WO2017018726A1 - Sensor signal detecting system and smart plug having same - Google Patents

Abstract

The present invention relates to a sensor signal detecting system and a smart plug having same, the system capable of differentiating various types of anomalous signals in real time and operating even with a small resource. The sensor signal detecting system of the present invention comprises both: an anomaly processing unit for primarily determining anomalies in signal data over a short term; and an anomaly determining unit for secondarily determining anomalies over a long term by calculating signal values, and can efficiently detect various types of anomalous signals. The sensor signal detecting system of the present invention can automatically create a normal pattern model and determine whether a model calculated from a newly acquired signal is the same pattern, and determine abnormality in real time. The smart plug of the present invention can prevent a delay by relaying a signal detected in real time or close to real time through another gateway. In addition, the smart plug of the present invention has a sensor signal detecting system including a sensor; a gateway; a USB charger; and a multi tab configured as a single product; so as to have the effect of reducing costs associated with increased transmission time, space, and product dimensions.

Description

Sensor signal detection system and smart plug having same

The present invention relates to a sensor signal detection system and a smart plug having the same. More particularly, the present invention relates to a sensor signal detection system and a smart plug including the same, capable of determining various kinds of abnormal signals in real time and operating with a small resource. It is about.

With the advent of the smart home and the Internet of Things (IOT) era, the importance of wireless sensor systems is emerging. Automation systems, which are becoming important in sensor systems, must clearly recognize the anomalies of the data or interesting parts of the collected data.

In addition, these wireless sensor systems must be able to derive various kinds of anomalies from signals from the sensors, must be robust, require relatively small resources, and can be used in real time or near real time. The signal must be detected.

Japanese Patent Laid-Open No. 20004-126732 discloses a temperature control device for controlling temperature in various temperature environments and the like. The temperature control device disclosed in the above-described patent discloses a control output (MW value) to be supplied to a heating / cooling device based on a temperature (PV value) and a set temperature (SP value) of an object detected by a temperature sensor such as a thermocouple of a resistance thermometer. Control the temperature of the object, while providing a numerical indication of the detected temperature.

The sensor system of the disclosed patent is a very simple method of controlling the device by determining that it is an abnormal state only when the temperature exceeds a set temperature value. That is, in the sensor system of the disclosed patent, it is impossible to detect data that is incorrectly read by the sensor, it is difficult to apply it in a condition where the set temperature value is to be changed in real time, and it is difficult to analyze the data in real time.

The present invention provides a sensor signal detection system that can derive various kinds of abnormalities in real time.

The present invention provides a sensor signal detection system operable with a small resource (Computational complexity, memory).

The present invention provides a smart plug that efficiently combines a sensor signal detection system with a power strip.

The present invention provides an abnormal signal to a user without delay without going through a separate gateway.

In one aspect the invention

One or more sensors;

An abnormality processing unit (slave) for detecting an abnormal signal among the signals received from the sensor;

An abnormality determination unit (master) that calculates a signal for a longer time than a detection time in the abnormal signal unit by receiving the abnormal signal received from the abnormality processing unit and finally determines abnormality; And

It relates to a sensor signal detection system including a wireless transceiver for data communication between the abnormality determination unit and the outside.

In another aspect the invention

A power strip with one or more terminals formed;

The sensor signal detection system located inside the multi-tap;

It is related to a smart plug including a Wi-Fi network gateway module located inside the power strip and connected to an Ethernet port.

The sensor signal detection system of the present invention includes an abnormality processing unit for determining abnormality of signal data for a short time as a first time and an abnormality determining unit for calculating abnormality by calculating a signal value for a long time in a second time. The sex signal can be detected efficiently.

The sensor signal detection system of the present invention automatically generates a normal pattern model, and determines abnormality in real time by determining whether a model calculated from a newly acquired signal is the same pattern.

The sensor detection system of the present invention has more than 90% of false false positives performance from a real life data set, no malfunction within a parameter selection range, and operates with a small resource (for example, a CPU or a memory). Provide a viable system.

The smart plug of the present invention can transmit an abnormality signal detected by the sensor signal detection system to the user through an external internet network without having to go through a separate Wi-Fi gateway by embedding a sensor signal detection system and a network gateway in a power strip. Accordingly, the smart plug of the present invention can prevent a delay from occurring through a process of passing through another gateway to a signal detected in real time or near real time.

In addition, the smart plug of the present invention can reduce the cost by increasing the transmission time, space, product volume by implementing a sensor signal detection system including a sensor, a gateway, a USB charger, a power strip in a single product.

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows the structure of the sensor signal detection system of this invention.

2 shows a hierarchical structure (hierarchical structure) for the signal detection system of the present invention.

3 shows an algorithm of the abnormality processing unit 20.

4 shows an algorithm of the abnormality determining unit 30.

5 shows an experimental result indicating that the abnormality determination unit detects the abnormality of the mean.

FIG. 6 shows that the abnormality determination unit 30 detects and removes a noise among signals received by the abnormality signal unit.

Figure 7 shows the sensor value graph change when the abnormal data is excluded from the reference model.

8 shows a smart plug of the present invention.

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. However, the scope of the present invention is not limited to the description or the drawings for the following embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows the structure of the sensor signal detection system of this invention. Referring to FIG. 1, the sensor signal detection system of the present invention includes a sensor 10, an abnormality processing unit 20, an abnormality determination unit 30, and a wireless transceiver 40.

The one or more sensors 10 may be selected from the group consisting of temperature sensors, humidity sensors, gas detection sensors, power sensors, and dust sensors.

In the present invention, the sensor may be located inside the main module or attached to the outer surface of the multi-tap, but is not necessarily limited thereto.

The abnormality processing unit 20 detects an abnormal signal among the signals received from the sensor. The abnormality processing unit 20 transmits the received sensor signal as well as the detected abnormal signal to the abnormality determining unit.

The abnormality determination unit 30 determines not only the abnormal signal received from the abnormality processing unit, but also a sensor signal for a longer time than a detection time of the abnormal signal unit to determine abnormality.

2 shows a hierarchical structure (hierarchical structure) for the signal detection system of the present invention. That is, the signal detection system of the present invention undergoes two steps of hierarchical signal analysis detection and discrimination, and calculates abnormal signal data detected primarily by the abnormality processing unit having a lower structure in a second order by an abnormality determining unit having a higher structure. The error of the abnormal signal detected by the difference can be eliminated. In addition, the signal detection system of the present invention calculates the signal processing time and data amount differently from the abnormality processing unit 20 and the abnormality determination unit 30 so that various types of abnormal signals can be efficiently detected. .

The abnormality processing unit 20 first configures a lower layer that calculates signal data generated over a short time range (corresponding to a small amount of data).

The abnormality processing unit 20 is provided with a micro control unit (MCU) for a few seconds to several minutes, preferably 10 minutes to 30 minutes of the spike signal value of the detection signal (sequentially the same value) Determine the maintained signal value or linear regression error value.

3 shows an algorithm of the abnormality processing unit 20.

First, data initialization is performed (S10). In the system of the present invention, data initialization is performed as follows. For example, first, the abnormality determination unit 30 performs pattern modeling with data for about 30 minutes after power on. The abnormality processing unit 20 performs pattern modeling with 10-20 data within a few seconds after power is applied.

The abnormality processor performs a rule check (R_ δ_s) for calculating whether the maximum allowable value of the continuous sensor value difference using the variance change value is exceeded (S20). The maximum allowable value can be set by the user.

Through the rule check, a signal that splashes or a signal exceeding a maximum allowable value may be detected in a short time unit.

Subsequently, the abnormality processing unit 20 determines whether the calculated value is greater than the linear regression error rate ε (S30). The linear regression error rate (ε) determination step detects the same value continuously, the value splashing like a spike. It also detects sudden changes in short term patterns.

Subsequently, the abnormality processing unit 20 determines whether the signal value satisfies the linearity range through a model check (S40). The model check can use a known method.

For example, the model check may be a reference model read by the sensor for t hours.

) And the sensor data read for the next T hours (

, t = 1, 2,. . . , T, called the current model value).

More specifically, the model check may model a signal value detected in time series in units of a predetermined time (T) (

), Reference model (

) And the linearity comparison (eg, the change in the shape of the waveform) and the linearity comparison (e.g., the difference in the slope value) can be derived whether or not the tolerance value is larger than the allowable value.

For example, the reference model (

Time series pattern L1 and current model

When the new time series pattern derived by L2 is L2, the difference between the two patterns represents L2 to L2.

Wow

It can be determined by the sample difference value of.

If you look more closely, different Time series

(

) ← {(X [1], Y [1]), (X [2], Y [2]), ...,, (X [m], Y [m])} and the new data

(

) ← {(~ X [1], ~ Y [1]), {~ X [2], ~ Y [2]), ..., (~ X [k], ~ Y [2])} Since this is a linear time series, the a * X [t] + b lines obtained from these two sets of data values can be obtained and judged using the difference between them.

Also, the abnormality processing unit may generate a new reference model when the new sensor data value obtained at a predetermined time T exceeds the linear regression error rate ε.

(Updated new reference model).

The anomaly handling section uses a new reference model (

(New normal pattern model) is continuously updated, and an abnormal signal can be detected or discriminated by determining in real time whether the model calculated from the new reference model and the newly acquired signal is the same pattern.

Where the new reference model (

) Can be obtained by the following formula.

<δ (model standard deviation)

Where Y (x) is the measured sensor value (temperature, etc.) and x is the sample number. The bigger X is, the more time progresses. The linearity determination is described above.

(

),

(

) Can be used to determine linearity.

Linearity is satisfied when the S (D, D) value is less than the standard deviation δ. If larger than this, it can be determined as an abnormal signal.

The abnormality determination unit 30 calculates the abnormal signal received from the abnormality processing unit 20 for a longer time than the detection time of the abnormal signal unit, and finally determines abnormality.

The abnormality determining unit 30 configures an upper layer that collects a larger amount of data than the abnormality processing unit 20, that is, a larger amount of data, and calculates the data based on the abnormality processing unit 20.

The abnormality determining unit 30 includes a central processor unit (CPU) 31 and a storage unit 32, and has the abnormality over a period of several minutes to several hours, and preferably for about 30 minutes to 1 hour. The linear regression error value is determined whether there is a sudden change in the average value of the abnormal signals received from the processor.

4 shows an algorithm of the abnormality determining unit 30. Referring to FIG. 4, the abnormality determination unit 20 performs data initialization (S110), rule check (S120), linear regression error rate (ε) determination (S130), and linearity category determination (S140).

First, the abnormality determination unit 30 calculates an average change of more data for a longer time period from the data of the signal values received from the abnormal processing unit, so that the sensor incorrectly reads the data, that is, the error of the sensor and the abnormal processing unit. Can be detected (S110 rule check). That is, whether there is a sudden change in the average value of the abnormal signals received from the abnormality processor may be determined based on whether the currently measured sensor signal value exceeds the maximum allowable value.

In addition, the abnormality determination unit 30 determines whether the operation value is greater than the linear regression error rate ε (S130).

Subsequently, the abnormality processing unit 20 determines whether the signal value satisfies the linearity range through a model check (S140). For the model check, reference may be made to the above description.

5 shows an experimental result indicating that the abnormality determination unit detects the abnormality of the mean. Referring to FIG. 5, the abnormality determination unit shows that the average abnormality is detected in the range of 3500 to 4300 samples.

Referring to FIG. 6, the linearity of the straight line represented by Y (j) and the straight line represented by Y (j + 1) are almost similar. In other words, the linearity values of these two straight lines are within δ (model standard deviation).

Figure 7 shows the sensor value graph change when the abnormal data is excluded from the reference model. Referring to FIG. 7, removing the abnormal data shows a trajectory of Y (x) values that are almost similar to the original reference model (red). That is, since the sensor signal detection system of the present invention determines after removing the abnormal pattern, it shows that the signal pattern of Figure 7 is normal

Detection categories of the abnormality processing unit and the abnormality determination unit in the present invention are shown in Table 1 below.

Abnormality	Explanation	Detection layer
Average change	Abnormal sensor data usually exhibits a different value in the average of the sensor data.	Abnormality Processing Unit
Variance change	Dispersion	Both an abnormality processing unit and an abnormality determination unit
Short spike	Short fault data type and equality	Abnormality Processing Unit
Constant reading	The sensor reports the same value over time.	Both an abnormality processing unit and an abnormality determination unit
Change in shape	Abnormal values differ in mean and / or variance compared to normal values, but are shorter than mean and variance changes.	Both an abnormality processing unit and an abnormality determination unit

The signal detection system of the present invention includes two signal detection layers and can detect abnormal signals having various lengths, widths, and patterns through feedback therebetween.

The system of the present invention is capable of detecting all kinds of abnormalities (more than about 90%) as well as normal operation (low false negative, false positive rates), and not only changes in the pattern of the data set but also parameter settings. It is relatively insensitive and consumes relatively small resources (Computational complexity, memory) due to the nature of the sensor system, and allows real-time abnormality to be derived.

Referring to FIG. 1, the sensor signal detection system includes a wireless transceiver 40 and a network gateway 50 for data communication between the abnormality determination unit and the outside.

The wireless transceiver may be selected from the group consisting of Wi-Fi, Bluetooth, and ZigBee, and preferably, may be a Wi-Fi module.

The Wi-Fi module 40 may include a transmitter and a receiver for wirelessly transmitting and receiving data to and from the external terminal, and include a communication unit capable of transmitting and receiving data with the network gateway 50.

The network gateway 50 may include an Ethernet port to which an Internet network of a local area network (LAN) is connected, and a memory for temporarily storing various data exchanged with the Wi-Fi module 40.

The network gateway 50 may be a router.

In another aspect, the present invention relates to a smart plug including the sensor signal detection system. 8 shows a smart plug of the present invention.

Referring to FIG. 8, the smart plug includes a multi-tap 210 having one or more terminals 211 and a PCB board 220 on which the sensor signal detection system is formed. Gateways such as the abnormality processing unit, the abnormality determination unit, the Wi-Fi module, and the router described above are installed in the PCB module.

The sensor signal detection system may be connected to an external internet network through the network gateway and the Wi-Fi module.

The smart plug of the present invention does not go through a separate Wi-Fi gateway by incorporating a sensor signal detection system and a gateway into a multi-tap, The abnormal signal detected by the sensor signal detection system may be transmitted to the user through an external internet network. In addition, the smart plug of the present invention can reduce the cost by increasing the transmission time, space, product volume by implementing a sensor signal detection system including a sensor, a gateway, a USB charger, a power strip in a single product.

Although the above has been described in detail with reference to a preferred embodiment of the present invention, this description is merely to describe and disclose an exemplary embodiment of the present invention. Those skilled in the art will readily recognize that various changes, modifications and variations can be made from the above description and the accompanying drawings without departing from the scope and spirit of the invention.

According to the present invention, a sensor signal detection system may be implemented in a plug to determine abnormality of a sensor signal, and a signal may be transmitted to a user through an external internet network without a separate Wi-Fi gateway by embedding a network gateway in a power strip.

Claims (7)

1. One or more sensors;

   An abnormality processing unit for detecting an abnormal signal among the signals received from the sensor;

   An abnormality determination unit for calculating an abnormality signal received from the abnormality processing unit for a longer time than a detection time in the abnormal signal unit to finally determine the abnormality; And

   And a wireless transceiver for data communication between the abnormality determining unit and the outside.
2. The method of claim 1, wherein the abnormality processing unit and the abnormality determination unit continuously update a new reference model, and determine in real time whether the model calculated from the reference model and the newly acquired signal is the same pattern to detect or determine an abnormal signal. A sensor signal detection system characterized by the above.
3. The method of claim 1, wherein the abnormality processing unit comprises a micro control unit (MCU), a spike signal value of the detection signal over a few seconds to several minutes, a signal value or a linear regression error that continuously maintains the same value Sensor signal detection system characterized in that it determines the value.
4. The apparatus of claim 1, wherein the abnormality determining unit comprises a central processor unit to determine whether there is a sudden change in an average value of abnormal signals received from the abnormality processing unit over a few minutes to several hours, and a linear regression error value. Sensor signal detection system, characterized in that.
5. 2. The sensor signal detection system of claim 1, wherein the at least one sensor is selected from the group consisting of temperature sensor, humidity sensor, gas detection sensor, power sensor and dust sensor.
6. The system of claim 1, wherein the system includes a Wi-Fi network gateway to connect to an external internet network.
7. The smart plug of the sensor signal detection system according to any one of claims 1 to 6, wherein the power strip having one or more terminals is formed and located inside the power strip.

Images