KR102283884B1 - Artificial intelligence based fire and power anomaly prediction system and method - Google Patents

Abstract

An artificial intelligence based fire and power anomaly prediction system according to an embodiment of the present invention includes a smart EMS device that collects sensing data for checking power anomaly including at least one of overcurrent, effective arc and temperature state through a plurality of sensors and transmits the collected sensing data to an integrated server; and an integrated server that stores the collected sensing data to make them big data, classifies the stored sensing data by clustering according to data properties, and predicts whether a fire or arc occurs or whether there is a power anomaly through machine learning.

Description

Artificial intelligence based fire and power anomaly prediction system and method

The present invention relates to an electric fire prediction and predictive maintenance system, and more particularly, based on artificial intelligence for predicting power abnormalities such as effective arc, overcurrent, leakage current, etc. that can affect fire based on artificial intelligence in advance It relates to an electric fire prediction and predictive maintenance system.

In apartments, buildings, schools, factories, ports, airports, commercial and sewage treatment plants, substations, heavy industry plants, subways, chemical complexes, steel mills, commercial areas, traditional markets, etc. is needed

In particular, when a fire occurs in a facility that supplies electricity (eg, various power plants, switchboards, transformers, etc.), it becomes impossible to supply electricity to industrial complexes, factories, and homes, and enormous property damage is expected.

Accordingly, an overload circuit breaker or an earth leakage circuit breaker is installed as a power supply safety device for various buildings and facilities, and when an abnormality occurs in a power line or an electric device, the power supply is quickly cut off.

For this reason, although the development of a system for monitoring such a fire prevention device is actively being made, most of it is only made in a reactive manner depending only on actual data, so effective arc and overcurrent that can affect the fire based on artificial intelligence , there is a need to develop a system for predicting power abnormalities such as leakage current in advance.

Furthermore, in order to monitor power for fire prevention, we are aware of the problems that occur when using electric energy in the process of judging whether there is a power leak or power abnormality, but there is no standard data that can be an accurate energy management index, so power improvement plan The reality is that it is difficult to find.

Therefore, detailed electrical data, monitoring and analysis must be performed in real time to find a power improvement plan, and when an abnormality occurs, various sensors, IoT communication equipment, and safety management platform for collecting electrical data so that the relevant information can be immediately notified to the manager and related parties There is a need for an artificial intelligence-based electric fire prediction and predictive maintenance system composed of

Republic of Korea Patent Publication No. 10-1272665 (registered on June 03, 2013)

It is an object of the present invention to provide an artificial intelligence-based electric fire prediction and predictive maintenance system and method for predicting in advance whether there is a power abnormality such as an effective arc, overcurrent, leakage current, etc. that can affect a fire based on artificial intelligence .

The artificial intelligence-based electric fire prediction and predictive maintenance system according to an embodiment of the present invention collects sensing data capable of checking power abnormalities including at least one of overcurrent, effective arc and temperature state through a plurality of sensors and a smart EMS device that transmits the collected sensing data to the integrated server; An integrated server that stores the collected sensing data to make it big data, classifies the stored sensing data by clustering according to data properties, and predicts whether a fire or arc occurs or whether there is a power abnormality through machine learning.

The smart EMS device detects current, leakage current, breaker on/off state, phase-specific current/voltage/power, single-phase current/voltage/power, power factor, harmonics, accumulated power or active/reactive/apparent power, and the sensing data a power analyzer that generates; an arc analyzer that detects whether an effective arc is generated by analyzing a resistive component, an inductive component, and a capacitive component of a load receiving power, and generates sensing data using this; It further includes a temperature sensor for generating sensed data by sensing the ambient temperature of the load to which power is supplied.

The artificial intelligence-based electric fire prediction and predictive maintenance system is provided individually for each region to manage a small group in each region where the smart EMS device is located, and power control and monitoring by the integrated server linked to a plurality of local platform servers a local platform server that performs functions; further includes

The integrated server is characterized in that it performs clustering by a K-means algorithm in order to classify the sensed data of similar properties among the sensed data.

The integrated server is characterized in that based on whether the sensed data is greater than or equal to a set threshold value, predicts whether an arc or fire occurs or whether there is an electric power failure.

The learning model is characterized in that it is an anomaly detection learning model generated through unsupervised learning that does not inform an abnormal state by numerically setting a threshold value using the sensing data in a normal state and the change value of the currently collected sensing data.

The learning model collects sensing data in a normal state even if there is no initial threshold as a reference, and sets the threshold by generating a sample of sensing data in an abnormal state in contrast to this.

The integrated server is characterized in that it is possible to predict the subject of the arc, fire, or power abnormality based on the components of power included in the sensing data collected through the arc analyzer.

The smart EMS device provides an external alarm when a fire or arc occurs or a power abnormality occurs, and the external alarm is transmitted to a manager terminal of a store where a fire or arc occurs or a power abnormality occurs, and to a private management company or a management authority. do.

The artificial intelligence-based electric fire prediction and predictive maintenance method using the artificial intelligence-based electric fire prediction and predictive maintenance system checks the power abnormality including at least one of overcurrent, effective arc and temperature state through a plurality of sensors. Collecting possible sensing data; transmitting the collected sensing data to the local platform server and the integrated server; Storing the collected sensing data into big data, and classifying the stored sensing data by clustering according to data attributes; Predicting whether a fire or arc has occurred or whether power is abnormal through machine learning; setting an energy safety management policy for fire prevention by analyzing the prediction result; and providing an external alarm in case of fire or arc occurrence or power failure.

The artificial intelligence-based electric fire prediction and predictive maintenance system and method of the present invention collects various sensing data from a load to which power is supplied, and predicts electric fire or arc occurrence or power abnormality based on artificial intelligence in advance to prevent accidents and It has the advantage of being able to respond quickly and predictably and preserve it.

In addition, it has the advantage of being able to build various sensors, IoT communication equipment, and safety management platform for collecting electrical data so that detailed electrical data, monitoring and analysis are performed periodically, and related contents can be immediately notified to managers and related parties. .

In addition, in order to monitor electric power for fire prevention, problems that occur when using electric energy in the process of determining whether there is an electric leak or electric power abnormality are monitored through artificial intelligence-based prediction results, and reference data that serves as an accurate energy management index As such, there is an advantage in that an energy management policy can be set to prepare a power improvement plan.

1 is a block diagram showing the overall configuration of an artificial intelligence-based electric fire prediction and predictive maintenance system according to an embodiment of the present invention.
2 is a diagram for explaining in detail an example of the monitoring and alarming process of the arc analyzer.
3 is a flowchart illustrating a process of an artificial intelligence-based electric fire prediction and predictive maintenance method according to an embodiment of the present invention.
4 and 5 are diagrams exemplarily showing sensing data of a power analyzer.
6 to 8 are diagrams exemplarily showing sensing data of an arc detector.
9 and 10 are diagrams exemplarily showing a facility management control screen.

Hereinafter, specific embodiments of the present invention will be described in detail with reference to the drawings. However, the spirit of the present invention is not limited to the embodiments presented, and those skilled in the art who understand the spirit of the present invention may add, change, delete, etc. other components within the scope of the same spirit, through addition, change, deletion, etc. Other embodiments included within the scope of the invention may be easily proposed, but this will also be included within the scope of the invention. In addition, components having the same function within the scope of the same idea shown in the drawings of each embodiment will be described using the same reference numerals.

1 is a block diagram showing the overall configuration of an artificial intelligence-based electric fire prediction and predictive maintenance system according to an embodiment of the present invention.

The artificial intelligence-based electric fire prediction and predictive maintenance system of the present invention may include a smart EMS device 100 , an integrated server 200 , and a local platform server 300 .

The smart EMS device 100 collects sensing data from the sensing device 110 provided therein, and transmits the sensing data to the integrated server 200 of the Internet network through the local platform server 300 installed in the region.

The sensing device 110 may be provided with a plurality of sensors for collecting various kinds of sensing data internally, and a power analyzer 111, an arc analyzer 112, gas to check overcurrent, effective arc and temperature state, etc. A detector 113 , a smoke detector 114 , and a temperature sensor 115 may be included.

The power analyzer 111 may include an overcurrent sensor, a circuit breaker detection sensor, and the like, and using them, referring to FIGS. 4 and 5 , overcurrent, leakage current, breaker on/off state, phase-specific current/voltage/power, single-phase current Sensing data can be generated by detecting /voltage/power, power factor, harmonics, integrated power, and active/reactive/apparent power.

In addition, the power analyzer 111 has built-in wired and wireless communication functions such as Ethernet, RS485, LoRa, etc., and can directly transmit data to the integrated server 200 or the local platform server 300, and the sensing data collection cycle is For example, it may be periodically collected in a time unit of 10 to 30 seconds.

The arc analyzer 112 includes a load circuit for analyzing the component of the load supplied with power, and analyzes the resistive component (R), the inductive component (L), and the capacitive component (C) of the load to determine whether an effective arc is generated. It detects and generates sensing data using it. Arcs can be divided into effective arcs that cause electrical defects such as fire or deterioration and ineffective arcs that are not related to electrical defects. .

In addition, referring to FIG. 8, to determine the effective arc, a special high-frequency sensor for arc frequency detection is used, an RLC resonance circuit is used, a voltage or current arc pulse synchronization circuit is used, a low-frequency filter circuit is used, or an oscillation frequency generation combination circuit is used. can also be used.

Furthermore, the arc shape can be detected in the form of a fine electric spark or in the form of a complex parallel arc, and the complex parallel arc is an arc that can occur before the electric wire is completely short-circuited. .

In addition, the arc analyzer 112 may be installed inside the distribution panel with reference to FIG. 2, and is disposed inside the distribution panel to detect the temperature inside the distribution panel, as well as detection data for overcurrent and leakage current, and detection of wire temperature can do.

At this time, current data is collected in real time to determine whether there is overcurrent or leakage current, and the allowable current compared to the wire and the allowable current compared to the circuit breaker can be analyzed. Make it possible to check whether the wire temperature is detected above the allowable or not.

In addition, the sensing data sensed by the arc analyzer 112 is transmitted through a gateway connected to the local platform server or the integrated server in the local network (eg, wireless communication self-network), and in the integrated server, the safety manager's management terminal (not shown) or When sensing data is detected above a threshold value or predicted by learning to the operating PC (not shown), a warning pop-up message (text message) may be transmitted or location information about the point of occurrence may be transmitted.

The gas detector 113 includes a gas detection sensor to generate sensing data for detecting gas leakage, and the smoke detector 114 includes a smoke detection sensor to additionally detect and sense whether a fire has occurred through smoke. create data

The temperature sensor 115 generates sensing data by sensing the ambient temperature of a load to which power is supplied, and exemplary items of the sensing data collected in relation to the temperature are shown in FIG. 7 .

Furthermore, the sensing data collected by the sensing device 110 is continuously collected in a time period, and the signal strength (magnitude) may be collected in a period or by performing a root mean square (RMS) operation in a time period to collect the sensing data. In addition, when this period is received by the integrated server 200, the integrated server 200 or the local platform server 300 can be synchronized with the alarm cycle when predicting whether a fire or arc occurs or whether there is a power abnormality. there is.

In addition, the smart EMS device 100 provides an external alarm when a fire, arc generation or power abnormality occurs, and as a target for providing an alarm, not only the manager terminal (PC, mobile terminal, etc.) of the store where the fire, arc generation or power abnormality occurs. Instead, it may be transmitted in the form of a pop-up message or an alert to a private management company or a management authority such as a fire department, police station, city hall, etc.

Furthermore, the smart EMS device 100 temporarily stores the sensing data collected at that time in a buffer together with the sensing data between a specific time period in the past based on timestamp information when a fire, arc, or power abnormality occurs. It can be used to manage and identify the location of occurrence of fires and power abnormalities that occurred later or who is responsible.

The integrated server 200 receives all the sensing data collected from the sensing device 110 through the local platform server 300 and stores it in a database to make it big data, and classifies the stored sensing data by clustering it according to data properties, It is possible to predict the occurrence of a fire through machine learning.

In particular, in the clustering process, clustering is performed by the K-means algorithm to classify the sensed data with similar properties, and the K-means algorithm is an algorithm that groups the given data into k clusters. and is operated in such a way that the variance of the distance difference is minimized. Similar properties may vary depending on electrical characteristics (voltage, current, resistance, power, ambient temperature, etc.), such as current anomaly (overcurrent, leakage current, etc.), power anomaly (active, reactive, apparent power, power factor, etc.), arc It can be divided into abnormalities (effective arc, ambient temperature, etc.).

The average can be defined and calculated in a set of sensing data using the K-means algorithm, and data that is abnormally farther from the centroid than other data points are removed by attribute pattern recognition of the collected sensing data. can be clustered.

Sensing data classified among similar properties is learned by using a learning model based on a machine learning algorithm for each property, and as a result, whether an arc or fire occurs, or whether there is a power abnormality is predicted.

In the case of the above-described machine learning algorithm, a deep neural network (DNN) may be used, or various types of machine learning algorithms may be used, such as a convolutional neural network (CNN) or a recurrent neural network (RNN) method. For example, in the present invention, a support vector machine (SVM) technique, which will be described later, may be used as a prediction technique for accurately discriminating an abnormal pattern of sensing data for each attribute.

The SVM technique is one of supervised machine learning methods mainly used for classification, regression, and outliers detection. For example, among the various methods for classifying the datasets of two groups, the best way to increase the classification accuracy is to be able to accurately identify the midpoint at the maximum distance for each group.

In particular, SVM is known as an optimized method for finding an optimal decision boundary capable of distinguishing a plurality of dimensions for data having a plurality of dimensions.

In addition, when applying the SVM technique as described above to the sensing data obtained from the sensors, as the sensed data is accumulated more and more, the number of machine learning training increases, and as a result, the accuracy of the modeling obtained through the training gradually increases. do. This characteristic is that the prediction error rate can be improved more and more as the number of training increases compared to the analysis technique using a specific mathematical or statistical modeling always has a certain degree of prediction error rate, which is a major factor in modeling through machine learning techniques. advantage.

In addition, in order to predict whether an arc or fire occurs or whether power is abnormal, it may be made based on whether the sensed data is greater than or equal to a set threshold, and the threshold may be determined or changed by learning of the learning model.

For example, when the collected sensing data is sensing data for overcurrent, the overcurrent condition is periodically checked based on the overcurrent compared to the capacity of the breaker to predict whether an arc or fire or power abnormality occurs according to the overcurrent condition.

Similarly, leakage current, circuit breaker on/off state, phase-specific current/voltage/power, single-phase current/voltage/power, power factor, and harmonics can be predicted by learning from the learning model based on the collected sensing data. .

The learning model can generate an anomaly detection learning model through unsupervised learning that does not inform an abnormal state by numerically setting a threshold value using the sensing data in the normal state and the change value of the currently collected sensing data.

In addition, the learning model collects sensing data in a normal state even if there is no initial threshold as a reference, and generates a sample of sensing data in an abnormal state in contrast to this to determine the threshold.

In addition, the integrated server 200 may predict the occurrence of an arc or fire based on the sensing data, or whether there is a power abnormality, while predicting the occurrence.

Specifically, referring to FIG. 6 , a specific component value is set using the resistive component (R), inductive component (L) or capacitive component (C) value of the power included in the sensing data collected through the arc analyzer 112 . When the threshold is sensed or more, or by using the abnormality of a specific frequency band of the load voltage or current, the occurrence entity can be predicted. It can be predicted that this is the subject of its occurrence.

In addition, when the inductive component is greater than or equal to the threshold value, it can be predicted that a fan, a fan, a vacuum cleaner, a dishwasher, a washing machine, a refrigerator, and an air conditioner are the generating agents.

In addition, even when a complex component among the resistance component, the inductive component, or the capacity component is above a threshold value, the cause of the occurrence can be predicted.

In addition, the integrated server 200 may generate an alarm event immediately according to the prediction result, and transmit the alarm event to the manager terminal of the place where the power abnormality occurred or the place including the generating subject.

In addition, the alarm event provided according to the prediction result may be made in stages according to the level at which the prediction result is digitized, for example, may be subdivided into 'foreign', 'caution', and 'warning' stages. Here, the precursor stage may mean a stage in which a fire is unlikely to occur, the caution stage is a stage in which a fire is highly likely, and the warning stage may refer to a stage in which a fire is extremely likely to occur.

In addition, the integrated server 200 can analyze the prediction result to set an energy safety management policy for fire prevention, and the set policy is transmitted to the smart EMS device 100 installed in each region through the local platform server 300. can be managed The smart EMS device 100 may perform power control based on the energy safety management policy.

The local platform server 300 may be provided as a local platform for managing a small group in each region where the smart EMS device 100 is located, and is powered by an integrated server 200 linked to a plurality of local platform servers 300 . Make sure that control and monitoring are done systematically.

To this end, the local platform server 300 has a built-in Internet communication protocol interlocked with a connection device (communication protocol) interlocked with the smart EMS device 100 and the integrated server 200, and the smart EMS device 100 and the local communication Data transmission/reception is performed by this, and data transmission/reception is performed with the integrated server 200 through an Internet network or the like.

In addition, the local platform server 300 may perform the same function as the integrated server 200, for example, by collecting sensing data and determining whether or not an arc or fire has occurred or power by determining the sensing data above a reference threshold based on an artificial intelligence algorithm. It is possible to predict whether there is an abnormality, to create a learning model, and to improve the prediction rate through learning.

In addition, the local platform server 300 analyzes the prediction result, sets an energy safety management policy for fire prevention, or receives the energy safety management policy set from the integrated server 200 and installs the smart EMS device 100 in each region. ), a policy is provided based on the policy, and power control may be performed through the monitoring screen as shown in FIGS. 9 and 10 .

In addition, information (eg, sensing data) exchanged between the smart EMS device 100, the integrated server 200, and the local platform server 300 is sensed to protect information from risks such as hacking from the outside. Data encryption/decryption technology can be applied to data transmission/reception. More specifically, the integrated server 200 grants identification information, each capable of identification, to the smart EMS device 100 and the local platform server 300, each smart EMS device 100, the local platform A lightweight encryption algorithm using the identification information of the server 300 as a private key is performed. The lightweight encryption algorithm is an encryption technology designed to be implemented in limited environments such as smart devices such as management terminals. The symmetric key encryption algorithm HIGHT (HIGh security and liht weigHT), LEA (Lightweight Encryption), and the hash function LSH (Lightweight Secure) Hash) can be used. By using this lightweight encryption algorithm to encrypt/decrypt sensing data, etc., it is possible to prevent illegal control due to external leakage or external hacking of the corresponding data.

3 is a flowchart illustrating a process of an artificial intelligence-based electric fire prediction and predictive maintenance method according to an embodiment of the present invention.

First, sensing data capable of checking power abnormality including at least one of overcurrent, effective arc, and temperature state is collected through a plurality of sensors (S10).

The collected sensing data is transmitted to the local platform server 300 and the integrated server 200 (S20).

Thereafter, the local platform server 300 or the integrated server 200 stores the collected sensing data to make it big data, and classifies the stored sensing data by clustering it according to data properties (S30).

In addition, it predicts whether a fire or arc occurs or whether there is a power abnormality through machine learning (S40).

Also, by analyzing the prediction result, an energy safety management policy for fire prevention is set (S50).

Finally, when a fire or arc occurs or power is abnormal, an external alarm is provided to the relevant private management company or management authority (S60).

100 ; Smart EMS device
110 ; sensing device
111; power analyzer
112 ; arc analyzer
113 ; gas detector
114; smoke detector
115 ; temperature sensor
200 ; integrated server
300 ; local platform server

Claims (10)

A smart EMS device that collects sensing data capable of checking power abnormality including at least one of overcurrent, effective arc and temperature state through a plurality of sensors, and transmits the collected sensing data to an integrated server;
an integrated server that stores the collected sensing data to make it big data, classifies the stored sensing data by clustering according to data properties, and predicts whether a fire or arc occurs or whether power is abnormal through machine learning;
A local platform server that is individually provided for each region to manage a small group in each region where the smart EMS device is located, and performs power control and monitoring functions by the integrated server linked to a plurality of local platform servers;
The smart EMS device,
A power analyzer that detects current, leakage current, breaker on/off state, phase-specific current/voltage/power, single-phase current/voltage/power, power factor, harmonics, accumulated power or active/reactive/apparent power to generate the sensed data;
an arc analyzer that detects whether an effective arc is generated by analyzing a resistive component, an inductive component, and a capacitive component of a load receiving power, and generates sensing data using this;
Further comprising a temperature sensor for generating sensed data by sensing the ambient temperature of the load to which power is supplied,
The integrated server
Predicting whether an arc or fire or power abnormality occurs based on whether the sensing data is above a set threshold value through learning using a learning model based on a machine learning algorithm,
The learning model is an anomaly detection learning model generated through unsupervised learning that does not inform an abnormal state by numerically setting a threshold value using the sensing data in a normal state and the change value of the currently collected sensing data,
The integrated server gives identification information, each capable of identification, to the smart EMS device and the local platform server, and uses the identification information of each smart EMS device and the local platform server as a private key. It performs a lightweight cryptographic algorithm that
The lightweight encryption algorithm is based on artificial intelligence, characterized in that encryption/decryption is performed using HIGHT (HIGH security and light weigHT), LEA (Lightweight Encryption), and LSH (Lightweight Secure Hash), which are symmetric key encryption algorithms. Electric fire prediction and predictive maintenance system. delete delete According to claim 1,
The integrated server
An artificial intelligence-based electric fire prediction and predictive maintenance system, characterized in that clustering is performed by a K-means algorithm to classify the sensed data of similar properties among the sensed data. delete delete According to claim 1,
The learning model collects sensing data in a normal state even if there is no initial threshold as a reference, and generates a sample of sensing data in an abnormal state in contrast to this and sets the threshold value. conservation system. According to claim 1,
The integrated server
An artificial intelligence-based electric fire prediction and predictive maintenance system, characterized in that it is possible to predict an arc or fire or power anomaly based on the components of power included in the sensing data collected through the arc analyzer. According to claim 1,
The smart EMS device provides an external alarm when a fire or arc occurs or an electric power abnormality occurs,
The external alarm is an artificial intelligence-based electric fire prediction and predictive maintenance system, characterized in that it is transmitted to the manager terminal of the store where a fire or arc occurs, or a power failure occurs, and to a private management company or management authority. In the artificial intelligence-based electric fire prediction and predictive maintenance method using the artificial intelligence-based electric fire prediction and predictive maintenance system of claim 1, 4, 7, 8 or 9,
Collecting sensing data capable of checking power abnormality including at least one of overcurrent, effective arc, and temperature state through a plurality of sensors;
transmitting the collected sensing data to the local platform server and the integrated server;
Storing the collected sensing data into big data, and classifying the stored sensing data by clustering according to data attributes;
Predicting whether a fire or arc has occurred or whether power is abnormal through machine learning;
setting an energy safety management policy for fire prevention by analyzing the prediction result;
Artificial intelligence-based electric fire prediction and predictive maintenance method comprising the step of providing an external alarm in case of fire or arc occurrence or power failure.

Images