KR102545304B1 - Apparatus and method managing power based on predicting power peak guideline - Google Patents

Abstract

The present invention relates to a power management apparatus and method using a predicted power peak guideline capable of reducing electricity costs. The power management method using the predicted power peak guideline according to the present embodiment is a power management method performed by a processor of a power management device, and uses time-series power consumption by using sensor data collected from an IoT sensor provided in a power demand entity. Generating data, generating future predicted power usage data by applying the time-series power usage data to a Fourier series to which coefficients are fitted, and predicting a power upper limit value that determines a predicted power upper limit value using the predicted power usage data as an input Determining a predicted power upper limit value from the predicted power usage data using a model and generating a predicted power peak guideline, and generating an alarm that the electricity rate is exceeded based on the distance from the time-series power usage data to the predicted power peak guideline. steps may be included.

Description

Apparatus and method for power management using predicted power peak guidelines {APPARATUS AND METHOD MANAGING POWER BASED ON PREDICTING POWER PEAK GUIDELINE}

The present invention relates to a power management apparatus and method using a predicted power peak guideline capable of reducing electricity costs.

Recently, as the subject of power management changes from supply-oriented to demand-oriented, the importance of power management in the consumption part is emerging. Power demand management deals with loads of various scales, from large-scale regional loads such as microgrids to building units and even single consumer units. The most basic research to manage power demand is to understand the pattern of “power” consumption and to predict the “power” consumption. Research on predicting future 　power　use has been conducted in various ways. Looking at previous studies in Korea, electricity consumption is predicted using weather information. However, since the cycle of the temperature data trend is different from the cycle of the trend of power consumption, it is difficult to accurately predict the power consumption because the cycle of the two data is not reflected and the power consumption is predicted.

The foregoing background art is technical information that the inventor possessed for derivation of the present invention or acquired during the derivation process of the present invention, and cannot necessarily be said to be known art disclosed to the general public prior to filing the present invention.

Korean Registered Patent Publication No. 10-1355978 (2014.01.21)

An object of the present invention is to provide a power management device and method capable of reducing electric charges of a power demander by generating and providing predicted power consumption and predicted power peak guidelines for the power demander.

An object of the present invention is to provide a power management apparatus and method capable of generating and providing predicted power consumption and predicted power peak guidelines for the power demand entity, thereby helping the power demand entity to set an electric bill budget.

An object of the present invention is to provide a power management apparatus and method capable of generating and providing predicted power consumption and predicted power peak guidelines for power demand entities to assist in power supply scheduling to power demand entities.

The problem to be solved by the present invention is not limited to the above-mentioned problems, and other problems and advantages of the present invention that are not mentioned can be understood by the following description and more clearly understood by the embodiments of the present invention. It will be. In addition, it will be appreciated that the problems and advantages to be solved by the present invention can be realized by the means and combinations indicated in the claims.

The power management method using the predicted power peak guideline according to the present embodiment is a power management method performed by a processor of a power management device, and uses time-series power consumption by using sensor data collected from an IoT sensor provided in a power demand entity. Generating data, generating future predicted power usage data by applying the time-series power usage data to a Fourier series to which coefficients are fitted, and predicting a power upper limit value that determines a predicted power upper limit value using the predicted power usage data as an input Determining a predicted power upper limit value from the predicted power usage data using a model and generating a predicted power peak guideline, and generating an alarm that the electricity rate is exceeded based on the distance from the time-series power usage data to the predicted power peak guideline. steps may be included.

A power management device using predicted power peak guidelines according to the present embodiment is a power management device, and includes a processor and a memory operably connected to the processor and storing at least one code executed by the processor. When executed through, the processor generates time-series power usage data using sensor data collected from the IoT sensor provided in the power demand entity, and applies the time-series power usage data to the Fourier series to which the coefficients are fitted to predict future power usage. Generates data, determines the predicted power upper limit from the predicted power usage data by using a power cap prediction model that determines the predicted power cap using the predicted power usage data as an input, generates a predicted power peak guideline, and time-series power usage data code that causes an alarm to be exceeded based on the distance from the predicted power peak guideline.

In addition to this, another method for implementing the present invention, another system, and a computer readable recording medium storing a computer program for executing the method may be further provided.

Other aspects, features and advantages other than those described above will become apparent from the following drawings, claims and detailed description of the invention.

According to the present invention, it is possible to provide a power management apparatus and method capable of reducing electric charges of a power demander by generating and providing predicted power consumption and predicted power peak guidelines for the power demander.

In addition, it is possible to provide a power management device and method capable of assisting the power demander in setting a budget for electric charges by generating and providing predicted power consumption and predicted power peak guidelines for the power demander.

In addition, it is possible to provide a power management apparatus and method capable of assisting in power supply scheduling to a power demand entity by generating and providing predicted power consumption and predicted power peak guidelines for the power demand entity.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

1 is an exemplary diagram of a power management environment according to an exemplary embodiment.
2 is a block diagram schematically illustrating a configuration of a power management device according to an exemplary embodiment.
FIG. 3 is a block diagram schematically illustrating the configuration of a power management unit in the power management device of FIG. 2 .
4 is a diagram illustrating power per minute data generated using data collected from an IoT sensor according to an embodiment.
5 is a diagram showing the coefficients of the fitted Fourier series according to the present embodiment.
6 is a graph visualizing predicted average power consumption data generated by applying average power consumption to a Fourier series to which coefficients are fitted according to the present embodiment.
7 is a diagram illustrating determination of a reinforcement learning-based predicted power peak guideline according to an embodiment.
8 is a graph visualizing the predicted power peak guideline according to the present embodiment.
9 and 10 are exemplary diagrams of generating an alarm for power management according to the present embodiment.
11 is a block diagram schematically illustrating a configuration of a power management device according to another exemplary embodiment.
12 is a flowchart illustrating a power management method according to an exemplary embodiment.

Advantages and features of the present invention, and methods for achieving them will become clear with reference to the detailed description of embodiments in conjunction with the accompanying drawings. However, it should be understood that the present invention is not limited to the embodiments presented below, but may be implemented in a variety of different forms, and includes all conversions, equivalents, and substitutes included in the spirit and scope of the present invention. . The embodiments presented below are provided to complete the disclosure of the present invention and to fully inform those skilled in the art of the scope of the invention to which the present invention belongs. In describing the present invention, if it is determined that a detailed description of related known technologies may obscure the gist of the present invention, the detailed description will be omitted.

Terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, the terms "include" or "have" are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded. Terms such as first and second may be used to describe various components, but components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another.

In addition, in this application, “unit” may be a hardware component such as a processor or a circuit, and/or a software component executed by a hardware component such as a processor.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In the description with reference to the accompanying drawings, the same or corresponding components are assigned the same reference numerals, and overlapping descriptions thereof are omitted. I'm going to do it.

1 is an exemplary diagram of a power management environment according to an exemplary embodiment. Referring to FIG. 1 , a power management environment 1 may include a power management device 100, a power demand entity 200, an IoT sensor 300, a user terminal 400, and a network 500.

The power management device 100 may generate time-series power consumption data using sensor data collected from the IoT sensor 300 provided in the power demand entity 200 . In this embodiment, time-series power usage data may include power usage data corresponding to a sensing time from sensing data classified according to a predetermined classification criterion. Also, the time-series power usage data may include time-series power usage data during the first period.

The power management device 100 may generate future predicted power usage data by applying the time-series power usage data to a Fourier series to which coefficients are fitted. Here, the time-series power usage data has a first period, and the first period may consist of T first time steps having preset times. Specifically, the preset time may include 24 hours, the first time step may include 1 minute, and T may include 1440 times. In this embodiment, before generating predicted power usage data, the power management apparatus 100 may calculate a Fourier series as a periodic function from the predicted power usage data and fit coefficients of the Fourier series. The power management apparatus 100 may calculate a Fourier series from time-series power usage data having a period, and may fit coefficients of the Fourier series.

The power management device 100 determines the predicted power upper limit value from the predicted power usage data through operation of a Fourier series using a power upper limit prediction model that determines the predicted power upper limit value by taking the predicted power consumption data as an input and predicts the peak power guide line can be created. Here, the predicted power consumption data input to the power upper limit prediction model may be data of a first period composed of L second time steps having preset times. Specifically, the preset time may include 24 hours, the second time step may include 15 minutes, and L may include 96.

In this embodiment, the power upper limit prediction model may include an artificial intelligence model that outputs an upper power limit value by taking as input predicted power consumption data consisting of L second time steps, and the power management device 100 predicts the power upper limit value An AI model can be trained to create a model.

The power management device 100 may generate and store a power upper limit prediction model by training an artificial intelligence model using supervised learning, unsupervised learning, or reinforcement learning. As an example, the power management device 100 may generate and store a power upper limit prediction model by training an artificial intelligence model using reinforcement learning, and the reinforcement learning analyzes what action an agent should take every moment. Given the circumstances to do so, one could include the theory that the best path can be found by experience, without data. Reinforcement learning can be performed mainly by a Markov decision process (MDP).

To explain the Markov decision process, first, an environment in which the information necessary for the agent to take the next action is given, second, how the agent will behave in that environment, and third, if the agent does well, a reward ( Fourth, the optimal policy can be derived by repeating experience until the future reward reaches the highest point.

The power management device 100 may determine an average value of L predicted power upper limit values output through the power upper limit value prediction model at every second time step during the first period as a predicted power peak guideline during the first period.

When the predicted power peak guideline is determined, the power management device 100 may generate an alarm indicating that an electric charge is exceeded based on a distance from input current time-series power usage data to the predicted power peak guideline. The power management device 100 predicts a time period in which electricity rates may exceed according to a Euclidean distance calculated from input current time-series power usage data and predicted power peak guidelines that are less than a preset threshold. An alarm including may be generated and the alarm may be transmitted to the user terminal 400 . Upon receiving the alarm, the user terminal 400 may cut off and manage unnecessary power of the power demand entity 200 so that the power of the power demand entity 200 can be saved.

In this embodiment, the power management device 100 may exist independently in the form of a server or may implement functions provided by the power management device 100 in the form of an application and be installed in the user terminal 400 .

In addition, the power management device 100 may be a database server that provides data necessary for applying various artificial intelligence algorithms.

Here, artificial intelligence (AI) is a field of computer science and information technology that studies how to enable computers to do thinking, learning, and self-development that can be done with human intelligence. It can mean allowing you to imitate intelligent behavior.

Also, artificial intelligence does not exist by itself, but is directly or indirectly related to other fields of computer science. In particular, in modern times, attempts to introduce artificial intelligence elements in various fields of information technology and use them to solve problems in those fields are being actively made.

Machine learning is a branch of artificial intelligence that can include areas of study that give computers the ability to learn without being explicitly programmed. Specifically, machine learning can be said to be a technology that studies and builds a system that learns based on empirical data, makes predictions, and improves its own performance, as well as algorithms for it. Algorithms in machine learning can take the form of building specific models to derive predictions or decisions based on input data, rather than executing rigidly defined static program instructions.

The power demand entity 200 is a load that consumes power, and may include loads of various scales, from large-scale regional loads to building units and consumer units. In this embodiment, the power demand entity 200 may include a building or a factory, and machines (eg, factory equipment, air conditioners, etc.) consuming power in various ways are provided inside the building.

The IoT sensor 300 is a sensor applied to the Internet of Things environment, and is provided in the power demand entity 200 to sense power data consumed by the power demand entity 200 and transmit it to the power management device 100. The IoT sensor 300 may include a current sensor, a power sensor, a vibration sensor, an illuminance sensor, a temperature sensor, a proximity sensor, and the like.

The user terminal 400 is a terminal provided by a user or manager who manages power of the power demand entity 200, and provides a power management service by accessing a power management application and/or a power management site provided by the power management device 100. can receive

The user terminal 400 may include a communication terminal capable of performing functions of a computing device (not shown), and in addition to a desktop computer 401, a smartphone 402, and a laptop computer 403 operated by a user, Tablet PCs, smart TVs, mobile phones, PDAs (personal digital assistants), media players, micro servers, GPS (global positioning system) devices, e-readers, digital broadcasting terminals, navigation devices, kiosks, MP3 players, digital cameras, home appliances and It may be other mobile or non-mobile computing devices, but is not limited thereto. In addition, the user terminal 400 may be a wearable terminal such as a watch, glasses, hair band, and ring having a communication function and a data processing function. The user terminal 400 is not limited to the above, and a terminal capable of web browsing can be borrowed without limitation.

The network 500 may serve to connect the power management device 100 , the IoT sensor 300 and the user terminal 400 . Such a network 500 may be, for example, a wired network such as a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), an integrated service digital network (ISDN), a wireless LAN (WLAN), or a code network (CDMA). -division multiple access), satellite communication, etc., but the scope of the present invention is not limited thereto. In addition, the network 500 may transmit and receive information using short-range communication and/or long-distance communication. Here, short-range communication may include Bluetooth, radio frequency identification (RFID), infrared data association (IrDA), ultra-wideband (UWB), ZigBee, and Wi-Fi technologies, and long-distance communication may include code-division multiple access (CDMA). ), frequency-division multiple access (FDMA), time-division multiple access (TDMA), orthogonal frequency-division multiple access (OFDMA), and single carrier frequency-division multiple access (SC-FDMA) technologies.

The network 500 may include connections of network elements such as hubs, bridges, routers, and switches. Network 500 may include one or more connected networks, e.g., a multiple network environment, including a public network such as the Internet and a private network such as a secure corporate private network. Access to network 500 may be provided through one or more wired or wireless access networks.

The network 500 may include connections of network elements such as hubs, bridges, routers, and switches. Network 500 may include one or more connected networks, e.g., a multiple network environment, including a public network such as the Internet and a private network such as a secure enterprise private network. Access to network 500 may be provided through one or more wired or wireless access networks.

Furthermore, the network 500 includes CAN (controller area network) communication, V2I (vehicle to infrastructure) communication, V2X (vehicle to everything) communication, wave (wireless access in vehicular environment) communication technology, things, etc. It can support an Internet of Things (IoT) network and/or 5G communication that exchanges and processes information between distributed components. Here, NB (narrowband)-IoT is one of low-power/wide-area IoT technologies using LTE (Long-Term Evolution) frequencies, and can be used for intermittently transmitting low-capacity data for tracking, sensing, and meter reading.

2 is a block diagram schematically illustrating a configuration of a power management device according to an exemplary embodiment. In the following description, descriptions of overlapping portions with those of FIG. 1 will be omitted. Referring to FIG. 2 , the power management device 100 may include a communication unit 110, a storage medium 120, a program storage unit 130, a database 140, a power management unit 150, and a control unit 160. there is.

The communication unit 110 may provide a communication interface required to provide a transmission/reception signal between the power management device 100, the IoT sensor 300, and the user terminal 400 in the form of packet data in conjunction with the network 500. Furthermore, the communication unit 110 may serve to receive a predetermined information request signal from the user terminal 400 and may serve to transmit information processed by the power management unit 150 to the user terminal 400. . Here, the communication network is a medium that serves to connect the power management device 100, the IoT sensor 300, and the user terminal 400, and the user terminal 400 is connected to the power management device 100. It may include a path providing an access path so that information can be transmitted and received after access. In addition, the communication unit 110 may be a device including hardware and software necessary for transmitting and receiving signals such as control signals or data signals to and from other network devices through wired or wireless connections.

The storage medium 120 performs a function of temporarily or permanently storing data processed by the controller 160 . Here, the storage medium 120 may include magnetic storage media or flash storage media, but the scope of the present invention is not limited thereto. The storage medium 120 may include built-in memory and/or external memory, and may include volatile memory such as DRAM, SRAM, or SDRAM, one time programmable ROM (OTPROM), PROM, EPROM, EEPROM, mask ROM, and flash ROM. , NAND flash memory, or non-volatile memory such as NOR flash memory, SSD. It may include a compact flash (CF) card, a flash drive such as an SD card, a Micro-SD card, a Mini-SD card, an Xd card, or a memory stick, or a storage device such as an HDD.

The program storage unit 130 generates time-series power usage data using sensing data collected from the IoT sensor 300 provided in the power demand entity 200, and converts the time-series power usage data into a Fourier series to which coefficients are fitted. A task of generating future predicted power consumption data by application, a task of calculating a Fourier series as a periodic function from the predicted power consumption data before generating the predicted power consumption data and fitting the coefficients of the Fourier series, and a power cap prediction model. Determining the predicted power upper limit from the predicted power usage data and generating the predicted power peak guideline, creating and storing the power upper limit prediction model before determining the predicted power upper limit, predicting from the input current time-series power usage data Based on the distance to the power peak guideline, it is equipped with control software that generates an alarm for exceeding electricity rates.

The database 140 may include a management database that stores various information for power management of the power demand entity 200 . For example, the management database includes specification information for the IoT sensor 300 provided in the power demand entity 200, 3D map information for the power demand entity 200, and power consumption installed in the power demand entity 200. Information about the type of machine may be stored.

In addition, the management database may store a Fourier series for generating predicted power usage data and a power upper limit prediction model for determining the predicted power upper limit value.

In addition, the database 140 may include a user database for storing information of a user (or manager) to be provided with a power management service for the power demand entity 200 . Here, the information of the user includes basic information about the user such as the user's name, affiliation, personal information, gender, age, contact information, e-mail, address, image, etc., and user information such as ID (or e-mail) and password. It may include information about authentication (login), country of access, access location, information about the device used for access, and information related to access, such as the connected network environment.

In addition, the user database includes user-specific information, information and/or category history provided by the user accessing the power management application or power management site, environment setting information set by the user, resource usage information used by the user, and resource usage by the user. Billing and payment information corresponding to may be stored.

The power management unit 150 may generate time-series power usage data using sensing data collected from the IoT sensor 300 provided in the power demand entity 200 . The power management unit 150 may generate future predicted power usage data by applying the time-series power usage data to a Fourier series to which coefficients are fitted. The power management unit 150 may calculate a Fourier series as a period function from the predicted power usage data before generating the predicted power usage data, and may fit coefficients of the Fourier series. The power management unit 150 may input the predicted power usage data into an upper power limit prediction model to determine a predicted power upper limit value and generate a predicted power peak guideline. The power management unit 150 may generate and store an upper power limit prediction model before determining the predicted power upper limit value. The power management unit 150 may generate an alarm that an electric charge is exceeded based on a distance from input current time-series power usage data to a predicted power peak guideline.

The control unit 160, as a kind of central processing unit, may control the entire operation of the power management device 100 by driving control software loaded in the program storage unit 130. The controller 160 may include all types of devices capable of processing data, such as a processor. Here, a 'processor' may refer to a data processing device embedded in hardware having a physically structured circuit to perform functions expressed by codes or instructions included in a program, for example. As an example of such a data processing device built into hardware, a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated (ASIC) circuit), field programmable gate array (FPGA), etc., but the scope of the present invention is not limited thereto.

3 is a block diagram for schematically explaining the configuration of a power management unit in the power management device of FIG. 2, and FIG. 4 describes power data per minute generated using data collected from an IoT sensor according to the present embodiment. FIG. 5 is a diagram showing the coefficients of the fitted Fourier series according to the present embodiment, and FIG. 6 is the predicted average power usage generated by applying the average power usage according to the present embodiment to the Fourier series to which the coefficients are fitted. A graph visualizing data, FIG. 7 is a diagram explaining the determination of a reinforcement learning-based predicted power peak guideline according to this embodiment, and FIG. 8 is a graph visualizing the predicted power peak guideline according to this embodiment. 9 and 10 are exemplary diagrams of generating an alarm for power management according to the present embodiment. In the following description, descriptions of portions overlapping those of FIGS. 1 and 2 will be omitted.

Referring to FIGS. 3 to 10 , the power management unit 150 may include a collection unit 151 , a first processing unit 152 , a second processing unit 153 , and a third processing unit 154 .

The collection unit 151 may generate time-series power consumption data using sensing data collected from the IoT sensor 300 provided in the power demand entity 200 . In this embodiment, the collection unit 151 classifies the sensing data collected from the IoT sensor 300 by applying a preset classification criterion, and generates power consumption data corresponding to the sensing time from the sensing data classified according to the classification criterion. You can create time series data that includes

Here, the classification criteria may include first to fourth classification criteria. The first classification criterion may include a criterion for classifying the sensing data into a late-night time zone (eg, from 22:00 to 08:00), in which power consumption is less than during the daytime. Nighttime electricity is mainly used for power generation such as nuclear power generation and thermal power generation, which are responsible for the base load during the nighttime hours when usage is less than during the daytime. can

The second classification criterion may include a criterion for classifying the sensing data into a summer season (eg, June to August) in which power consumption is high. As concerns about the supply and demand of electricity in the summer are heightened due to abnormally high temperatures and power plant shutdown accidents, etc., it is possible to establish a unit plant shutdown program or a self-generation plan for each risk level of the power reserve ratio through the second classification criterion.

The third classification criterion may include a criterion for classifying the sensing data into a winter season (eg, from December to February) in which power usage is less than that of the summer season. Through the third classification criterion, it is possible to establish a power supply and demand plan necessary for the operation rate of heating and power demand subjects 200 due to cold weather.

The fourth classification criterion is a criterion for classifying sensing data according to time by minute, hour, day, week, month, and year. can include Power supply and demand plans can be established by separately classifying power levels that do not have an unusual impact.

FIG. 4 illustrates time-series power consumption data generated by applying a first classification criterion to sensing data collected by the collection unit 151 from the IoT sensor 300 . Specifically, the time-series power usage data shown in FIG. 4 shows an example in which power usage data in the late-night time zone among sensing data collected from the IoT sensor 300 from the past to the present is generated in units of a first time step.

The first processing unit 152 may generate future predicted power consumption data by applying the time-series power consumption data generated by the collecting unit 151 to a Fourier series to which coefficients are fitted.

The first processor 152 may calculate a Fourier series as a periodic function from the predicted power usage data and fit coefficients of the Fourier series before generating future predicted power usage data. In order to fit the coefficients of the Fourier series, the first processing unit 152 may calculate average power usage data for the previous time-series power usage data group during a first period consisting of T first time steps having a preset time. there is. For example, from time-series power usage data in units of a first time step during a first period (from 0:00 to 24:00) in the past (eg, January 1, 1980) to the present (eg, July 2022) Month 17) average power usage data up to time-series power usage data in units of a first time step during the first period may be calculated. The average power usage data may be average time-series power usage in units of a first time step during the first period. For example, in this embodiment, in order to predict average power usage data in units of the first time step during the first period of July 18, 2022, from the past (January 1, 1980) to the previous (July 2022 17) may use average power usage data calculated from power usage data in units of the first time step during the first period.

The first processing unit 152 performs curve fitting on the average power usage data based on a non-linear least square method and applies the Fourier series expressed in Equation 1 below to fit the coefficients of the Fourier series. can do.

는 미래의 평균 예측 전력 사용량 데이터를 나타내고,

,

및

은 이전 전력 사용량 데이터군의 평균값을 이용하여 피팅된 계수를 나타내고,

는 고정되어 있고,

는 시간(예를 들어, 제1 타임 스텝)을 나타낼 수 있다. here,

Represents future average predicted power usage data,

,

and

Represents a coefficient fitted using the average value of the previous power usage data group,

is fixed,

may represent time (eg, the first time step).

FIG. 5 shows coefficients of 17 Fourier series calculated by curve fitting average power consumption data based on a non-linear least square method.

,

및

)가 피팅된 푸리에 급수에, 예측할 미래의 시간 데이터(

: 제1 주기를 갖는 T개의 제1 타임 스텝)를 입력하여 다음(미래) 제1 주기의 T개의 제1 타임 스텝에 대한 제1 예측 평균 전력 사용량 데이터를 생성할 수 있다.The first processing unit 152 is a coefficient (

,

and

) is the fitted Fourier series, the predicted future time data (

: T first time steps having a first period) may be input to generate first predicted average power usage data for T first time steps of a next (future) first period.

FIG. 6 is a graph visualizing first predicted average power usage data generated by applying average power usage calculated from the time-series power usage data shown in FIG. 4 to a Fourier series to which coefficients are fitted, with the horizontal axis representing Time is represented, and the vertical axis represents the amount of power. 6, 610 is a graph visualizing average power usage data for time-series power usage data classified by the first classification criterion, and 620 is a graph visualizing predicted average power usage data calculated using a fitted coefficient. .

The second processing unit 153 may input the first predicted average power usage data generated by the first processing unit 152 to an upper power limit prediction model to determine a predicted power upper limit value and generate a predicted power peak guideline.

The second processing unit 153 may generate an upper power limit prediction model before determining an upper limit predicted power by using the upper limit power prediction model.

: 제1 주기를 갖는 T개의 제1 타임 스텝)를 입력하여 산출될 수 있다. 여기서 제1 예측 평균 전력 사용량 데이터의 개수는 1440개일 수 있다. 또한 제2 예측 평균 전력 사용량 데이터는, 제1 예측 평균 전력 사용량 데이터를 제1 주기를 같는 L개의 제2 타임 스텝으로 분류하고, 제2 타입스텝에 포함되는 제1 예측 평균 전력 사용량 데이터의 평균값으로 산출될 수 있다. 여기서 제2 예측 평균 전력 사용량 데이터의 개수는 96개 일 수 있다.The second processing unit 153 may calculate second predicted average power consumption data from the first predicted average power consumption data output from the first processing unit 152 when generating the power upper limit prediction model. Here, the first predicted average power usage data is future time data predicted to a Fourier series to which coefficients are fitted (

: It can be calculated by inputting T first time steps having a first period. Here, the number of first predicted average power usage data may be 1440. In addition, the second predicted average power usage data is an average value of the first predicted average power usage data included in the second type step by classifying the first predicted average power usage data into L second time steps having the same first period. can be derived. Here, the number of second predicted average power usage data may be 96.

The second processing unit 153 may generate an upper power limit prediction model for determining the predicted power upper limit value from the second predicted average power usage through training of the reinforcement learning-based neural network model.

7A shows a reinforcement learning-based neural network model including an agent and an environment (evnvironment), and FIG. 7B shows an action and a reward for determining an upper limit of prediction power by an agent. It is an exemplary diagram for a detailed explanation.

Referring to FIG. 7A , a plurality of power upper limit values may be preset in an agent. A plurality of power upper limit values may be set to a power value lower than the power value set in the previous year's power peak guideline (730 in FIG. 8). For example, as for the plurality of power upper limit values, 10 power upper limit values may be set in increments of 1000 kw from 1000 kw (first power upper limit value) to 10000 kw (tenth power upper limit value).

For a given state (state: s _t ) in which the second predicted average power usage data is input, the agent selects among a plurality of power upper limit values based on the power difference between the second predicted average power usage data and the plurality of power upper limit values An action of determining a certain power upper limit value as the predicted power upper limit value may be performed.

Here, the power difference may include a difference in area as shown in FIG. 7B. 7B , A is a state given as an input to an agent, and may be second predicted average power usage data. Here, the second predicted average power usage data may be an average value of the first predicted average power usage data during the second time step. Also, from FIG. 7B , B may be an area for a plurality of power upper limit values set in the agent, and includes a first area B1 for the first power upper limit value to a tenth area B10 for the tenth power upper limit value. can do.

The agent calculates a difference in area between the area (A) of the second predicted average power usage data and the area (B: B1-B10) of the plurality of upper limit power values, and calculates an upper power limit value having the smallest difference in area. It is possible to perform an action for determining the predicted power upper limit value.

In this embodiment, the neural network model may be trained to receive a reward for a difference in area corresponding to an upper limit of prediction power determined by an agent.

In this embodiment, the neural network model receives a given state (sate) and action (action) of an agent and returns a reward and the next state (state: s _t+1 ). The agent may be configured to perform an action in a given state to receive a maximum reward and update the ability to determine the upper limit of the predicted power.

In this embodiment, the reinforcement learning-based neural network may use a deep Q-nekwork (DQN) or Q-learning reinforcement learning-based neural network.

The gym package environment of OpenAI Gym can be used as an environment used in the practice of reinforcement learning to determine the upper limit of predicted power. OpenAI Gym is a library that helps with reinforcement learning and enables reinforcement learning in more general situations. In addition, other libraries suitable for artificial neural network-based reinforcement learning for determining the upper limit of prediction power, such as TensorFlow, can be used.

The second processing unit 153 may generate and store a power upper limit prediction model for determining the predicted power upper limit value from the second predicted average power usage data through training of the neural network model.

From this, when the predicted power consumption data is input to the power upper limit prediction model, a corresponding predicted power upper limit value may be output. The second processing unit 153 may generate an average value of L predicted power upper limit values output through the power upper limit value prediction model at every second time step during the first period as a predicted power peak guideline during the first period.

8 is a graph visualizing the predicted power peak guideline according to the present embodiment. Referring to FIG. 8 , 810 is a graph visualizing average power usage data for time-series power usage data classified by a first classification criterion, and 820 is a graph visualizing predicted average power usage data calculated using a fitted coefficient. , 830 is a graph visualizing the power peak guideline of the previous year, and 840 is a graph visualizing the predicted power peak guideline generated using the predicted average power consumption data of 820 . In this embodiment, the predicted power peak guideline of 840 may be determined to be smaller than the power peak guideline of 830 of the previous year.

When the power peak guideline is determined, the third processing unit 154 may generate an alarm indicating that an electric charge is exceeded based on a distance from input current time-series power usage data to the predicted power peak guideline. The third processing unit 154 determines the predicted time period during which electricity rates may be exceeded according to the Euclidean distance calculated from the current time-series power usage data and the predicted power peak guideline, which is less than a preset threshold. An alarm including the above may be generated and the alarm may be transmitted to the user terminal 400 . Here, the Euclidean distance is a measure for calculating the relative distance difference between two points (time-series power consumption data and the power value of the predicted power peak guideline). It can be seen that it is close to

9 and 10 are exemplary diagrams of generating an alarm for power management according to the present embodiment. Referring to FIG. 9 , the third processing unit 154 determines that the Euclidean distance calculated from the input current time-series power usage data and the predicted power peak guideline is less than a predetermined threshold, and the power demand An example of generating a pop-up window and an alarm including text information including predicted time zones in which electricity rates may be exceeded on a 3D map including the subject 200 and outputting the alarm to the user terminal 400 is shown.

Referring to FIG. 10, the third processing unit 154 determines that the Euclidean distance calculated from the input current time-series power usage data and the predicted power peak guideline is less than a predetermined threshold value, and the power demand On the 3D map including the subject 200, the power demand subject 200 to generate an alarm is displayed in a specific color (eg, red), and a graphic and an alarm are generated and output to the user terminal 400 An example is shown.

Upon receiving the alarm, the user terminal 400 blocks unnecessary power of the power demander 200 or power that can be cut off so that the power of the power demander 200 can be reduced to cancel the occurrence of the alarm, thereby reducing electricity bills. can be managed so that

11 is a block diagram schematically illustrating a configuration of a data classification apparatus according to another embodiment. In the following description, descriptions of portions overlapping with those of FIGS. 1 to 11 will be omitted. Referring to FIG. 12 , a power management device 100 according to another embodiment may include a processor 170 and a memory 180.

In this embodiment, the processor 170 includes the communication unit 110, the storage medium 120, the program storage unit 130, the database 140, the power management unit 150 and the control unit 160 disclosed in FIGS. 2 and 3. can handle the functions performed by

The processor 170 may control the entire operation of the power management device 100 . Here, a 'processor' may refer to a data processing device embedded in hardware having a physically structured circuit to perform functions expressed by codes or instructions included in a program, for example. As an example of the data processing device built into the hardware, processing devices such as microprocessors, central processing units, processor cores, multiprocessors, ASICs, and FPGAs may be covered, but the scope of the present invention is not limited thereto. .

The memory 180 is operatively connected to the processor 170 and may store at least one code associated with an operation performed by the processor 170 .

In addition, the memory 180 may perform a function of temporarily or permanently storing data processed by the processor 170 and may include data built into the database 140 . Here, the memory 180 may include a magnetic storage medium or a flash storage medium, but the scope of the present invention is not limited thereto. Such memory 180 may include internal memory and/or external memory, volatile memory such as DRAM, SRAM, or SDRAM, OTPROM, PROM, EPROM, EEPROM, mask ROM, flash ROM, NAND flash memory, or NOR It may include a non-volatile memory such as flash memory, a flash drive such as SSD, CF card, SD card, Micro-SD card, Mini-SD card, xD card, or memory stick, or a storage device such as HDD.

12 is a flowchart illustrating a power management method according to an exemplary embodiment. In the following description, descriptions of portions overlapping those of FIGS. 1 to 11 will be omitted. The power management method according to the present embodiment will be described on the assumption that the power management device 100 is performed by the processor 170 with the help of peripheral components.

Referring to FIG. 12 , in step S1210 , time-series power consumption data may be generated using sensor data collected from the IoT sensor 300 provided in the power demand entity 200 of the processor 170 . In this embodiment, the processor 170 classifies the sensing data collected from the IoT sensor 300 by applying a preset classification criterion, and includes power usage data corresponding to the sensing time from the sensing data classified according to the classification criterion. Time-series power consumption data can be generated. Here, the classification criteria include a first classification criterion for classifying the sensing data into a late-night time zone with less power consumption than daytime, a second classification criterion for classifying the sensing data into a summer period with high power consumption, and a second classification criterion for classifying the sensing data as a summer period with high power consumption. A third classification criterion for classifying the sensing data into fewer winters, and the sensing data according to time by minute, hour, day, week, month, and year ( year) may include a fourth classification criterion for classification.

In step S1220, the processor 170 may generate future predicted power usage data by applying the time-series power usage data to the Fourier series to which coefficients are fitted. Here, the processor 170 may calculate a Fourier series as a period function from the predicted power usage data and fit coefficients of the Fourier series before generating the predicted power usage data. When fitting the coefficients of the Fourier series, the processor 170 may calculate average power usage data for a previous power usage data group during a first period consisting of T first time steps having a preset time. The processor 170 may fit the coefficients of the Fourier series by curve fitting the average power usage based on a non-linear least square method. Accordingly, the processor 170 applies each of the T first time steps during the next first period to the Fourier series to which the coefficients are fitted, and the first prediction average corresponding to each of the T first time steps during the next first period. Power usage data can be generated.

In step S1230, the processor 170 determines a predicted power upper limit from the predicted power usage data and generates a predicted power peak guideline by using a power upper limit prediction model for determining the predicted power upper limit using the predicted power usage data as an input. there is. In this embodiment, the processor 170 may generate an upper power limit prediction model before determining the predicted power upper limit value. When the processor 170 generates the power upper limit value prediction model, the first prediction average power usage data during the first period consisting of L second time steps having a time longer than the first time step is calculated as an average of the first predicted average power usage data. 2 Estimated average power consumption data can be calculated. The processor 170 determines the power difference between the second predicted average power usage data and the plurality of upper power limit values for a given state in which the second predicted average power usage data is input by an agent in which a plurality of upper power limit values are preset. Based on , the action of determining one of the upper power limits as the upper limit of the predicted power is performed, and the neural network model is trained to receive a reward for the power difference corresponding to the upper limit of the predicted power determined by the agent. can In this embodiment, the neural network model receives a given state and action of the agent and returns a reward and a next state (evnvironment), the agent performs an action in the given state to receive the maximum reward, and the predicted power upper limit It may be a reinforcement learning-based neural network model configured to update the decision performance of In addition, the reinforcement learning-based neural network may use a deep Q-network (DQN), a reinforcement learning-based neural network. The processor 170 may generate and store a power upper limit value prediction model for determining the predicted power upper limit value from the second predicted average power usage data through training of the neural network model. Thereafter, the processor 170 may generate an average value of L predicted power upper limit values output through the power upper limit value prediction model at every second time step during the first period as a predicted power peak guideline during the first period.

In step S1240, the processor 170 may generate an alarm indicating that the electricity rate is exceeded based on the distance from the time-series power usage data to the predicted power peak guideline. The processor 170 includes a predicted time period during which electricity rates may be exceeded according to the Euclidean distance calculated from the input current time-series power usage data and the predicted power peak guideline being less than a preset threshold. may generate an alarm and transmit the alarm to the user terminal 400 . Upon receiving the alarm, the user terminal 400 may cut off and manage unnecessary power of the power demand entity 200 so that the power of the power demand entity 200 can be saved.

Embodiments according to the present invention described above may be implemented in the form of a computer program that can be executed on a computer through various components, and such a computer program may be recorded on a computer-readable medium. At this time, the medium is a magnetic medium such as a hard disk, a floppy disk and a magnetic tape, an optical recording medium such as a CD-ROM and a DVD, a magneto-optical medium such as a floptical disk, and a ROM hardware devices specially configured to store and execute program instructions, such as RAM, flash memory, and the like.

Meanwhile, the computer program may be specially designed and configured for the present invention, or may be known and usable to those skilled in the art of computer software. An example of a computer program may include not only machine language code generated by a compiler but also high-level language code that can be executed by a computer using an interpreter or the like.

In the specification of the present invention (particularly in the claims), the use of the term "above" and similar indicating terms may correspond to both singular and plural. In addition, when a range is described in the present invention, it includes an invention in which individual values belonging to the range are applied (unless there is a description to the contrary), and each individual value constituting the range is described in the detailed description of the invention Same as

The steps constituting the method according to the present invention may be performed in any suitable order unless an order is explicitly stated or stated to the contrary. The present invention is not necessarily limited according to the order of description of the steps. The use of all examples or exemplary terms (eg, etc.) in the present invention is simply to explain the present invention in detail, and the scope of the present invention is limited due to the examples or exemplary terms unless limited by the claims. it is not going to be In addition, those skilled in the art can appreciate that various modifications, combinations and changes can be made according to design conditions and factors within the scope of the appended claims or equivalents thereof.

Therefore, the spirit of the present invention should not be limited to the above-described embodiments and should not be determined, and all scopes equivalent to or equivalently changed from the claims as well as the claims described below are within the scope of the spirit of the present invention. will be said to belong to

100: power management device
200: demand subject
300: IoT sensor
400: user terminal

Claims (8)

A power management method performed by a processor of a power management device,
Generating time-series power consumption data using sensor data collected from an IoT sensor provided in a power demand entity;
generating future predicted power usage data by applying the time-series power usage data to a Fourier series to which coefficients are fitted;
determining a predicted power upper limit value from the predicted power usage data using a power upper limit prediction model that determines an upper predicted power limit value using the predicted power usage data as an input and generating a predicted power peak guideline; and
Generating an alarm that an electricity rate is exceeded based on a distance from the time-series power usage data to the predicted power peak guideline;
Generating the time-series power usage data,
Classifying the sensing data collected from the IoT sensor by applying a preset classification criterion; and
Generating time-series power usage data including power usage data corresponding to a sensing time from sensing data classified according to the classification criterion;
The classification criteria are,
A first classification criterion for classifying the sensing data into a late-night time zone with less power consumption than during the daytime, a second classification criterion for classifying the sensing data into a summer season with higher power consumption than other seasons, and a power usage of the sensing data A third classification criterion for classifying the sensing data as winter season less than summer season, and the sensing data according to time by minute, hour, day, week, month Including a fourth classification criterion for classifying by star and year,
Prior to generating the predicted power usage data, further comprising calculating a Fourier series as a periodic function from the predicted power usage data and fitting coefficients of the Fourier series,
Fitting the coefficients of the Fourier series,
Calculating average power usage data for a previous power usage data group during a first period consisting of T first time steps having a preset time; and
Further comprising the step of fitting the coefficients of the Fourier series by curve fitting the average power usage based on a non-linear least square method,
Generating the predicted power usage data,
First predicted average power usage data corresponding to each of the T first time steps during the next first period by applying each of the T first time steps during the next first period to a Fourier series to which the coefficient is fitted Including the step of generating,
Before determining the predicted power upper limit value, further comprising generating the power upper limit prediction model;
Generating the power upper limit prediction model,
Calculating second predicted average power usage data as an average of the first predicted average power usage data during a first period consisting of L second time steps having a time longer than the first time step;
An agent in which a plurality of power upper limit values are preset is based on the second predicted average power usage data and the power difference between the plurality of upper power limit values for a given state in which the second predicted average power usage data is input. To perform an action of determining any one of the plurality of power upper limits as the predicted power upper limit value, and to compensate for the power difference corresponding to the predicted power upper limit value determined by the agent Training the neural network model to receive a reward step; and
Generating and storing a power upper limit prediction model for determining a predicted power upper limit value from the second predicted average power usage data through training of the neural network model, and
The neural network model,
In an environment (evnvironment) that receives the given state and the action of the agent and returns the reward and the next state (evnvironment), the agent performs an action in the given state to receive the maximum reward and determines the upper limit of predicted power A reinforcement learning-based neural network model configured to update
Power management method using predicted power peak guidelines. delete delete delete According to claim 1,
The reinforcement learning-based neural network uses a reinforcement learning-based neural network that is a deep Q-network (DQN).
A power management method using predicted power peak guidelines. According to claim 1,
Generating the predicted power peak guidelines,
Generating an average value of L predicted power upper limit values output through the power upper limit value prediction model for each second time step during the first period as a predicted power peak guideline during the first period,
A power management method using predicted power peak guidelines. A computer-readable recording medium storing a computer program for executing the method of any one of claims 1, 5, and 6 using a computer. As a power management device,
processor; and
a memory operatively connected to the processor and storing at least one code executed by the processor;
When the memory is executed through the processor, the processor generates time-series power usage data using sensor data collected from an IoT sensor provided in a power demand entity;
Applying the time-series power usage data to a Fourier series to which coefficients are fitted to generate future predicted power usage data;
Determining a predicted power upper limit from the predicted power usage data using a power upper limit prediction model that determines the predicted power upper limit by taking the predicted power usage data as an input and generating a predicted power peak guideline;
storing code that causes an alarm to be exceeded based on the distance from the time-series power usage data to the predicted power peak guideline;
The memory causes the processor to:
When generating the time-series power consumption data, the sensing data collected from the IoT sensor is classified by applying a preset classification criterion,
Cause time-series power usage data including power usage data corresponding to a sensing time to be generated from the sensing data classified according to the classification criterion;
The classification criteria are,
A first classification criterion for classifying the sensing data into a late-night time zone with less power consumption than during the daytime, a second classification criterion for classifying the sensing data into a summer season with higher power consumption than other seasons, and a power usage of the sensing data A third classification criterion for classifying the sensing data as winter season less than summer season, and the sensing data according to time by minute, hour, day, week, month It includes a fourth classification criterion for classifying by star and year,
The memory causes the processor to:
before generating the predicted power usage data, further storing code that causes a Fourier series as a periodic function to be calculated from the predicted power usage data, and coefficients of the Fourier series to be fitted;
The memory causes the processor to:
When fitting the coefficients of the Fourier series, calculating average power usage data for a previous power usage data group during a first period consisting of T first time steps having a predetermined time period;
further storing code that causes the average power usage to curve fit based on a non-linear least square to fit the coefficients of the Fourier series;
The memory causes the processor to:
When the predicted power consumption data is generated, each of the T first time steps during the next first period is applied to a Fourier series to which the coefficient is fitted, and each of the T first time steps during the next first period Generating first predicted average power usage data corresponding to;
before determining the predicted power upper limit, further storing code that causes generating the power upper limit prediction model;
The memory causes the processor to:
When the power upper limit value prediction model is generated, the second prediction as an average of the first predicted average power usage data during a first period consisting of L second time steps having a time longer than the first time step Calculate average power usage data;
An agent in which a plurality of power upper limit values are preset is based on the second predicted average power usage data and the power difference between the plurality of upper power limit values for a given state in which the second predicted average power usage data is input. to perform an action of determining one of the plurality of power upper limits as the predicted power upper limit value, and train a neural network model to receive a reward for a power difference corresponding to the predicted power upper limit value determined by the agent, ,
Store code that causes generation and storage of a power upper limit prediction model for determining a predicted power upper limit value from the second predicted average power usage data through training of the neural network model;
The neural network model,
In an environment (evnvironment) that receives the given state and the action of the agent and returns the reward and the next state (evnvironment), the agent performs an action in the given state to receive the maximum reward and determines the upper limit of predicted power A reinforcement learning-based neural network model configured to update
Power management devices using predicted power peak guidelines.

Images