KR102599139B1 - The method and apparatus for generating anomaly detection reference signal in electric vehicle charging station distribution system using artificial neural network based predictor - Google Patents

Abstract

This specification provides a method for generating a reference signal for diagnosing problems in an electric vehicle charging station distribution system using an artificial neural network predictor, comprising the steps of acquiring sensing data from at least one sensor and determining input values of an artificial neural network based on the sensing data. , the output value of the artificial neural network is set to the contact temperature value as a single value, the input values are set to values related to the contact temperature value, and the input values are applied to the artificial neural network to derive the predicted contact temperature value as the output value. It may include the step of comparing the predicted contact temperature value and the measured contact temperature value to determine whether the power distribution system is abnormal.

Description

Method and device for generating reference signals for diagnosing abnormalities in electric vehicle charging station distribution system using artificial neural network predictor

This specification relates to a method and device for generating an abnormality diagnosis reference signal in an electric vehicle charging station distribution system. Specifically, this specification relates to a method and device for generating an abnormality diagnosis reference signal to determine whether there is an abnormality in a power distribution system for electric vehicle charging station power supply that receives power from the grid.

A power distribution system may refer to a system that distributes needed power after receiving produced power. Electric vehicle charging stations include a power distribution system that receives generated power and distributes the supplied power to electric vehicles. Demand for electric vehicles has recently been increasing, and considering environmental issues, it is expected that electric vehicles will replace internal combustion engine vehicles in the future. Along with the spread of electric vehicles, there is a need to increase the number of electric vehicle charging stations. Here, the electric vehicle charging station includes the above-described power supply and distribution system, and if an abnormality occurs in the power supply and distribution system, an accident such as a fire may occur or a problem of not smoothly providing charging services may occur. Accordingly, a method for determining abnormalities within the electric vehicle charging station distribution system may be necessary, which will be described later.

Korean registered patent 10-2066948 B1

The purpose of this specification is to provide a method for generating an abnormality diagnosis reference signal to determine whether an electric vehicle charging station distribution system is abnormal.

The purpose of this specification is to provide a method of setting diagnostic variables related to deterioration diagnosis to determine whether there is an abnormality in the electric vehicle charging station distribution system.

The purpose of this specification is to provide a method of setting the contact temperature as a diagnostic variable related to deterioration diagnosis.

The purpose of this specification is to provide a method for reducing the cost of sensors for determining abnormalities in electric vehicle charging station distribution systems and minimizing sensors through feature engineering.

The purpose of this specification is to improve the economic feasibility of the device by generating a reference signal for determining abnormalities in the power distribution system based on machine learning (or machine learning) using an artificial neural network (ANN).

According to an embodiment of the present specification, in a method of generating a reference signal for diagnosing an abnormality in an electric vehicle charging station distribution system using an artificial neural network predictor, the method includes acquiring sensing data from at least one sensor, inputting an artificial neural network based on the sensing data. As a step of determining the values, the output value of the artificial neural network is set to the contact temperature value as one value, the input values are set to values associated with the contact temperature value, and the contact temperature value predicted by applying the input values to the artificial neural network. It may include deriving an output value and comparing the predicted contact temperature value with the measured contact temperature value to determine whether there is an abnormality in the power distribution system.

In addition, according to an embodiment of the present specification, in an apparatus for generating a reference signal for diagnosing an abnormality in an electric vehicle charging station power supply and distribution system using an artificial neural network predictor, a sensor unit for acquiring at least one sensing data, a database unit for storing data, and power distribution system. It includes a standard setting unit that sets standards for determining whether there is a system abnormality, an artificial neural network prediction unit and sensor unit that sets diagnostic variables based on an artificial neural network, a database unit, a standard setting unit, and a control unit that controls the artificial neural network prediction unit, The control unit acquires sensing data through the sensor unit and determines input values of the artificial neural network based on the sensing data. The output value of the artificial neural network is set to the contact temperature value as one value, and the input values are related to the contact temperature value. Values are set, and the input values are applied to the artificial neural network through the artificial neural network prediction unit to derive the predicted contact temperature value as an output value, and compare the predicted contact temperature value with the measured contact temperature value to determine whether there is an abnormality in the power distribution system. can be determined.

Additionally, the following matters can be commonly applied.

According to an embodiment of the present specification, the input values include sensing data, additional data derived from the sensing data through feature engineering based on the contact temperature value, external data input from the outside, and feedback derived from the output value of the artificial neural network. It may contain at least one of the data.

In addition, according to an embodiment of the present specification, the input values are categorized into electrical data related to electrical operation, environmental data related to the surrounding environment of the electric vehicle charging station distribution system, external data input from the outside, and feedback data to the artificial neural network. can be provided.

In addition, according to an embodiment of the present specification, the electrical data includes at least one of a busbar current value, a cumulative operating time value related to the rated current, and a ground leakage current value, and the environmental data includes a contact part fine dust value and an external temperature value. , includes at least one of an external humidity value and a vibration value, the external data includes at least one of a breaker on/off operation accumulated count value and a critical failure accumulated count value, and the feedback data includes a diagnosis result value and a diagnosis result accuracy. It may contain at least one of the values.

In addition, according to an embodiment of the present specification, at least one sensor includes general-purpose sensors that acquire sensing data related to input values of the artificial neural network and a temperature sensor that measures the temperature value of the contact portion, where the temperature sensor measures the temperature value of the contact portion. By sensing the temperature of the contact part spaced apart by a set distance, the measured contact temperature value can be derived.

In addition, according to an embodiment of the present specification, when determining whether the power distribution system is abnormal by comparing the predicted contact temperature value and the measured contact temperature value, the difference between the predicted contact temperature value and the measured contact temperature value After deriving and removing the white noise included in the difference value, it is possible to determine whether there is an abnormality.

Additionally, according to an embodiment of the present specification, setting an initial contact temperature value based on experimental data, obtaining a diagnosis result value based on the initial contact temperature value, and constructing a database based on the diagnosis result value. and deriving a contact temperature value based on an artificial neural network when the database is constructed, wherein when the contact temperature value is derived through an artificial neural network based on the constructed database, sensing data is acquired from at least one sensor. And, input values of the artificial neural network may be determined based on the sensing data.

This specification has the effect of generating an abnormality diagnosis reference signal to determine whether there is an abnormality in the electric vehicle charging station distribution system.

This specification has the effect of providing highly intuitive diagnostic variables related to deterioration diagnosis to determine whether there is an abnormality in the electric vehicle charging station distribution system.

This specification has the effect of providing a method of performing feature engineering on data sensed from a low-cost sensor or general-purpose sensor.

This specification has the effect of improving the economic feasibility of the device by generating a reference signal to determine whether there is an abnormality in the power distribution system based on machine learning using an artificial neural network (ANN).

The effects that can be obtained in this specification are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below. will be.

1 is a diagram showing the configuration of a device according to an embodiment of the present specification.
Figure 2 is a diagram showing a network environment according to an embodiment of the present specification.
Figure 3 is a diagram showing a reference signal generating device for diagnosing abnormalities in the electric vehicle charging station distribution system according to an embodiment of the present specification.
Figure 4 is a diagram showing a reference signal generating device for diagnosing abnormalities in the electric vehicle charging station distribution system according to an embodiment of the present specification.
Figure 5 is a diagram showing a reference signal generating device for diagnosing abnormalities in the electric vehicle charging station distribution system according to an embodiment of the present specification.
Figure 6 is a diagram showing a method for determining abnormality through an artificial neural network, according to an embodiment of the present specification.
Figure 7 is a diagram showing a method of performing categorization on artificial neural network input according to an embodiment of the present specification.
FIG. 8 is a diagram illustrating the operation of a reference signal generating device for diagnosing abnormalities in an electric vehicle charging station distribution system according to an embodiment of the present specification.
FIG. 9 is a diagram illustrating the operation of a reference signal generating device for diagnosing abnormalities in an electric vehicle charging station distribution system according to an embodiment of the present specification.
Figure 10 is a diagram showing a method of performing diagnosis and comparison according to an embodiment of the present specification.
Figure 11 is a diagram showing the effect of applying a Kalman filter according to an embodiment of the present specification.
Figure 12 is a flowchart showing a method of generating a reference signal for diagnosing an abnormality in an electric vehicle charging station distribution system using an artificial neural network predictor according to an embodiment of the present specification.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings. The detailed description set forth below in conjunction with the accompanying drawings is intended to illustrate exemplary embodiments of the invention and is not intended to represent the only embodiments in which the invention may be practiced. The following detailed description includes specific details to provide a thorough understanding of the invention. However, one skilled in the art will appreciate that the present invention may be practiced without these specific details.

The following embodiments combine the components and features of the present invention in a predetermined form. Each component or feature may be considered optional unless explicitly stated otherwise. Each component or feature may be implemented in a form that is not combined with other components or features. Additionally, some components and/or features may be combined to form an embodiment of the present invention. The order of operations described in embodiments of the present invention may be changed. Some features or features of one embodiment may be included in other embodiments or may be replaced with corresponding features or features of other embodiments.

Specific terms used in the following description are provided to aid understanding of the present invention, and the use of such specific terms may be changed to other forms without departing from the technical spirit of the present invention.

In some cases, in order to avoid ambiguity of the concept of the present invention, well-known structures and devices are omitted or are shown in block diagram form focusing on the core functions of each structure and device. In addition, the same components are described using the same reference numerals throughout this specification.

Additionally, in this specification, terms such as first and/or second may be used to describe various components, but the components should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another component, for example, without departing from the scope of rights according to the concepts of the present specification, a first component may be named a second component, and similarly , the second component may also be named the first component.

In addition, throughout the specification, when a part is said to "include" a certain component, this means that it may further include other components rather than excluding other components, unless specifically stated to the contrary. In addition, terms such as "?? unit" and "?? unit" described in the specification refer to a unit that processes at least one function or operation, which may be implemented through a combination of hardware and/or software.

1 is a diagram showing the configuration of a device according to an embodiment of the present specification. Referring to FIG. 1, the device 110 may include a memory 111, a control unit 112, and a transmitting/receiving unit 113. The device 110 may further include other configurations other than those described above, and is not limited to a specific embodiment. As an example, the device 110 of FIG. 1 may be a device that transmits and receives data through communication with another device 120, and may be a user device. As an example, the device 110 may refer to a smartphone, smart pad, laptop, PC, and other communication-enabled devices, and may not be limited to a specific device.

Here, the memory 111 of the device 110 is a non-transitory computer-readable recording medium, such as random access memory (RAM), read only memory (ROM), disk drive, solid state drive (SSD), and flash memory. It may include non-permanent mass storage devices such as memory, and is not limited to a specific form. Additionally, the memory 111 may store instructions or program codes related to driving or operation of the device 110. Additionally, the memory 111 may store the operating system or other software of the device 110. Additionally, as an example, the device 110 may use software loaded from a recording medium that can be read by a separate computer. Here, computer-readable recording media may include floppy drives, disks, tapes, DVD/CD-ROM drives, memory cards, and other recording media, and are not limited to specific embodiments.

The control unit 112 of the device 110 may be configured to execute and process commands related to driving or operation of the device 110. Additionally, the control unit 112 may be a logical entity that controls the transceiver 113 and other components, and is not limited to a specific embodiment. The control unit 112 of the device 110 may be configured to process commands of a computer program by performing arithmetic, logic, and input/output operations. As an example, a command may be provided to the control unit 112 based on a signal stored in the memory 111 or obtained through the transceiver 113, and the control unit 112 may perform an operation based on it.

The transceiver 113 may provide a function for communication with at least one of another device 120 and a server through a network. As an example, another device 120 may also include a memory 121, a control unit 122, and a transceiver 123. Additionally, the other device 120 may be the above-described smart device, PC, or other communication capable device, and is not limited to a specific embodiment.

The transmitting and receiving unit 113 may be a means for an input unit or an output unit. For example, the input unit (not shown) may be a component that provides a keyboard, mouse, touchpad, camera, and other input signals, and the output unit (not shown) may be a component that provides a display, speaker, and other output signals. You can. However, it may not be limited to this. Devices such as, and external output devices may include devices such as displays, speakers, haptic feedback devices, etc.

Figure 2 is a diagram showing a network environment according to an embodiment of the present specification.

Referring to FIG. 2 , at least one device or server 210, 220, 230, or 240 may be connected through a network 200. As an example, at least one device or server 210, 220, 230, or 240 may be connected through a wired or wireless network 200, and may transmit and receive data through this. The device or server 210, 220, 230, or 240 may be the device 110 of FIG. 1, but is not limited to a specific form. Additionally, the devices or servers 210, 220, 230, and 240 may be fixed devices implemented as computer systems. As another example, the devices or servers 210, 220, 230, and 240 may be mobile terminals. As a specific example, the devices or servers 210, 220, 230, and 240 include a smart phone, mobile phone, navigation, computer, laptop, tablet PC, wearable device, IoT (internet of things) device, It may be a virtual reality (VR)/augmented reality (AR) device or other device, and is not limited to a specific embodiment.

Additionally, a server may be a device that is connected to one or more devices through a network and has functions for providing code, content, and services. As a specific example, the server may provide a service or a response according to a request to at least one device connected through the network 200. Here, the server may be linked based on software or applications installed on each of at least one device, and may provide services through this.

The above-mentioned devices or servers 210, 220, 230, and 240 may exchange data with each other through the network 200 or provide services by operating interconnected applications, software, and other programs. As an example, an artificial neural network (ANN) based on artificial intelligence may be built and operated in at least one of the above-described devices or servers 210, 220, 230, and 240. As an example, an artificial neural network is built on a specific server (e.g. artificial intelligence server), and the specific server acquires data from each device (e.g. sensor) through the network and provides the obtained data as input to the artificial neural network. After deriving the output, the output information can be transmitted to at least one device (e.g. corresponding device). As another example, the artificial neural network may be a deep learning-based neural network with multiple layers. Here, all of the plurality of layers can be built on one server. As another example, multiple layers may be distributed and built across multiple devices. A plurality of devices may operate by exchanging their respective data and weights. As an example, ANN may be a statistical learning algorithm built by imitating the central nervous system of animals through machine learning. ANN can operate based on machine learning, which is the machine learning described above. As another example, ANN may operate based on deep learning, which is built with multiple neural networks. As an example, the following describes a method of using an artificial neural network to generate reference signals for diagnosing abnormalities in the electric vehicle charging station distribution system. The artificial neural network used below may be an artificial neural network driven based on FIGS. 1 and 2 described above, and may not be limited to a specific type of artificial neural network.

As an example, the following describes a method of generating a reference signal for diagnosing abnormalities in an electric vehicle charging station distribution system through feature engineering and intuitive diagnostic parameter settings for data acquired through a general-purpose sensor. Artificial neural networks can be applied to abnormality diagnosis and determination devices in electric vehicle charging station distribution systems. Here, the abnormality diagnosis and determination device of the electric vehicle charging station distribution system can generate input values suitable for the artificial neural network by performing feature engineering through physical data and other data sensed from a low-cost sensor (or general-purpose sensor). there is. Additionally, the output value of the artificial neural network can be set to the contact temperature value as an intuitive diagnostic variable based on the input values. As an example, the electric vehicle charging station distribution system may include a plurality of contact parts based on the electrical opening and closing part. Here, the artificial neural network can provide the contact temperature value for each contact portion as output. That is, individual output values for each contact part can be derived by an artificial neural network, which will be described later. Through the above-described information, abnormality diagnosis and determination of the electric vehicle charging station distribution system can be performed only with low-cost general-purpose sensors, and economic efficiency can be improved.

More specifically, electric vehicle charging stations are built with a power distribution system, and if an error occurs in the power distribution system, smooth power supply may be difficult. In consideration of the above, a method for determining whether there is an abnormality in the power distribution system may be necessary for smooth power supply.

For example, a method of determining whether an abnormality exists in a power distribution system may be a method of detecting abnormality-related signals and diagnosing deterioration through partial discharge detection. Here, the determination of whether there is an abnormality in the power distribution system is determined by a TEV (Transient Earth voltage) sensor that detects an instantaneous transient ground voltage signal in the 1 to 30 [MHz] band, and a 40 [MHz] sound pressure generated by partial discharge (PD). Abnormality-related signals are detected using an AA (Airborne-Acoustic) acoustic emission sensor that detects the [kHz] component signal and an AE (Airborne-Emission) acoustic vibration sensor that detects vibration in the 150 [kHz] band caused by partial discharge. This can be done by diagnosing deterioration through partial discharge detection.

However, the TEV sensor, AA sensor, AE sensor, and other sensors described above have a relatively high frequency band compared to general-purpose voltage sensors or acoustic sensors and may be relatively expensive sensors. In addition, when determining whether an abnormality in the power distribution system is based on a high frequency band, the abnormality determining device may need to be equipped with a high-performance AD converter. Additionally, when configuring an analog circuit for signal processing based on the above, the system needs to perform an operation for noise processing. Additionally, the system can also consider PCB pattern design. Accordingly, system development costs increase, which creates a disadvantage in the process of lowering the price of the product.

Another way to determine whether there is an abnormality in the power distribution system is a VHF (Very High Frequency) sensor in the 30 to 300 [MHz] band, a UHF (Ultra High Frequency) sensor in the 300 to 3000 [MHz] band, and a heat sensor for temperature sensing of the electrical switch. It may be a partial discharge diagnosis method using an image temperature sensor. Since the above-described method also uses a sensor based on a high frequency band, the above-described problem may exist. Additionally, the above-described method requires image processing when using a thermal imaging sensor, and thus the signal processing method may become more complicated. Thermal imaging cameras have the advantage of enabling non-contact sensing that ensures insulation between the temperature sensing circuit and the electrical contact part. However, the temperature measurement method using a thermal imaging camera can be expensive and has the disadvantage of being disadvantageous in terms of economic efficiency. For example, the business feasibility of the main device may increase as a device for diagnosing the main device (e.g. power distribution system) is built into the main device, and the smaller the impact on the overall price increase of the device. Therefore, simplicity may be required in the sensors, related circuits, and signal processing methods used to determine whether there is an abnormality in the power distribution system.

Additionally, in general, partial discharge (PD) may be a phenomenon that mainly occurs at ultra-high voltages of 22.9 [kV] or higher. Here, the distribution system for electric vehicle charging stations can be implemented at a voltage level of 22.9 [kV] level (13 [kV] level overseas, such as in the United States). Therefore, the deterioration diagnosis method through partial discharge may not be suitable for the electric vehicle charging station distribution system, and a deterioration diagnosis method considering the electric vehicle charging station distribution system may be necessary. In the following, this is described.

Figure 3 is a diagram showing a reference signal generating device for diagnosing abnormalities in the electric vehicle charging station distribution system according to an embodiment of the present specification. Referring to FIG. 3, the reference signal generating device 300 for diagnosing abnormalities in the electric vehicle charging station distribution system includes at least one of a control unit 310, a database unit 320, a sensor unit 330, and a standard setting unit 340. However, it is not limited to this and may further include other configurations. Here, the control unit 310 may be an entity that controls the database unit 320, the sensor unit 330, and the standard setting unit 340, and performs monitoring based on information obtained from each entity. That is, the control unit 310 performs diagnosis (or monitoring) by comparing the information sensed through the sensor unit 330 with each standard set through the standard setting unit 340, derives the diagnosis result, and stores the information in the database 320. It can be saved in . Additionally, the control unit 310 may be able to operate in conjunction with other entities, but is not limited thereto.

Additionally, the database 320 may store data related to the reference signal generating device 300 for diagnosing problems in the electric vehicle charging station distribution system. As an example, the database 320 may store diagnostic criteria information and diagnosis result information. Additionally, the database 320 may store input and output values of an artificial neural network or other data, which will be described later. Additionally, the sensor unit 330 may be an entity composed of hardware and software for data sensing. Here, the sensor unit 330 may be a low-cost sensor or a general-purpose sensor, and through this, the economic feasibility of the reference signal generating device 300 for diagnosing abnormalities in the electric vehicle charging station distribution system can be improved. Additionally, the standard setting unit 340 may be an entity that sets a diagnostic standard system for diagnosis (or monitoring). As an example, the standard setting unit 340 may define diagnostic variables and perform comparison and diagnosis to determine quantitative abnormalities. Additionally, the standard setting unit 340 may set abnormality determination conditions to perform subsequent operations after abnormality determination.

Figure 4 is a diagram showing a reference signal generating device for diagnosing abnormalities in the electric vehicle charging station distribution system according to an embodiment of the present invention. Referring to FIG. 4 , the reference signal generating device 300 for diagnosing abnormalities in the electric vehicle charging station power supply and distribution system may operate based on the above-described FIG. 3 . Here, each operation of the electric vehicle charging station distribution system abnormality diagnostic reference signal generating device 300 monitors and databases include a diagnostic standard system, an operation level standard setting unit (or standard setting unit), and hardware and software (or sensors) for data sensing. It can operate in conjunction with (part). As an example, the electric vehicle charging station distribution system abnormality diagnostic reference signal generating device 300 sets a diagnostic standard system, sets an operation level standard based on this, and compares the set standard with the data measured through the sensor to determine whether there is an abnormality. It can be determined.

Here, diagnostic variables can be set to determine whether there is an abnormality. Additionally, a diagnostic standard level may be set to determine whether the corresponding diagnostic variable is abnormal. Afterwards, anomaly detection can be performed based on the difference value by comparing the actual physical quantity corresponding to the diagnostic variable and the measured actual quantity.

As an example, the reference level of a diagnostic variable can be derived through a deterministic model method obtained using a physical model of the system being diagnosed. Additionally, the reference level of the diagnostic variable can be derived through a method of determining an abnormal diagnosis reference signal through experimentation under an appropriate driving environment. In deterministic models, the process of acquiring and tuning variables to calculate the model may be complicated or an appropriate model may not exist. On the other hand, the experimental reference signal determination method has the disadvantage of reducing the accuracy of abnormality diagnosis when the actual environment and the experimental environment are different.

In the following, a method and device for generating a diagnostic reference signal based on a driving DB (Data Base) using machine learning or deep learning based on an artificial neural network will be described in consideration of the above-mentioned points. As an example, the electric vehicle charging station distribution system abnormality diagnosis reference signal generating device 300 secures a database using the abnormality diagnosis reference level determined by experiment in the initial stage, and can reset the diagnosis reference signal level when sufficient data is accumulated. and this will be described later.

Additionally, as an example, in the reference signal generating device 300 for diagnosing abnormalities in the electric vehicle charging station distribution system, the diagnostic variable needs to be set to the temperature of the contact part. In other words, the reference signal generating device 300 for diagnosing abnormalities in the electric vehicle charging station power supply and distribution system may set the diagnostic variable to the temperature value of the contact part, considering that deterioration is accelerated when the temperature of the contact part of the electric switch exceeds the design standard. Here, there may be a plurality of contact parts in the electric vehicle charging station distribution system, and the contact temperature value for each contact part can be set as a diagnostic variable. For example, actual standards such as IEC and ANSI regulate the temperature rise of each component that makes up products such as molded case circuit breakers, and "Seong-gyu Park, Jong-cheol Lee, Yun-je Kim, "The effect of contact resistance on the internal temperature rise of molded case circuit breakers," Journal of Energy Engineering 13(1), 2004.2, 12-19". Therefore, it may be appropriate to set the diagnostic variable to the temperature of the contact part. In addition, the temperature of the contact area is a highly versatile element and can be measured using a low-cost temperature sensor compared to the existing thermal imaging camera method, thereby improving economic efficiency. In particular, considering that the occurrence of partial discharge may be minimal in low-voltage switchboards below 22.9 [kV], the diagnostic variable can be set to the temperature of the contact part to diagnose abnormalities.

In the following, the diagnostic variable can be set as the contact temperature value, and this can be set as the output value of the artificial neural network. Additionally, the input values of the artificial neural network can be set to at least one value related to the temperature of the contact part. Here, at least one value related to the contact temperature value can be derived as an input value of the artificial neural network by performing a pre-processing process based on feature engineering on data sensed through a low-cost general-purpose sensor. In other words, the input values of the artificial neural network are created through feature engineering of the values obtained through the general-purpose sensor, and the input values are applied to the artificial neural network to derive the contact temperature value as the output value and compare it with the measured temperature value to determine whether there is an abnormality. can be judged.

Figure 5 is a diagram showing a reference signal generating device 300 for diagnosing problems in an electric vehicle charging station distribution system according to an embodiment of the present invention. Referring to FIG. 5 , the reference signal generating device 300 for diagnosing abnormalities in the electric vehicle charging station distribution system may further include an artificial neural network prediction unit 350 for the artificial neural network in the configuration of FIG. 3 . The control unit 310 may derive values sensed through the sensor unit 330 into values suitable as input values for the artificial neural network prediction unit 350 through feature engineering and transmit them to the artificial neural network prediction unit 350. The artificial neural network prediction unit 350 may derive the contact temperature value as an output based on the input values and transmit it to the control unit 310. The control unit 310 may compare the temperature value of the contact part with the actual measured temperature value to determine whether there is an abnormality and store the corresponding information in a database.

Figure 6 is a diagram showing a method for determining abnormality through an artificial neural network, according to an embodiment of the present specification. Referring to FIG. 6, the input values of the artificial neural network 610 include three busbar currents, fine dust in the contact area, external humidity, external temperature, cumulative operation time at 50% or more of the rated current, cumulative operation time at 70% or more of the rated current, and rating. At least one of the accumulated operating time of more than 90% of the current, the accumulated number of breaker on/off operations, the ground leakage current, the accumulated number of important failures, and vibration may be provided. However, this is only one example and is not limited to the above-described embodiment. Here, ground leakage current and vibration can be measured through actual physical sensors. In addition, the accumulated operating time of 50% or more of the rated current, the accumulated operating time of 70% or more of the rated current, and the accumulated operating time of 90% or more of the rated current may be data created after processing using data of the measured busbar current. That is, the above-mentioned values may be values derived through feature engineering. Therefore, no actual measurement cost is incurred in measuring the above-mentioned information. Additionally, other artificial neural network 610 input values can also be derived through feature engineering, which creates meaningful random variables using physically sensed data.

As a specific example, a current sensor may be included within the electric vehicle distribution system for current sensing. Additionally, for the above-described input values of the artificial neural network 610, an external temperature sensor, an external humidity sensor, a contact part fine dust sensor, a ground leakage current sensor, and other sensors may be included. However, the above-mentioned sensors are low-cost general-purpose sensors and may be inexpensive. In other words, the above-mentioned sensor can be used for general purposes in the industry, so it can be possible to significantly reduce costs compared to existing methods.

In addition, as an example, the cumulative number of circuit breakers on/off is counted using the value of a voltage sensor inherently installed inside the switchboard, and the cumulative number of important failures can be provided by a switchboard manager entering the information through a keyboard into the diagnostic system. Since the diagnostic device automatically counts the number of circuit breakers on/off, number of major failures, and other phenomena that occur during operation, a separate sensor may not be used.

Additionally, as an example, contact temperature data, which is a diagnostic variable, can be installed in a form that is firmly fixed near the contact area using a general temperature measurement sensor (e.g. PT 100). For example, the electrical stability of the temperature sensing circuit can be ensured by not attaching the temperature sensor closely to the contact part. If a temperature sensor is attached in close contact with a contact area to which a voltage of several kV or more is applied, the contact voltage may be connected to the sensing circuit and affect it. Therefore, the contact part and the temperature sensor can be spaced apart considering the insulation distance. The above-described method may be an indirect contact point temperature measurement method that uses the contact part as a heat source and senses the surrounding temperature. Here, the temperature sensor uses the characteristic that when the temperature of the heat source changes, the surrounding temperature also changes linearly. As the sensor moves away from the heat source, the effect of temperature change at the measuring point due to fluctuations in the heat source weakens, so the required insulation distance can be minimized. A temperature sensor can be installed in the structure. Thereafter, the electric vehicle charging station distribution system abnormality diagnosis reference signal generating device 300 may compare and diagnose the contact temperature value derived as an output and the actual measured contact temperature value to derive a diagnosis result.

Here, as an example, the input values of the artificial neural network 610 include busbar current, contact part fine dust, external humidity, external temperature, accumulated operating time over 50% of the rated current, accumulated operating time over 70% of the rated current, and 90% of the rated current. Learning may be performed by considering the relationship between at least one of abnormal operation accumulated time, breaker on/off operation accumulated number, ground leakage current, critical failure accumulated number, and vibration and the contact temperature value as the output. That is, at least one input value of the artificial neural network 610 may be related to the contact temperature value as one output value, and learning is performed in the artificial neural network 610 by considering the correlation between each input value and the contact temperature value. It can be done.

As a specific example, different weights may be applied to each input value within the artificial neural network 610 based on the number and type of input values applied to the artificial neural network 610. As an example, the artificial neural network 610 may determine each weight for deriving the contact temperature value based on the number and type of input values. As a specific example, when the number of input values is large, the artificial neural network 610 may set the weight for each input value to be small. On the other hand, when the number of input values is small, the artificial neural network 610 may set the weight for each input value to be high.

As another example, the artificial neural network 610 may consider the input value type. As a specific example, busbar current and leakage current as input value types may be factors that have a large influence on the contact temperature value. On the other hand, external temperature and external humidity as input values may be factors that have a small effect on the contact temperature value. Accordingly, the artificial neural network 610 may assign a high weight to current-related input values and a low weight to temperature and humidity-related information as biased information. That is, the artificial neural network 610 can set the level of importance differently.

As another example, the artificial neural network 610 may perform overall learning by reflecting all input values, and may perform partial learning by reflecting some input values. That is, the artificial neural network 610 can perform full learning based on all input values and partial learning based on partial input values. As an example, the artificial neural network 610 may obtain correlation information between input values and contact temperature values as one output value based on full learning and partial learning. Additionally, the artificial neural network 610 can determine the correlation between input values through the above-described learning.

If fine dust increases with a specific input value, leakage current may increase. Here, the contact temperature value may increase as the leakage current increases. Accordingly, the artificial neural network 610 can learn the extent to which the contact part temperature value is influenced by considering the amount of fine dust in the contact part, and determine the weight accordingly. Additionally, busbar current is a source of heat due to Joule loss generated in the busbar and can affect the temperature of the contact area. As an example, the artificial neural network 610 can learn the extent to which the contact temperature value is influenced by considering the amount of busbar current and determine a weight accordingly.

Additionally, the rated current X% may be a load factor. Here, the accumulated operating time according to the load rate can be obtained through the main load rate index information of the electric vehicle distribution system. For example, the artificial neural network 610 may determine that the degree of relative deterioration may be high if the accumulated operating time in a high load rate state is high based on the main load rate index information, and may set the weight high accordingly. That is, the artificial neural network 610 can determine the corresponding weight based on the load ratio indicator information. Additionally, the artificial neural network 610 may consider ambient temperature and humidity information in relation to the contact temperature value. For example, ambient temperature and humidity information may be bias factors that can affect busbar temperature even under the same load conditions. As a specific example, under the same load conditions, the higher the ambient temperature, the higher the busbar temperature may be. On the other hand, under the same load conditions, the higher the humidity, the more moisture to cool the heat, and the lower the busbar temperature. In consideration of the above, the artificial neural network 610 may assign weights to the corresponding input values. As another example, the cumulative number of failures may be system reliability information. Here, as the cumulative number of failures increases, reliability may decrease, and the weight may reflect that information.

As another example, the accumulated number of circuit breaker on/off operations may also have an effect. For example, as the cumulative number of times the circuit breaker is turned on/off increases, the fastening of the fastener becomes loose due to vibration, which may generate heat. As a specific example, if the fastening part is not tight, heat may increase due to an increase in spark/contact resistance at the contact part due to vibration occurring in the fastening part. In addition, as it is a type of switch device, an increase in the number of circuit breaker operations can generate heat due to an increase in resistance at the breaker electrodes (contact between fixed and movable contacts). The generated heat can increase the temperature of the busbar connected to the breaker electrode and conductor. As an example, a circuit breaker may include a fixed contact point and a movable contact point, and each contact point is connected to an external busbar. Additionally, when the breaker is turned off, the movable contact point of the breaker may move and separate from the fixed contact point. On the other hand, when the breaker is turned on, the movable contact point of the breaker may move and be connected to the fixed contact point. Here, arcing (a type of spark that occurs when contact points are separated by a short distance) may occur due to on/off operation, and damage to contact points may occur. At this time, damage to the contact point may lead to heat generation at the contact point. That is, as the cumulative number of turns on/off of the circuit breaker increases, the contact temperature value may increase, and the artificial neural network 610 may assign weights by reflecting the corresponding information. Additionally, if the fastening part is shaken by vibration, heat may be generated. Additionally, an increase in ground leakage current may result in an overcurrent condition throughout the water supply. Here, if overcurrent occurs, the busbar temperature may increase. Here, the artificial neural network 610 may assign weights by considering not only the busbar current but also the leakage current, and the leakage current may be information complementary to the busbar current.

Here, as an example, when the artificial neural network 610 performs learning based on the above-described input values, the artificial neural network 610 may perform learning by distinguishing between biased input values and input values that are correlated. there is. As an example, the above-described temperature and humidity information may be biased input values that have little correlation with load operation. Here, the artificial neural network 610 can perform learning while independently adjusting temperature and humidity information using biased input values. On the other hand, the above-described busbar current and ground leakage current may be input values that are correlated. Accordingly, the artificial neural network 610 does not perform learning by setting the busbar current and ground leakage current as independent conditions, but can perform learning by simultaneously controlling the above-described input values. For example, as the busbar current increases, the ground leakage current may also increase, and the artificial neural network 610 may assign weights that are synchronized to the busbar current and ground leakage current with the associated input value rather than assigning individual weights. there is. As another example, there may be a correlation between the on/off number of circuit breakers, the cumulative number of failures, and the vibration input value. As a specific example, as the number of on/off times of the breaker increases, vibration may also increase, so synchronized weighting can be applied rather than individual weighting. In other words, the artificial neural network 610 can increase accuracy by considering and learning input values that are related to each other among the above-mentioned input values together, and learning biased information by applying individual weights. Based on the above, the artificial neural network 610 can derive the contact temperature value as an output through learning based on the input values.

Figure 7 is a diagram showing a method of performing categorization on artificial neural network input according to an embodiment of the present specification. Referring to FIG. 7, data provided as input values of the artificial neural network 610 may be the data described above in FIG. 6. However, the data in FIG. 6 may be an example, and input values of the artificial neural network 610 may be determined based on each available data. Here, the input values of the artificial neural network 610 may be determined as output values of the artificial neural network 610 and are related to the contact temperature value. Specifically, the artificial neural network 610 input values may be values that can affect the contact temperature value and may be values to which respective weights are set in consideration of the influence on the contact temperature value. As a specific example, considering the external temperature value measured as the input value in FIG. 6, if the external temperature is in the room temperature range and does not significantly affect the contact part, the weight applied to the artificial neural network 610 may be set small. . On the other hand, when the external temperature value is below the threshold value in mid-winter or above the threshold value in mid-summer, the weight may be set high because the influence on the contact area may be significant, but this is only an example and is not limited to this.

Additionally, the input values of the artificial neural network 610 may be categorized based on configurations that may affect the contact temperature value and provided as input values of the artificial neural network 610. As an example, referring to FIG. 7, categorizing may consist of electrical data 710, environmental data 720, external data 730, and feedback data 740, but this is only an example and is limited thereto. It may not work out.

As an example, the electrical data 710 may be a busbar current measured based on a current sensor, a cumulative rated current operation time derived based thereon, ground leakage current, and other electrical data. Additionally, the environmental data 720 may be data about the external environment, such as external temperature, external humidity, and vibration, and is not limited to a specific form. Additionally, the external data 730 may be data derived from the outside, such as the accumulated number of circuit breakers on/off or the accumulated number of important failures, and since the data is not sensed data, additional costs may not be incurred. Finally, feedback data 740 may be provided as input based on the artificial neural network 610. As an example, the temperature of the measurement unit may be derived as an output of the artificial neural network 610, and a diagnosis result may be derived after comparing the temperature of the measurement unit with the actual temperature of the measurement unit. Here, the accuracy information of the diagnosis result or the diagnosis result information itself may be provided as input to the artificial neural network 610. That is, the input values of the artificial neural network 610 can be used to derive the contact temperature value as an output value by considering physical data measured by a sensor, data measured by an environment-related sensor, data acquired from the outside, and feedback data. .

Figures 8 and 9 are diagrams showing the operation of the reference signal generating device 300 for diagnosing problems in the electric vehicle charging station power supply and distribution system according to an embodiment of the present specification. Specifically, referring to FIG. 8, the input values of the artificial neural network 610 may be determined based on at least one of electrical data 710, environmental data 720, external data 730, and feedback data 740. and this is the same as described above. Here, the above-mentioned sensor unit 330 may include a sensor that measures physical quantities related to electrical data (e.g. current sensor) and a sensor that measures data related to environmental data (e.g. temperature/humidity sensor). Additionally, as described above, the sensor unit 330 may include a temperature sensor that measures the temperature value of the contact portion at a fixed position and spaced at a preset interval in consideration of insulation from the contact portion. That is, the values measured in the sensor unit 330 are provided as the input value 810 of the artificial neural network 810 through feature engineering, and the contact temperature value and sensor derived as the output value 820 of the artificial neural network 610 Diagnosis may be performed by comparing contact temperature values measured through unit 330. Additionally, the diagnosis result information where the diagnosis was performed may be fed back as the input value 810.

More specifically, referring to FIG. 9, the sensor may be a low-cost sensor described above as a general-purpose sensor 920. At this time, data measured by a general-purpose sensor and data 930 on which preprocessing has been performed based on the data may be derived as a plurality of categorized input values 940, which may be as shown in FIG. 7.

Afterwards, the corresponding values are applied as input values of the artificial neural network 950, and one output value 960 can be derived as a temperature value for each contact part based on the artificial neural network 610. That is, the output value of the artificial neural network 610 is determined as a contact temperature value as one output value, and the input values of the artificial neural network 610 are one output value and a plurality of categorized inputs related to the contact temperature value. It can be composed of values. Thereafter, the derived contact temperature value may be compared with the temperature value measured by the temperature sensor 970, and the diagnosis result 980 may be derived through the difference value based on the comparison.

Here, the input values of the artificial neural network 610 described above generate reference signals for diagnostic variables using current, fine dust, temperature sensors, and vibration sensors, so the frequency band of the sensor is low, so the implementation of the signal processing circuit and related algorithms is simple. It can happen. Therefore, development costs may be low. In addition, since the device uses a general-purpose sensor, if appropriate production volume is secured in the commercialization stage after development, it can be advantageous in terms of self-ization, minimizing the effect on the price formation of the main device. Additionally, by using an intuitive diagnostic variable (contact temperature value), it can be easier for a system designer or system operator to intuitively understand and explain the diagnostic results. Therefore, it is easy to train related maintenance engineers needed for large-scale commercialization, providing the advantage of stable business operation.

Figure 10 is a diagram showing a method of performing diagnosis and comparison according to an embodiment of the present specification. Referring to FIG. 10, the comparison and diagnosis unit 1010 can diagnose an abnormality based on Equation 1 below.

[Equation 1]

As an example, in the reference signal generating device 300 for diagnosing abnormalities in the electric vehicle charging station distribution system, there may be a plurality of contact parts, and a temperature value may be derived as an output value for each contact part based on the artificial neural network 610. In Equation 1, n can be a value from 1 to j, which means the number of contact parts. Here, as an example, the positions of each contact part in the electric vehicle charging station distribution system may be different, and information about the position of each contact part and the configurations around each contact part may also be provided as input values of the artificial neural network 610 described above. However, it is not limited to specific examples.

In equation 1 is the predicted contact temperature, may be the measured contact temperature. Here, the residual error in Equation 1 may exhibit white noise characteristics if the predictor operates ideally and the system being diagnosed has no problems. As an example, diagnostic performance can be improved by removing white noise included in the residual derived by Equation 1, and Equation 2 and Equation 3 below can be considered. Equation 2 and Equation 3 may be a method of applying j Kalman filters using a state variable model to each residual to remove noise included in the residuals and then compare them. In equation 2, x is the residual, may be system noise. Additionally, in Equation 3 may be measurement noise. Here, the initial value of Equation 2 Can be set to the average value of the diagnostic variables used in training. Additionally, in Equation 3 is N(0, ) can be. Here, measurement noise can be modeled as a normal random variable with a mean of zero and a standard deviation of k, as shown in Figure 10.

Additionally, Figure 11 is a diagram showing the effect of applying a Kalman filter according to an embodiment of the present specification. Referring to FIG. 11, in the process of comparing and diagnosing the contact temperature value and the measured temperature value, white noise filtering may be performed based on Equation 2 and Equation 3. Afterwards, the size of the variance may become smaller as the area where the residual (error) is distributed decreases, and may be as shown in FIG. 11. As described above, if the predictor is designed so that the residuals show the characteristics of white noise, the Kalman filter can improve the performance of the predictor by reducing the variance of the residuals.

[Equation 2]

[Equation 3]

Figure 12 is a flowchart showing a method of generating a reference signal for diagnosing an abnormality in an electric vehicle charging station distribution system using an artificial neural network predictor according to an embodiment of the present specification. Referring to FIG. 12, the device can determine an experiment-based diagnostic standard level (S1210). Here, the diagnostic standard level may be a value related to determining whether something is abnormal or may be the contact temperature value described above. As an example, the device may set the initial value through experiment-based diagnostic criteria because no data exists initially when setting the diagnostic method. Thereafter, the device may obtain a diagnosis result based on the diagnosis reference level. (S1220) As an example, the device obtains a diagnosis result by comparing the contact temperature value determined through experiment with the temperature value measured through the temperature sensor. can do. Here, the device can accumulate data in a database through the diagnosis result information. (S1230) As an example, the device can accumulate information about the accuracy of the diagnosis result or information about factors affecting the diagnosis result in the database. . Afterwards, the device can reset the diagnostic standard level with the accumulated data based on the artificial neural network (S1240). That is, the device performs feature engineering of the data sensed from the above-described general-purpose sensors based on the accumulated data. Thus, the input values of the artificial neural network can be derived. Afterwards, the contact temperature value can be derived as the output value of the artificial neural network using the above-mentioned input values. Afterwards, the device can obtain a diagnosis result based on the diagnosis standard level. (S1250) That is, the device can obtain a diagnosis result using the contact temperature value as the artificial neural network output value, and the corresponding information is returned to the artificial neural network. Through feedback, the artificial neural network can be continuously updated.

Embodiments of the present invention described above can be implemented through various means. For example, embodiments of the present invention may be implemented by hardware, firmware, software, or a combination thereof.

In the case of hardware implementation, the method according to embodiments of the present invention uses one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), and Programmable Logic Devices (PLDs). , can be implemented by FPGAs (Field Programmable Gate Arrays), processors, controllers, microcontrollers, microprocessors, etc.

In the case of implementation by firmware or software, the method according to embodiments of the present invention may be implemented in the form of a module, procedure, or function that performs the functions or operations described above. Software code can be stored in a memory unit and run by a processor. The memory unit is located inside or outside the processor and can exchange data with the processor through various known means.

Detailed description of preferred embodiments of the invention disclosed above is provided to enable any person skilled in the art to make or practice the invention. Although the present invention has been described above with reference to preferred embodiments, those skilled in the art may make various modifications and changes to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You will understand that you can do it. Accordingly, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein. In addition, although the preferred embodiments of the present specification have been shown and described above, the present specification is not limited to the specific embodiments described above, and the technical field to which the invention pertains without departing from the gist of the present specification as claimed in the claims. Of course, various modifications can be made by those skilled in the art, and these modifications should not be understood individually from the technical ideas or perspectives of the present specification.

In addition, both the product invention and the method invention are described in the specification, and the description of both inventions can be applied supplementally as needed.

110: device
111: Device memory
112: Control unit of the device
113: Transmitter and receiver of device
120: device
121: Device memory
122: Control unit of the device
123: Transmitter and receiver of device
300: Electric vehicle charging station distribution system abnormality diagnosis reference signal generation device
310: Control unit of electric vehicle charging station distribution system abnormality diagnosis reference signal generation device
320: Database unit of electric vehicle charging station distribution system abnormality diagnosis reference signal generation device
330: Sensor unit of the electric vehicle charging station distribution system abnormality diagnosis reference signal generating device
340: Standard setting unit of electric vehicle charging station distribution system abnormality diagnosis reference signal generation device
350: Artificial neural network prediction unit of electric vehicle charging station distribution system abnormality diagnosis reference signal generation device

Claims (9)

In the method of generating a reference signal for diagnosing an electric vehicle charging station distribution system abnormality using an artificial neural network predictor,
A step of acquiring sensing data from at least one sensor, wherein the at least one sensor includes general-purpose sensors and a temperature sensor that is spaced apart by a preset distance and measures a temperature value of the contact portion by sensing the temperature of the contact portion;
A step of determining input values and output values of an artificial neural network based on the sensing data, wherein the input values are set using general-purpose sensing data obtained from the general-purpose sensors, and the output value of the artificial neural network is the temperature. Based on the contact temperature value measured through the sensor, the predicted contact temperature value is set to one value,
The input values include input value groups categorized based on data characteristics after data preprocessing and feature engineering are performed on the general-purpose sensing data, and each of the categorized input value groups is the contact temperature value. The values associated with each have weights based on their correlation with the contact temperature value;
deriving the predicted contact temperature value as the output value by applying the input values considering the weight of each of the categorized input value groups to the artificial neural network;
Comprising the step of comparing the predicted contact temperature value and the measured contact temperature value to determine whether there is an abnormality in the power distribution system,
The input value group is categorized into electrical data related to electrical operation, environmental data related to the surrounding environment of the electric vehicle charging station distribution system, external data input from the outside, and feedback data,
The general-purpose sensors are low-cost general-purpose sensors and include a current sensor, an external temperature sensor, an external humidity sensor, a contact part fine dust sensor, and a ground leakage current sensor,
The electrical data includes busbar current value, rated current-related accumulated operating time value, and ground leakage current value derived by performing the data pre-processing process and the feature engineering on the data acquired through the current sensor and the ground leakage current sensor. Contains,
The environmental data includes a contact part fine dust value, an external temperature value, and an external humidity value derived by performing the data preprocessing process and the feature engineering on data acquired through the external temperature sensor, the external humidity sensor, and the contact part fine dust sensor. and vibration values,
The external data includes the accumulated number of circuit breaker on/off operations and the accumulated number of important failures,
The feedback data includes a diagnostic result value and a diagnostic result accuracy value,
The weight of each of the categorized input value groups of the artificial neural network is set differently based on the input value type, and the weight for the electrical data is set higher than the weight for the environmental data,
The artificial neural network provides the output value based on learning using all of the electrical data, the environmental data, the external data, and the feedback data, and learning using some of the electrical data, the environmental data, the external data, and the feedback data. A reference signal generation method that acquires correlation information with the predicted contact temperature value and reflects it on the weight of each of the categorized input value groups of the artificial neural network.
delete delete delete delete According to claim 1,
When determining whether the power distribution system is abnormal by comparing the predicted contact temperature value with the measured contact temperature value, a difference value between the predicted contact temperature value and the measured contact temperature value is derived, and the difference is derived. A method of generating a reference signal that determines whether the abnormality exists after removing the white noise included in the value.
According to claim 1,
Setting an initial contact temperature value based on experimental data and obtaining a diagnostic result value based on the initial contact temperature value;
Building a database based on the diagnostic results; and
Once the database is established, it further includes deriving the contact temperature value based on the artificial neural network,
When the contact temperature value is derived through the artificial neural network based on the constructed database, sensing data is obtained from the at least one sensor, and the input values of the artificial neural network are determined based on the sensing data. , Method of generating reference signals.
A computer program stored in a computer-readable medium that is combined with hardware to generate a reference signal for diagnosing abnormalities in an electric vehicle charging station distribution system using the artificial neural network predictor according to any one of claims 1, 6 to 7.
In the reference signal generation device for diagnosing abnormalities in the electric vehicle charging station distribution system using an artificial neural network predictor,
A sensor unit that acquires at least one sensing data;
A database unit that stores data;
A standard setting unit that sets standards for determining whether the distribution system is abnormal;
An artificial neural network prediction unit that sets diagnostic variables based on an artificial neural network; and
A control unit that controls the sensor unit, the database unit, the standard setting unit, and the artificial neural network prediction unit,
The control unit
The sensing data is acquired through the sensor unit, wherein the at least one sensor includes general-purpose sensors and a temperature sensor that is spaced apart by a preset distance and measures the temperature value of the contact part by sensing the temperature of the contact part,
The input values and output values of the artificial neural network are determined based on the sensing data, wherein the input values are set using general-purpose sensing data obtained from the general-purpose sensors, and the output value of the artificial neural network is set using the temperature sensor. Based on the contact temperature value measured through the contact temperature value, the predicted contact temperature value is set to one value, and the input values are categorized based on data characteristics after data preprocessing and feature engineering are performed on the general-purpose sensing data. Includes a group of raised input values, wherein each of the categorized input value groups is values associated with the contact temperature value and has a respective weight based on the correlation with the contact temperature value,
Deriving the predicted contact temperature value as the output value by applying the input values considering the weight of each of the categorized input value groups to the artificial neural network through the artificial neural network prediction unit,
Compare the predicted contact temperature value with the measured contact temperature value to determine whether there is an abnormality in the power distribution system,
The input value group is categorized into electrical data related to electrical operation, environmental data related to the surrounding environment of the electric vehicle charging station distribution system, external data input from the outside, and feedback data,
The general-purpose sensors are low-cost general-purpose sensors and include a current sensor, an external temperature sensor, an external humidity sensor, a contact part fine dust sensor, and a ground leakage current sensor,
The electrical data includes busbar current value, rated current-related accumulated operating time value, and ground leakage current value derived by performing the data pre-processing process and the feature engineering on the data acquired through the current sensor and the ground leakage current sensor. Contains,
The environmental data includes a contact part fine dust value, an external temperature value, and an external humidity value derived by performing the data preprocessing process and the feature engineering on data acquired through the external temperature sensor, the external humidity sensor, and the contact part fine dust sensor. and vibration values,
The external data includes the accumulated number of circuit breaker on/off operations and the accumulated number of important failures,
The feedback data includes a diagnostic result value and a diagnostic result accuracy value,
The weight of each of the categorized input value groups of the artificial neural network is set differently based on the input value type, and the weight for the electrical data is set higher than the weight for the environmental data,
The artificial neural network provides the output value based on learning using all of the electrical data, the environmental data, the external data, and the feedback data, and learning using some of the electrical data, the environmental data, the external data, and the feedback data. A reference signal generating device that acquires correlation information with the predicted contact temperature value and reflects it on the weight of each of the categorized input value groups of the artificial neural network.