Anomaly detection method and anomaly detection system

Abstract

To provide an anomaly detection method and an anomaly detection system that can detect moisture intrusion into electrical equipment and prevent insulation deterioration due to condensation and moisture adhesion.SOLUTION: An anomaly detection method acquires measurement data from temperature and humidity sensors arranged in electrical equipment, calculates volumetric absolute humidity (306), calculates deviation values (308), calculates moving average values for each pair of temperature and humidity sensors (310), determines K1, K2, and K3 from time-series data of the deviation values and moving average values during a learning period, calculates multiple threshold values from K1, K2, and K3 for the deviation values and moving average values (312), and after the learning period, compares the deviation values and moving average values calculated from the measurement data with the multiple threshold values to determine whether an anomaly has occurred in moisture intrusion in the electrical equipment (320). This allows the threshold values for determining an anomaly in moisture intrusion to be determined automatically and accurately through relatively simple calculations.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to an abnormality detection method and an abnormality detection system capable of detecting the intrusion of moisture into electrical equipment such as switchboards, and preventing in advance the occurrence of malfunctions such as operation stoppage of electrical equipment, and the occurrence of serious accidents such as short circuits and ground faults. Background Art

[0002] Electrical equipment such as switchboards is disposed outdoors exposed to wind and rain, may be installed in harsh environments such as high temperature and high humidity, and is used for a long period of time in the same environment after installation. When an abnormality such as insulation deterioration occurs in electrical equipment such as a switchboard, all equipment at the power supply destination will be affected. In addition, the occurrence of an abnormality may also lead to serious accidents such as fire. Therefore, as a countermeasure against dew condensation and moisture adhesion that cause insulation deterioration in electrical equipment, a humidity control method is implemented in which relative humidity is measured by a sensor, and when the humidity exceeds a threshold value (for example, 60% RH), a space heater is energized to reduce the relative humidity.

[0003] For example, the following Patent Document 1 discloses a dew condensation detection unit that predicts the occurrence of dew condensation inside a housing that accommodates electrical equipment. This dew condensation detection unit measures the surface temperature of the housing of the electrical equipment, and the temperature and humidity inside the electrical equipment, to improve the accuracy of controlling temperature control equipment such as heaters or fans disposed in the electrical equipment, and controls the temperature control equipment based on the measurement results.

[0004] The following Patent Document 2 discloses an abnormality detection method capable of detecting an abnormality of electrical equipment without being affected by the external environment such as environmental changes at the installation site of a plurality of connected and arranged switchboards (row switchboards). This abnormality detection method detects temperature abnormality by relatively evaluating the measurement data of temperature sensors disposed on each of the plurality of switchboards. Specifically, a representative value (average value) is obtained from the measurement data of each temperature sensor, the difference between the measurement data and the representative value is calculated, the moving average value and moving standard deviation value of the time-series data of the difference are calculated, and a temperature abnormality is detected based on these calculation results. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2017-32281 [Patent Document 2] Japanese Patent Publication No. 2020-9184 [Overview of the project] [Problems that the invention aims to solve]

[0006] In electrical equipment configured as a series of panels (e.g., switchboards), moisture intrusion from pits and rainwater intrusion (e.g., through ventilation openings or gaps caused by deterioration of gaskets and housings) can cause condensation or moisture adhesion in specific panels, leading to insulation deterioration and malfunctions. Neither Patent Document 1 nor 2 can detect the risk of moisture intrusion into specific panels in such a series of panels. Considering that the temperature dependence of saturated water vapor is nonlinear (i.e., an exponential curve), even if the method using the difference between measured data and representative values ​​is applied to relative humidity measurement data, as in Patent Document 1, moisture intrusion cannot be detected with high accuracy.

[0007] Furthermore, it is preferable to be able to detect condensation or moisture accumulation on specific equipment in electrical equipment where multiple devices are used in the same environment, not limited to a grid consisting of multiple switchboards. Note that moisture intrusion refers to the intrusion of water into electrical equipment, whether in liquid or gaseous form. In other words, it is not limited to the intrusion of liquid water; for example, if humid air enters the electrical equipment and condenses, this is also included in moisture intrusion.

[0008] Therefore, the present invention aims to provide an abnormality detection method and system that can detect moisture intrusion in electrical equipment (including electrical devices) and prevent insulation deterioration due to condensation and moisture adhesion. [Means for solving the problem]

[0009] (1) An anomaly detection method according to the first aspect of the present invention includes: a first step of acquiring temperature data and relative humidity data measured at the same time from N pairs of temperature sensors and humidity sensors arranged in electrical equipment, where N is an integer of 2 or more; a second step of calculating volumetric absolute humidity using the temperature data and the relative humidity data corresponding to the temperature data; a third step of calculating an average value from at least a portion of the N volumetric absolute humidity values ​​calculated in the second step; a fourth step of calculating the deviation value for each of the N volumetric absolute humidity values ​​by volumetric absolute humidity / average value - 1; a fifth step of calculating a moving average value for each pair of temperature sensors and humidity sensors using the deviation values ​​obtained by repeatedly executing the first, second, third, and fourth steps for a predetermined period; and during the learning period, the first, second, third, Step 6 sorts the time-series data of deviation values ​​and moving averages generated by repeating Steps 4 and 5 in descending order; Step 7 determines, for each of the deviation values ​​and moving averages, the data corresponding to the L1%, L2%, and L3% of the total number of time-series data, starting from the top, as K1, K2, and K3, respectively, from the data sorted in Step 6; Step 8 calculates multiple thresholds from the corresponding K1, K2, and K3 for each of the deviation values ​​and moving averages; and Step 9 determines whether or not an abnormality of moisture intrusion has occurred in the electrical equipment by comparing the deviation values ​​and moving averages calculated from temperature data and relative humidity data measured by temperature and humidity sensors after the learning period with the multiple thresholds, wherein L2 is greater than L1 and L3 is greater than L2.

[0010] This enables accurate detection of moisture intrusion abnormalities in electrical equipment. Furthermore, the threshold for determining moisture intrusion abnormalities can be automatically and accurately determined using relatively simple calculations. Therefore, human judgment is unnecessary for determining the threshold, simplifying the operation of the moisture intrusion abnormality detection system. Additionally, because the threshold can be determined using relatively simple calculations, high-performance computing elements are unnecessary, reducing the manufacturing cost of the system.

[0011] (2) In (1) above, the average value calculated in the third step can be calculated by subtracting the volume absolute humidity corresponding to the deviation value calculated in the fourth step from the N volume absolute humidity values. This makes it possible to determine a more appropriate value as a threshold for detecting abnormal moisture intrusion.

[0012] (3) In (1) or (2) above, step 9 may include step 10 of setting a point by comparing the deviation value and moving average value calculated from temperature data and relative humidity data measured after the learning period with a plurality of thresholds, and step 11 of determining that if the point is above a predetermined value, an abnormality of moisture intrusion has occurred in the electrical equipment. This makes it easier to detect abnormalities of moisture intrusion in electrical equipment.

[0013] (4) In any one of (1) to (3) above, L1 may be greater than 0.1 and less than 0.3, L2 may be greater than 1 and less than 5, and L3 may be greater than 10 and less than 20. This allows for more accurate detection of abnormal moisture intrusion.

[0014] (5) An anomaly detection system according to the second aspect of the present invention includes, where N is an integer of 2 or more, N pairs of temperature sensors and humidity sensors arranged in electrical equipment in a one-to-one correspondence, and a calculation device, wherein the calculation device includes: a first process of acquiring temperature data and relative humidity data measured at the same time by each of the N pairs of temperature sensors and humidity sensors; a second process of calculating volumetric absolute humidity using the temperature data and the relative humidity data corresponding to the temperature data; a third process of calculating an average value from at least a portion of the N volumetric absolute humidity values ​​calculated by the second process; a fourth process of calculating the deviation value of each of the N volumetric absolute humidity values ​​by volumetric absolute humidity / average value-1; a fifth process of calculating a moving average value for each pair of temperature sensors and humidity sensors using the deviation values ​​obtained by repeatedly executing the first, second, third, and fourth processes for a predetermined period; and during the learning period, the first The process involves a sixth process that sorts the time-series data of deviation values ​​and moving averages generated by repeating the first, second, third, fourth, and fifth processes in descending order; a seventh process that determines, for each deviation value and moving average, the data corresponding to the L1%, L2%, and L3% of the total time-series data, starting from the top, as K1, K2, and K3, respectively, from the data sorted by the sixth process; an eighth process that calculates multiple thresholds from the corresponding K1, K2, and K3 for each deviation value and moving average; and a ninth process that, after the learning period, determines whether or not an abnormality of moisture intrusion has occurred in the electrical equipment by comparing the deviation values ​​and moving averages calculated from the temperature data and relative humidity data measured by the temperature sensor and humidity sensor with the multiple thresholds, and L2 is greater than L1, and L3 is greater than L2.

[0015] This enables accurate detection of moisture intrusion abnormalities in electrical equipment. Furthermore, the threshold for determining moisture intrusion abnormalities can be automatically and accurately determined using relatively simple calculations. Therefore, human judgment is unnecessary for determining the threshold, simplifying the operation of the moisture intrusion abnormality detection system. Additionally, because the threshold can be determined using relatively simple calculations, high-performance computing elements are unnecessary, reducing the manufacturing cost of the system.

[0016] (6) In any one of (1) to (5) above, the plurality of threshold values may be calculated by adding, to K1 or K2, an integer multiple of any one of a first difference value which is a difference between K1 and K2, a second difference value which is a difference between K2 and K3, a third difference value which is a difference between K1 and K3, or any one of the first difference value, the second difference value and the third difference value.

[0017] (7) In any one of (1) to (6) above, the learning period may be a period of one month or more before the start of the determination of whether an abnormality has occurred, or may be a period in the previous year corresponding to a detection period in which the determination of whether an abnormality has occurred is performed.

[0018] (8) In any one of (1) to (7) above, the learning period may be not less than 2 months and less than 6 months.

[0019] (9) In any one of (1) to (8) above, N may be 4 or more.

[0020] (10) In any one of (1) to (9) above, the electrical equipment may include a plurality of racks, and when the plurality of racks include a rack having a temperature control function, a temperature sensor and a humidity sensor do not need to be arranged on said rack.

[0021] (11) In any one of (1) to (5) above, the plurality of threshold values may include a value obtained by adding a constant to K2 or K3. Effects of the Invention

[0022] According to the present invention, it is possible to provide an abnormality detection method and an abnormality detection system that can detect moisture intrusion into electrical equipment (including electrical devices) and prevent insulation degradation due to condensation and moisture adhesion in advance. Brief Description of the Drawings

[0023] [Figure 1]FIG. 1 is a front view showing the schematic configuration of an anomaly detection system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a block diagram showing the internal configuration of the sensor unit shown in FIG. 1. [Figure 3] FIG. 3 is a block diagram showing the internal configuration of the arithmetic device shown in FIG. 1. [Figure 4] FIG. 4 is a flowchart showing processing executed by the arithmetic device. [Figure 5] FIG. 5 is a graph for explaining a method of calculating a moving average value. [Figure 6] FIG. 6 is a diagram showing the relationship between threshold values and points. [Figure 7] FIG. 7 is a graph showing temperatures measured by five sensor units disposed on each panel of an outdoor switchboard. [Figure 8] FIG. 8 is a graph showing relative humidity measured by five sensor units disposed on each panel of an outdoor switchboard. [Figure 9] FIG. 9 is a graph showing volumetric absolute humidity calculated from the measurement data of FIGS. 7 and 8. [Figure 10] FIG. 10 is a graph showing deviation values related to the five volumetric absolute humidity values shown in FIG. 9. [Figure 11] FIG. 11 is a graph showing a 12-hour moving average value calculated from the deviation values shown in FIG. 10. [Figure 12] FIG. 12 is a graph showing a 24-hour moving average value calculated from the deviation values shown in FIG. 10. [Figure 13] FIG. 13 is a graph showing a 72-hour moving average value calculated from the deviation values shown in FIG. 10. [Figure 14] FIG. 14 is a diagram showing, in a table format, the relationship between threshold values and points determined respectively from the deviation values shown in FIG. 10 and the moving average values shown in FIGS. 11 to 13. [Figure 15] FIG. 15 is a graph showing points determined for the deviation values shown in FIG. 10 using the threshold values shown in FIG. 14. [Figure 16]Figure 16 is a graph showing the points determined using the threshold values ​​shown in Figure 14, with respect to the moving average values ​​shown in Figure 11. [Figure 17] Figure 17 is a graph showing the points determined using the threshold values ​​shown in Figure 14, with respect to the moving average values ​​shown in Figure 12. [Figure 18] Figure 18 is a graph showing the points determined using the threshold values ​​shown in Figure 14, with respect to the moving average values ​​shown in Figure 13. [Figure 19] Figure 19 is a graph showing the analysis results when the point calculation method is changed. [Modes for carrying out the invention]

[0024] In the following embodiments, identical parts are assigned the same reference numeral. Their names and functions are also identical. Therefore, detailed descriptions of them will not be repeated.

[0025] (Configuration of the anomaly detection system) Referring to Figure 1, the anomaly detection system 100 according to an embodiment of the present invention includes a first sensor unit 110, a second sensor unit 112, a third sensor unit 114, a fourth sensor unit 116, and a fifth sensor unit 118, respectively, which are arranged in the first, second, third, fourth, and fifth distribution panels 120, 122, 124, 126, and 128, and a computing device 130. The first to fifth distribution panels 120 to 128 are distribution panels of the same type or similar distribution panels, and are a plurality of distribution panels (hereinafter also referred to as "row panels") that are connected and installed side by side. The first to fifth sensor units 110 to 118 are each located in the lower part of the first to fifth distribution panels 120 to 128, and are located in approximately the same position (for example, at the same height). The computing device 130 is located outside the fifth distribution panel 128.

[0026] Referring to Figure 2, the first sensor unit 110 includes a control unit 140, a memory unit 142, a communication unit 144, a timer 146, a bus 148, an A / D converter 150, a temperature sensor 152, and a humidity sensor 154. The control unit 140 controls each component of the first sensor unit 110, and is, for example, a CPU (Central Processing Unit), a microcomputer, etc. The memory unit 142 stores data and is, for example, a rewritable non-volatile semiconductor memory. The communication unit 144 is a wireless communication module for communicating with the outside. The timer 146 receives a request from the control unit 140 and transmits information representing the current time (hereinafter simply referred to as the current time) to the control unit 140.

[0027] The temperature sensor 152 is an element for measuring temperature and outputs an analog signal corresponding to the detected temperature. The temperature sensor 152 is, for example, a resistance thermometer or a thermocouple. The humidity sensor 154 is an element for measuring relative humidity (%RH) and outputs an analog signal corresponding to the detected relative humidity. The humidity sensor 154 is, for example, a resistive or capacitive humidity sensor. As will be described later, in order to accurately calculate absolute humidity, it is preferable that the temperature sensor 152 and the humidity sensor 154 be placed in close proximity. For example, a single-chip type digital temperature and humidity sensor such as the SHT-31 manufactured by Sensirion can be used.

[0028] The A / D conversion unit 150 converts the analog signals output from the temperature sensor 152 and humidity sensor 154 into digital data at predetermined timings and outputs it. The digital data (temperature and relative humidity measurement data) output from the A / D conversion unit 150 is stored in the storage unit 142. Note that if the temperature sensor 152 and humidity sensor 154 can output the measured values ​​as digital data, the A / D conversion unit 150 may not be necessary. Data exchange between each unit is performed via the bus 148. Power is supplied to each unit by built-in batteries or by service outlets provided in each distribution panel.

[0029] The second sensor unit 112 to the fifth sensor unit 118 are configured in the same way as the first sensor unit 110. The communication units of the first sensor unit 110 to the fifth sensor unit 118 each have information (such as a communication address) to uniquely distinguish them. Furthermore, the timer times of each of the first sensor unit 110 to the fifth sensor unit 118 are assumed to be synchronized (the same time) with the timer time of the arithmetic unit 130, which will be described later. Note that "same time" (the times are synchronized) means not only that they are exactly the same time, but also that they match within a predetermined tolerance range.

[0030] Referring to Figure 3, the arithmetic unit 130 includes a control unit 170, a storage unit 172, a communication unit 174, a timer 176, a display unit 178, an operation unit 180, and a bus 182. The control unit 170 controls each component of the arithmetic unit 130, and is, for example, a CPU. The storage unit 172 stores data and is, for example, a rewritable non-volatile semiconductor memory. The communication unit 174 is a communication module for communicating from the first sensor unit 110 to the fifth sensor unit 118. The communication unit 174 has a wireless communication function for communicating from the first sensor unit 110 to the fifth sensor unit 118. The timer 176 receives a request from the control unit 170 and transmits the current time to the control unit 170. The display unit 178 displays visual information (text, images, etc.). The display unit 178 includes, for example, a display panel such as a liquid crystal display panel and a drive circuit that drives each pixel of the display panel. The operation unit 180 is for inputting instructions to the arithmetic unit 130, and is, for example, a touch panel. Data exchange between the various components of the arithmetic unit 130 is performed via the bus 182. The arithmetic unit 130 may be, for example, a computer. Alternatively, the arithmetic unit 130 may be, for example, a sequencer.

[0031] (Operation of the anomaly detection system) In the following description, the first sensor unit 110 will be explained as a representative of the multiple sensor units. The second sensor units 112 to the fifth sensor units 118 operate in the same manner as the first sensor unit 110. (Sensor unit operation)

[0032] The first sensor unit 110 transmits temperature data and relative humidity data (hereinafter both referred to as measurement data) measured by the temperature sensor 152 and humidity sensor 154 at regular time intervals to the arithmetic unit 130. This function of the first sensor unit 110 is realized by the control unit 140 reading and executing a predetermined program stored in the storage unit 142. Specifically, the control unit 140 obtains the current time from the timer 146, determines whether it is time to measure temperature and humidity (hereinafter also referred to as measurement timing), and if it determines that it is measurement timing, controls the A / D conversion unit 150 to convert the analog signal (measurement signal of the sensor element 150) input to the A / D conversion unit 150 into digital data and stores it in the storage unit 142. Subsequently, the control unit 140 reads the measurement data from the storage unit 142 and transmits it to the arithmetic unit 130 via the communication unit 144. At this time, the transmitted measurement data includes temperature data and relative humidity data, so it is transmitted in a way that allows the arithmetic unit 130 to distinguish between the two. For example, a code representing temperature or humidity and the measurement data can be transmitted as a set. The order of data included in the wireless communication packet may also be predetermined (for example, temperature data followed by relative humidity data).

[0033] For example, if the first sensor unit 110 is to measure and transmit the temperature and humidity inside the first power distribution panel 120 every hour, the measurement timing information, such as the start time of measurement and the measurement interval (e.g., 60 minutes), can be stored in the storage unit 142 in advance. The control unit 140 can read the measurement timing information from the storage unit 142 and refer to the current time using the timer 146 to perform the above processing.

[0034] Here, we assume that the first sensor unit 110 to the fifth sensor unit 118 have the same measurement timing set. As described above, since the timers of the first sensor unit 110 to the fifth sensor unit 118 are synchronized, the first sensor unit 110 to the fifth sensor unit 118 transmit temperature and humidity measurement data from the first distribution panel 120 to the fifth distribution panel 128, respectively, at the same timing. Note that "same timing" does not only mean completely identical timing, as is clear from the above explanation regarding "same time," but also includes cases where they coincide within a predetermined tolerance range.

[0035] (Operation of the arithmetic unit) The arithmetic unit 130 receives measurement data (i.e., temperature data and relative humidity data) transmitted from each of the first sensor unit 110 to the fifth sensor unit 118, and stores the received measurement data in the storage unit 172 in time series for each of the temperature data and relative humidity data for each of the first sensor unit 110 to the fifth sensor unit 118 (i.e., adds it to the measurement data that has been received and stored in the past). As a result, a total of 10 (=5 × 2) time-series data are stored in the storage unit 172. As described above, measurements are taken and measurement data is transmitted from each of the first sensor unit 110 to the fifth sensor unit 118 at the same time, so the control unit 170 can store the measurement data measured at the same time in the storage unit 172 in correspondence. After storing the measurement data in the storage unit 172 for each of the first sensor unit 110 to the fifth sensor unit 118 for a predetermined period (hereinafter referred to as the learning period), the arithmetic unit 130 performs analysis processing as described later and calculates a threshold value used for detecting abnormal moisture intrusion. After the learning period, the computing unit 130 determines, based on the analysis results (i.e., threshold values), whether or not a moisture intrusion abnormality has occurred in any of the first to fifth distribution panels 120 to 128, or whether or not there are signs of a moisture intrusion abnormality occurring.

[0036] The arithmetic unit 130 performs, for example, the processes shown in Figure 4. Each process in Figure 4 is realized by the control unit 170 reading a predetermined program from the storage unit 172 and executing it.

[0037] In step 300, the control unit 170 determines whether it has received measurement data (i.e., temperature data and relative humidity data) transmitted from any of the first sensor unit 110 to the fifth sensor unit 118 via the communication unit 174. If it determines that the data has been received, the control proceeds to step 302. Otherwise, the process in step 300 is repeated.

[0038] In step 302, the control unit 170 stores the received measurement data in the storage unit 172. As described above, the control unit 170 stores the measurement data in the storage unit 172 in chronological order for each of the temperature data and relative humidity data for each of the first sensor unit 110 to the fifth sensor unit 118.

[0039] In step 304, the control unit 170 determines whether or not to perform the analysis. Specifically, the control unit 170 determines whether or not it was possible to acquire measurement data during a predetermined learning period. If the period at which each sensor transmits measurement data is constant, it may be determined whether or not a predetermined number of measurement data points were acquired.

[0040] In step 306, the control unit 170 reads a set of measurement data (temperature data and relative humidity data) from the storage unit 172 for each of the first sensor unit 110 to the fifth sensor unit 118, and uses it to calculate the volumetric absolute humidity. For example, the control unit 170 uses a set of temperature data Tm (°C) and relative humidity data Hm (%RH) to calculate the volumetric absolute humidity Ym (g / m³) using the following formula 1. 3 The following is calculated: m is used to identify the sensor unit, and m=1 to 5 correspond to the first sensor unit 110 to the fifth sensor unit 118, respectively. Ym = 217 × (6.1078 × 10 K ) / (Tm+273.15)×Hm / 100 (Formula 1) Here, K = 7.5 × Tm / (Tm + 237.3). The calculated volumetric absolute humidity Ym is stored in the memory unit 172 in chronological order for each sensor unit.

[0041] In step 308, the control unit 170 calculates a deviation value with respect to the volumetric absolute humidity Ym calculated in step 306. Specifically, the control unit 170 reads out the volumetric absolute humidity Ym (i.e., five data points from m=1 to m=5) calculated from the measurement data at the same timing for each of the first sensor unit 110 to the fifth sensor unit 118 from the storage unit 172. Let Ym be the volumetric absolute humidity for the mth sensor unit (m=1 to m=5), and with respect to Ym, the control unit 170 calculates the average value Yavm of the data other than Ym, Yn (n≠m), and calculates the deviation value Xm using Equation 2. Xm=Ym / Yavm-1...(Formula 2) This allows one set of deviation values ​​Xm (m=1 to 5) to be calculated from one set of volumetric absolute humidity (i.e., 5 data points). The control unit 170 calculates the deviation values ​​Xm (m=1 to 5) for all volumetric absolute humidity data calculated during the learning period and stores them in the storage unit 172. The deviation values ​​Xm (m=1 to 5) are stored as time-series data for each sensor.

[0042] The amount of water vapor that can exist per unit volume of air (saturation water vapor amount (g / m³) 3 The absolute humidity (1m³) increases exponentially with increasing temperature. Therefore, even if the change in relative humidity is the same, the change in water vapor content differs greatly depending on the temperature. That is, when the relative humidity fluctuates by a predetermined amount, the change in water vapor content at high temperatures is greater than at low temperatures. For this reason, when evaluating humidity, if the evaluation is done using only the difference in relative humidity, the sensitivity to fluctuations in relative humidity becomes low at low temperatures, and conversely, the sensitivity becomes high at high temperatures, making it impossible to uniformly evaluate humidity fluctuations. For this reason, a deviation amount including a multiplier (Ym / Yavm) to the average value calculated as described above is used. Note that the subtraction of 1 is to shift the evaluation value to a value close to zero. Volumetric absolute humidity (1m³) of electrical equipment 3The amount of water vapor per unit (g) is determined by the amount of water vapor in the outside air unless other moisture is supplied, so it is almost constant for all panels, and if there is no abnormality in moisture intrusion, the deviation amount Xm calculated by Equation 2 will be close to zero.

[0043] In step 310, the control unit 170 calculates a moving average value Xavm (m=1 to 5) from the deviation values ​​Xm (m=1 to 5) calculated in step 308 for each of the first sensor unit 110 to the fifth sensor unit 118. A predetermined period (hereinafter also called a window) T for identifying the data for which the moving average value Xavm is calculated is, for example, 12 hours, 24 hours, and 72 hours. Referring to Figure 5, the horizontal axis represents time, and the vertical axis represents the deviation value Xm of the m-th temperature sensor. The deviation value Xm is schematically shown by a dashed line. The actual deviation value Xm is digital data, and in Figure 5, a part of it is shown by a black circle. For the m-th temperature sensor, the process of calculating the average value of multiple deviation values ​​Xmi (i=1 to n) over the predetermined period T is repeated while shifting the data for which the average value is calculated (i.e., data within the predetermined period T) (for example, shifting from data within t1 to t3 to data within t2 to t4). The control unit 170 uses the deviation value Xm (m=1 to 5) calculated during the learning period to calculate the moving average value Xavmj (m=1 to 5, j=1 to 3) for each of three predetermined periods Tj (j=1 to 3), and stores it in the storage unit 172. For example, T1=12 (hours), T2=24 (hours), and T3=72 (hours). The moving average value Xavmj (m=1 to 5, j=1 to 3) is stored as time-series data for each combination of sensor and predetermined period.

[0044] In step 312, the control unit 170 uses the time-series data of the deviation value Xm (m=1 to 5) calculated in step 308, and the time-series data of the moving average value Xavmj (m=1 to 5, j=1 to 3) calculated in step 310, to determine a threshold for detecting an anomaly. The threshold is determined for each of the 20 time-series data points.

[0045] Specifically, for each piece of time-series data, the control unit 170 sorts the data in descending order of numerical value (that is, the first data is the maximum value), and sets the 4th, 55th, and 382nd data as K1, K2, and K3 respectively. The 4th, 55th, and 382nd data are located approximately at 0.135%, 2.275%, and 15.865% from the top in each set of time-series data, respectively. These numerical values respectively correspond to 3σ, 2σ, and 1σ when each piece of time-series data follows a normal distribution (the standard deviation is σ).

[0046] Next, the control unit 170 determines thresholds Th1 to Th4 for determining points respectively corresponding to each of the deviation values Xm (m = 1 to 5) and the moving average values Xavmj (m = 1 to 5, j = 1 to 3). That is, for each piece of time-series data, the control unit 170 sets K2 as the threshold Th1, K1 as the threshold Th2, 2×K1-K2 as the threshold Th3, and 2×K1-K3 as the threshold Th4. The K1, K2, and K3 used are those determined as described above for each piece of time-series data. The control unit 170 stores the thresholds Th1 to Th4 determined for each piece of time-series data in the storage unit 172.

[0047] As an example, with reference to FIG. 6, the correspondence between the four thresholds Th1 to Th4 and the four points from 0 to 4 is shown for the time-series data of the deviation values Xm. That is, if Xm<Th1(=K2), 0 points are assigned to the Xm. If Th1(=K2)≦Xm<Th2(=K1), 1 point is assigned to the Xm. If Th2(=K1)≦Xm<Th3(=2×K1-K2), 2 points are assigned to the Xm. If Th3(=2×K1-K2)≦Xm<Th4(=2×K1-K3), 3 points are assigned to the Xm. If Xm>Th4(=2×K1-K3), 4 points are assigned to the Xm.

[0048] Similarly, using thresholds Th1 to Th4 for each of the moving average values ​​Xavmj (m=1 to 5, j=1 to 3), the points corresponding to the moving average value Xavmj are determined. That is, in Figure 6, the deviation value Xm is replaced with the moving average value Xavmj, and K1 to K3 are those determined as described above for the time series data of the moving average value Xavmj.

[0049] Based on the above, the threshold values ​​Th1 to Th4 for determining the points have been determined from the volumetric absolute humidity calculated from the measurement data during the learning period. Therefore, the control unit 170 will then use the determined threshold values ​​to perform anomaly detection processing.

[0050] In step 314, the control unit 170 determines whether it has received a set of measurement data (i.e., five temperature data points and five relative humidity data points) transmitted from each of the first sensor unit 110 to the fifth sensor unit 118 via the communication unit 174. If it determines that the data has been received, the control proceeds to step 316. Otherwise, the process in step 314 is repeated. The control unit 170 stores the received measurement data in the storage unit 172.

[0051] In step 316, the control unit 170 calculates the volumetric absolute humidity for each sensor unit from the measurement data received in step 314 (i.e., five temperature data points and five relative humidity data points) using Equation 1 described above.

[0052] In step 318, the control unit 170 uses the volumetric absolute humidity calculated in step 316 and the time-series data of volumetric absolute humidity stored in the memory unit 172 up to that point to calculate the deviation value Xm using Equation 2 as described above, and calculates the moving average value Xavmj of the deviation value Xm. Furthermore, the control unit 170 uses threshold values ​​K1 to K4 stored in the memory unit 172 to determine points corresponding to each of the deviation value Xm and the moving average value Xavmj. A total of 20 points are determined. As will be described later, since step 318 is repeated, the control unit 170 may store the determined points as time-series data in the memory unit 172. After that, the control proceeds to step 320.

[0053] In step 320, the control unit 170 determines whether each point determined in step 318 is "3" or greater. If it is determined that at least one of the 20 points is "3" or greater, the control proceeds to step 322. Otherwise, the control proceeds to step 324.

[0054] In step 322, the control unit 170 displays a message indicating an abnormality. For example, the control unit 170 displays an image containing a predetermined message on the display unit 178. For example, if a point corresponding to either the deviation value Xm or the moving average value Xavmj is "3" or greater, the control unit displays a message indicating that a moisture intrusion abnormality has occurred (or is possible to occur) in the power distribution panel where the m-th sensor unit is located. As will be described later in the embodiment, the state in which the points are "3" or "4" occurs rarely. Therefore, if a point of "3" or greater occurs, it can be determined that there is a suspicion of moisture intrusion. Alternatively, as will be described later, it may be determined whether the frequency of occurrence of points of "3" or greater is high or low.

[0055] In step 324, the control unit 170 determines whether or not it has received a termination instruction. A termination instruction is given, for example, by operating the operation unit 180. If it is determined that a termination instruction has been received, the program terminates. Otherwise, control returns to step 314, and the control unit 170 repeats the above process.

[0056] As a result, the anomaly detection system 100 can accurately detect anomalies in moisture intrusion in the panel. Furthermore, the threshold for determining an anomaly in moisture intrusion can be automatically and accurately determined using relatively simple calculations. Specifically, the deviation value Xm and the moving average value Xavmj are calculated, and their time-series data are rearranged to automatically calculate the threshold (i.e., Th1 to Th4) for determining the point for determining an anomaly in moisture intrusion. This eliminates the need for human judgment to determine the threshold, making the operation of the anomaly detection system for moisture intrusion easier. In addition, since the threshold can be determined using relatively simple calculations, high-performance computing elements are not required, reducing the manufacturing cost of the system.

[0057] Furthermore, by adjusting the learning period, a more appropriate threshold can be determined for determining the point at which abnormal moisture intrusion is detected. Therefore, the accuracy of detecting abnormal moisture intrusion in the panel can be improved.

[0058] In the above, step 316 determines whether a single point is "3" or higher, but is not limited to this. It is also possible to determine whether the occurrence frequency of points of "3" or higher is high. "High occurrence frequency" means, for example, a state in which points of "3" or higher occur a predetermined number of times (e.g., 10 times) or more during a predetermined period (e.g., the most recent 3 days (i.e., 72 hours) from the last measurement). The predetermined period may be a period from the most recent 2 days to 10 days from the last measurement. Instead of the predetermined number of occurrences, the probability of occurrence may be used, and for example, if the probability of points of "3" or higher occurring during the predetermined period is above a predetermined value, it may be determined that the "occurrence frequency is high". The predetermined value can be set to a value of, for example, 10% or more and 30% or less.

[0059] As described above, by using a sensor unit having a wireless communication function, the sensor unit can be easily installed in an existing electrical installation (without causing a power outage depending on conditions).

[0060] In the above description, an explanation has been given of a case where, for each piece of time-series data, data are arranged in descending order of numerical value, and the 4th, 55th, and 382nd data are set as K1, K2, and K3, respectively; however, the present invention is not limited to this. For example, for each piece of time-series data, data may be arranged in descending order of numerical value, and values at ranks corresponding to L1%, L2%, and L3% (where L1<L2<L3) of the total number of data, counting from the top, may be set as K1, K2, and K3. Note that the value at the rank corresponding to L% of the total number, counting from the top, means the data when the count value obtained by counting the number of pieces of data from the top data exceeds L% of the total number, or the data immediately before that. L1, L2, and L3 may satisfy 0.1<L1<0.3, 1<L2<5, and 10<L3<20, respectively. The aforementioned values of 0.135%, 2.275%, and 15.865% satisfy these conditions.

[0061] In the above description, an explanation has been given of a case where, for deviation values Xm (m = 1 to 5) and moving average values Xavmj (m = 1 to 5, j = 1 to 3), four thresholds Th1 to Th4 are used to correspond to five levels of points (that is, 0 to 4); however, the present invention is not limited to this. Six or more levels of points may be made to correspond. For example, if 2×K1-K3+(K1-K2) (that is, a value equivalent to 6σ) is used as the threshold Th5, six levels of points (that is, 0 to 5) can be made to correspond.

[0062] In the above, thresholds Th1 to Th4 are determined from K1, K2, K3 and their calculated values, but this is not limited to this. For example, a deviation value constant that indicates the occurrence of moisture intrusion may be added to K2 or K3 to set the threshold. For example, if the deviation value when actual moisture intrusion occurs is judged to be around 0.1, the thresholds Th1 to Th3 for points 0 to 3 may be set as described above, and only the threshold Th4 for point 4 may be set to a value obtained by adding a constant to K1, for example K1 + 0.08, instead of the above 2 × K1 - K3. When the data variability is small (i.e., when the values ​​of K1 to K3 are small), determining the threshold based only on K1 to K3 may lead to oversensitivity in detecting anomalies. Setting Th4 = K1 + 0.08 can avoid oversensitivity. The constant added to K1 should be in the range of 0.05 to 0.15.

[0063] The above describes a case where, after a learning period, points are set using multiple thresholds for the deviation value Xm (m=1 to 5) and moving average value Xavmj (m=1 to 5, j=1 to 3) calculated from the temperature sensor measurement data, but it is not limited to this. Setting points is to make the judgment process easier to understand, and it is not necessary to set points. The deviation value Xm (m=1 to 5) and moving average value Xavmj (m=1 to 5, j=1 to 3) may also be compared with a threshold that determines whether moisture intrusion is abnormal, for example, threshold Th3 (i.e., a threshold that corresponds to 3 points if it is above this threshold) to determine whether moisture intrusion has occurred. For example, if either the deviation value Xm (m=1 to 5) or the moving average value Xavmj (m=1 to 5, j=1 to 3) is above threshold Th3, or if the frequency of occurrences of threshold Th3 or above is high, it can be determined that moisture intrusion has occurred.

[0064] The number of sensor units placed on the panel row should be at least four, preferably five or more, for the purpose of identifying the panel where an anomaly occurred. If the total number of sensor units is small, an anomaly in the measurement data due to moisture intrusion in a specific panel will greatly affect the representative value, and thus easily affect the differential data of other panels. If the total number of temperature sensors is five or more, such an effect can be suppressed.

[0065] Furthermore, some of the panel layouts may include panels with factors that cause temperature and humidity fluctuations, such as air conditioners or cooling fans. Installing temperature sensors on such panels and performing detection may unnecessarily increase the deviation amount Xm, potentially reducing the sensitivity of detecting moisture intrusion abnormalities. Therefore, it is desirable to exclude such panels from detection (i.e., not install sensor units on them).

[0066] The above example shows a grid consisting of five switchboards, but it is not limited to this. It may consist of four or fewer switchboards, or six or more switchboards. The elements that make up the grid are not limited to switchboards. Other electrical equipment may also be included.

[0067] Furthermore, it is not limited to a grid of identical or similar electrical equipment arranged in parallel. It may also be multiple electrical equipment located in close proximity. Moreover, it may be a single electrical piece of equipment with multiple sensor units.

[0068] The above description assumes that each sensor unit is located at the bottom of each distribution panel, and that the mounting heights of each temperature and humidity sensor are approximately equal. However, the mounting positions of each temperature and humidity sensor are arbitrary. It is desirable to place the sensor units at the bottom of the distribution panel for relative humidity evaluation to detect moisture intrusion into the panel. Furthermore, by placing each temperature sensor at approximately the same height, the influence of the surrounding environment on each sensor unit can be standardized, making it easier to detect abnormal moisture intrusion into the distribution panel.

[0069] The above describes a case where all deviation values ​​Xm (m=1 to 5) and moving average values ​​Xavmj (m=1 to 5, j=1 to 3) of volumetric absolute humidity are calculated, and threshold values ​​Th1 to Th4 are determined for each, but the method is not limited to this. Alternatively, at least one of the deviation values ​​of volumetric absolute humidity and the moving average value over multiple periods may be calculated, and threshold values ​​Th1 to Th4 may be determined for that calculated value (for example, Xm). Subsequently, volumetric absolute humidity is calculated from the measurement data of the temperature sensor and humidity sensor, and the calculated value Xm corresponding to the threshold is calculated. By using the threshold to determine the point corresponding to the calculated value Xm, it is possible to detect anomalies in moisture intrusion at that point.

[0070] Furthermore, in the above, the average value Yav, which serves as the basis for calculating the deviation value, is calculated by excluding the monitored measurement from all measured values, but this is not the only way. The average value of all measured values, including the monitored measurement, may be calculated, and the deviation value may be calculated based on that average value. Even if the deviation value is calculated in this way, the same moving average value Xavmj (m=1 to 5, j=1 to 3) as above can be calculated, and the same effect (i.e., the accuracy of detecting anomalies in water intrusion) can be obtained.

[0071] The above describes the case where the calculation of the deviation value Xm (m=1 to 5) and the moving average value Xavmj (m=1 to 5, j=1 to 3) by the arithmetic unit 130 is performed after all measurement data for the learning period has been acquired, but it is not limited to this. The calculation may also be performed each time the arithmetic unit 130 receives a set of measurement data measured simultaneously from each sensor during the learning period.

[0072] Communication between each sensor unit and the arithmetic unit 130 may be via wired communication. Alternatively, the arithmetic unit 130 may be separated into a data acquisition device and an analysis device, with the data acquisition device placed near the grid and the analysis device placed away from the grid. The data acquisition device acquires measurement data from each sensor unit, stores it temporarily, and then transmits the measurement data to the analysis device. The analysis device calculates the volumetric absolute humidity and calculates the volumetric absolute humidity deviation value Xm (m=1 to 5) and the moving average value Xavmj (m=1 to 5, j=1 to 3). Measurement data may also be transmitted from the data acquisition device to the analysis device via wireless communication. If the distance between the data acquisition device and the analysis device is long, a long-distance transmission method such as RS422 or RS485 may be used in the case of wired communication. Furthermore, the data acquisition device and the analysis device do not necessarily communicate directly. For example, the collected measurement data may be stored on a predetermined data server or cloud via radio waves from a mobile phone or the internet, and the analysis device may read the data from the data server or cloud and perform the analysis. Furthermore, the analysis results may be communicated to the facility manager via email or other means.

[0073] The above describes a case where the first sensor unit 110 to the fifth sensor unit 118 transmit measurement data each time they measure temperature and humidity, but the system is not limited to this. Multiple measurement data can be stored and transmitted to the arithmetic unit 130 all at once. Since the measurement timing of each of the first sensor unit 110 to the fifth sensor unit 118 is the same, when transmitting multiple data together, if the measurement order between the multiple data transmitted can be determined in the arithmetic unit 130, a representative value can be determined using the measurement data from the first sensor unit 110 to the fifth sensor unit 118 measured at the same timing.

[0074] In the above configuration, each sensor unit has a timer and a memory unit, and sends measurement data to the arithmetic unit 130 at regular intervals, but this is not required. For example, the arithmetic unit 130 may have a timer and communicate with each sensor unit at regular intervals (for example, every hour) to obtain the current values ​​of temperature and humidity (for example, by polling each sensor unit to send a request to send measurement data), and log the data (store it as time-series data) in the memory unit 172 of the arithmetic unit 130.

[0075] The above describes the case in which a message is displayed on the display unit 178, but it is not limited to this. The message may also be displayed by sound (including voice) or by lighting an LED (Light Emitting Diode), and furthermore, the occurrence of a moisture intrusion abnormality may be indicated by display on a remote device such as a central monitoring panel or by sending an email.

[0076] Regarding the learning period, especially for outdoor enclosures, the external environment (solar radiation intensity and angle, outside temperature, precipitation, wind speed and direction, etc.) changes with the seasons, so it is desirable to include a season close to the period for detecting abnormal moisture intrusion (hereinafter referred to as the detection period). For example, as in the example described later, a three-month learning period is used, and the following month is used as the detection period. Furthermore, if measurement data from the previous year is available, it is desirable to use the measurement data for the period of the previous year corresponding to the detection period as the measurement data for the learning period.

[0077] The training period should ideally be at least two months, preferably around three months, to ensure sufficient data volume. However, if the training period is too long, it will include data with significantly different conditions from the detection period, leading to decreased detection accuracy. A maximum of around six months is desirable. The timing of recording measurement data (e.g., time intervals) can be determined appropriately between 10 minutes and one hour. Typically, recording measurement data once every hour is sufficient. [Examples]

[0078] The experimental results below demonstrate the effectiveness of the present invention. Similar to Figure 1, sensor units were placed on each panel of a grid consisting of five electrical equipment units installed outdoors, and temperature and humidity were measured. The measured temperature data and relative humidity data are shown in Figures 7 and 8, respectively, and the results of the analysis as described above are shown in Figures 9 to 13. In Figures 7 to 13, S1 to S5 shown in the leftmost column labeled "Sensor" correspond to sensor units arranged in the same way as the first sensor unit 110 to the fifth sensor unit 118 in Figure 1.

[0079] Figure 7 shows the data measured by the temperature sensors of each sensor unit. The vertical axis represents temperature, and the horizontal axis represents days. Temperature measurements were taken every hour from October 23rd to March 31st of the following year. The learning period mentioned above was from October 23rd to January 31st of the following year (see the arrow in the graph for sensor unit S1 in Figure 7). The total number of data points measured during the learning period was 2407. Using the measurement data from the learning period, the threshold for determining the points was determined as described above. In addition, a heating source (i.e., a heater) was installed on the panel where sensor unit S3 (corresponding to the third sensor unit) was set up, and heating was performed after the learning period (i.e., after the threshold was determined) (see the arrow in the graph for sensor unit S3 in Figure 7). Specifically, heating was performed with a 54W heater from February 2nd to February 10th. On February 10th, it was replaced with a 90W heater and heating was performed from February 10th to February 17th. The heater was replaced with a 144W heater on February 17th and heated from February 17th to February 24th. On February 24th, it was replaced with a 36W heater and heated from February 24th to March 3rd.

[0080] Figure 8 shows the data measured by the humidity sensors of each sensor unit. The vertical axis represents relative humidity, and the horizontal axis represents days. As described above, temperature measurements were taken every hour from October 23rd to March 31st of the following year.

[0081] Figure 9 shows the volumetric absolute humidity calculated using Equation 1 above, based on the temperature data shown in Figure 7 and the relative humidity data shown in Figure 8.

[0082] Figure 10 shows the trend of the deviation value Xm calculated as described above, with respect to the volumetric absolute humidity trend (i.e., graph) shown in Figure 9. Figure 11 shows the trend of the moving average value Xavm1 (m=1 to 5) calculated with a 12-hour window for the deviation value Xm shown in Figure 10. Figure 12 shows the trend of the moving average value Xavm2 (m=1 to 5) calculated with a 24-hour window for the deviation value Xm shown in Figure 10. Figure 13 shows the trend of the moving average value Xavm3 (m=1 to 5) calculated with a 72-hour window for the deviation value Xm shown in Figure 10.

[0083] The threshold values ​​were calculated as described above using the data from the learning period for the deviation value Xm shown in Figure 10 and the moving average value Xavmj (j=1 to 3) shown in Figures 11 to 13. The results are shown in Figure 14. In Figure 14, the threshold values ​​Th1 to Th4 are shown in table format. That is, for each table of sensor unit Sm (m=1 to 5), the threshold values ​​Th1 to Th4 for the deviation value Xm and the moving average value Xavmj (j=1 to 3) are shown. As described above, the threshold values ​​Th1 to Th4 are the boundary values ​​for points 0 to 4.

[0084] As described above, thresholds Th1 (=K2) and Th2 (=K1) correspond to deviation values ​​Xm at 1 point and 2 points, respectively, and the deviation values ​​corresponding to these thresholds are those corresponding to the top 2.275% and 0.135%. Thresholds Th3 (=2×K1-K2) and Th4 (=2×K1-K3) correspond to deviation values ​​Xm at 3 points and 4 points, respectively, and are calculated from K1, K2, and K3. The deviation values ​​corresponding to thresholds Th3 (=2×K1-K2) and Th4 (=2×K1-K3) are those corresponding to 4σ and 5σ, respectively.

[0085] Comparing the thresholds shown in Figure 14 with respect to the period (i.e., window) for calculating the moving average, we see that the threshold decreases as the window size increases. Therefore, it can be seen that as the window size increases, the response to detecting moisture intrusion anomalies decreases, but the sensitivity to detecting moisture intrusion anomalies increases. Note that response refers to the continuous time required to detect an anomaly (i.e., moisture intrusion), and the response decreases as the continuous time increases. If the window for calculating the moving average is relatively long, even if a moisture intrusion anomaly occurs, if it is a sudden intrusion, the moving average will be smaller than when the window is relatively short. Therefore, if the window for calculating the moving average is relatively long, it is difficult to detect moisture intrusion anomalies unless they occur continuously (i.e., the continuous time required for detection is long), and the response is low. Conversely, by using the instantaneous deviation value Xm, it is possible to achieve moisture intrusion anomaly detection with low sensitivity but high response.

[0086] Using the threshold values ​​shown in Figure 14, points corresponding to the deviation values ​​Xm (m=1 to 5) shown in Figure 10 and the moving average values ​​Xavmj (j=1 to 3) shown in Figures 11 to 13 were determined. The results are shown in Figures 15 to 18. Specifically, Figure 15 shows the points corresponding to the deviation values ​​Xm (m=1 to 5) shown in Figure 10. Figure 16 shows the points corresponding to the moving average value Xavm1 (m=1 to 5) with a 12-hour window, as shown in Figure 11. Figure 17 shows the points corresponding to the moving average value Xavm2 (m=1 to 5) with a 24-hour window, as shown in Figure 12. Figure 18 shows the points corresponding to the moving average value Xavm3 (m=1 to 5) with a 72-hour window, as shown in Figure 13. Note that in Figures 15 to 18, as described above, the heating period of the panel on which the sensor unit S3 is placed is indicated by arrows.

[0087] Since 3 and 4 points are rare cases, the occurrence or increased frequency of such points suggests a possible abnormality. Looking at Figures 15 to 18, which show points related to deviation values ​​and moving averages, sensor units S1, S2, S4, and S5 show a low frequency of points exceeding "2" even after the learning period. The maximum point value is "3," and the frequency of points being "3" is low, and they do not occur consecutively. In contrast, for sensor unit S3, it can be seen in all of Figures 16 to 18 that high points (i.e., points of "3" or higher) occurred during the period when overheating was simulated (i.e., the heating period). This is thought to be because the overheating source promoted convection within the panel and the accompanying introduction of outside air. Furthermore, in Figure 15, which shows points related to deviation values, no high points of "3" or higher occurred for sensor unit S3, which is thought to be because the type of abnormality during the experiment was difficult to detect using deviation values. In cases of significant moisture intrusion into the distribution panel (such as rainwater intrusion from the outside), abnormalities can also be detected using deviation values. By using deviation values, the responsiveness of abnormality detection can be increased, and the time from the occurrence of an abnormality to detection can be shortened.

[0088] By comparing the changes in the points of sensor unit S3 during the heating period shown in Figures 16 to 18, we can compare the sensitivity of anomaly detection depending on the period (i.e., window) for calculating the moving average. In this case, heating was performed with a constant power (i.e., 54W, 90W, 144W, and 36W, respectively) for one week at a time. The points corresponding to the moving average calculated with a larger window (see Figures 17 and 18) showed higher sensitivity, and were able to detect anomalies in overheating caused by heaters of 54W or higher. Specifically, when the window was set to 24 hours and 72 hours, points of "3" or more occurred during the period from February 2nd to February 24th, which was the period when heating was performed with 54W, 90W, and 144W. The frequency of these occurrences was highest with a 72-hour window among the three types of windows. Furthermore, it can be seen that the duration of high points increases with increasing heat output (i.e., heater power consumption) for the points corresponding to the moving average with a 72-hour window (see Figure 18). In this way, by using points corresponding to multiple moving average values ​​with different windows for detecting anomalies in water intrusion, it is possible to implement both detection that prioritizes detection sensitivity and detection that prioritizes response.

[0089] Figure 19 shows the results of analyzing data related to sensor unit S3 after changing the method for calculating the threshold Th4 for the four points. Specifically, instead of Th4 = 2 × K1 - K3 as shown in Figure 14, Th4 = K1 + 0.08 was used. The thresholds Th1 to Th3 are the same as those shown in Figure 14. Comparing the graph in Figure 19 with the graphs related to sensor unit S3 in Figures 16 to 18, in Figure 19, there are a maximum of 3 points even during the heating period, suggesting the possibility of some kind of abnormality occurring. However, it can be estimated that the possibility of moisture intrusion is low.

[0090] The present invention has been described above by describing embodiments, but the embodiments described above are illustrative, and the present invention is not limited to the embodiments described above. The scope of the present invention is given with reference to the description in the detailed description of the invention, and includes all modifications within the meaning and scope equivalent to the wording contained herein. [Explanation of symbols]

[0091] 100 Anomaly Detection Systems 110 First Sensor Unit 112 Second Sensor Unit 114 Third Sensor Unit 116. Fourth Sensor Unit 118 Fifth Sensor Unit 120 1st switchboard 122 2nd switchboard 124 3rd switchboard 126 4th switchboard 128 No. 5 switchboard 130 Arithmetic equipment 140, 170 Control Unit 142, 172 Storage section 144, 174 Communications Department 146, 176 timers Buses 148 and 182 150 A / D conversion unit 152 Temperature Sensor 154 Humidity Sensor 178 Display section 180 Operation section

Claims

1. The first step involves acquiring temperature data and relative humidity data measured at the same time from N pairs of temperature and humidity sensors, each corresponding to a specific pair, which are installed in electrical equipment, where N is an integer of 2 or more. A second step involves calculating the volumetric absolute humidity using the temperature data and the relative humidity data corresponding to the temperature data. A third step involves calculating an average value from at least a portion of the N volumetric absolute humidity values ​​calculated in the second step, A fourth step in which the deviation value of each of the N volumetric absolute humidity values ​​is calculated by the volumetric absolute humidity / the average value - 1, A fifth step involves calculating a moving average value for each pair of temperature sensors and humidity sensors using the deviation values ​​obtained by repeatedly performing the first, second, third, and fourth steps for a predetermined period of time. A sixth step involves sorting the time-series data of the deviation value and the moving average value, generated by repeating the first, second, third, fourth, and fifth steps, in descending order. With respect to each of the aforementioned deviation values ​​and moving average values, the seventh step involves determining, from the data sorted in the sixth step, the data with ranks corresponding to L1%, L2%, and L3% of the total number of time-series data, starting from the highest rank, as K1, K2, and K3, respectively. The eighth step involves calculating multiple threshold values ​​from K1, K2, and K3 corresponding to each of the aforementioned deviation values ​​and moving average values, After the plurality of thresholds have been calculated, a ninth step is to determine whether or not an abnormality of moisture intrusion has occurred in the electrical equipment by comparing the deviation value and moving average value calculated by executing the second, third, fourth, and fifth steps using the temperature data and relative humidity data newly measured by the temperature sensor and the humidity sensor with the plurality of thresholds, L2 is larger than L1. An anomaly detection method characterized in that L3 is greater than L2.

2. The anomaly detection method according to claim 1, characterized in that the average value calculated in the third step is calculated by excluding the volume absolute humidity corresponding to the deviation value calculated in the fourth step from the N volume absolute humidity values.

3. Step 9 is, A tenth step in which points are set by comparing the deviation value and the moving average value calculated using the temperature data and relative humidity data measured after the plurality of thresholds have been calculated with the plurality of thresholds, An abnormality detection method according to claim 1 or claim 2, characterized by including an eleventh step of determining that an abnormality of moisture intrusion has occurred in the electrical equipment if the aforementioned point is above a predetermined value.

4. L1 is greater than 0.1 and less than 0.

3. L2 is greater than 1 and less than 5. The anomaly detection method according to claim 1 or claim 2, characterized in that L3 is greater than 10 and less than 20.

5. Let N be an integer of 2 or more, and the electrical equipment is equipped with N sets of temperature sensors and humidity sensors that correspond one-to-one. Including the computing unit, The aforementioned computing device is A first process to acquire temperature data and relative humidity data measured at the same time by each of the N sets of temperature sensors and humidity sensors, A second process for calculating volumetric absolute humidity using the temperature data and the relative humidity data corresponding to the temperature data, A third process for calculating an average value from at least a portion of the N volumetric absolute humidity values ​​calculated by the second process, A fourth process in which the deviation value of each of the N volumetric absolute humidity values ​​is calculated by the volumetric absolute humidity / the average value - 1, A fifth process, which calculates a moving average value for each pair of temperature sensors and humidity sensors using the deviation values ​​obtained by repeatedly performing the first, second, third, and fourth processes for a predetermined period, A sixth process sorts the time-series data of the deviation value and the moving average value, generated by repeating the first, second, third, fourth, and fifth processes, in descending order. With respect to each of the aforementioned deviation values ​​and moving average values, a seventh process is performed in which, from the data sorted by the sixth process, data with ranks corresponding to L1%, L2%, and L3% of the total number of time-series data, counting from the highest rank, are determined as K1, K2, and K3, respectively. An eighth process which calculates multiple threshold values ​​from K1, K2, and K3 corresponding to each of the aforementioned deviation value and moving average value, After the multiple thresholds are calculated, a ninth process is performed to determine whether or not an abnormality in moisture intrusion has occurred in the electrical equipment by comparing the deviation value and moving average value calculated by the second, third, fourth, and fifth processes using the temperature data and relative humidity data newly measured by the temperature sensor and humidity sensor with the multiple thresholds. L2 is larger than L1. An anomaly detection system characterized by L3 being greater than L2.

Images