KR20230068453A - Abnormality detection device and method for heating control system - Google Patents

Abstract

Disclosed is an abnormal detection device and method for a heating control system. The anomaly detection apparatus of the present invention collects the communication unit and the received data that receives data measured from a sensor node including at least one sensor, inputs the collected data to a learning model in which past system states are learned, and learns the learning model. and a control unit that detects abnormality by diagnosing whether or not there is an abnormality in the current system state based on the output result value.

Description

Abnormality detection device and method for heating control system

The present invention relates to an anomaly detection technology, and more particularly, to an anomaly detection device and method for a heating control system that detects a temperature anomaly using a learning model learned by an unsupervised learning technique of machine learning.

The abnormality detection of the conventional southern control system is performed only by diagnosis using a predefined reference value. In the conventional case using such a predefined reference value, it is difficult to reflect individual characteristics according to the configuration of the heating control system, and a problem of resetting the reference value according to the characteristics of the heating control system may occur.

In particular, in the conventional method, since the reference value is set by the expert's knowledge and experience, and performance may be limited due to this, it is necessary to set a new reference value as the characteristics of the heating control system change, and improvement is also required in terms of efficiency. .

Korean Registered Patent Publication No. 10-2153913 (2020.09.10.)

An object of the present invention is to provide an anomaly detection device and method for a heating control system that accurately diagnoses temperature anomalies and increases operational stability and reliability of the system.

In order to achieve the above object, an anomaly detection device for a heating control system according to the present invention collects the communication unit and the received data for receiving measured data from a sensor node including at least one sensor, and the collected data and a control unit for inputting a past system state into a learned learning model, diagnosing whether or not there is an anomaly in the current system state based on a result value output from the learning model, and detecting an anomaly.

In addition, the control unit uses at least one statistical technique of maximum value, minimum value, average, median value, standard deviation, variance, moving average, skewness and kurtosis for the received data, Fourier transform, wavelet Performing preprocessing using at least one signal processing technique among wavelet transforms, or at least one dimensionality reduction technique among principal component analysis (PCA) and t-Stochastic Nearest Neighbor (t-SNE) characterized by

In addition, the control unit is characterized in that the learning model is modeled using an auto-encoder.

In addition, the auto-encoder builds a neural network reflecting characteristics of the data of the past system state through encoding, and calculates a difference between the modeled data and the data of the current system state through decoding.

In addition, the control unit is characterized in that the diagnosis is determined through a difference between data generated during data regeneration, and a discrimination criterion is set using a distribution of residuals with the learned learning data.

In addition, the control unit is characterized in that the determination criterion is set based on data for a predetermined specific period (Δt).

An anomaly detection method for a heating control system according to the present invention includes the steps of an anomaly detection device collecting data, and the anomaly detection device inputting the collected data to a learning model in which a past system state has been learned, and outputting from the learning model and diagnosing whether or not there is an abnormality in the current system state based on the obtained result value and detecting an abnormality.

In addition, the step of detecting the anomaly is characterized in that the learning model is modeled using an auto-encoder.

An anomaly detection apparatus and method for a heating control system of the present invention detects anomaly adaptive to data rather than a limited anomaly detection such as a reference value by detecting a temperature anomaly using a learning model learned by an unsupervised learning technique of machine learning. It is possible to improve stability and reliability for maintenance by performing

In addition, it is not necessary to acquire abnormal data separately, so it can be easily applied to the existing system and the cost of data acquisition can be reduced.

1 is a configuration diagram for explaining a heating control system according to an embodiment of the present invention.
2 is a block diagram for explaining an abnormality detection apparatus according to an embodiment of the present invention.
3 is a block diagram for explaining a control unit according to an embodiment of the present invention.
4 is a diagram for explaining an auto encoder according to an embodiment of the present invention.
5 is a flowchart for explaining an anomaly detection method according to an embodiment of the present invention.
6 is a flowchart for explaining a model learning method according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. First, in adding reference numerals to the components of each drawing, it should be noted that the same components have the same numerals as much as possible even if they are displayed on different drawings. In addition, in describing the present invention, if it is determined that a detailed description of a related known configuration or function is obvious to those skilled in the art or may obscure the gist of the present invention, the detailed description will be omitted.

1 is a configuration diagram for explaining a heating control system according to an embodiment of the present invention.

Referring to FIG. 1 , the heating control system 400 increases operational stability and reliability of the system by performing an accurate temperature abnormality diagnosis. The heating control system 400 includes an abnormal detection device 100, a sensor node 200 and a user terminal 300.

The abnormality detection device 100 detects the temperature abnormality of the heating control system 400 . The anomaly detection device 100 may detect a temperature anomaly by applying the received data to a learning model learned through an unsupervised learning technique of machine learning. Through this, the anomaly detection device 100 helps the system operator to take immediate action to prevent loss and additional damage by supporting early recognition of an abnormal state of the system. The abnormality detection apparatus 100 may be a computing system including a laptop computer, a desktop computer, a tablet PC, a handheld PC, a server computer, a cluster computer, and the like.

The sensor node 200 includes at least one sensor, and the sensor is installed in an area where the heating control system 400 is operated. The sensor node 200 measures environment information of an installed area through sensors. Here, the sensor includes a temperature sensor, a pressure sensor, a differential pressure sensor, a humidity sensor, and the like, and the environment information may include temperature, pressure, differential pressure, humidity, and the like. Preferably, the sensor node 200 may be implemented such that a plurality of sensors are evenly installed in an area where the heating control system 400 is operated. The sensor node 200 transmits data including the measured environment information to the anomaly detection device 100 . To this end, the sensor node 200 may include a communication module.

The user terminal 300 is a terminal used by a user or a system operator, and receives monitoring information about a result of abnormality detection from the abnormality detection device 100 . The user terminal 300 outputs the received monitoring information. Here, the monitoring information may include information on whether or not there is an abnormality in temperature, and may include information on a region where an abnormality occurs when an abnormality occurs. The user terminal 300 may be a personal computing system including a smart phone, laptop, desktop, tablet PC, handheld PC, and the like.

Meanwhile, the heating control system 400 builds a communication network 450 to perform communication between the abnormal detection device 100, the sensor node 200, and the user terminal 300. The communication network 450 may be composed of a backbone network and a subscriber network. The backbone network may be composed of one or a plurality of integrated networks among an X.25 network, a Frame Relay network, an ATM network, a Multi Protocol Label Switching (MPLS) network, and a Generalized Multi Protocol Label Switching (GMPLS) network. Subscriber networks include FTTH (Fiber To The Home), ADSL (Asymmetric Digital Subscriber Line), cable network, zigbee, Bluetooth, and wireless LAN (IEEE 802.11b, IEEE 802.11a, IEEE 802.11g, IEEE 802.11n ), Wireless Hart (ISO/IEC62591-1), ISA100.11a (ISO/IEC 62734), COAP (Constrained Application Protocol), MQTT (Multi-Client Publish/Subscribe Messaging), WIBro (Wireless Broadband), Wimax, 3G, It may be High Speed Downlink Packet Access (HSDPA), 4G and 5G. In some embodiments, the communication network 450 may be an internet network or a mobile communication network. In addition, the communication network 450 may include all other well-known wireless communication or wired communication methods to be developed in the future.

2 is a block diagram for explaining an abnormality detection apparatus according to an embodiment of the present invention.

Referring to FIGS. 1 and 2 , the abnormality detection apparatus 100 includes a communication unit 10 and a control unit 30 , and may further include an output unit 50 and a storage unit 70 .

The communication unit 10 communicates with the sensor node 200 and the user terminal 300 . The communication unit 10 receives data including environment information from the sensor node 200 and transmits monitoring information to the user terminal 300 .

The control unit 30 performs overall control of the abnormality detection device 100 . The control unit 30 collects data received from the communication unit 10 . The control unit 30 inputs the collected data to a learning model in which the past system state has been learned, and diagnoses whether the current system state is abnormal based on a result value output from the learning model. The control unit 30 detects an abnormality based on the diagnosis result, and generates monitoring information using the detected result. The control unit 30 may control the generated monitoring information to be output or transmitted to the user terminal 300 . Here, the monitoring information may include information on whether or not there is an abnormality in temperature, and may include information on a region where an abnormality occurs when an abnormality occurs.

Meanwhile, the controller 30 learns a learning model based on the collected data. At this time, the control unit 30 first performs preprocessing to extract an index that can best represent the characteristics of the system. The control unit 30 models the past system state of the heating control system 400 using an auto-encoder based on the preprocessed data. That is, the controller 30 may learn the learning model using an auto-encoder, but is not limited thereto and may learn the learning model through various learning techniques.

The output unit 50 outputs data received from the communication unit 10 and outputs monitoring information generated from the control unit 30 . To this end, the output unit 50 may be a display, such as a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), or an organic light-emitting diode. It may include at least one of diode, OLED), flexible display, and 3D display.

The storage unit 70 stores a program or algorithm for driving the anomaly detection device 100 . The storage unit 70 stores data received from the communication unit 10 and monitors information generated from the control unit 30 . The storage unit 70 is a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (eg SD or XD memory, etc.), RAM (Random Access Memory, RAM), SRAM (Static Random Access Memory), ROM (Read-Only Memory, ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, It may include at least one storage medium of a magnetic disk and an optical disk.

3 is a block diagram for explaining a control unit according to an embodiment of the present invention, and FIG. 4 is a diagram for explaining an auto encoder according to an embodiment of the present invention.

Referring to FIGS. 2 to 4 , the control unit 30 includes a data collection unit 31 , a model learning unit 33 and an anomaly detection unit 35 .

The data collection unit 31 collects at least one piece of data received from the communication unit 10 . Here, the data means data including environment information. The data collection unit 31 stores the collected data in the storage unit 70 .

The model learning unit 33 preprocesses the data stored in the storage unit 70 . The model learning unit 33 uses at least one statistical technique of maximum value, minimum value, average, median value, standard deviation, variance, moving average, skewness and kurtosis, Fourier transform, wavelet transform ), or at least one dimensionality reduction technique among principal component analysis (PCA) and t-Stochastic Nearest Neighbor (t-SNE), preprocessing may be performed. The model learning unit 33 models a learning model using an auto-encoder. In other words, the model learning unit 33 searches for a feature combination that can best represent the driving characteristics of the heating control system 400, and performs modeling for a learning model capable of efficiently diagnosing abnormalities based on the searched feature combination. do. Here, the autoencoder learns the encoding of data for dimensionality reduction in an unsupervised learning method. In an autoencoder, a reconstruction neural network is trained along with a reduced neural network, and an attempt is made to decode a representation as close as possible to the original input from the reduced representation. That is, the auto-encoder builds a neural network that reflects the characteristics of the data of the past system state through encoding, and calculates the difference between the modeled data and the data of the current system state through decoding.

The anomaly detection unit 35 inputs the collected data to the learning model learned from the model learning unit 33, diagnoses whether or not there is an anomaly in the current system state based on the result value output from the learning model, and detects the anomaly. Here, the anomaly detection unit 35 determines diagnosis through a difference between data generated during data regeneration, and sets a criterion for determination using a distribution of residuals with the learned learning data. Preferably, the abnormality detection unit 35 may use the largest value among the error values as the determination criterion, but is not limited thereto and may be flexibly set according to the user. In addition, the anomaly detection unit 35 may set a criterion for determination based on data for a predetermined specific period (Δt) rather than data for a single time. The abnormality detection unit 35 generates monitoring information using the detected result. Here, the monitoring information may include information on whether or not there is an abnormality in temperature, and may include information on a region where an abnormality occurs when an abnormality occurs. The anomaly detection unit 35 may control the generated monitoring information to be output or transmitted to the user terminal 300 .

5 is a flowchart for explaining an anomaly detection method according to an embodiment of the present invention.

Referring to FIGS. 1 and 5 , the anomaly detection method detects a temperature anomaly using a learning model learned by an unsupervised learning technique of machine learning, thereby detecting anomaly adaptive to data rather than a limited anomaly detection such as a reference value. It can improve stability and reliability for maintenance by performing The anomaly detection method does not require separate acquisition of anomaly data, so it can be easily applied to an existing system and can reduce the cost of data acquisition.

In step S110, the abnormality detection device 100 collects data. The anomaly detection device 100 receives and collects at least one piece of data including environmental information from the sensor node 200 in real time. At this time, the abnormality detection device 100 may store the collected data.

In step S120, the anomaly detection device 100 detects an anomaly in temperature using the collected data. The anomaly detection device 100 inputs the collected data to a learning model in which a past system state has been learned, and detects an anomaly by diagnosing whether or not there is an anomaly in the current system state based on a result value output from the learning model. Here, the learning model is a model modeling the past system state of the heating control system 400 using an auto encoder. The abnormality detection device 100 generates monitoring information using the detected result. The abnormal detection device 100 outputs the generated monitoring information or transmits it to the user terminal 300 .

6 is a flowchart for explaining a model learning method according to an embodiment of the present invention.

Referring to FIGS. 1 and 6 , the model learning method learns a learning model using an auto-encoder. Here, the model learning method may precede the anomaly detection method shown in FIG. 5 .

In step S210, the anomaly detection device 100 collects and stores data. The anomaly detection device 100 receives and collects at least one piece of data including environmental information from the sensor node 200 in real time, and stores the collected data.

In step S220, the anomaly detection device 100 learns the learning model. The abnormality detection device 100 pre-processes the stored data. The anomaly detection device 100 models a learning model using an auto-encoder. That is, the anomaly detection device 100 searches for a feature combination that can best represent the driving characteristics of the heating control system 400, and models a learning model capable of efficiently diagnosing anomaly based on the searched feature combination. .

The method according to an embodiment of the present invention may be provided in the form of a computer readable medium suitable for storing computer program instructions and data. Such a computer-readable recording medium may include program commands, data files, data structures, etc. alone or in combination, and includes all types of recording devices storing data that can be read by a computer system. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs (Compact Disk Read Only Memory) and DVDs (Digital Video Disks). Optical media), magneto-optical media such as floptical disks, and program instructions such as ROM (Read Only Memory), RAM (RAM, Random Access Memory), flash memory, etc. and a hardware device specially configured to do so. In addition, the computer-readable recording medium is distributed in computer systems connected through a network, so that computer-readable codes can be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the present invention can be easily inferred by programmers in the technical field to which the present invention belongs.

Although the above has been described and illustrated in relation to preferred embodiments for illustrating the technical idea of the present invention, the present invention is not limited to the configuration and operation as shown and described in this way, without departing from the scope of the technical idea. It will be readily apparent to those skilled in the art that many changes and modifications can be made to the present invention. Accordingly, all such appropriate changes and modifications and equivalents should be regarded as falling within the scope of the present invention.

10: Ministry of Communications
30: control unit
50: output unit
70: storage unit
100: anomaly detection device
200: sensor node
300: user terminal
400: heating control system
450: communication network

Claims (8)

A communication unit for receiving measured data from a sensor node including at least one sensor; and
Collecting the received data, inputting the collected data to a learning model in which the past system state has been learned, and diagnosing anomaly in the current system state based on the result value output from the learning model to detect anomaly control unit;
Abnormality detection device for a heating control system comprising a. According to claim 1,
The control unit,
Using at least one statistical technique of maximum value, minimum value, average, median value, standard deviation, variance, moving average, skewness and kurtosis for the received data, Fourier transform, wavelet transform Heating characterized in that preprocessing is performed using at least one signal processing technique among, or at least one dimensionality reduction technique among principal component analysis (PCA) and t-Stochastic Nearest Neighbor (t-SNE) An anomaly detection device for control systems. According to claim 1,
The control unit,
An abnormality detection device for a heating control system, characterized in that the learning model is modeled using an auto-encoder. According to claim 3,
The autoencoder,
An anomaly detection device for a heating control system, characterized in that for constructing a neural network reflecting the characteristics of the data of the past system state through encoding, and calculating the difference between the modeled data and the data of the current system state through decoding. According to claim 1,
The control unit,
An anomaly detection device for a heating control system, characterized in that the diagnosis is determined through a difference between data generated during data regeneration, and a discrimination criterion is set using a distribution of residuals with the learned learning data. According to claim 5,
The control unit,
An anomaly detection device for a heating control system, characterized in that for setting a discrimination criterion based on data for a predetermined specific period (Δt). collecting data by an anomaly detection device; and
inputting, by the anomaly detection device, the collected data to a learning model in which a past system state is learned, and detecting an anomaly by diagnosing whether or not there is an anomaly in a current system state based on a result value output from the learning model;
Abnormal detection method for a heating control system comprising a. According to claim 7,
The step of detecting the abnormality is,
An abnormality detection method for a heating control system, characterized in that for modeling the learning model using an auto-encoder.

Images