KR20240044324A - Anomaly Detection Method, Power Control Method Using the Anomaly Detection Method, and Computing Device for Performing the Methods - Google Patents

Abstract

An anomaly detection method and a power control method using the anomaly detection method are disclosed. The anomaly detection method includes collecting power usage data for different applications through smart metering; Splitting the collected power usage data into a training set, validation set, and test set to apply to an anomaly detection model for detecting anomalies in the power usage data; In response to inputting power usage data included in the training set into the anomaly detection model, training the anomaly detection model to detect outliers or missing values in the power usage data; and applying power usage data included in the verification set to the learned anomaly detection model to verify the performance of the learned anomaly detection model.

Description

Anomaly detection method, power control method using the anomaly detection method, and computing device for performing the methods {Anomaly Detection Method, Power Control Method Using the Anomaly Detection Method, and Computing Device for Performing the Methods}

The present invention relates to a method and device for detecting abnormalities in metering data collected through smart metering and controlling the charging/discharging of batteries and electric vehicles by considering the detected abnormalities.

Recently, smart metering is receiving attention again as it meets the diverse needs of the grid market for high-efficiency systems, digital loads, and low-carbon green growth due to technological advancements, and the paradigm is changing significantly. In smart metering, outliers may be included in metering data for various reasons, such as aging meters, defective communication lines, and the influence of the surrounding environment.

Even though buildings are designed to be highly energy efficient, there are often cases where energy saving goals are not achieved due to outliers in metering data on energy consumption. Various methods, such as deep learning techniques, statistical methods, and information theory, are being used to recognize these outliers.

Analysis and detection of abnormal phenomena in the power system are important because they can collect and supply highly reliable information about the metering data generation process, but existing systems lack artificial intelligence-based energy management techniques in this field.

The present invention seeks to provide a method and device for detecting outliers and missing values included in quantitative data using an autoencoder implemented as a graph convolutional network (GCN) - bidirectional long short-term memory (LSTM) network.

In addition, the present invention maximizes the profits of renewable energy suppliers by removing outliers and missing values of metering data detected through an autoencoder, and adjusts peak loads and reduces electricity costs through a multi-purpose optimization model that minimizes peak loads. We would like to provide a method and device for doing so.

However, technical challenges are not limited to the above-mentioned technical challenges, and other technical challenges may exist.

An abnormality detection method according to an embodiment of the present invention includes collecting power usage data for different applications through smart metering; Splitting the collected power usage data into a training set, validation set, and test set to apply to an anomaly detection model for detecting anomalies in the power usage data; In response to inputting power usage data included in the training set into the anomaly detection model, training the anomaly detection model to detect outliers or missing values in the power usage data; and applying power usage data included in the verification set to the learned anomaly detection model to verify the performance of the learned anomaly detection model.

The method may further include performing preprocessing on the collected power usage data, and the dividing step may divide the preprocessed power usage data into a training set, a validation set, and a test set.

The performing step includes classifying the collected power usage data by generation and resampling it at regular intervals; removing noise by applying a discrete wavelet transform algorithm to the resampled power usage data; And it may include performing data normalization by standardizing the power usage data from which the noise has been removed.

The learning step includes identifying an output sequence output in response to inputting an input sequence composed of power usage data included in the training set into an anomaly detection model in the form of an Ocoincoder combining an encoder and a decoder; and training the anomaly detection model to minimize reconstruction error between the input sequence and the output sequence.

The Ocoincoder-type anomaly detection model can be implemented as a graph convolutional network (GCN) - bidirectional long short-term memory (LSTM) network.

The verifying step may verify the performance of the learned anomaly detection model through a mean square error or error matrix identified by applying power usage data included in the verification set to the learned anomaly detection model.

A power control method using an anomaly detection model according to an embodiment of the present invention includes the steps of identifying outliers and missing values by applying power usage data collected through smart metering to the anomaly detection model; Applying the power usage data from which the identified outliers and missing values have been removed to a multi-objective optimization model including a first objective function for maximizing the profits of a renewable energy supplier and a second objective function for minimizing the maximum load; And it may include controlling the charging/discharging of the battery or electric vehicle using the output of the multi-objective optimization model.

The first objective function is time (i) grid power consumption for load demand, (ii) switch function for grid power consumption, (iii) solar power generation power consumption for load demand, (iv) switch function for solar power consumption power, (v) battery charge/discharge power consumption, (vi) battery charge/discharge switch function, time It can be determined by (vii) the power consumption for charging/discharging of the electric vehicle and (viii) the switch function for charging/discharging of the electric vehicle.

The second objective function is time in (i) Power consumption for the schedulable load, (ii) switch function for the schedulable load, (iii) Power consumption for non-schedulable loads, (iv) Switch function and time for unschedulable loads (v) can be determined as a time-based demand response (DR) program.

The controlling step is performed when the difference between system power consumption and solar power generation power consumption exceeds the maximum limit of power consumption, and If the state of charge of the battery and the electric vehicle exceeds the minimum state of charge, the battery and the electric vehicle may be set to discharge mode.

The controlling step is such that the difference between system power consumption and solar power generation power consumption is less than the maximum limit of power consumption, and the time If the charge state of the battery and the electric vehicle is less than the maximum charge state, the battery and the electric vehicle can be set to charging mode.

A computing device according to an embodiment of the present invention includes one or more processors; and a memory for loading or storing a program executed by the processor, wherein the program collects power usage data for different applications through smart metering and detects anomalies in the power usage data. An operation of dividing the collected power usage data into a training set, validation set, and test set to apply to an anomaly detection model for the power usage data, in response to inputting the power usage data included in the training set into the anomaly detection model, An operation of training the anomaly detection model to detect outliers or missing values in usage data and an operation of verifying the performance of the learned anomaly detection model by applying power usage data included in the verification set to the learned anomaly detection model. It may include instructions to perform .

The processor may perform preprocessing on the collected power usage data and divide the preprocessed power usage data into a training set, validation set, and test set.

The processor classifies the collected power usage data by generation, resamples it at regular intervals, removes noise by applying a discrete wavelet transform algorithm to the resampled power usage data, and generates the power usage data from which the noise has been removed. Data normalization can be performed by standardizing the data.

The processor identifies an output sequence output in response to inputting an input sequence consisting of power usage data included in the training set into an anomaly detection model in the form of an occoin coder combining an encoder and a decoder, and identifies an output sequence between the input sequence and the output sequence. The anomaly detection model can be trained so that the reconstruction error of is minimized.

The Ocoincoder-type anomaly detection model can be implemented as a graph convolutional network (GCN) - bidirectional long short-term memory (LSTM) network.

The processor may verify the performance of the learned anomaly detection model through a mean square error or error matrix identified by applying the power usage data included in the verification set to the learned anomaly detection model.

According to an embodiment of the present invention, outliers and missing values included in metering data can be detected using an autoencoder implemented as a GCN-bidirectional LSTM network.

In addition, according to an embodiment of the present invention, the profits of renewable energy suppliers are maximized by removing outliers and missing values of metering data detected through an autoencoder, and the peak load is reduced through a multi-objective optimization model that minimizes the peak load. You can control it and save on electricity costs.

Figure 1 is a diagram showing the configuration of a smart metering system according to an embodiment of the present invention.
Figure 2 is a diagram showing the configuration of a smart metering server system according to an embodiment of the present invention.
Figure 3 is a diagram showing the configuration of a computing device included in a central access server according to an embodiment of the present invention.
Figure 4 is a flowchart showing a method for learning an anomaly detection model according to an embodiment of the present invention.
Figure 5 is a diagram showing a deep learning-based anomaly detection model using an autoencoder according to an embodiment of the present invention.
Figure 6 is a flowchart showing a power control method according to an embodiment of the present invention.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be changed and implemented in various forms. Accordingly, the actual implementation form is not limited to the specific disclosed embodiments, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical idea described in the embodiments.

Terms such as first or second may be used to describe various components, but these terms should be interpreted only for the purpose of distinguishing one component from another component. For example, a first component may be named a second component, and similarly, the second component may also be named a first component.

When a component is referred to as being “connected” to another component, it should be understood that it may be directly connected or connected to the other component, but that other components may exist in between.

Singular expressions include plural expressions unless the context clearly dictates otherwise. As used herein, “A or B”, “at least one of A and B”, “at least one of A or B”, “A, B or C”, “at least one of A, B and C”, and “A Each of phrases such as “at least one of , B, or C” may include any one of the items listed together in the corresponding phrase, or any possible combination thereof. In this specification, terms such as “comprise” or “have” are intended to designate the presence of the described features, numbers, steps, operations, components, parts, or combinations thereof, and are intended to indicate the presence of one or more other features or numbers, It should be understood that this does not exclude in advance the possibility of the presence or addition of steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art. Terms as defined in commonly used dictionaries should be interpreted as having meanings consistent with the meanings they have in the context of the related technology, and unless clearly defined in this specification, should not be interpreted in an idealized or overly formal sense. No.

Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the accompanying drawings, identical components will be assigned the same reference numerals regardless of the reference numerals, and overlapping descriptions thereof will be omitted.

Figure 1 is a diagram showing the configuration of a smart metering system according to an embodiment of the present invention.

The smart metering system 100 can measure power usage data in real time and then provide the measured power usage data to a plurality of market participants through two-way data communication. The smart metering system 100 can provide a variety of sophisticated services such as demand data analysis, service quality measurement, power outage management, distribution network analysis/planning, and customer billing through two-way data communication.

This smart metering system 100 may provide multiple interfaces for data transmission to facilitate communication between system components and multiple market participants.

Referring to FIG. 1, the smart metering system 100 may include four different interfaces: P1 interface 102, P2 interface 104, P3 interface 106, and P4 interface 108. More specifically, the P1 interface 102 can be used to connect the meter to an external device (corresponding to another service module 110 in FIG. 1). At this time, the other service module 110 may refer to various services provided by the smart metering system, and may provide various services to customers based on data collected by the smart metering system 100.

For example, services that other service modules 110 can provide to customers may include energy usage analysis, energy saving suggestions, and energy usage prediction. However, this type of service is only an example and is not limited to the above example.

Other service modules 110 may be connected to the independent service provider 122 and may provide more services to customers by utilizing services provided by the independent service provider 122. At this time, the energy gateway 112 may collect metering data from metering equipment connected to another service module 110 through the read-only P1 102 interface.

The P2 interface 104 can be used to connect different types of meters 114, 116, , 118 to the energy gateway 112 via wired or wireless. That is, the energy gateway 112 can collect metering data measured in the electricity meter 114, electric vehicle meter 116, and battery meter 118 through the P2 interface 104.

The P3 interface 106 may be used to connect the energy gateway 112 to a distribution system operator (DSO), i.e., a central access server 120. That is, the energy gateway 112 can transmit smart metering information including metering data collected through meters, status of meters, power quality, and power outage measurements to the central access server 120 through the P3 interface 106. there is. As an example, the P3 interface 106 for communication between the energy gateway 112 and the central access server 120 may be set to a communication protocol based on the international standard IEC 62056.

P4 interface 108 may be used to connect central connection server 120 to multiple types of providers. For example, the central access server 120 connects independent service providers such as Internet Service Providers (ISPs) 122, energy suppliers 124, and transmission line operators 126 through the P4 interface 108. can be connected to

At this time, the transmission line network operator 126 obtains measured values from the P3 interface 106, and the web service for accessing the central access server 120 can be provided through the P4 interface 108, so an independent service provider ( 122) and the energy supplier 124 can obtain metering data directly from the customer with the consent of the customer, regardless of the central access server 120. In other words, the smart metering system 100 can provide various services to customers by connecting and cooperating with various companies.

For example, the transmission line network operator 126 can establish a power supply and demand plan using data measured at the P3 interface 106, and through this, power supply stability can be improved by efficiently establishing a power supply and demand plan.

Figure 2 is a diagram showing the configuration of a smart metering server system according to an embodiment of the present invention.

Referring to FIG. 2, the data concentrator 200 generally receives metering data related to energy consumption and other customer information from the energy gateway 112 of the electric vehicle operator 201 and the battery operator 202 through a wireless or power line communication network. Relevant data can be collected and the collected data can be transmitted to the smart metering server system 210 in real time. At this time, the smart metering server system 210 of FIG. 2 may correspond to the central access server of FIG. 1.

The smart metering server system 210 can store data transmitted from the data concentrator 200 in the big data server 211, and provides metering data in conjunction with the metering data confirmation server 212 and the application service server 213. It can be managed.

Various application services installed in the energy gateway based on metering data can be provided to the smart metering server system 210 in real time through the advanced metering service server. Additionally, management services such as Home Energy Management System (HEMS) can be applied to the power system through the application service server 213 according to a demand response (DR) program.

Customers can access metering data in real time through the user portal 214 of the smart metering server system 210. The cloud-based smart metering system provided by the present invention can share metering data through the server 220, mobile 221, and web 222 even remotely and in other areas by linking with the smart metering server system 210.

Figure 3 is a diagram showing the configuration of a computing device included in a central access server according to an embodiment of the present invention.

As shown in FIG. 3, the computing device 300 may include one or more processors 310 and a memory 320 that loads or stores a program 330 executed by the processor 310. . The components included in the computing device 300 of FIG. 3 are only examples, and those skilled in the art will recognize that other general-purpose components other than those shown in FIG. 3 may be included. Able to know.

The processor 310 controls the overall operation of each component of the computing device 300. The processor 310 may be a Central Processing Unit (CPU), Micro Processor Unit (MPU), Micro Controller Unit (MCU), Graphic Processing Unit (GPU), Neural Processing Unit (NPU), Digital Signal Processor (DSP), or the present invention. It may be configured to include at least one of any type of processor well known in the art. Additionally, the processor 310 may perform operations on at least one application or program to execute methods/operations according to various embodiments of the present invention. Computing device 300 may include one or more processors.

Memory 320 stores one or a combination of two or more of various data, instructions, and information used by components included in computing device 300 (e.g., processor 310). Memory 320 may include volatile memory and/or non-volatile memory.

The program 330 may include one or more actions implementing methods/operations according to various embodiments of the present invention, and may be stored in the memory 320 in software form. Here, the operations correspond to instructions implemented in the program 330. For example, the program 330 collects power usage data for different applications through smart metering, and sets the collected power usage data to a training set to apply it to an anomaly detection model to detect anomalies in the power usage data. , the operation of splitting into a validation set and a test set, the operation of training an anomaly detection model to detect outliers or missing values in the power usage data in response to inputting the power usage data included in the training set into the anomaly detection model, and the validation set. It may include instructions for performing an operation to verify the performance of the learned anomaly detection model by applying the included power usage data to the learned anomaly detection model.

When the program 330 is loaded into the memory 320, the processor 310 can perform methods/operations according to various embodiments of the present invention by executing a plurality of operations to implement the program 330.

The execution screen of the program 330 may be displayed through the display 340. In the case of FIG. 3, the display 340 is represented as a separate device connected to the computing device 300, but in the case of the computing device 300 such as a terminal that the user can carry, such as a smartphone or tablet, the display 340 may be a component of the computing device 300. The screen displayed on the display 340 may be before information is input into the program or may be the result of executing the program.

Figure 4 is a flowchart showing a method for learning an anomaly detection model according to an embodiment of the present invention.

The learning method of the anomaly detection model shown in FIG. 4 can be performed by the processor of the computing device shown in FIG. 3. In step 410, the processor may collect power usage data for different applications through smart metering. For example, through smart metering, processors can collect power usage data for various applications within a household, such as appliances, batteries, or electric vehicles.

In step 420, the processor may perform preprocessing on the collected power usage data. More specifically, the processor may perform (i) data cleaning and sampling, (ii) data noise removal, and (iii) data normalization on the collected power usage data. First, the processor may (i) classify metering data by generation in response to data organization and sampling, provide tag values, and perform resampling at regular intervals (for example, 15-minute intervals). Afterwards, the processor can (ii) apply a discrete wavelet transform algorithm to the resampled power usage data in response to data noise removal to remove noise, which is the cause of performance degradation of the anomaly detection model. Finally, the processor can obtain the best training results for the anomaly detection model by performing normalization on the power usage data from which noise has been removed in response to (iii) data normalization.

In step 430, the processor may divide the preprocessed power usage data into a training set, a validation set, and a test set. At this time, the training set can be used to learn an anomaly detection model, the validation set can be used to verify the performance of the learned anomaly detection model, and the test set can be used to evaluate the verified anomaly detection model.

In step 440, the processor may train the anomaly detection model to detect outliers or missing values in the power usage data in response to inputting the power usage data included in the training set into the anomaly detection model. In smart metering, outliers and missing values can occur randomly in the real world and are usually caused by meters not delivering measurements or faulty meters.

First, a meter not delivering a reading can occur for a variety of reasons. For example, if the meter's battery is discharged and the power supply is interrupted, the meter's communication module is broken, or the meter's communication network is unstable, the measured value may not be transmitted.

Additionally, a faulty meter refers to a meter that measures the measured value incorrectly. For example, if the measurement sensor of the meter is broken, if the meter is not properly calibrated, or if the meter is installed in an inappropriate location, an incorrect measured value may be transmitted.

In addition, outliers and missing values may occur due to failure of the measuring device, poor performance of the measuring sensor, poor installation and protection, intentional damage, power outage, flickering, and phase loss.

These outliers and missing values can cause gaps or discontinuities in power usage data, and if outliers and missing values are not detected and processed, the accuracy of the power management process can be reduced, ultimately leading to energy waste.

Therefore, it is essential to identify and remove outliers and missing values before applying a power management method, and the deep learning-based anomaly detection model provided by the present invention can detect outliers and missing values using an autoencoder. At this time, the autoencoder can use unsupervised learning, which is a method of predicting results for new data by clustering quantitative data without a correct answer label.

As an example, Figure 5 is a diagram showing a deep learning-based anomaly detection model using an autoencoder according to an embodiment of the present invention. Referring to Figure 5, the autoencoder is a GCN - bidirectional LSTM network consisting of a Graph Convolutional Network (GCN) layer 520, a bidirectional Long Short-Term Memory (LSTM) encoder layer 530, and a bidirectional LSTM decoder layer 540. It can be implemented as:

The anomaly detection model regenerates the input sequence through an autoencoder implemented as a GCN-bidirectional LSTM network in which the input sequence consisting of power usage data included in the training set is input through the input layer 510, or predicts the target sequence and outputs it. It can be output through layer 550.

At this time, the processor may repeatedly train the anomaly detection model so that the reconstruction error between the input sequence and output sequence of the anomaly detection model is minimized.

In step 450, the processor may verify the performance of the learned anomaly detection model by applying the power usage data included in the verification set to the learned anomaly detection model. At this time, the prediction performance and classification performance of the learned anomaly detection model can be evaluated by mean square error (MSE) and confusion matrix, respectively.

At this time, the prediction performance of the anomaly detection model is an indicator that evaluates the difference between the value predicted by the anomaly detection model and the actual value, and generally the mean squared error (MSE) can be used. MSE is the average of the squared difference between the predicted value and the actual value. The smaller the value, the better the prediction performance of the anomaly detection model.

Meanwhile, the classification performance of an anomaly detection model is an indicator that evaluates the degree of agreement between the results classified by the anomaly detection model and the actual results, and an error matrix can generally be used. The error matrix represents the results classified by the anomaly detection model and the actual results as a 2x2 matrix, and based on this, indicators such as accuracy, precision, and recall can be calculated.

The processor can perform power management efficiently and robustly according to the anomaly detection results identified through the verified anomaly detection model.

Figure 6 is a flowchart showing a power control method according to an embodiment of the present invention.

The power control method shown in FIG. 6 may be performed by a processor of the computing device shown in FIG. 3. In step 610, the processor may identify outliers and missing values by applying power usage data collected through smart metering to an anomaly detection model.

In step 620, the processor converts the power usage data from which the identified outliers and missing values have been removed into a multi-objective function including a first objective function to maximize the profit of the renewable energy provider and a second objective function to minimize the maximum load. It can be applied to optimization models.

The first objective function included in the multi-objective optimization model aims to reduce grid dependency, increase renewable energy utilization and maximize the profits of renewable energy suppliers, and the second objective function allows customers to use three types of DR programs. The goal is to minimize the maximum load while taking into account.

For example, the first objective function can be expressed as Equation 1 below, and it can be proposed to reduce grid power consumption and increase solar power generation power utilization with the goal of maximizing the profits of renewable energy suppliers.

<Equation 1>

here is the time Grid power consumption for load demand of is the time switch function for grid power consumption, is the time PV power consumption for load demand, is the time A switch function for the power consumption of solar power generation, is the time Indicates the charging/discharging power consumption of the battery.

also, is the time The charge/discharge switch function of the battery, is the time Power consumption for charging/discharging of electric vehicles, is the time shows the switch function for charging/discharging an electric vehicle. in other words, , , and A switch function such as is controlled by the first objective function on the supply side, and Can be determined using two methods: the existing method according to Equation 3 and the proposed method according to Equation 4.

Renewable energy suppliers act as intermediaries between wholesale power markets and consumers in the energy market. The profit of a renewable energy supplier is calculated as income from energy sales and the cost of purchasing energy from the wholesale market or providing services. The renewable energy supplier's cost function is estimated at the optimal retail price and used to encourage customers to adopt certain electricity consumption habits.

The second objective function can be expressed as Equation 2 below, and the DR program aims to minimize peak load while considering electricity charges. Because energy gateways have low load capacity, maintenance costs for the entire smart metering system can become high within a few hours at peak load. Therefore, reducing the peak load by a few hours can reduce the maintenance cost of the entire smart metering system and ensure a robust design of the smart metering system. Smart metering systems can schedule movable loads according to the proposed power management, but cannot schedule non-schedulable and non-interruptible loads and must power these loads immediately when needed.

<Equation 2>

here, is the time at Power consumption for schedulable loads, is the time at Switch function for the schedulable load, is the time at Power consumption for non-schedulable loads, Is Switch function for unschedulable loads, is the time Indicates a time-based DR program.

In step 630, the processor may control charging/discharging of the battery or electric vehicle using the output of the multi-objective optimization model. In power management, the charging/discharging mode of the battery and electric vehicle is determined by the charging state of the battery and electric vehicle ( ) value, as well as a comparison of grid power consumption and solar power generation power consumption.

The existing battery and electric vehicle charging/discharging methods can be expressed as Equation 3 below. More specifically time The charging/discharging modes of the battery and electric vehicle each use a switch function. and It is expressed as, and the battery owner and electric vehicle owner determine the desired charging time interval and the minimum minimum time when charging is completed. The value can be determined.

Smart metering operators can determine when to charge/discharge batteries and electric vehicles, increasing the overall benefits and energy efficiency of smart metering, and the switch function and The value of is given as follows: grid power consumption of batteries and electric vehicles, solar power generation power consumption, and It can be determined by charging/discharging conditions related to .

<Equation 3>

here, is the maximum limit of power consumption, is the time Battery and charging status of electric vehicles, The minimum state of charge of the battery and electric vehicle, represents the maximum limit state of charge.

Meanwhile, for batteries and electric vehicles, the operating mode can be displayed as -1 (discharge), +1 (charge), and 0 (idle). In the existing charging/discharging method, the difference between grid power consumption and solar power generation power consumption go Greater than and charge state value go If larger, the battery and electric vehicle may be assigned to 'discharge mode'. Additionally, the difference between system power consumption and solar power generation power is is less than and the state of charge value is If it is less than that, the battery and electric vehicle can be assigned to 'charging mode'. In other cases, batteries and electric vehicles can be assigned to ‘idle mode’.

The charging/discharging method of batteries and electric vehicles proposed in the present invention can be expressed as Equation 4 below. More specifically, the proposed charging/discharging method for batteries and electric vehicles uses anomaly detection for stable charging/discharging and can perform power management efficiently and robustly by finding abnormal states of predicted future values as follows.

<Equation 4>

here, is the time grid power consumption, is the time GCN - the current value of the grid power consumption of the autoencoder implemented as a bidirectional LSTM network, is the time Power consumption of solar power in is the time Shows the abnormality detection switch function by the autoencoder in .

In the proposed charging/discharging method for batteries and electric vehicles, the operation modes consisting of -1 (charge), +1 (discharge), and 0 (idle) can be determined through comparison of system power. GCN - If no outliers or missing values are detected in the process of measuring system power consumption through an autoencoder implemented in a bidirectional LSTM network, in Equation 4 The existing charging/discharging method can be applied using .

However, if outliers or missing values are detected in the process of measuring system power consumption through an autoencoder implemented with a GCN-bidirectional LSTM network, in Equation 4 second You can determine the charging/discharging of batteries and electric vehicles by substituting . In this proposed charging/discharging method for batteries and electric vehicles, the charging/discharging of batteries and electric vehicles can be controlled more safely and robustly by adding two functions of outlier detection and missing value replacement.

The embodiments described above may be implemented with 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 include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, and a field programmable gate (FPGA). It may be implemented using a general-purpose computer or a special-purpose computer, such as an array, programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and software applications running on the operating system. Additionally, a processing device may access, store, manipulate, process, and generate data in response to the execution of software. For ease of understanding, a single processing device may be described as being used; however, those skilled in the art will understand that a processing device includes multiple processing elements and/or multiple types of processing elements. It can be seen that it may include. For example, a processing device may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are possible.

Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing unit to operate as desired, or may be processed independently or collectively. You can command the device. Software and/or data may be used on any type of machine, component, physical device, virtual equipment, computer storage medium or device to be interpreted by or to provide instructions or data to a processing device. It can be saved as . Software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on a computer-readable recording medium.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. A computer-readable medium may store program instructions, data files, data structures, etc., singly or in combination, and the program instructions recorded on the medium may be specially designed and constructed for the embodiment or may be known and available to those skilled in the art of computer software. there is. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -Includes optical media (magneto-optical media) and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

The hardware devices described above may be configured to operate as one or multiple software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited drawings, those skilled in the art can apply various technical modifications and variations based on this. For example, the described techniques are performed in a different order than the described method, and/or components of the described system, structure, device, circuit, etc. are combined or combined in a different form than the described method, or other components are used. Alternatively, appropriate results may be achieved even if substituted or substituted by an equivalent.

Therefore, other implementations, other embodiments, and equivalents of the claims also fall within the scope of the claims described below.

100: Smart metering system
102: P1 interface
104: P2 interface
106: P3 interface
108: P4 interface
110: Other service modules
112: Energy Gateway
114: Electric meter
116: Electric vehicle meter
118: Battery meter
120: Central access server
122: Independent service provider
124: Energy supplier
126: Transmission line network operator

Claims (1)

collecting power usage data for different applications through smart metering;
Splitting the collected power usage data into a training set, validation set, and test set to apply to an anomaly detection model for detecting anomalies in the power usage data;
In response to inputting power usage data included in the training set into the anomaly detection model, training the anomaly detection model to detect outliers or missing values in the power usage data; and
Verifying the performance of the learned anomaly detection model by applying power usage data included in the verification set to the learned anomaly detection model.
An anomaly detection method comprising:

Images