KR101875329B1 - Forecasting apparatus and method for power consumption - Google Patents

Abstract

Provided is a power prediction apparatus implemented by a processor. The power prediction apparatus can comprise: a selection part selecting a first prediction model among K power load prediction models based on a power usage pattern for a selected electronic device; and a prediction part predicting a power consumption of a customer after a predetermined time by using the selected first prediction model.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a power consumption estimating apparatus,

The following discussion relates to an apparatus and method for predicting power usage, and more particularly to an apparatus and method for predicting power usage applied to a microgrid.

As the Internet of Things (IoT) -based sensors, such as smart meters that support bi-directional communication between power suppliers and listeners, and smart plugs that control electricity usage remotely via wireless communications, have become popular, It is becoming possible to acquire data on power consumption of each electric power consumer and electric power equipment. There is a growing need for measures to improve the overall stability and balance of the power system by predicting the amount of power used for small-scale power loads by effectively utilizing the collected power data.

Most of today's power-load prediction technologies are being developed for large-scale power loads in the wide area. At the current level of technology, small-scale power-load predictions show low accuracy due to high variability and complexity of the consumption form of the target power load.

Korean Patent No. 10-1588851 (January 28, 2016) discloses the structural characteristics of a building, the information on the inside and outside of a building, the characteristics of energy consumption by loads, Power consumption information for each load, and the like. Korean Patent No. 10-1678857 (published on Nov. 23, 2016) calculates a shortage or a surplus power amount of a specific micro grid and compares the power generation capacity of the micro grid with the power shortage amount of the surrounding micro grid Lt; RTI ID = 0.0 > sharing power. ≪ / RTI >

According to one aspect, a power prediction apparatus implemented by a processor is provided. Wherein the power prediction apparatus includes a selection unit that selects a first prediction model among K power load prediction models and K is a natural number based on a power usage pattern of the selected electronic device, And a prediction unit for predicting a power consumption of the customer after the time.

According to an embodiment, the selector may select the first prediction model according to a time interval in which a peak value of the power usage amount with respect to the electronic apparatus exists.

According to another embodiment, the selector may select the first prediction model according to an electric power interval in which an average power consumption for a predetermined time period is present with respect to the electronic apparatus.

According to another embodiment, the K power load prediction models are machine-learned using K pieces of learning data clustered according to the power usage pattern of the electronic device, and the learning data is divided into a plurality of And may include power consumption of the customer. More specifically, the K is sized such that the predicted accuracy calculated through the test data is greater than or equal to a predetermined threshold value, and the K power load prediction models are computed using the clustered learning data corresponding to the determined K, .

According to another aspect, a power prediction method is provided. The power prediction method comprising the steps of: identifying an electronic device having a correlation coefficient equal to or greater than a threshold value, based on learning data; and determining the learning data according to a predetermined measurement period on the basis of the identified electronic device as K Cluster into a cluster.

According to an embodiment, the power prediction method may further include generating K power load prediction models using learning data corresponding to each of the K clusters, The magnitude can be determined such that the prediction accuracy of the power load prediction model becomes a predetermined threshold value or more.

According to another embodiment, the step of generating the K power load prediction models includes repeatedly generating the K power load prediction models using a prediction period according to the power load variability of the customer, May represent the customer to which the generated power load prediction model is applied.

According to yet another embodiment, the step of generating the K power load prediction models may include calculating a prediction accuracy of K power load prediction models generated according to a first prediction period, And generating new K power load prediction models repeatedly using the two prediction periods.

According to another embodiment, the power prediction method may further include setting a measurement period in accordance with a power load variability of a customer to which the power load prediction model is applied, And collecting the learning data.

According to another embodiment, the step of setting the measurement period includes calculating a prediction accuracy of K power load prediction models corresponding to learning data collected according to a first measurement period, 2 measurement period may be newly set.

According to yet another embodiment, identifying the electronic device may include identifying a plurality of electronic devices that are within a predetermined range in order of priority from the learning data associated with a particular customer.

1 is an exemplary diagram illustrating a process of generating a power load prediction model.
2A and 2B are diagrams for explaining a process of calculating a correlation between a power load and an electronic device.
3 is a flow chart illustrating a method for generating a power load prediction model.
4 is a block diagram showing a power prediction apparatus.
5 is a diagram illustrating a process of selecting a first prediction model among K power load prediction models according to an embodiment of the present invention.
6 is an exemplary diagram illustrating a process of predicting the amount of power used by the power prediction apparatus.
7 is a flowchart illustrating a process of adjusting a prediction period and a data measurement period based on prediction accuracy, according to another embodiment of the present invention.

Specific structural or functional descriptions of embodiments are set forth for illustration purposes only and may be embodied with various changes and modifications. Accordingly, the embodiments are not intended to be limited to the specific forms disclosed, and the scope of the disclosure includes changes, equivalents, or alternatives included in the technical idea.

The terms first or second, etc. may be used to describe various elements, but such terms should be interpreted solely for the purpose of distinguishing one element from another. For example, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

It is to be understood that when an element is referred to as being "connected" to another element, it may be directly connected or connected to the other element, although other elements may be present in between.

The singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, the terms "comprises ", or" having ", and the like, are used to specify one or more of the features, numbers, steps, operations, elements, But do not preclude the presence or addition of steps, operations, elements, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the meaning of the context in the relevant art and, unless explicitly defined herein, are to be interpreted as ideal or overly formal Do not.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the following description of the present invention with reference to the accompanying drawings, the same components are denoted by the same reference numerals regardless of the reference numerals, and a duplicate description thereof will be omitted.

Explanation of Terms

In the following description, a consumer may represent a customer purchasing electricity for his or her intended use. For example, the consumer can be classified into general consumer, business consumer, and industrial consumer depending on the usage type of electric power. An example of a consumer may be a microgrid.

The microgrid represents an independent distributed power source separate from the global power system. Specifically, the micro grid refers to a next-generation power system that is a combination of renewable energy sources such as solar and wind power and energy storage devices so that electricity can be self-sufficient in a small area. The power prediction apparatus according to the present embodiment can be used to predict the power consumption of each microgrid.

Machine learning represents a technology for predicting the future by using various types of data such as amount, cycle, and format of data. Today, various types of algorithms, such as K-means, Self Organizing Map (SOM) and Support Vector Machine (SVM) algorithms, which are used as examples of unsupervised learning, Can be used.

1 is an exemplary diagram illustrating a process of generating a power load prediction model. Referring to FIG. 1, a process by which a power prediction apparatus generates a power load prediction model according to an embodiment is illustrated. The power predictor may perform a data preprocessing step 110, a data clustering step 120 and a machine learning step 130 to generate a power load prediction model corresponding to each cluster.

The power prediction apparatus according to the present embodiment can calculate the variability of the load with respect to each customer. The variability of the load can be defined as the ratio of the actual variation of the power usage to the average predicted value of the power usage. A detailed description of the process in which the variability of the load is calculated will be described in detail below with reference to the drawings to be added. In step 110, the power prediction apparatus can set a prediction cycle for generating the power load prediction model and a measurement cycle of the learning data according to the variability of the load. Further, at step 110, the power prediction apparatus can identify an electronic device having a correlation with the power load that is equal to or greater than a threshold value.

In step 120, the power prediction apparatus may allocate the entire learning data set to K clusters based on the power usage data on the identified electronic devices. Accordingly, the power prediction apparatus can generate a power load prediction model using the K clusters allocated in step 130. [ Hereinafter, specific processes performed by the power prediction apparatus for each of the steps 110, 120, and 130 will be described in more detail.

Data preprocessing process

In step 110, the power predictor may calculate the correlation between the power load and the electronic device using the raw data. In one embodiment, the raw data may represent power load data over time for a plurality of consumers. More specifically, the row data may include power load data over time for a plurality of micro-grids. In addition, the row data may include power load data for each electronic device. Illustratively, the low data may be data indicating how the amount of power consumption generated by the first electronic device varies with time, and how the amount of power consumption generated by the second electronic device varies with time.

In another embodiment, the power prediction device may obtain the raw data from a sensor that provides power load data. Illustratively, the sensor may be either a smart meter or a smart plug. In yet another embodiment, the power prediction device may obtain the raw data from a designated power management server.

2A and 2B are diagrams for explaining a process of calculating a correlation between a power load and an electronic device. Referring to Figures 2A and 2B, there are two graphs illustrating power usage over time. The X-axis of the two graphs represents time (minutes), and the Y-axis represents power (kW). Referring to FIG. 2A, the total power usage 211 and the amount of power usage 212 by the first electronic device are shown in a specific time zone. More specifically, when the total power consumption 211 and the power consumption 212 by the first electronic device are similar (for example, from about 0 minutes to about 350 minutes and from about 800 minutes to about 1450 minutes) .

Similarly, referring to FIG. 2B, the total power usage 221 associated with a particular time zone and the power usage 222 by a second electronic device are shown together. In the embodiment of FIG. 2B, the total power consumption 221 for a specific time period (ex. About 5 minutes to about 120 minutes) and the power consumption 222 by the second electronic device are similar.

As described above, a specific electronic device can exhibit a correlation with a total electric power consumption of the customer more than a threshold value. The power prediction apparatus according to the present embodiment can calculate the correlation between the power load and the electronic device using Equation 1 below.

In Equation (1), X represents the total power consumption of a specific customer, and Y represents the power consumption associated with a specific electronic device among the total power consumption. Shows a correlation with the electric power load and the electronic apparatus, μ _X and μ _Y represents an average value of the X and Y, respectively, σ _X and σ _Y represents a distribution of the X and Y, respectively, E is to indicate the expected function . In one embodiment, the power prediction apparatus can calculate a correlation corresponding to each of an air conditioner, an electric car, a microwave oven, a refrigerator, and a dishwasher according to Equation (1) using the power consumption of the first consumer as low data . The above-described electronic apparatuses are only exemplary descriptions for the sake of understanding, and should not be construed as limiting or limiting the scope of other embodiments.

3 is a flow chart that generally illustrates a method for generating a power load prediction model. Referring to FIG. 3, a method for generating a power load prediction model includes identifying (310) an electronic device having a correlation coefficient equal to or greater than a threshold value, based on learning data, A step 320 of allocating the learning data to K clusters, and a step 330 of generating K power load prediction models using learning data corresponding to each of the K clusters.

In step 310, the power prediction apparatus may identify an electronic device having a correlation coefficient equal to or greater than a threshold value, based on the learning data. Illustratively, the learning data may represent power usage data transmitted using the Internet. The power prediction device can identify an electronic device having a correlation coefficient equal to or greater than a threshold value with the total power load of the customer.

In one embodiment, the power prediction device can identify an electronic device having a correlation coefficient of 0.5 or greater with the total power load of the first consumer. In another embodiment, the power prediction apparatus can identify electronic devices existing within a predetermined order based on the order of correlation coefficients. For example, the power predictor may calculate a correlation coefficient with the total power load of the first customer and identify three electronic devices in three orders in order of the correlation coefficient. As for the specific process of calculating the correlation coefficient by the power prediction apparatus, the description of Equation (1) described above can be applied as it is, and a duplicate description will be omitted.

Also, in the data preprocessing process, the power prediction apparatus can set the measurement cycle S of data used as learning data according to the power load fluctuation of the customer. The customer may indicate a particular customer to which the power load prediction model generated by the power prediction device is applied. More specifically, the power prediction apparatus can calculate the power load variability? _L of the customer according to Equation (2) below.

은 예측 주기 T에 대응하는 전력 사용량의 평균 예측값을 나타낼 수 있다. 위와 같이, 전력 예측 장치는 기설정된 측정 주기 S에 따라 복수의 수용가의 전력 사용량에 관한 학습 데이터를 수집할 수 있다. 예시적으로, 전력 예측 장치는 제1 수용가의 전력 부하 변동성이 증가하는 경우, 측정 주기를 짧게 조절하여 보다 빈번하게 데이터를 수집하여 예측의 정확도를 향상시킬 수 있다.In Equation 2, γ _L (i + 1) is the suyongga corresponding to show the power load variability of suyongga P _L (i + 1) is the i + 1 measuring cycle corresponding to the i + 1 measuring cycle the actual P _L (i) is the actual power usage of the customer corresponding to the i th measurement period,

May represent an average predicted value of the power consumption corresponding to the prediction cycle T. As described above, the power prediction apparatus can collect learning data related to the power consumption of a plurality of customers in accordance with the predetermined measurement cycle S. Illustratively, the power prediction apparatus can improve the accuracy of prediction by collecting data more frequently by shortening the measurement period when the power load variability of the first consumer is increased.

Hereinafter, a process of generating the K power load prediction models through the data clustering process 120 and the machine learning process 130 will be described in detail.

Clustering process

In step 320, the power prediction apparatus may allocate the learning data to K clusters based on the electronics identified in step 310. [ In the following description, the term " allocated " can be used as a concept corresponding to the meaning of performing data clustering. The learning data may represent power consumption of a plurality of customers collected according to a predetermined measurement cycle S. In one embodiment, the power prediction apparatus may allocate the learning data to K clusters according to a power usage pattern for the identified electronic device. In addition, the power prediction apparatus may determine a K value that ensures maximum reliability or accuracy, and may cluster learning data into K clusters according to the determined K value.

Machine learning course

In step 330, the power prediction apparatus may generate K power load prediction models using corresponding learning data of each of the K clusters. Also, the power prediction apparatus can calculate the prediction accuracy of each of the power load prediction models using the generated power load prediction model and the test data. The power prediction apparatus may adjust the size of the K so that the calculated prediction accuracy is greater than or equal to a predetermined threshold value. The learning data corresponding to each of the K clusters can generate the K power load prediction models according to a machine learning method such as Artificial Neural Network (ANN: Artificial Neural Network) or SVM (Support Vector Machine). The detailed description of artificial intelligence neural network and SVM is straight forward to experts in the technical field, so detailed description will be omitted. In addition, the artificial intelligence network and SVM are only one of various embodiments of machine learning methods that can be selected in the process of generating K power load prediction models, but should not be construed as limiting or limiting the scope of other embodiments .

In another embodiment, the power prediction apparatus may repeatedly generate the K power load prediction models using the prediction cycle S determined according to the power load variance? _L of the customer. The customer may represent the customer to which the generated power load prediction model is applied. For example, the first audience may have a power load variability of 20% and a prediction period of 30 minutes. In this case, the power prediction apparatus can generate K power load prediction models repeatedly at intervals of 30 minutes using the transmitted learning data. When the power load variability of the first consumer price changes, the power prediction apparatus can adjust the prediction period based on the changed power load variability. Illustratively, the power prediction device can generate a power load prediction model by setting a shorter prediction period when the power load variability of a particular customer is greater.

Application Process (Power Consumption Forecast Process)

4 is a block diagram showing a power prediction apparatus. Referring to FIG. 4, a power prediction apparatus 400 implemented with a processor is shown. The power prediction apparatus 400 may include a selection unit 410 and a prediction unit 420. The selection unit 410 can select the first prediction model among the K power load prediction models based on the power usage pattern for the selected electronic device. K, which defines the number of power load prediction models, may represent a natural number.

5 is a diagram illustrating a process of selecting a first prediction model among K power load prediction models according to an embodiment of the present invention. Referring to FIG. 5, the selection unit 410 may receive a power usage pattern 511 related to a specific electronic device. In addition, the selection unit 410 can select a power load prediction model that performs power load prediction for the first customer price based on the input power usage pattern 511. [ The selecting unit 410 can select any one prediction model having the pattern of the highest similarity with the power usage pattern 511 among the K power load prediction models.

As an example, the selection unit 410 may select a prediction model according to a time interval in which a peak value of the power usage amount with respect to the electronic device exists. For example, the first consumer may have the highest power usage pattern by the air conditioner during the daytime. In this case, the selection unit 410 can select the first prediction model generated based on the learning data having the highest power consumption by the air conditioner during the daytime.

In another embodiment, the selection unit 410 may select a prediction model according to an electric power interval in which an average power consumption for a predetermined time period of the electronic equipment exists. For example, the second consumer may have a power usage pattern, which is indicated by a 4 kW power consumption by charging an electric car in the night time zone. In this case, the selector 410 can select the second prediction model generated based on the learning data nearest to 4 kW in the amount of power consumption by charging the electric vehicle in the night time zone.

The predicting unit 420 can estimate the power consumption of the customer after a predetermined time using the selected prediction model. More specifically, the predicting unit 420 may output the power consumption of the customer by inputting the power usage amount of the electronic device selected in the selected prediction model as a parameter. A detailed description of the process of the predictor 420 inputting a plurality of parameters to the selected prediction model and outputting the predicted power usage amount for the customer can be explained with reference to FIG. 6 below.

6 is an exemplary diagram illustrating a process of predicting the amount of power used by the power prediction apparatus. Illustratively, there may be a case where the first power load prediction model is selected by the selector 410 among the K power load prediction models. In this case, the predicting unit 420 adds the first parameter 611 related to time, the second parameter 612 related to the power consumption, the third parameter 613 related to the selected electronic device to the selected first power load prediction model, And a fourth parameter 614 relating to climate data. In one embodiment, the first parameter 611 on time includes a time index associated with power usage during the day (24 hours), a day index associated with power usage during the week (7 days) And a month index associated with power usage during a year (12 months).

In another embodiment, the second parameter 612 regarding the power consumption may include time-specific power consumption data within a predetermined period. As another example, the third parameter 613 related to the selected electronic device may include power consumption data by an individual electronic device such as an air conditioner, an electric car, a microwave oven, and the like. In yet another embodiment, the fourth parameter 614 for climate data may include an apparent temperature within a predetermined period of time. The process of calculating the apparent temperature using the atmospheric temperature, the wind speed, the dew point, and the humidity is straight forward to those skilled in the art, so a detailed description thereof will be omitted.

The predicting unit 420 may input parameters previously specified in the selected power load prediction model to output the power usage prediction amount 620 related to a specific customer. More specifically, the predicting unit 420 may calculate the power usage prediction amount 620 of a specific microgrid after a predetermined time. The power prediction apparatus 400 according to the present embodiment can effectively use the big data collected from sensors such as a smart meter and a smart plug to provide a more accurate power consumption prediction value. Also, since the electric power predicting device 400 can preferentially use the electronic devices having a high correlation with the electric power load such as an electric car, it can expect the effect of grasping the influence on the electric power network in advance. The power prediction apparatus 400 according to the present embodiment can provide the effect of effectively improving the prediction accuracy with respect to the power load showing severe variability.

7 is a flowchart illustrating a process of adjusting a prediction period and a data measurement period based on prediction accuracy, according to another embodiment of the present invention. In step 710, the power prediction apparatus may set a prediction cycle T for generating a power load prediction model according to a predetermined initial value and a measurement cycle S for learning data. Also, in step 720, the power prediction apparatus may calculate the prediction accuracy using the generated K power load prediction models. Specifically, the power prediction apparatus calculates the difference between the actual power consumption and the predicted value as a prediction error, and calculates the prediction accuracy by comparing the calculated prediction error with the difference of the _target accuracy? _Target . More specifically, the power prediction apparatus can use RMSE (Root Mean Square Error), MRE (Mean Relative Error) and R ² (R squared value) values in calculating the _target accuracy ξ _target . The detailed calculation process of RMSE, MRE and R ² is straightforward to the experts in the technical field, so a detailed description will be omitted.

In one embodiment, when the prediction error is smaller than the target accuracy? _Target , the power prediction apparatus may terminate the related control process without performing the adjustment process of the prediction cycle T and the data measurement cycle S.

In another embodiment, if the prediction error is greater than or equal to the target accuracy? _Target , the power prediction apparatus may perform step 730 of adjusting the prediction period T and the data measurement period S. Illustratively, the power prediction apparatus may be reduced by a predetermined amount of each of the prediction period T and the data measurement period S such that the prediction error becomes smaller than the target accuracy? _Target . The power prediction apparatus according to the present embodiment can provide a flexible prediction technique by adjusting the prediction cycle so as to satisfy the target prediction error. In addition, an effect of providing the target accuracy by adjusting the measurement interval of the collected learning data to satisfy the target prediction error can be expected.

The embodiments described above may be implemented in hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented within a computer system, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, such as an array, a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be permanently or temporarily embodied in a transmitted signal wave. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer readable medium may include program instructions, data files, data structures, and the like, alone or in combination. Program instructions to be recorded on a computer-readable medium may be those specially designed and constructed for an embodiment or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Although the embodiments have been described with reference to the drawings, various technical modifications and variations may be applied to those skilled in the art. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced.

Claims (12)

Implemented by the processor:
A selection unit for selecting a first predictive model among K power load prediction models and K is a natural number based on a power usage pattern of the selected electronic device; And
A prediction unit for predicting a power consumption of a customer after a predetermined time using the selected first prediction model,
Lt; / RTI >
The processor repeatedly generates the K power load prediction models using the predicted cycle determined according to the variability of the load of the customer,
Wherein the variability of the load is defined as a ratio of an actual variation of the power usage to an average predicted value of the power usage.
The method according to claim 1,
Wherein the selection unit selects the first prediction model according to a time interval in which a peak value of the power usage amount related to the electronic device exists.
The method according to claim 1,
Wherein the selection unit selects the first predictive model according to an electric power interval in which an average power consumption for a predetermined time period is present in the electronic apparatus.
The method according to claim 1,
Wherein the K power load prediction models are machine-learned using learning data clustered in K according to a power usage pattern of the electronic device, and the learning data includes power consumption including power consumption of a plurality of customers in a predetermined measurement cycle Prediction device.
5. The method of claim 4,
Wherein K is determined such that the predicted accuracy calculated through the test data is greater than or equal to a predetermined threshold, and the K power load prediction models are computed using a clustered learning data corresponding to the determined K, .
Identifying an electronic device having a correlation coefficient equal to or greater than a threshold power load based on the learning data;
Allocating the learning data according to a predetermined measurement cycle to K clusters based on the identified electronic device; And
Generating K power load prediction models using learning data corresponding to each of the K clusters
Lt; / RTI >
Wherein generating the K power load prediction models comprises:
Calculating the variability of the load with respect to each customer; And
Generating the K power load prediction models repeatedly using a prediction cycle determined according to the variability of load of the customer;
Lt; / RTI >
Wherein the variability of the load is defined as a ratio of an actual variation of the power usage to an average predicted value of the power usage.
The method according to claim 6,
Wherein K is determined such that the prediction accuracy of each of the power load prediction models calculated through the test data is greater than or equal to a predetermined threshold value.
The method according to claim 6,
Wherein generating the K power load prediction models comprises:
Reducing the measurement period in case the load variability of the customer is increased
≪ / RTI >
The method according to claim 6,
Wherein generating the K power load prediction models comprises:
Calculating predicted accuracy of the K power load prediction models generated according to the first prediction cycle and generating new K power load prediction models repeatedly using the second prediction cycle changed according to the calculated prediction accuracy
≪ / RTI >
The method according to claim 6,
Setting a measurement period according to a power load variability of a customer to which the K power load prediction models are applied; And
Collecting learning data related to power consumption of a plurality of customers according to the predetermined measurement period
≪ / RTI >
11. The method of claim 10,
Wherein the setting of the measurement period comprises:
Calculating a prediction accuracy of K power load prediction models corresponding to the learning data collected according to the first measurement period and newly setting a second measurement period according to the calculated prediction accuracy
≪ / RTI >
8. The method of claim 7,
Wherein identifying the electronic device comprises:
Identifying a plurality of electronic devices existing within a predetermined range in order of priority from learning data associated with a specific customer;
≪ / RTI >

Images