KR102778904B1 - Short-circuit fire monitoring system for power facility - Google Patents

Abstract

The single-wire fire monitoring system for power facilities of the present invention comprises: a fire image monitoring module for determining whether a fire has occurred and its relevance to the power facility from a captured image of the power facility and a nearby area; a single-wire monitoring module for monitoring a single-wire state of the power facility; and a fire hazard alarm device for determining fire-related risk factors for the power facility from the determination of whether a fire has occurred and the single-wire state, and alerting the fire-related risk factors for the power facility, wherein the fire hazard alarm device can estimate that a fire has occurred due to the single-wire if the time at which the fire image detection module detects the occurrence of a fire in the power facility and the time at which the single-wire monitoring module detects the single-wire state are similar.

Description

{SHORT-CIRCUIT FIRE MONITORING SYSTEM FOR POWER FACILITY}

The present invention relates to a fire monitoring system for power facilities capable of preventing fires caused by abnormalities in power facilities, particularly, forest fires caused by power line disconnection, and a fire cause analysis method.

In Korea, over 70% of the entire land area is forest. Since the 1970s, as the economy has improved, the forests have become denser, but on holidays, due to the increased carelessness of hikers, many forest fires occur. In addition to the carelessness of hikers, forest fires are also caused by natural disasters such as burning of rice field ridges or lightning in rural areas. In addition, the causes of forest fires are becoming more diverse.

Forest fires caused by various causes are becoming increasingly large-scale through dense forests, and if early suppression fails, it is not easy to suppress, which causes many casualties and property damage. Forest fires not only cause damage/loss to forests, but also cause, for example, the destruction of Naksansa Temple and cultural properties, and pose a major threat to the nation's power transmission network. Therefore, early detection and early suppression of forest fires should be given top priority from the perspectives of forest preservation, human life protection, and property protection, and there is a growing demand for bold investment in advance for this purpose.

The conventional forest fire monitoring method is mainly based on visual surveillance and control by managers. That is, forest surveillance posts are installed at various points of the mountain where people frequently enter or where forest fire monitoring is easy, and when the manager detects a forest fire, he or she uses a communication means to notify the central control center.

Meanwhile, the conventional forest fire surveillance system using cameras is a system that installs cameras on major elements of a mountain, captures images received through the cameras, and allows the central control center's operating personnel to visually check the images to detect forest fires.

However, the conventional forest fire surveillance system had the problem that a small number of operators could not simultaneously monitor images received from a large number of cameras installed in major elements of the mountain, and that visual detection was not accurate because multiple images were displayed and confirmed on a single screen. In addition, the conventional forest fire surveillance system had the problem that rapid detection and response to forest fires were impossible when the operators were absent.

Technologies for automating forest fire detection using camera footage, etc. are also being attempted. These fire detection and monitoring systems enable early response to fires by detecting the point of ignition and issuing a fire alarm when a fire is detected through CCTV cameras or fire detection sensors.

The forest fire surveillance method utilizing somewhat improved existing video surveillance devices captures general images and thermal images using CCTV, infrared, ultraviolet cameras, and various sensors, and determines temperature distribution information or fire images from the acquired images. Accordingly, the installation cost of surveillance devices per location is excessive, and additional labor and service costs are required for maintenance of various devices and parts. In addition, since the presence of a fire must be determined solely by video images, the sensitivity of various signals cannot be lowered to prevent failure to recognize fire accidents, making it vulnerable to misrecognition by ultraviolet rays, fog, etc. Power companies need to detect and determine short circuit situations in power lines with a high risk of fire by fusing video surveillance information acquired from CCTV with power data such as voltage and current, which are status information of power facilities, but the related technology system is insufficient.

The forest fire surveillance system using CCTV installed by power companies is installed in the passage of distribution lines vulnerable to forest fires, and the surveillance work is performed at the Distribution Control Center where the operator is stationed 24 hours a day. The system operator monitors the surveillance system at all times while operating the Distribution Automation System (DAS). However, in the case of such a system based on the perception of the supervisor, it is difficult to continuously monitor the surveillance screen monitor, and the frequent acquisition of incorrect information in the field significantly reduces the work efficiency of the system operator.

In addition, the power line disconnection detection system through the simple linkage of the existing proposed distribution automation system (DAS) and the intelligent power metering system (AMI Advanced Metering Infrastructure) has the problem that the power to the modem of the power meter without auxiliary power is cut off when a power outage occurs, so the phase loss and power outage information of the power meter cannot be transmitted to the server, making it difficult to apply in the field in practice.

In particular, high impedance faults (HIFs) in 22.9kV extra-high voltage distribution lines caused by power line breakage are difficult to clearly detect where they occur, so they often depend on customer reports and line patrols. This means that it takes a long time to identify and repair the accident location, and there are high concerns about fire risks and safety accidents. In addition, there is a risk that breakage accidents in low-voltage distribution lines in mountainous areas may develop into forest fires due to the lack of an early detection method using existing power grid monitoring systems, including the DAS system. Therefore, in order to quickly detect breakage accidents in power facilities operated in mountainous, dry, and windy areas vulnerable to fire, it is necessary to establish technical measures to quickly detect breakage status regardless of high or low voltage systems and prevent the spread of fire.

To illustrate the problem situation specifically, let's assume that the video surveillance system installed in the mountain power line route that is currently monitoring the forest fire risk area is simply monitoring the line route using CCTV. In this environment, the operator of the power distribution system operation center must check the monitor screen of the system regularly to check for forest fires.

Currently, in order to prevent repetitive occurrence of arcs and the spread of fire due to the reclosing operation of circuit breakers, domestic and foreign power companies are implementing a policy to stop the reclosing function of circuit breakers (CBs) and reclosers installed in fire-vulnerable areas such as mountainous regions during dry and strong winds. In order to determine whether or not a power line installed in the field is open, the system operation manager checks the fault indicator and open phase information in the DAS system that operates the 22.9kV distribution system to estimate whether or not a disconnection has occurred, and then takes measures to cut off the power in the relevant section and stop test transmission to prevent fire or safety accidents caused by arcing in the open-circuit atmosphere. However, if a high impedance fault (HIF) occurs when the power line is open, which is not completely grounded to the ground, the circuit breaker will not operate, so continuous arcing may occur at the open-circuit area, which may develop into a fire or cause safety accidents such as electric shock.

In addition, the DAS system monitors the 22.9kV high-voltage distribution system, so it has a blind spot in that it cannot monitor the low-voltage distribution system supplied at 220V and 380V. Accordingly, it is difficult to detect and prevent accidents in which power lines are cut off in the low-voltage distribution system in a fire-vulnerable section and develop into fires.

Republic of Korea Public Notice No. 10-2020-0080579

The present invention aims to provide a short-circuit fire monitoring system for power facilities that can accurately and quickly monitor/analyze whether a fire has occurred due to a short-circuit accident of a power line installed in a fire-vulnerable location such as a mountainous area.

A fire surveillance system for power facilities according to one aspect of the present invention may include a fire image surveillance module that determines whether a fire has occurred and its relevance to the power facility from a captured image of the power facility and a nearby area; a disconnection surveillance module that monitors a disconnection status of the power facility; and a fire hazard alarm device that determines fire-related risk factors for the power facility from the determination of whether a fire has occurred and the disconnection status, and issues an alarm for fire-related risk factors for the power facility.

Here, the short circuit detection module can periodically check the operation of AMI devices within the range affected by the power facility to determine whether there is a short circuit related to the power facility.

Here, the disconnection monitoring module can recognize a disconnection when the number of transmission failures for the Alive message, which is status information of a small-capacity modem that the AMI devices periodically transmit separately from the measurement data transmission cycle, accumulates and is counted to a critical number.

Here, the fire hazard alarm device can estimate that the fire is caused by a short circuit if the time at which the fire image detection module detects a fire in the power facility and the time at which the short circuit monitoring module detects a short circuit are similar.

Here, the fire hazard alarm device can perform a method for analyzing a short circuit as a cause of a fire, including the steps of: checking for a short circuit or phase failure alarm within the DAS section for the power facility if a fire is estimated to be caused by the short circuit; determining that there is a short circuit or phase failure alarm within the DAS section for the power facility; determining that there is a short circuit or phase failure alarm within the DAS section for the power facility if there is no short circuit or phase failure alarm within the DAS section for the power facility; determining that there is a short circuit or phase failure alarm within the AMI low-voltage meters within the affected range for the power facility if there is a short circuit or phase failure alarm among the AMI low-voltage meters; and determining that there is a fire caused by a cause other than a short circuit or phase failure among the AMI low-voltage meters.

Here, the method for analyzing a short circuit as the cause of the fire may further include, after the step of determining that the low-voltage power line is short-circuited, a step of confirming the grid connection location and power outage scale of the AMI low-voltage instrument that failed the confirmation; and a step of estimating the short circuit section based on the grid connection location and power outage scale.

Here, the fire video surveillance module can estimate the location of the ignition point and whether it is close to power equipment by using a deep learning model that has learned images of fire elements such as smoke and flames and power equipment such as poles, transformers, insulators, and wires as training data.

Here, the fire video surveillance module may perform a preprocessing step of removing noise including light reflection from the captured video; or a step of removing clouds and fog that are likely to be falsely detected as smoke through a difference image of consecutive frames.

Here, the fire video surveillance module can classify a fire as a ‘fire near power equipment’ when it detects fire elements such as smoke and flames overlapping on the power equipment image.

Here, the fire hazard warning device can output a screen that provides information on fire factors, proximity to power equipment, and power line disconnection occurrence sections at once.

Here, the fire hazard warning device can define data on a time period with a low possibility of fire occurrence and weather information based on machine learning and provide this to the manager on a screen.

According to another aspect of the present invention, a method for analyzing the cause of a fire for a power facility may include the steps of: recognizing a fire near the power facility from a surveillance video; checking a short-circuit or phase failure alarm within a DAS section for the power facility, and determining that there is a short-circuit of a high-voltage power line if an alarm exists; checking the operation status of AMI low-voltage meters within a range affected by the power facility if there is no short-circuit or phase failure alarm within the DAS section; determining that there is a low-voltage power line short-circuit if there is a confirmation failure among the AMI low-voltage meters; and determining that the fire is caused by a cause other than a power line short-circuit if there is no confirmation failure among the AMI low-voltage meters.

Here, the step of recognizing a fire from the surveillance video may include the step of detecting the occurrence of a fire in the surveillance video of the power facility; the step of calculating the separation distance between the power facility and the fire component that occurred; and the step of sending an alarm according to the separation distance.

Here, after the step of determining that the low-voltage power line is disconnected, the step of confirming the grid connection location and power outage scale of the AMI low-voltage instrument that failed the confirmation may be further included; and the step of estimating the disconnection section based on the grid connection location and power outage scale may be further included.

By implementing a single-wire fire monitoring system for power facilities according to the invention of the present invention having the above-described configuration, there is an advantage in that it is possible to accurately and quickly monitor whether a fire has occurred due to a single-wire accident of a power line installed in a fire-vulnerable location such as a mountainous area.

The single-line fire monitoring system for power facilities of the present invention automatically detects fire elements on a CCTV screen using deep learning technology and then sends an alarm, thereby reducing the CCTV monitoring burden on system operation managers and improving the effectiveness of monitoring work.

The single-wire fire monitoring system for power facilities of the present invention comprehensively analyzes fire element image detection information, single-wire/phase failure alarms of the DAS system, and Alive message reception information of the AMI system, so that a system operation manager can determine whether a fire has occurred in a 22.9 kV extra-high voltage distribution line as well as a 220 V/380 V low voltage distribution line, whether a power line has been disconnected, and where the fire has occurred, and accordingly, follow-up measures such as preemptive blocking of the distribution system where an accident has occurred, on-site dispatch instructions, and fire reporting can be promptly implemented, so that there is an advantage in efficiently preventing the spread of fire.

The short-circuit fire monitoring system for power equipment of the present invention has the advantage of being able to effectively prevent fires caused by short-circuit accidents in power lines installed in fire-vulnerable locations at low cost.

FIG. 1 is a block diagram illustrating a fire monitoring system for power facilities according to the invention.
Figure 2 is a block diagram illustrating a detailed configuration of the fire video surveillance module of Figure 1.
Figure 3 is a block diagram showing a detailed configuration of the short-circuit monitoring module and fire hazard alarm device of Figure 1.
Figure 4 is a flow chart illustrating a method for analyzing the cause of a fire in a power facility according to the invention.
FIG. 5 is a conceptual diagram illustrating the advantages of implementing a fire surveillance system according to the invention in a field where a CCTV system for surveillance has already been constructed.

In describing the present invention, terms such as first, second, etc. may be used to describe various components, but the components may not be limited by the terms. The terms are only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component.

When it is said that a component is connected or coupled to another component, it can be understood that it may be directly connected or coupled to that other component, but there may also be other components in between.

The terminology used in this specification is only used to describe specific embodiments and is not intended to limit the invention. The singular expression may include the plural expression unless the context clearly indicates otherwise.

In this specification, terms such as “include” or “have” are intended to specify the presence of a feature, number, step, operation, component, part or combination thereof described in the specification, and can be understood as not excluding in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

Additionally, the shape and size of elements in the drawing may be exaggerated for clearer explanation.

The idea of the present invention can support the construction of a fire detection system using camera (CCTV) images that also apply detection of power line disconnection sections.

For example, the present invention can build a power facility fire monitoring system that can accurately and quickly detect whether a fire has occurred due to a short circuit in a power line installed in a fire-vulnerable area such as a mountainous region by utilizing artificial intelligence-based image recognition technology and metering information from an advanced metering infrastructure (AMI), and can provide a timely warning of danger to power facility operators, including system operation managers.

In addition, the present invention can provide a power line disconnection detection and fire detection algorithm in the form of a learning model that detects fire elements such as flames/smoke from CCTV surveillance images, issues an alarm according to the distance/proximity between power facilities and the fire elements, and determines whether power lines are disconnected in the area and the disconnection section.

FIG. 1 is a block diagram illustrating a fire monitoring system for power facilities according to the invention.

A city power facility fire surveillance system (1000) may include a fire image surveillance module (200) that determines whether a fire has occurred from a captured image of a power facility and a nearby area; a disconnection surveillance module (400) that monitors a disconnection status of the power facility; and a fire hazard alarm device (600) that determines fire-related risk factors for the power facility from the determination of whether a fire has occurred and the disconnection status, and issues an alarm for fire-related risk factors for the power facility.

The above fire video surveillance module (200) is intended to automatically detect suspicious areas when smoke, flames, etc. occur at or near power facilities by using an artificial intelligence-based fire surveillance algorithm on existing CCTV footage.

The above-mentioned short-circuit monitoring module (400) can detect, for example, a phase failure or no voltage in a high- or low-voltage power meter connected to a distribution system in a fire video surveillance area through an AMI system.

The above fire hazard alarm device (600) is intended to send an alarm to the system operation manager when a power line disconnection or fire hazard is detected.

The above-described artificial intelligence-based fire video surveillance module (200) is a module that estimates the location of the ignition point and whether it is close to the power facility by using a deep learning model that has learned fire elements such as smoke and flames and images of power facilities such as electric poles, transformers, insulators, and wires as training data to estimate the ignition point near the power facility under video surveillance.

The above fire video surveillance module (200) can be implemented in the form of software to monitor video acquired from a CCTV system without replacing video equipment or surveillance software installed on site by utilizing an existing CCTV system, and can be installed in an existing system to detect the occurrence of a fire in real time.

For example, the fire video surveillance module (200) first receives a video obtained through CCTV as input and goes through a preprocessing step such as removing noise such as light reflection. In the preprocessed video, a deep learning-based object detection algorithm (Scaled YOLOV4, YOLOv4, etc.) can be used to detect power equipment, fire elements (smoke/flame), and ignition locations.

During the fire detection process, clouds or fog with similar characteristics to smoke may be intermittently misdetected as smoke. To compensate for this shortcoming, Optical flow information obtained through the difference image of consecutive frames is used to remove clouds and fog with a high possibility of being misdetected as smoke, thereby preventing false alarms transmitted to the manager in advance. In addition, meteorological information is fused with the image-based fire detection algorithm (deep learning-based object detection/Optical Flow) to further enhance fire detection performance. That is, data on periods with a low possibility of fire occurrence and meteorological information are trained/estimated based on machine learning and fused with image-based results to increase the accuracy of fire detection while reducing false detections.

The above fire image surveillance module (200) outputs a signal for determining whether there is a fire and the location of the ignition based on the result obtained by integrating and analyzing the entire algorithm operation. If a fire is recognized to have occurred in a power facility, that is, if fire elements such as smoke and flames are detected overlapping the power facility image, it is classified as a 'fire near the power facility' and an alarm is notified to the user, and the risk of fire occurrence can be determined by linking with the power line disconnection surveillance information of the 'AMI power line disconnection surveillance module'.

The above-mentioned single-line monitoring module (400) may be an AMI-based power line single-line monitoring module, and in this case, the period for communicating metering information from an AMI power meter to a server is typically 15 minutes for high voltage and 60 minutes for low voltage, and when the power of the distribution system driving the power meter is cut off, the power of the power meter and its modem are also turned off. Accordingly, when a power outage occurs in a high-voltage distribution line or a single-phase power line single-line disconnection occurs in a low-voltage distribution system, the power of the AMI power meter and modem is turned off, and thus the status cannot be transmitted to the server.

This proposed technology configures a data transmission algorithm to periodically transmit Alive messages, which are status information of a small modem, to the AMI server separately from the transmission cycle of metering data (LP, Load Profile) to determine whether the power meter is powered off. The transmission cycle is set to 30 seconds, and in the event of a power outage due to a power line break, etc., the number of failed message transmissions in which the Alive message does not reach the AMI server is accumulated and counted 5 times to recognize the power outage. In addition, the Alive message transmission cycle and the accumulated count of power outage recognition can be varied and applied depending on the requirements of the field.

Since AMI power meters have connection information to the distribution system, if a power outage or phase loss event occurs in a power meter connected to a distribution system in a wildfire-prone area monitored by CCTV, the event can be sent as an alarm to the user, allowing the system operator to take initial measures such as remotely shutting off the risk area or issuing a dispatch order for on-site inspection.

If a downed power line crosses a tree or the like and progresses to a high-resistance ground fault (HIF), the power to the single-phase power meters of AMI on the load side of the line will be cut off simultaneously, so the phenomenon of Alive messages not reaching the AMI server can be utilized to estimate the expected downed line section and provide an alert to the user.

Figure 2 is a block diagram illustrating a detailed configuration of the fire video surveillance module (200) of Figure 1.

The fire video surveillance module (200) may include an image input buffer (210) for collecting CCTV images of power facilities; an image preprocessing unit (220) for removing noise image elements that generally occur in captured images, such as light reflection noise; an image learning model (230) for detecting the occurrence of fire elements based on machine learning in the preprocessed captured images and identifying power facilities; a primary fire determination unit (240) for primarily determining a fire according to the detected fire elements; a smoke determination unit (250) for applying a motion vector to an image object of the captured images to smoke among the fire elements to determine whether smoke exists; a final fire determination unit (260) for finally determining whether fire exists according to the smoke determination; a power facility location confirmation unit (270) for confirming the locations of power facilities related to the finally determined fire; and an ignition point determination unit (280) for determining the positional relationship between the ignition point of the finally determined fire and the power facilities.

The above image preprocessing unit (220) can remove light reflection noise, diffraction interference patterns, night photography noise, high-sensitivity noise, etc., which are commonly known in landscape photography, etc., by processing them in a way that suppresses high-frequency components or frequency bands for the corresponding noise, for example, based on FFT.

The urban image learning model (230) may include an image object (element) identification unit (232) that identifies fire element image objects such as flames and smoke, and identifies power facility image objects such as poles, wires, transformers, and insulators; an image object DB preprocessing unit (234) that performs preprocessing for data storage for the image objects; and a learning unit (236) that performs learning for identifying the image objects based on deep learning.

Depending on the implementation, the image object (element) identification unit (232) can obtain installation location information of the corresponding power facility from the DB for the power facility and match it to the identified power facility image object.

For example, the image object DB preprocessing unit (234) can perform labeling, class classification, box coordinate extraction, and image augmentation for fire images.

For example, the learning unit (236) can perform learning to identify image objects of fire elements such as flames and smoke, and image objects of power facilities such as electric poles, wires, transformers, and insulators by applying a deep learning module for image object recognition based on YOLOV4.

The above first fire judgment unit (240) is for immediate fire judgment from real-time input shooting images. For example, the fire judgment determined by the first fire judgment unit (240) can only be visually output as a caution on a screen for the manager. This is in accordance with the policy of already giving an alarm at the caution level, considering that it takes time to utilize the motion vector described later.

The above smoke determination unit (250) can remove fog and clouds with characteristics similar to smoke by using the motion vectors of image objects identified based on Optical Flow. Specifically, smoke is dominated by upward motion vectors and/or motion vectors according to wind direction, and in the case of clouds, which are objects at considerable distances, horizontal motion vectors are dominant, and in the case of fog, motion vectors are not clearly revealed. According to these characteristics, clouds and fog with characteristics similar to smoke can be removed. Of course, smoke is the only fire element among them.

Depending on the implementation, the smoke judgment unit (250) can perform processing that can distinguish flames using motion vectors.

For example, the final fire judgment unit (260) can remove fire elements (flames, smoke) judged by the first fire judgment unit (240) that match those judged as clouds or fog by the smoke judgment unit (250).

The above power facility location confirmation unit (270) can confirm the type of each facility of adjacent power facilities and the distance from the fire element determined by the final fire judgment unit (260).

The above ignition point determination unit (280) can determine through image analysis whether the fire element determined by the final fire determination unit (260) originated from the power facility or whether it is a forest fire caused by another cause near the power facility. To this end, the type of each facility confirmed by the power facility location confirmation unit (270) and the distance to the fire element can be reflected.

Meanwhile, in Fig. 2, a weather learning model (650) is illustrated together. It is advantageous to implement the weather learning model (650) to belong to the fire hazard warning device (600) rather than to the fire image surveillance module (200). However, for other purposes, it may be implemented to belong to the fire image surveillance module (200).

The urban weather learning model (650) may include a weather information collection unit (652) that collects weather information (such as measured values and weather conditions) when a fire occurs; a weather DB preprocessing unit (654) that performs preprocessing (e.g., labeling) for data storage of the collected weather information; and a learning unit (236) that performs learning to calculate the probability of fire occurrence in a situation according to the weather information.

Figure 3 is a block diagram showing the detailed configuration of the short-circuit monitoring module (400) and fire hazard alarm device (600) of Figure 1.

The city-based short-circuit monitoring module (400) includes an AMI instrument information collection unit (420) that collects status information from the AMI instruments in charge, and a short-circuit location determination unit (440) that determines the short-circuit location from the status information of the AMI instruments.

The fire hazard warning device (600) may include a weather learning model (650) that calculates the probability of fire occurrence according to weather conditions at the time of fire occurrence; an alarm information generation unit (660) that generates alarm information for an administrator by combining the judgment result of the short-circuit monitoring module (400), the judgment result of the fire image monitoring module (200), and the probability of fire occurrence; and an administrator interface (680) that outputs the alarm information to the administrator.

For example, the AMI instrument information collection unit (620) can receive Alive messages, which are status information of a modem equipped in each AMI instrument, periodically (e.g., every 30 seconds), and accumulate and record the reception results. The purpose of accumulating and recording is to recognize a power outage by accumulating the number of times the Alive message fails to arrive, for example, in the event of a power outage due to a power line disconnection, by accumulating the number of times the message transmission fails to arrive, by 5 times.

In addition, the AMI instrument information collection unit (620) can set the Alive message transmission cycle and the number of accumulated counts for power outage recognition according to the administrator's designation through the administrator interface (680).

The method for determining the short-circuit position performed by the short-circuit position determining unit (640) will be described later.

For example, if the first fire judgment unit (240) of FIG. 2 judges a fire, the alarm information generation unit (660) displays the border of the captured image of the CCTV judged as the first fire in a color on a screen that displays multiple CCTVs together, and if the final fire judgment unit (260) of FIG. 2 judges a fire, the captured image of the CCTV judged as the final fire can be switched to the full screen, the fire location and related information can be displayed, and an alarm can also be output by voice.

The above-mentioned administrator interface (680) outputs visual/audible alarms to the administrator according to the alarm information generated by the above-mentioned alarm information generation unit (660), and receives instructions for various settings from the administrator.

Table 1 below shows an example of implementing a fire hazard alarm module for a power line short-circuit section.

The above fire hazard warning device (600) is a power line disconnection section fire hazard warning module, and allows a power system operator to comprehensively determine whether a fire has occurred and whether a power line has been disconnected within the monitoring section by identifying information collected from an artificial intelligence fire image monitoring module and an AMI power line disconnection monitoring module. For example, the warning information generating unit (660) can generate a screen containing information on fire factors, proximity to power equipment, and power line disconnection occurrence section information, as shown in Table 1, and display the screen on the administrator interface (680) to collectively provide the screen to the administrator.

For example, a short circuit accident occurred in the A-phase power line at Gangwon D/L 123 in the CCTV video surveillance area, and a spark occurred due to an arc at the short circuited power line. The artificial intelligence fire video surveillance module recognizes the fire, calculates the proximity to the power facility, and immediately sends a fire element alarm to the system operation manager. Since all AMI power meters on the low-voltage line on the load side of the short circuit section of A-phase experience a power outage, the power line short circuit location can be estimated based on the AMI power outage information and provided to the manager (operator).

Figure 4 is a flow chart illustrating a method for analyzing the cause of a fire in a power facility according to the invention.

A method for analyzing the cause of a fire in an urban power facility may include the steps of: recognizing a fire near the power facility from a surveillance video (S100); checking for a short-circuit or phase failure alarm within a DAS section for the power facility (S210), and determining that there is an alarm as a high-voltage power line short-circuit (S220); checking the operation status of AMI low-voltage meters within a range affected by the power facility if there is no short-circuit or phase failure alarm within the DAS section (S230); determining that there is a low-voltage power line short-circuit if there is a confirmation failure among the AMI low-voltage meters (S240); and determining that there is a fire caused by a cause other than a power line short-circuit if there is no confirmation failure among the AMI low-voltage meters (S250).

According to the implementation, the step (S100) of recognizing a fire from the surveillance video may include the steps of: detecting a fire element (flame, smoke, etc.) in the surveillance video of the power facility (S110); calculating a distance between the power facility and the generated fire element (S140); and sending an alarm according to the distance (S160). At this time, if the fire element is not detected, the surveillance state is continuously maintained (S120).

Depending on the implementation, after the step of determining the low-voltage power line disconnection (S250), the step of confirming the grid connection location and power outage scale of the AMI low-voltage instrument that failed the confirmation (S260); and the step of estimating the disconnection section based on the grid connection location and power outage scale (S270).

The operation status of the above AMI low-voltage meters can be performed by checking the Alive message of the AMI meter, for example, determining that the operation status is abnormal if the Alive message of a specific AMI meter is not received five or more consecutive times. In this case, in the step S260, if the Alive message is not confirmed only in one or two specific AMI low-voltage meters, i.e., if the operation status is abnormal, there is a high possibility that it is due to a problem with the meter or the communication line, but if the Alive messages of all AMI low-voltage meters belonging to a specific partial area are not confirmed (i.e., if it is a state similar to a large-scale power outage), there is a high possibility that it is due to a low-voltage line disconnection accident in the specific partial area, and a determination can be made based on this possibility.

The step (S100) of recognizing a fire from the above surveillance video can be performed as a first step immediately upon occurrence of the incident using the captured video received in real time, and the step (S270) of estimating the possible short-circuit section after performing the steps S230 to S260, which are completed by counting the number of unconfirmed Alive messages of the AMI device for a predetermined number of times (time), can be performed as a final judgment.

Among the analysis methods for the cause of fire in power equipment in the illustrated flow diagram, steps S110 to S160 may be performed in the fire image surveillance module (200) of FIG. 1, and steps S210 to S270 may be performed in the short-circuit location determination unit (640) of FIG. 3.

FIG. 5 is a conceptual diagram illustrating the advantages of implementing a fire surveillance system according to the invention in a field where a CCTV system for surveillance has already been constructed.

In the conventional distribution operation system (AS-IS), the system operation manager must monitor and operate the system using the DAS system in a 24-hour shift, so the CCTV forest fire surveillance system that requires visual confirmation increases the workload of the staff in charge and reduces effectiveness due to the difficulty of continuous surveillance. In such a situation, fires that occur in vulnerable areas such as mountainous areas, strong winds, and dry areas can develop into large forest fires and cause massive damage if early detection and initial response are delayed.

By building a fire surveillance system (TO-BE) based on this proposed technology, it automatically detects fire elements on CCTV screens using deep learning technology and sends out an alarm, thereby reducing the CCTV surveillance workload of system operation managers and improving the effectiveness of surveillance work. Accordingly, follow-up measures such as preemptive shutdown of the distribution system where an accident occurred, on-site dispatch instructions, and fire reporting can be quickly implemented, efficiently preventing the spread of fire damage.

*The above description is merely an illustrative description of the technical idea of the present invention, and those skilled in the art will appreciate that various modifications and variations may be made without departing from the essential characteristics of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain it, and the scope of the technical idea of the present invention is not limited by these embodiments. The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within a scope equivalent thereto should be interpreted as being included in the scope of the rights of the present invention.

200 : Fire video surveillance module 210 : Video input buffer
220: Image preprocessing unit 230: Image learning model
240: 1st fire judgment section 250: Smoke judgment section
260: Final fire judgment department 270: Power equipment location confirmation department
280: Ignition point determination unit 400: Short circuit monitoring module
420: AMI instrument information collection unit 440: Short circuit location determination unit
600 : Fire hazard alarm device 650 : Weather learning model
660: Alarm information generation unit 680: Administrator interface
1000 : Power Plant Fire Monitoring System

Claims (3)

This is a fire surveillance system that monitors and analyzes whether a fire occurs due to a short circuit in a power line using a deep learning model on video.
A fire video surveillance module that determines whether a fire has occurred and its relevance to the power facility from video footage of the power facility and its surrounding area;
A short circuit monitoring module for monitoring the short circuit status of the above power equipment; and
A fire hazard alarm device that determines whether a fire has occurred and determines fire-related risk factors for the power facility from the short-circuit status, and warns of fire-related risk factors for the power facility
Including, but not limited to,
The above single line monitoring module,
A method for periodically checking the operation of AMI devices within the range affected by the above power facility to determine whether there is a short circuit related to the above power facility.
When the number of transmission failures for the Alive message, which is a small-capacity modem status information that the above AMI instruments transmit to the AMI server at a specified transmission cycle (tens of seconds) separately from the metering data transmission cycle (15 minutes for high pressure and 60 minutes for low pressure), accumulates to a critical number and is counted, it is recognized as a disconnection.
The above fire hazard alarm device,
A short-circuit fire monitoring system for power equipment, wherein the fire is presumed to be caused by a short-circuit if the time at which the fire image detection module detects a fire in the power equipment and the time at which the short-circuit monitoring module detects a short-circuit condition are similar.
In the first paragraph,
The above fire hazard alarm device, if a fire is suspected to be caused by the above short circuit,
A step of checking for a short circuit or phase failure alarm within the DAS section for the above power facility, and determining that there is an alarm as a short circuit of a high-voltage power line;
If there is no open circuit or phase alarm in the above DAS section, a step of checking the operating status of AMI low-voltage meters within the range affected by the power facility;
If there is a failure in verification among the above AMI low-voltage meters, a step for determining that there is a low-voltage power line disconnection; and
If none of the above AMI low-voltage devices fail to be verified, then it is time to determine that the fire is due to a cause other than a power line disconnection.
A short-circuit fire monitoring system for power equipment performing a short-circuit analysis method as a fire cause including.
In the second paragraph,
The method of short-circuit analysis as the cause of the above fire is as follows:
After the step of determining that the above low-voltage power line is disconnected,
Step for checking the grid connection location and power outage scale of the AMI low-voltage instrument that failed the above verification; and
Step for estimating the disconnection section based on the above-mentioned grid connection location and power outage scale
Single-wire fire surveillance system for power installations including: