Method for providing power forecasting service based on time-series data and apparatus for performing same

Abstract

The present invention relates to a method for providing a power forecasting service and an apparatus for providing same. More specifically, the present invention relates to a method for providing a service that collects power consumption in real time on the basis of time-series data, forecasts power consumption by means of a deep learning model, and provides an expected power consumption, and to an apparatus for providing same. To resolve the above technical issue, the method for providing a power forecasting service performed in a computing device, according to the present invention, comprises the steps of: collecting the power consumption of a building or facility in real time; generating power consumption data corresponding to a predetermined period on the basis of the collected power consumption; calculating an expected power consumption by using a first model that has learned a time-series change in power consumption by inputting the power consumption data; and providing the calculated expected power consumption by time corresponding to a predetermined period, wherein the expected power consumption is preferably updated and provided in real time according to a predetermined cycle. According to the present invention, it is possible to more accurately forecast and provide power consumption by using a deep learning model that has learned a time-series change in power consumption data generated by collecting, in real time, the power consumption of a building or facility, environmental information, weather information, and power usage information.

Description

Method for providing power consumption prediction service based on time series data and device for performing the same

[0001] The present invention relates to a method for providing a service for power consumption prediction and a device for providing the same. More specifically, the present invention relates to a method for providing a service for collecting power consumption in real time based on time-series data, predicting power consumption using a deep learning model, and providing the predicted power consumption, and a device for providing the same.

[0002] The demand for energy continues to increase worldwide due to economic growth, urbanization, and industrialization.

[0003] In particular, electric energy, which can be obtained through various power plants such as thermal, nuclear, wind, solar, and hydroelectric power, is used for various purposes in various fields of daily life, so it can be said to be an energy field that is essential for human survival.

[0004] However, this electrical energy is subject to seasonal, temperature, weather, and time-of-day fluctuations, making its production unstable and subject to significant fluctuations. Therefore, accurate power forecasting is essential to maintain the stability of the power grid and formulate electricity production plans.

[0005] Accordingly, in Korea, the short-term power demand forecasting program (KSFT: KPX Short-Term Load Forecaster) developed by the Korea Power Exchange is currently used to forecast power consumption, and power consumption is forecasted through the exponential smoothing method, which predicts the 24-hour power demand pattern of the forecast date based on the power demand pattern of the past 3 days and 24 hours.

[0006] However, the exponential smoothing method used in short-term power demand forecasting programs generates 24-hour power demand patterns without considering seasonality and temperature effects, making it difficult to respond appropriately to exceptional temperature fluctuations. Furthermore, exponential smoothing struggles to reflect trends in power consumption during data analysis, making it inaccurate for predicting power consumption that varies over time. Furthermore, it's not suitable for long-term power forecasting.

[0007] Therefore, to solve the aforementioned problem, it is necessary to devise a method for providing a power consumption prediction service that reflects environmental information while being able to predict the expected power consumption more accurately using a deep learning model.

[0008] The purpose of the present invention is to provide a scientific power consumption prediction method based on data according to changes in power consumption over time of a building or facility using a deep learning model.

[0009] More specifically, the present invention aims to provide a service that predicts the amount of power corresponding to a predetermined period by using a deep learning model that has learned changes in power data, and provides a service that updates the predicted expected power amount in real time according to a predetermined cycle and provides data visualization.

[0010] A method for providing a power consumption prediction service performed in a computing device according to the present invention for solving the above technical problem comprises the steps of: collecting the power consumption of a building or facility in real time; generating power consumption data corresponding to a predetermined period based on the collected power consumption; calculating an expected power consumption using a first model that learns a time-series change in the power consumption by inputting the power consumption data; and providing the calculated expected power consumption by time corresponding to the predetermined period, wherein it is preferable that the expected power consumption be provided by being updated in real time according to a predetermined cycle.

[0011] The step of generating power data corresponding to a predetermined period based on the collected power amount further includes a step of collecting at least one more of environmental information, weather information, and power usage information, and it is preferable that the content of the power usage information is determined according to the purpose of the building or facility.

[0012] It is desirable that the above power data be stored by time and used for learning.

[0013] The above first model is preferably composed of a combination of at least two of a Long Short-Term Memory (LSTM), a Temporal Convolutional Network (TCN), and a Temporal Fusion Transformer (TFT) model, and the selected model preferably includes learning changes in the power data at different times.

[0014] If, within a given period of providing the expected power quantity, there is a point in time when the maximum demand power quantity is predicted to exceed the preset contracted demand power quantity, it is desirable to provide a guidance message via a notification pop-up.

[0015] The above-mentioned collected power amount is provided in the form of a visualization of one of a chart, table, and graph for each hour corresponding to a predetermined period along with the above-mentioned expected power amount, and it is preferable that the visualization form is time series data generated based on a time interval.

[0016] It is desirable that the visualized form above be provided differently in the form of the amount of collected power corresponding to a previous point in time and in the form of the amount of expected power corresponding to a later point in time based on the time point provided.

[0017] According to the present invention for solving the above technical problem, a computing device comprises: a processor; a memory for loading a computer program executed by the processor; and a storage for storing the computer program, wherein the computer program comprises: an operation for collecting the power amount of a building or facility in real time; an operation for generating power amount data corresponding to a predetermined period based on the collected power amount; an operation for calculating an expected power amount using a first model that has learned a time-series change in the power amount by inputting the power amount data; and an operation for providing the calculated expected power amount by time corresponding to the predetermined period, wherein it is preferable that the expected power amount be updated in real time according to a predetermined cycle and provided.

[0018] The operation of generating power data corresponding to a predetermined period based on the collected power amount further includes an operation of collecting at least one more of environmental information, weather information, and power usage information, and it is preferable that the content of the power usage information is determined according to the purpose of the building or facility.

[0019] It is desirable that the above power data be stored by time and used for learning.

[0020] The above first model is preferably composed of a combination of at least two of a Long Short-Term Memory (LSTM), a Temporal Convolutional Network (TCN), and a Temporal Fusion Transformer (TFT) model, and the selected model preferably includes learning changes in the power data at different times.

[0021] If, within a given period of providing the expected power quantity, there is a point in time when the maximum demand power quantity is predicted to exceed the preset contracted demand power quantity, it is desirable to provide a guidance message via a notification pop-up.

[0022] The above-mentioned collected power amount is provided in the form of a visualization of one of a chart, table, and graph for each hour corresponding to a predetermined period along with the above-mentioned expected power amount, and it is preferable that the visualization form is time series data generated based on a time interval.

[0023] It is desirable that the visualized form above be provided differently in the form of the amount of collected power corresponding to a previous point in time and in the form of the amount of expected power corresponding to a later point in time based on the time point provided.

[0024] It may also include program code for executing the power prediction method.

[0025] According to the present invention, power consumption can be more accurately predicted and provided by using a deep learning model that learns time-series changes in power consumption data generated by collecting power consumption, environmental information, weather information, and power usage information of a building or facility in real time.

[0026] Additionally, by providing the expected power amount corresponding to a given period in the form of data visualization, it can help in obtaining and understanding the necessary information.

[0027] Furthermore, by more accurately predicting power consumption, it is possible to operate the electric energy supply system efficiently and promote environmentally friendly electric energy production, which not only enables accurate power consumption but also has positive and innovative effects across the energy sector.

[0028] Figure 1 is an exemplary diagram showing a method for providing a power amount prediction service according to one embodiment of the present invention.

[0029] FIG. 2 is a flowchart illustrating a method for providing a power amount prediction service according to one embodiment of the present invention.

[0030] FIG. 3 is an exemplary diagram showing a power data generation process for providing a power prediction service according to one embodiment of the present invention.

[0031] Figures 4 to 5 are exemplary diagrams showing a process for calculating an expected power amount for providing a power amount prediction service according to one embodiment of the present invention.

[0032] Figures 6 and 7 are exemplary diagrams showing screens on which a power prediction service is provided according to one embodiment of the present invention.

[0033] FIG. 8 is an exemplary diagram illustrating an implementation of a computing device of a server according to one embodiment of the present invention.

[0034] The following merely exemplifies the principles of the invention. Therefore, those skilled in the art will be able to implement the principles of the invention and invent various devices within the scope and spirit of the invention, even if not explicitly described or illustrated herein. Furthermore, all conditional terms and embodiments listed herein are expressly intended, in principle, to facilitate understanding of the invention, and are not intended to be limiting to the specifically enumerated embodiments and conditions.

[0035] The above-described purposes, features and advantages will become clearer through the following detailed description with reference to the attached drawings, so that a person having ordinary skill in the art to which the invention pertains can easily practice the technical idea of ​​the invention.

[0036] Furthermore, when describing the invention, detailed descriptions of known technologies related to the invention will be omitted if they are deemed to unnecessarily obscure the gist of the invention. Below, preferred embodiments of the present invention will be described in detail with reference to the attached drawings.

[0037] FIG. 1 is an exemplary diagram showing a system in which a method for providing a power consumption prediction service based on a deep learning model according to one embodiment of the present invention is performed.

[0038] Referring to FIG. 1, a system according to the present embodiment may be configured with a building or facility (200) that is to output an expected power amount, a local server (10) that collects the power amount collected from the building or facility (200) in real time, and a cloud server (400) that provides the expected power amount to a user terminal (100). Inside the cloud server (400), a first model (470), which is a deep learning model that learns time-series changes in power amount by inputting power amount data generated based on the collected power amount, and calculates an expected power amount, may be configured.

[0039] In addition, in this embodiment, the cloud server (400) can collect global observation information, such as weather data, through an API (Application Programming Interface) provided by an external public data server (see FIG. 3) to more accurately calculate the adjustment value required for the expected power output.

[0040] In this embodiment, the local server (10) can collect in real time the amount of power recorded in a power meter installed in a building or facility (200). At this time, the power collection medium is not limited to the power meter and may include a voltage sensor or a smart meter.

[0041] A power meter is a device used to measure and bill electricity consumption in homes, commercial buildings, or facilities (200). It is externally mounted and accurately records electricity usage. Additionally, a voltage sensor is used to measure the voltage of an electrical circuit, and both voltage and current are measured to calculate electricity consumption. Furthermore, a smart meter is a modernized version of a traditional power meter. It monitors electricity usage in real time and collects data remotely, improving power efficiency and providing consumers with information about their energy use.

[0042] In addition, the local server (10) may also serve as a relay device that collects power from a building or facility (200), collects environmental information from a public data server (see FIG. 3), and collects power usage information from a power usage information collection module (see FIG. 3) and provides the collected information to a cloud server (400).

[0043] In this embodiment, a local server (10) is installed in each building or facility (200), and may be divided into unit zones for the zones where power is to be predicted in order to generate expected power information, and for example, multiple local servers may be installed in each building.

[0044] In this embodiment, the user terminal (100) can be provided with the expected power amount output from the cloud server (400) as a means for providing the expected power amount to the user terminal (100) through a service platform such as a website or an executable program, including an application.

[0045] Hereinafter, a method for providing a power consumption prediction service according to the present embodiment will be described in more detail with reference to FIG. 2.

[0046] FIG. 2 is a flowchart illustrating a method for providing a power amount prediction service according to one embodiment of the present invention.

[0047] First, the local server (10) collects power in real time from a building or facility (200) through a power collection medium (S100).

[0048] Next, the cloud server (400) generates power data corresponding to a predetermined period based on the power amount collected from the local server (10) (S200).

[0049] Power data may refer to data that reflects information such as power recorded over a set period of time and environmental information additionally collected over a set period of time.

[0050] Specifically, the local server (10) additionally collects at least one of environmental information, weather information, and power usage information to generate power data. Rather than simply collecting power and outputting an estimated power amount, the local server collects a variety of information data and utilizes this to more accurately predict power amounts.

[0051] Hereinafter, the process of generating power data for providing a power prediction service according to the present embodiment will be described in more detail with reference to FIG. 3.

[0052] Referring to FIG. 3, the local server (10) can collect environmental information and weather information from a public data server (250). Here, the public data server (250) preferably refers to a portal server operated by the Korea Meteorological Administration, and can provide global weather observation information such as ground, ocean, high-rise, aerial observation, satellite, radar, and numerical forecast model data to the local server (10).

[0053] Furthermore, the provided observation information can be used as climate statistical information for calculating the expected power amount by being provided in chronological order, and can be used to provide and predict statistical analysis information on the overall domestic weather, such as temperature analysis, precipitation analysis, climate averages, rainy season, number of yellow dust days, number of heat wave days, and number of tropical nights.

[0054] In addition, the local server (10) can collect power usage information from the power usage information collection module (260). Here, the power usage information collection module (260) can be located inside or outside the local server (10), and can perform the role of collecting power usage information of a building or facility (200), and the content of the information can be determined according to the purpose of the building or facility (200).

[0055] For example, a residential building may exhibit a regular power usage pattern, characterized by higher power consumption in the morning and evening, depending on residential activity. Power consumption may also fluctuate with the seasons. Furthermore, an industrial facility may exhibit higher power consumption depending on machinery and production processes, and power consumption characteristics may vary depending on the industrial sector. The power usage information collection module (260) can collect power usage information by identifying and analyzing the purpose of the building or facility (200).

[0056] Here, the power data can be generated by combining collected power and one or more pieces of information from among environmental information, weather information, and power usage information, and the generated power data can be stored by time and used for learning the first model (470).

[0057] More specifically, the collected power amount may refer to the power amount recorded over a set period of time, such as a day, and may be a value that reflects additional information such as environmental information collected over a set period of time to generate power amount data.

[0058] Next, the cloud server (400) calculates the expected power amount using the first model (470) that has learned the time-series change in power amount as input of the generated power amount data (S300).

[0059] Here, the predicted power amount can be learned by a process of minimizing the error between the predicted power amount data for the first predetermined period and the actual power amount data for the first predetermined period based on the time point provided.

[0060] The first model is structured as a time-series neural network. Features that have relatively little impact on the predicted power output are forgotten for subsequent power forecasts, while key features are remembered for future predictions. This process allows weights to be learned. The specified time period can be varied at the discretion of the user or service.

[0061] If the forecasted power amount is predicted in 15-minute units based on the current time provided, the forecasted power amount at the current time can be calculated by inputting the power amount data for the last 2 hours and 1 hour and 45 minutes before the current time, and the power amount data for the last 1 hour and 45 minutes before the current time.

[0062] The process of calculating the expected power amount will be described in detail with reference to FIG. 4. For example, when the current time is 3:00 PM and the time at which the power amount is desired is 15 minutes later, the first model (470) can calculate the expected power amount at 3:00 PM using the power amount data at 12:45 PM, which is 2 hours ago, and 1:00 PM, which is 1 hour and 45 minutes ago, and the power amount data at 1:45 PM, which is 1 hour ago, and 2:00 PM, which is 45 minutes ago.

[0063] Next, referring to Fig. 5, the actual power consumption at 3:00 PM can be collected from the database inside the cloud server (400), and the error calculated by comparing them can be learned by back-propagating.

[0064] To put it simply again, learning can be achieved by comparing the data predicted by the first model (470) with actual power data and backpropagating the calculated error back to the first model (470).

[0065] At this time, weights for power data can be calculated through the backpropagation process, and the predicted power can be output using the first model (470) updated with the weights at this time. In this embodiment, a neural network model that can further learn the degree of change in the comparison group over time in power data as weights enables more accurate prediction.

[0066] Referring again to FIG. 4, the first model (470) can be configured in the form of a heterogeneous time series network trained to produce predicted data for an expected power amount by inputting two types of power amount data for a given period of time.

[0067] Preferably, the first model (470) is a 1-1 model, which may be an LSTM (Long Short-Term Memory models) model with a structure that adds a cell state to the hidden state of an existing RNN (Recurrent Neural Networks).

[0068] First, the LSTM model handles temporal dependence in time series data well and has the characteristic of handling long-term dependence well.

[0069] In addition, the first model (470) may be a Temporal Convolution Network (TCN) model that processes time series data using a convolution operation that extracts patterns and features of various sizes by applying a filter to input time series data, as in the first-second model.

[0070] The TCN model can learn topological characteristics by using filters to efficiently detect temporal patterns through convolution operations.

[0071] Alternatively, the 1-3 model may be composed of a combination of at least two TFT (Temporal Fusion Transformer) models that are effective for data modeling that considers temporal patterns by adding attention to specific parameters based on the Transformer model structure.

[0072] The first model (470) of the present embodiment can predict the expected power amount more accurately and precisely by utilizing the ensemble effect that takes advantage of the strengths of each model.

[0073] Next, the cloud server (400) provides the calculated expected power amount to the user terminal (100) by time corresponding to a predetermined period (S400).

[0074] Here, the estimated power amount can be provided in real time by updating at a predetermined interval. For example, referring to FIG. 6, the estimated power amount can be provided at 15-minute or 1-hour intervals based on the current time, and this can be provided at a cycle selected by the user terminal (100).

[0075] In addition, the amount of power collected in this embodiment can be provided in the form of a visualization of data by time corresponding to a predetermined period along with the expected amount of power, for example, in the form of a chart, a table, or a graph, which can be time series data generated based on a time interval.

[0076] Next, in providing the collected power amount and the expected power amount of the cloud server (400), the visualized form of charts, tables, and graphs provided hourly may be provided in a different form for the collected power amount corresponding to a previous point in time and the expected power amount corresponding to a subsequent point in time based on the time of provision. Here, the time of provision may be the current time.

[0077] For example, continuing with reference to Figure 6, a baseline may be drawn based on the current time point, 11:15 AM. The collected power corresponding to the previous time point may be represented as a solid line in the graph, and the expected power corresponding to the subsequent time point may be represented as a dotted line in the graph. The visualization of the power amount shown here is not limited to this example and may be represented in various other visualization forms.

[0078] Next, in this embodiment, if there is a point in time when the maximum demand power amount exceeding the preset contract demand power amount is predicted within a predetermined period of time during which the cloud server (400) provides the expected power amount to the user terminal (100), the cloud server (400) can provide a guidance message through a notification pop-up.

[0079] Specifically, referring to Figure 7, if the expected power exceeds the preset contracted power, a notification pop-up message can be provided stating, “Power demand exceeding the contracted power is predicted. Facility inspection is required.”

[0080] Furthermore, the wording of the guidance message is not limited to the above-mentioned wording and may be provided in other wordings, and the means of provision is not limited to a notification pop-up and may be provided through various channels such as a banner area or specific text or icon.

[0081] Here, the preset contracted power demand and maximum demand are set based on the power contracted for the user during the specified period of service use. The maximum demand may refer to the amount of power exceeding this amount. This allows users to establish reasonable power consumption plans.

[0082] Additionally, the projected power amount provided may be provided in 15-minute increments for a total of three hours into the future, or in daily increments for a total of four days into the future. The projected power amount information provided may include both time-series data and power amount data. For daily forecasts, the projected power amount information may include maximum demand power information.

[0083] Accordingly, the cloud server (400) according to the present embodiment can collect additional information according to the power consumption and purpose of a building or facility (200), and output the expected power consumption more accurately using a deep learning model, thereby enabling accurate power consumption and providing positive and innovative effects in the overall energy sector.

[0084] Hereinafter, with reference to FIG. 8, a specific hardware implementation of a cloud server (400) as a computing device according to one embodiment of the present invention will be described.

[0085] Referring to FIG. 8, in some embodiments of the present invention, a cloud server (400) may be implemented in the form of a computing device (300). One or more of each module constituting the computing device (300) is implemented on a general-purpose computing processor and thus may include a processor (308), an input / output I / O (302), a memory (340), an interface (306), and a bus (314). The processor (308), the input / output device (302), the memory (340), and / or the interface (306) may be coupled to each other via the bus (314). The bus (314) corresponds to a path through which data is transferred.

[0086] Specifically, the processor (308) may include at least one of a CPU (Central Processing Unit), an MPU (Micro Processor Unit), an MCU (Micro Controller Unit), a GPU (Graphics Processing Unit), a microprocessor, a digital signal processor, a microcontroller, an application processor (AP), and logic elements capable of performing functions similar thereto.

[0087] The input / output device (302) may include at least one of a keypad, a keyboard, a touchscreen, and a display device. The memory device (340) may store data and / or programs, etc.

[0088] The interface (306) may perform a function of transmitting data to or receiving data from a communication network. The interface (306) may be wired or wireless. For example, the interface (306) may include an antenna or a wired / wireless transceiver. The memory (340) may further include high-speed DRAM and / or SRAM, etc., as a volatile operating memory that enhances the operation of the processor (308) while protecting personal information.

[0089] Additionally, the memory (340) stores programming and data configurations that provide the functionality of some or all of the modules described herein. For example, it may include logic for performing selected aspects of the verification method described above.

[0090] A program or application is loaded as a set of instructions including each step of performing the above-described power prediction method stored in the memory (340), and the processor is enabled to perform each step.

[0091] Furthermore, the various embodiments described herein may be implemented in a recording medium readable by a computer or similar device, for example, using software, hardware, or a combination thereof.

[0092] Furthermore, the various embodiments described herein may be implemented in a recording medium readable by a computer or similar device, for example, using software, hardware, or a combination thereof.

[0093] In terms of hardware implementation, the embodiments described herein can be implemented using at least one of ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), processors, controllers, micro-controllers, microprocessors, and other electrical units for performing functions. In some cases, the embodiments described herein can be implemented as a control module itself.

[0094] In a software implementation, the embodiments described herein, such as the procedures and functions, may be implemented as separate software modules. Each of the software modules may perform one or more of the functions and operations described herein. The software code may be implemented as a software application written in a suitable programming language. The software code may be stored in a memory module and executed by a control module.

[0095] The above description is merely an example of the technical idea of ​​the present invention, and those skilled in the art will appreciate that various modifications, changes, and substitutions can be made without departing from the essential characteristics of the present invention.

[0096] Accordingly, the embodiments disclosed in the present invention and the accompanying drawings are intended to illustrate, rather than limit, the technical concept of the present invention, and the scope of the technical concept of the present invention is not limited by these embodiments and the accompanying drawings. The protection scope of the present invention should be interpreted by the following claims, and all technical concepts within the scope equivalent thereto should be interpreted as being included within the scope of the rights of the present invention.

Claims

1. A method for providing a power consumption prediction service performed on a computing device, A step of collecting the power consumption of a building or facility in real time; A step of generating power data corresponding to a predetermined period based on the collected power amount; A step of calculating an expected power amount using a first model that learns the time-series change in power amount by inputting the above power amount data; Including a step of providing the above-mentioned calculated expected power amount by time corresponding to a predetermined period, A method for providing a power forecasting service, characterized in that the above predicted power amount is provided by being updated in real time according to a predetermined cycle.

2. In paragraph 1, The step of generating power data corresponding to a predetermined period based on the collected power amount is as follows: Further comprising a step of collecting at least one of environmental information, weather information, and power usage information; A method for providing a power consumption prediction service, characterized in that the content of the power usage information is determined according to the purpose of the building or facility.

3. In paragraph 1, The above power data is, A method for providing a power consumption prediction service characterized in that it is stored by time and used for learning.

4. In paragraph 1, The above first model is, It consists of a combination of at least two of the following models: LSTM (Long Short-Term Memory), TCN (Temporal Convolutional Network), and TFT (Temporal Fusion Transformer). A method for providing a power forecasting service, characterized in that the selected model includes learning changes in the power data at different times.

5. In paragraph 1, Within a given period of time to provide the expected power amount, A method for providing a power forecasting service characterized in that, when there is a point in time when the maximum demand power exceeding the preset contract demand power is predicted, a guidance message is provided through a notification pop-up.

6. In paragraph 1, The above collected power amount is provided in the form of a visualization of one of a chart, table, or graph for each hour corresponding to a given period along with the above expected power amount, A method for providing a power consumption prediction service, characterized in that the above visualized form is time series data generated based on time intervals.

7. In paragraph 6, The above visualized form is, A method for providing a power forecasting service, characterized in that the form of the power amount corresponding to a previous point in time and the form of the expected power amount corresponding to a subsequent point in time are provided differently based on the time of provision.

8. Processor; a memory that loads a computer program executed by the processor; and Including storage for storing the above computer program, The above computer program, The act of collecting the power consumption of a building or facility in real time; An operation of generating power data corresponding to a predetermined period based on the power amount collected above; An operation of calculating an expected power amount using a first model that has learned the time-series change in power amount by inputting the above power amount data. Including an operation of providing the above-mentioned calculated expected power amount by time corresponding to a predetermined period, A computing device characterized in that the above predicted power amount is provided by being updated in real time according to a predetermined cycle.

9. In paragraph 8, The operation of generating power data corresponding to a predetermined period based on the power amount collected above is as follows: Further comprising an action of collecting at least one of environmental information, weather information, and power usage information; A computing device characterized in that the content of the power usage information is determined according to the purpose of the building or facility.

10. In paragraph 8, The above power data is, A computing device characterized by being stored by time and used for learning.

11. In paragraph 8, The above first model is, It consists of a combination of at least two of the following models: LSTM (Long Short-Term Memory), TCN (Temporal Convolutional Network), and TFT (Temporal Fusion Transformer). A computing device characterized in that the selected model includes learning changes in the power data at different times.

12. In paragraph 8, Within a given period of time to provide the expected power amount, A computing device characterized in that it provides a guidance message through a notification pop-up when there is a point in time when the maximum demand power amount is predicted to exceed the preset contract demand power amount.

13. In paragraph 8, The above collected power amount is provided in the form of a visualization of one of a chart, table, or graph for each hour corresponding to a given period along with the above expected power amount, A computing device characterized in that the above visualized form is time series data generated based on time intervals.

14. In paragraph 13, The above visualized form is, A computing device characterized in that, based on the time point provided, the form of the collected power corresponding to a previous time point and the form of the expected power corresponding to a subsequent time point are provided differently.

Images