclosure.

[0036] The processor can obtain total power data (210) representing the amount of power consumed by multiple home appliances in a household. Specifically, the processor can obtain total power data (210) (i.e., daily power data) representing the amount of power consumed by multiple home appliances in a household during a day. In Fig. 2, a process of performing learning using a single total power data (210) is illustrated, but the present invention is not limited thereto, and the processor can perform learning using multiple total power data (i.e., multiple daily power data).

[0037] The processor can obtain a plurality of segment data (210_1 to 210_n) by dividing the total power data (210) into regular time intervals. For example, the processor can obtain a plurality of segment data (210_1 to 210_n) by dividing the total power data (210) into 15-minute intervals. Then, the processor can perform learning on the first machine learning module (110) using the plurality of segment data (210_1 to 210_n). Hereinafter, the learning process performed by the processor in the first machine learning module (110) will be described.

[0038] The processor can obtain a plurality of sequence data (220_1 to 220_n) corresponding to each of the plurality of segment data (210_1 to 210_n) by embedding the plurality of segment data (210_1 to 210_n). Specifically, the processor can decompose the first segment data (210_1) by frequency to obtain a plurality of individual power consumption data corresponding to each of the plurality of home appliances installed in the corresponding household. For example, the processor can perform a Fourier transform on the first segment data (210_1) to obtain a plurality of individual power consumption data corresponding to each of the plurality of home appliances installed in the corresponding household. In this case, the plurality of individual power consumption data can be configured in the form of a Mel-spectrogram.

[0039] The processor can obtain one sequence data based on a plurality of individual power data. For example, the processor can obtain first sequence data (220_1) corresponding to first segment data (210_1) based on a plurality of individual power data. In this case, one sequence data can be configured as discretized data by performing K-means clustering based on the plurality of individual power data. For example, when the plurality of individual power data are vector data, one sequence data can be configured as data in the form of integers. By this configuration, noise collected in the process of measuring power through a measuring device is removed, so that the power consumption pattern appearing in the total power data (210) (or, the plurality of segment data (210_1 to 210_n)) can be analyzed more clearly.

[0040] The processor can mask any sequence data (hereinafter referred to as “target sequence data”) included in the plurality of sequence data (220_1 to 220_n). Then, the processor can perform unsupervised learning on a machine learning model (specifically, the first machine learning module (110)) to obtain at least one context data (230_1 to 230_n) based on a correlation between the target sequence data and each of the neighboring sequence data among the plurality of sequence data (220_1 to 220_n). For example, the processor can mask second sequence data (220_2) included in the plurality of sequence data (220_1 to 220_n) as target sequence data. Then, the processor can perform unsupervised learning on the machine learning model to obtain second context data (230_2) based on the correlation between the second sequence data (220_2) and each of the neighboring sequence data (220_1, 220_3 to 220_n) and the second sequence data (220_2).

[0041] The processor can fine-tune the second machine learning module (e.g., the second machine learning module (120) of FIG. 1) based on the acquired at least one context data (230_1 to 230_n). Here, the second machine learning module may mean a part of a machine learning model configured to output expected power data after any power data input to the machine learning model (e.g., the machine learning model (100)) based on the at least one context data (230_1 to 230_n). Accordingly, the machine learning model can be optimized to predict the power demand of a specific household based on the result of learning the power consumption pattern (i.e., the at least one context data (230_1 to 230_n)).

[0042] FIG. 3 is a schematic diagram of a power demand prediction model learning system (300) according to one embodiment of the present disclosure. As illustrated, the system (300) may include at least one of an information processing system (310), a database (320), and a user terminal (330).

[0043] The information processing system (310) can perform learning and / or inference of a power demand prediction model (e.g., a machine learning model (100)). That is, the information processing system (310) can represent a device configured to perform the operation of the 'machine learning model' described above in FIG. 1 and / or FIG. 2.

[0044] The information processing system (310) can obtain total power data representing the amount of power consumed by multiple home appliances in a household from the database (320). Additionally or alternatively, the information processing system (310) can also obtain any power data representing the amount of power consumed by multiple home appliances in a household in real time from the database (320). In this case, the 'power data' can be obtained through a measuring device (e.g., a smart meter) connected to the multiple home appliances. Furthermore, the 'power data' can also be obtained through a server (e.g., an IoT server) connected to the multiple home appliances.

[0045] The information processing system (310) may provide the results learned and / or inferred through the machine learning model to the user terminal (330). For example, the information processing system (310) may transmit the expected power data acquired from the machine learning model to the user terminal (330). Additionally or alternatively, the information processing system (310) may analyze the results learned and / or inferred through the machine learning model and provide the analyzed results to the user terminal (330). For example, the information processing system (310) may obtain information related to the occurrence of an outlier (e.g., peak power) from the expected power data acquired from the machine learning model and provide the information to the user terminal (330). As another example, the information processing system (#10) may identify a home appliance associated with an outlier from the expected power data acquired from the machine learning model and provide information related to the identified home appliance to the user terminal (330).

[0046] FIG. 4 is a block diagram showing the internal configuration of an information processing system (310) and a user terminal (330) according to one embodiment of the present disclosure.

[0047] The information processing system (310) may include a memory (412), a processor (414), a communication module (416), and an input / output interface (418). As illustrated in FIG. 4, the information processing system (310) and the user terminal (330) may be configured to communicate information and / or data via a network using the respective communication modules (416, 426). Similarly, the user terminal (330) may refer to any computing device capable of executing a program for controlling a power consumption management method and capable of wired / wireless communication. As illustrated, the user terminal (330) may include a memory (412), a processor (414), a communication module (416), and an input / output interface (418). In addition, the input / output device (430) may be configured to input information and / or data to the user terminal (330) or output information and / or data generated from the user terminal (330) via the input / output interface (428).

[0048] The memory (412, 422) may include any non-transitory computer-readable recording medium. According to one embodiment, the memory (412, 422) may include a non-permanent mass storage device such as a random access memory (RAM), a read only memory (ROM), a disk drive, a solid state drive (SSD), a flash memory, etc. As another example, a non-permanent mass storage device such as a ROM, an SSD, a flash memory, a disk drive, etc. may be included in the information processing system (310) or the user terminal (330) as a separate permanent storage device distinct from the memory. In addition, the memory (412, 422) may store an operating system and at least one program code (e.g., a code for executing a power consumption management program installed and run on the user terminal (330).

[0049] These software components may be loaded from a computer-readable recording medium separate from the memory (412, 422). This separate computer-readable recording medium may include a recording medium directly connectable to the information processing system (310) and the user terminal (330), and may include, for example, a computer-readable recording medium such as a floppy drive, a disk, a tape, a DVD / CD-ROM drive, a memory card, etc. As another example, the software components may be loaded into the memory (412, 422) through a communication module (416, 426) other than a computer-readable recording medium. For example, at least one program may be loaded into the memory (412, 422) based on a computer program that is installed by files provided by developers or a file distribution system that distributes installation files of applications over a network.

[0050] The processor (414, 424) may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to the processor (414, 424) by a memory (412, 422) or a communication module (416, 426). For example, the processor (414, 424) may be configured to execute instructions received according to program code stored in a storage device such as the memory (412, 422).

[0051] The communication modules (416, 426) may provide a configuration or function for the information processing system (310) and the user terminal (330) to communicate with each other via a network, and may provide a configuration or function for the information processing system (310) and / or the user terminal (330) to communicate with another user terminal or another system (e.g., a separate cloud system, etc.). For example, a request or data generated by the processor (424) of the user terminal (330) according to a program code stored in a recording device such as a memory (422) may be transmitted to the information processing system (310) via a network under the control of the communication module (426). Conversely, a control signal or command provided under the control of the processor (414) of the information processing system (310) may be received by the user terminal (330) via the communication module (426) of the user terminal (330) via the communication module (416) and the network.

[0052] The input / output interface (428) may be a means for interacting with an input / output device (430). Specifically, the input / output device (430) may include an input device such as a camera including an audio sensor and / or an image sensor, an Inertial Measurement Unit (IMU) sensor, a keyboard, a microphone, a mouse, etc. Additionally, the input / output device (430) may include an output device such as a display, a speaker, a haptic feedback device, etc. As another example, the input / output interface (418) may be a means for interfacing with a device that integrates a configuration or function for performing input and output, such as a touch screen.

[0053] In FIG. 4, the input / output device (430) is illustrated as not being included in the user terminal (330), but is not limited thereto, and may be configured as a single device with the user terminal (330). In addition, the input / output interface (418) of the information processing system (310) may be a means for interfacing with a device (not shown) for input or output that is connected to the information processing system (310) or that the information processing system (310) may include. In FIG. 4, the input / output interfaces (418, 428) are illustrated as elements configured separately from the processors (414, 424), but is not limited thereto, and the input / output interfaces (418, 428) may be configured to be included in the processors (414, 424).

[0054] The information processing system (310) and the user terminal (330) may include more components than those in FIG. 4. However, it is not necessary to clearly illustrate most of the conventional components. According to one embodiment, the user terminal (330) may be implemented to include at least some of the input / output devices (430) described above. In addition, the user terminal (330) may further include other components such as a transceiver, a global positioning system (GPS) module, a camera, various sensors, a database, etc. For example, if the user terminal (330) is a smartphone, it may include components that a smartphone generally includes, and for example, various components such as an acceleration sensor, a gyro sensor, a microphone module, a camera module, various physical buttons, buttons using a touch panel, input / output ports, and a vibrator for vibration may be implemented to be further included in the user terminal (330).

[0055] According to one embodiment, the processor (424) of the user terminal (330) may be configured to operate a program of the user terminal (330) that provides a power consumption management service. At this time, code associated with the program may be loaded into the memory (422) of the user terminal (330). While the program is being operated, the processor (424) of the user terminal (330) may receive information and / or data provided from the input / output device (430) through the input / output interface (428) or may receive information and / or data from the information processing system (310) through the communication module (426), and may process the received information and / or data and store the information and / or data in the memory (422). In addition, such information and / or data may be provided to the information processing system (310) through the communication module (426). Additionally, the processor (424) can receive data and / or instructions related to power (or amount of power) through an information processing system (310) connected via a network, and store the received data in the memory (422).

[0056] The processor (414) of the information processing system (310) may be configured to manage, process, and / or store information and / or data related to power (or amount of power) received from a plurality of user terminals and / or a plurality of external systems (e.g., database (130)). According to one embodiment, the processor (414) may manage, process, and / or store user input and data according to the user input received from the user terminal (330). Additionally or alternatively, the processor (414) may be configured to store and / or update a program for executing an algorithm used to provide a power consumption management service for the user terminal (330) from a separate cloud system, database, etc. connected to a network.

[0057] FIG. 5 is a flowchart of a method (500) for learning a power demand prediction model according to one embodiment of the present disclosure. The method (500) may be performed by at least one processor of an information processing system (e.g., information processing system (310)). Furthermore, the method (500) may begin with a step (S510) of acquiring total power data representing the amount of power consumed by multiple home appliances in a household.

[0058] The processor can obtain a plurality of segment data by dividing the total power data into regular time intervals (S520). Then, the processor can obtain a plurality of sequence data corresponding to the plurality of segment data by embedding each of the plurality of segment data (S530). Specifically, the processor can obtain a plurality of individual power data representing the amount of power consumed by each of the plurality of home appliances by decomposing any segment data included in the plurality of segment data by frequency, and obtain one sequence data based on the plurality of individual power data. In this case, one sequence data can be included in the plurality of sequence data. Additionally or alternatively, one sequence data can be configured as discretized data by performing K-means clustering based on the plurality of individual power data.

[0059] The processor may mask target sequence data included in the plurality of sequence data (S540). Then, the processor may perform unsupervised learning on the machine learning model to obtain at least one context data based on a correlation between the target sequence data and each of the neighboring sequence data among the plurality of sequence data (S550). Additionally, the processor may fine-tune the machine learning model based on the at least one context data. Meanwhile, the machine learning model may be configured to output predicted power data after the arbitrary power data based on the at least one context data in response to inputting arbitrary power data.

[0060] The processor may input arbitrary power consumption data into a machine learning model to acquire predicted power consumption data, and may also decompose the predicted power consumption data by frequency to acquire multiple individual predicted power consumption data corresponding to multiple home appliances. In this case, in response to detecting an outlier in the predicted power consumption data, the processor may extract one home appliance associated with the outlier from among the multiple home appliances.

[0061] As illustrated, the power control method may include a step of receiving power data associated with one household, a step of generating a prediction model for predicting power demand of one household using the power data, a step of generating an anomaly detection model for detecting power anomalies of one household using the power data, and a step of outputting a predicted power demand and anomaly detection data from the prediction model and the anomaly detection model.

[0062] The preceding description of the present disclosure is provided to enable those skilled in the art to make or use the present disclosure. Various modifications to the present disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to various modifications without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the examples described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0063] Although exemplary implementations may refer to utilizing aspects of the presently disclosed subject matter in the context of one or more standalone computer systems, the present subject matter is not so limited, but rather may be implemented in conjunction with any computing environment, such as a network or distributed computing environment. Furthermore, aspects of the presently disclosed subject matter may be implemented on or across multiple processing chips or devices, and storage may be similarly affected across multiple devices. Such devices may include personal computers, network servers, and handheld devices.

[0064] While the present disclosure has been described in connection with certain embodiments herein, it should be understood that various modifications and variations may be made without departing from the scope of the present disclosure, which would be apparent to those skilled in the art. Furthermore, such modifications and variations should be considered to fall within the scope of the claims appended to this specification.

Claims

1. A step of obtaining total power data representing the amount of power consumed by multiple home appliances in a household; A step of dividing the total power data into regular time intervals to obtain multiple segment data; A step of obtaining a plurality of sequence data corresponding to the plurality of segment data by embedding each of the plurality of segment data; A step of masking target sequence data included in the plurality of sequence data; and A step of performing unsupervised learning on a machine learning model to obtain at least one context data based on a correlation between each of the sequence data neighboring the target sequence data among the plurality of sequence data and the target sequence data. Including, A method for learning a power demand prediction model, performed by at least one processor, wherein the machine learning model is configured to output predicted power data following the arbitrary power data based on the at least one context data in response to inputting arbitrary power data.

2. In paragraph 1, The step of obtaining a plurality of sequence data corresponding to the plurality of segment data by embedding each of the plurality of segment data is as follows: A step of obtaining a plurality of individual power amount data representing the amount of power consumed by each of the plurality of home appliances by decomposing any segment data included in the plurality of segment data by frequency; and A step of obtaining one sequence data based on the above multiple individual power data Including, A method for learning a power demand prediction model, wherein the one sequence data is included in the plurality of sequence data, and is performed by at least one processor.

3. In paragraph 2, A method for learning a power demand prediction model performed by at least one processor, wherein the above one sequence data is composed of discretized data by performing K-means clustering based on the plurality of individual power data.

4. In paragraph 1, A step of fine-tuning the machine learning model based on at least one context data. A method for learning a power demand forecasting model, the method comprising:

5. In paragraph 1, A step of obtaining the expected power data by inputting the arbitrary power data into the machine learning model; and A step of decomposing the above expected power data by frequency to obtain a plurality of individual expected power data corresponding to each of the plurality of home appliances. A method for learning a power demand forecasting model, the method comprising:

6. In paragraph 5, In response to detecting an outlier in the above predicted power data, a step of extracting one home appliance associated with the outlier among the plurality of home appliances. A method for learning a power demand forecasting model, the method comprising:

7. A computer program recorded on a computer-readable recording medium to execute the power demand prediction model learning method described in any one of Articles 1 to 6.

8. Memory; At least one processor connected to said memory and configured to execute a computer-readable program contained in said memory Including, The above program Obtaining total power data representing the amount of power consumed by multiple home appliances in a household; Dividing the above total power data into regular time intervals to obtain multiple segment data, Obtaining a plurality of sequence data corresponding to the plurality of segment data by embedding each of the plurality of segment data, Masking the target sequence data included in the above plurality of sequence data Performing unsupervised learning on a machine learning model to obtain at least one context data based on a correlation between the target sequence data and each of the sequence data neighboring the target sequence data among the plurality of sequence data is configured to run, A power demand prediction model learning device, wherein the machine learning model is configured to output predicted power data following the arbitrary power data based on the at least one context data in response to inputting at least one sequence data based on the arbitrary power data.

Images