Temperature monitoring system, machine learning device, and temperature monitoring device

Abstract

The objective is to provide a temperature monitoring system, a machine learning device, and a temperature monitoring device that can accurately predict the highly complex changes in internal temperature of equipment in equipment where the internal temperature changes due to the influence of the outside air temperature. [Solution] The system includes: a first input data acquisition unit 2100 that acquires training data including time-series data of the ambient temperature of the equipment and time-series data of the internal temperature of the equipment; a first learning model generation unit 240 that uses the training data to generate a trained model for inferring the internal temperature of the equipment under normal conditions from the time-series data of the ambient temperature of the equipment; a second input data acquisition unit 310 that acquires time-series data of the ambient temperature of the equipment; and a second temperature inference unit 330 that uses the trained model to output the internal temperature of the equipment under normal conditions from the time-series data of the ambient temperature of the equipment.

Description

Technical Field

[0001] The present disclosure relates to a temperature monitoring system, a machine learning device, and a temperature monitoring device.

Background Art

[0002] Conventionally, there has been disclosed a device for monitoring the presence or absence of temperature abnormalities inside an electrical panel, temperature measuring means attached to any one of the exhaust port of the electrical panel, the inner side of the upper surface of the electrical panel, and the outer side of the upper surface of the electrical panel, and an intake port to the electrical panel, and a data logger for recording temperature data measured by the temperature measuring means. (See Patent Document 1 below)

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, since the temperature change inside the electrical panel is complex, it is difficult to accurately predict the temperature inside the electrical panel by simply comparing the difference or response delay in temperature change between any one of the exhaust temperature, the temperature on the upper surface inside the panel, and the temperature on the upper surface outside the panel, which are the indicators for detecting overheating inside the electrical panel in Patent Document 1, and the intake temperature.

[0005] The present disclosure discloses a technique for solving the above problems, and an object thereof is to provide a temperature monitoring system, a machine learning device, and a temperature monitoring device that can accurately predict the temperature inside a facility that changes complexly in a facility where the temperature inside the facility changes due to the influence of the outside air temperature.

Means for Solving the Problems

[0006] The temperature monitoring system disclosed herein is A first input data acquisition unit acquires learning data including time-series data of the outside temperature of the equipment and time-series data of the internal temperature of the equipment. A first learning model generation unit generates a trained model for inferring the internal temperature of the equipment under normal conditions from time-series data of the external temperature of the equipment using the aforementioned training data, A second input data acquisition unit acquires time-series data of the outside temperature of the aforementioned equipment, The system includes a second temperature inference unit that uses the trained model to output the normal internal temperature of the equipment from time-series data of the outside temperature of the equipment.

[0007] Furthermore, the machine learning device described herein is A first input data acquisition unit acquires learning data including time-series data of the outside temperature of the equipment and time-series data of the internal temperature of the equipment. The system includes a first learning model generation unit that generates a trained model for inferring the normal internal temperature of the equipment from time-series data of the outside temperature of the equipment using the aforementioned training data.

[0008] Furthermore, the temperature monitoring device disclosed herein, A second input data acquisition unit acquires time-series data of the outside temperature of the equipment, The system includes a second temperature inference unit that outputs the normal internal temperature of the equipment from the time-series data of the outside temperature of the equipment, using a trained model for inferring the normal internal temperature of the equipment from the time-series data of the outside temperature of the equipment. [Effects of the Invention]

[0009] According to the temperature monitoring system, machine learning device, and temperature monitoring device disclosed herein, it is possible to accurately predict the complexly changing internal temperature of equipment in equipment where the internal temperature changes due to the influence of the outside air temperature. [Brief explanation of the drawing]

[0010] [Figure 1]It is a diagram showing the relationship between the time-series data of the outside air temperature and the time-series data of the temperature inside the facility. [Figure 2] It is a block diagram showing the schematic configuration of the facility according to Embodiment 1. [Figure 3] It is a block diagram showing the overall configuration of the temperature monitoring system according to Embodiment 1. [Figure 4] It is a block diagram showing the configuration of the machine learning device according to Embodiment 1. [Figure 5] It is a diagram showing the dataset of the outside air temperature and the temperature inside the facility of the machine learning device according to Embodiment 1. [Figure 6] It is a diagram showing the internal structure of the LSTM cell according to Embodiment 1. [Figure 7] It is a diagram showing the internal structure of the GRU cell according to Embodiment 1. [Figure 8] It is a flowchart for explaining the operation of the machine learning device according to Embodiment 1. [Figure 9] It is a block diagram showing the configuration of the temperature prediction unit according to Embodiment 1. [Figure 10] It is a flowchart showing the operation of the temperature prediction unit according to Embodiment 1. [Figure 11] It is a block diagram showing the configuration of the abnormal overheat determination unit according to Embodiment 1. [Figure 12] It is a flowchart showing the operation of the abnormal overheat determination unit of Embodiment 1. [Figure 13] It is a block diagram showing the overall configuration of the temperature monitoring system according to Embodiment 2. [Figure 14] It is a block diagram showing the configuration of the machine learning device according to Embodiment 2. [Figure 15] It is a flowchart showing the operation of the machine learning device according to Embodiment 2. [Figure 16] It is a block diagram showing the configuration of the temperature prediction unit according to Embodiment 2. [Figure 17] It is a flowchart showing the operation of the temperature prediction unit according to Embodiment 2. [Figure 18]It is a block diagram showing the configuration of the abnormal overheat determination unit according to Embodiment 2. [Figure 19] It is a flowchart showing the operation of the abnormal overheat determination unit according to Embodiment 2. [Figure 20] It is a block diagram showing the schematic configuration of the equipment according to Embodiment 3. [Figure 21] It is a block diagram showing an example of the hardware of the machine learning device, temperature prediction unit, and abnormal overheat determination unit according to the embodiment.

Embodiments for Carrying Out the Invention

[0011] [Basic Concept of the Present Disclosure] The present disclosure relates to a temperature monitoring system for equipment in which the temperature inside the equipment changes under the influence of the outside air temperature. Using time-series data of the outside air temperature as input data and time-series data of the temperature inside the equipment as label data, a trained model learned by machine learning based on the input data and the label data is used to output the normal temperature inside the equipment from the time-series data of the outside air temperature. The equipment of the present disclosure means electrical equipment that generates heat, such as a switchboard, a machine room where the switchboard is arranged, a motor, a boiler, air conditioning equipment where a heat pump, etc. are arranged, a machine room where an uninterruptible power supply device is arranged, and an electric vehicle charging station outdoors, and the equipment where the electrical equipment is arranged.

[0012] FIG. 1 is a diagram showing the relationship between time-series data A of the outside air temperature (Ambient Temperature) and time-series data B of the temperature (Internal Temperature of Outdoor-Installed Power Distribution System Enclosure) inside the equipment (power distribution equipment) installed outdoors.

[0013] Generally, due to heat generation from the energized parts inside the equipment, as shown in FIG. 1, the temperature inside the equipment is higher than the outside air temperature, and since the equipment housing etc. has a heat capacity, when the outside air temperature rises, the temperature inside the equipment also rises with a delay, and when the outside air temperature drops, the temperature inside the equipment also drops with a delay. However, the relationship between ambient temperature and internal equipment temperature is very complex. The rate of change in internal equipment temperature is affected by factors such as the temperature difference with ambient temperature or the rate of change in ambient temperature, and is also influenced by factors other than ambient temperature, such as fluctuations in the load current values ​​of equipment inside the facility. As mentioned above, changes in internal equipment temperature are influenced by various factors and are therefore very complex. It is difficult to derive accurate internal equipment temperature based on simple comparisons such as the difference with the outside air temperature or the rate of temperature change. Consequently, accurately capturing internal equipment temperature has been challenging in the past.

[0014] To address the above-mentioned problems, this disclosure constructs a trained model that predicts the normal internal temperature of equipment by learning the relationship between time-series data of ambient temperature and time-series data of internal temperature of equipment using machine learning, deep learning, etc. Then, by calculating the difference between the predicted internal temperature of equipment under normal conditions predicted by the trained model and the actual internal temperature of equipment, and determining whether this difference is above a threshold, it is possible to accurately predict abnormal overheating of the internal temperature of equipment.

[0015] The embodiments of this disclosure will be described in detail below.

[0016] [Description of the mechanism] Embodiment 1. Figure 2 is a block diagram showing the schematic configuration of electrical equipment according to Embodiment 1. As shown in Figure 2, a first temperature sensor 10 is located inside the equipment (electrical equipment such as power distribution equipment) 100 to measure the temperature inside the equipment. In addition, a second temperature sensor 20 is located outside the equipment 100 to measure the ambient temperature. The first temperature sensor 10, which measures the temperature inside the equipment, is located, for example, on the top surface of the charging part inside the equipment, on the top surface of the equipment control room, etc. The second temperature sensor 20, which measures the outside air temperature, measures the outside air temperature unaffected by the temperature inside the equipment. For example, it is desirable to measure the ambient temperature in an environment identical to the equipment installation environment, such as sunlight, wind direction, and wind speed, at a distance of 1 meter or more from the equipment 100. However, if the target is multiple pieces of equipment located in a factory, for example, the outside air temperature may be measured at one representative location on the roof of any factory building. The first temperature sensor 10 is used to acquire time-series data of the temperature inside the facility (measurement date and time and temperature), and the second temperature sensor 20 is used to acquire time-series data of the ambient temperature (measurement date and time and temperature). The second temperature sensor 20, which measures the outside air temperature, can be replaced with, for example, time-series data of outside air temperature every 10 minutes for various locations, which is published by the Japan Meteorological Agency.

[0017] Figure 3 is a block diagram showing the overall configuration of the temperature monitoring system according to Embodiment 1. As shown in Figure 3, the temperature monitoring system according to Embodiment 1 comprises a machine learning device 200 and a temperature monitoring device 500 having a temperature prediction unit 300 and an abnormal overheating determination unit 400. The machine learning device 200, the temperature prediction unit 300, and the abnormal overheating detection unit 400 are connected in a way that enables data transmission. This connection may be via a network other than a wired cable, such as a LAN (Local Area Network), the Internet, a public telephone network, or a combination thereof. Furthermore, the connection may be wireless, regardless of whether it is wired or wireless.

[0018] Figure 4 is a block diagram showing the configuration of the machine learning device according to Embodiment 1. As shown in Figure 4, the machine learning device 200 includes a first input data acquisition unit 2100, a first training data construction unit 230, a first training model generation unit 240, and a first data storage unit 250. The first input data acquisition unit 2100 includes an input data acquisition unit 210 and a label data acquisition unit 220. The input data acquisition unit 210 acquires input data A1, which is time-series data of outside temperature. The label data acquisition unit 220 acquires label data B1, which is time-series data of the temperature inside the equipment.

[0019] The first training data construction unit 230 constructs training data (training dataset) based on the time-series data A1 of ambient temperature acquired by the input data acquisition unit 210 and the time-series data B1 of the temperature inside the equipment acquired by the label data acquisition unit 220. The time-series data of ambient temperature from training time T1 to a predetermined past time is used as input data A1a, and the temperature inside the equipment at training time T1 is used as label data B1a. In other words, as shown in Figure 5, the first training data construction unit 230 uses time-series data of outside air temperature from training time T1 to a predetermined past time, for example, measurement data of outside air temperature every 5 minutes from training time T1 to 24 hours prior, as input data A1a, and the temperature inside the equipment at training time T1 as label data B1a, to create one set of training data. Then, this training data is used for training, for example, one year's worth, to construct training data. While the length of the time series data for input data A1a can be arbitrary, considering that the ambient temperature cycle is 24 hours, it is desirable to use time series data of 12 hours or more (half a 24-hour cycle) in order to appropriately predict the rate of change before and after the inflection point.

[0020] The first learning model generation unit 240 performs machine learning based on the learning data constructed by the first learning data construction unit 230 and generates a trained model M1 capable of predicting the temperature inside the equipment.

[0021] For example, using a regression analysis model in machine learning, when time-series data of outside temperature from the prediction point to a predetermined past point is input, model parameters (regression coefficients, intercepts) for accurately calculating the predicted value of the internal temperature of the equipment at the prediction point can be obtained as the learning result.

[0022] Furthermore, deep learning can be used as a machine learning technique. For example, a recurrent neural network (RNN) can be used as a deep learning technique because it can handle an unlimited number of input data points, long input data sequences, and preserve the order of the data sequences. Furthermore, it is also possible to utilize networks with a structure called LSTM (Long Short-Term Memory), for example. LSTM performs calculations to determine how much of the previous information to discard (forget gate), how much new information to input (input gate), and how much information to output (output gate). An example of the internal structure of an LSTM cell is shown in Figure 6. Alternatively, you can use a model called GRU (Gated Recurring Unit), which is a simplified version of the LSTM structure. GRU uses a two-stage calculation process: determining how much information to discard (reset gate) and how much information to add (update gate). An example of the internal structure of a GRU cell is shown in Figure 7. In addition, algorithms such as CNN (Convolutional Neural Network) can also be used. By using deep learning algorithms like those described above, the computational load increases, but the prediction accuracy can be improved.

[0023] The conditions for terminating or repeating the machine learning process can be arbitrarily defined. For example, the decision could be based on whether the RMSE (Root Mean Square Error) exceeds a threshold. Alternatively, the maximum or average error between the predicted value and the measured value, obtained by inputting time-series data of factors affecting the temperature inside the equipment that were not used in the training data, could be used as an indicator.

[0024] The first data storage unit 250 stores the trained model M1 generated by the first learning model generation unit 240 by performing machine learning based on time-series data of ambient temperature. Then, the trained model M1 stored in the first data storage unit 250 of the machine learning device 200 is transmitted to the temperature prediction unit 300 of the temperature monitoring device 500.

[0025] Figure 8 is a flowchart summarizing the operation of the machine learning device according to Embodiment 1 described above. As shown in Figure 8, in step S101, the input data acquisition unit 210 acquires input data A1, which is time-series data of outside temperature. In step S102, the label data acquisition unit 220 acquires label data B1, which is time-series data of the temperature inside the equipment. In step S103, the first training data construction unit 230 constructs training data (training dataset) based on the time-series data A1 of the outside temperature acquired by the input data acquisition unit 210 and the time-series data B1 of the temperature inside the equipment acquired by the label data acquisition unit 220. The time-series data A1a is the outside temperature for the period from training time T1 to a predetermined past time, and B1a is the temperature inside the equipment at training time T1. In step S104, the first learning model generation unit 240 performs machine learning based on the training data constructed in step S103. In step S105, the first learning model generation unit 240 determines whether or not to terminate the machine learning process. If machine learning is completed in step S105, the first learning model generation unit 240 saves the trained model M1 constructed by machine learning to the first data storage unit 250. In step S106, the trained model M1 constructed by the first learning model generation unit 240 is transmitted to the temperature prediction unit 300 of the temperature monitoring device 500.

[0026] Figure 9 is a block diagram showing the configuration of the temperature prediction unit according to Embodiment 1. As shown in Figure 9, the temperature prediction unit 300 includes a second input data acquisition unit 310, a second input data construction unit 320, a second temperature inference unit 330, and a second data storage unit 340.

[0027] Next, the operation of the temperature prediction unit in Embodiment 1 will be described based on the flowchart in Figure 10. As shown in Figure 10, in step S310, the second input data acquisition unit 310 acquires input data A2, which is time-series data of outside temperature. In step S320, the second input data construction unit 320 constructs inference data using time series data A2a of outside temperature from the prediction time to a predetermined past time, based on the outside temperature time series data A2 acquired by the second input data acquisition unit 310. In step S330, the second temperature inference unit 330 inputs the inference data, designated as input data A2a, constructed by the second input data construction unit 320, into the trained model M1 generated by the machine learning device 200. In step S340, the second temperature inference unit 330 performs calculations based on the trained model M1 to obtain the predicted result value C2 (prediction target date and time and temperature prediction value) of the normal temperature inside the equipment at the prediction time, and stores it in the second data storage unit 340. In step S350, the predicted value C2 of the normal internal temperature of the equipment, which has been stored in the second data storage unit 340, is transmitted to the abnormal overheating determination unit 400.

[0028] Figure 11 is a block diagram showing the configuration of the abnormal overheating detection unit according to Embodiment 1. As shown in Figure 11, the abnormal overheating determination unit 400 includes a third sensor data acquisition unit 410, a third prediction data acquisition unit 420, a third difference calculation unit 430, a third data storage unit 440, a third determination unit 450, and a third display unit 460.

[0029] Next, the operation of the abnormal overheating detection unit of Embodiment 1 will be described based on the flowchart in Figure 12. As shown in Figure 12, in step S410, the third sensor data acquisition unit 410 acquires time-series data (measurement date and time and temperature) of the actual measured temperature inside the equipment from the first temperature sensor 10 which measures the temperature inside the equipment as shown in Figure 2. In step S420, the third prediction data acquisition unit 420 acquires the predicted result value C2 (target date and time and predicted temperature) of the normal temperature inside the equipment from the temperature prediction unit 300. In step S430, the third difference calculation unit 430 combines the time-series data of the actual temperature inside the facility acquired by the third sensor data acquisition unit 410 with the time-series data of the predicted temperature inside the facility acquired by the third prediction data acquisition unit 420, calculates the difference between the actual temperature value and the predicted temperature value, and stores it in the third data storage unit 440. In step S440, the third determination unit 450 determines whether the difference between the measured temperature value and the predicted temperature value calculated by the third difference calculation unit 430 exceeds a preset threshold. In step S450, the third display unit 460 displays that if the difference between the measured temperature and the predicted temperature exceeds a preset threshold, it indicates that the temperature difference indicates abnormal overheating. In this case, it may be set to sound an alarm. Alternatively, for example, if remote monitoring is being performed, it may be displayed as an alert on the monitoring software, or an alarm may be sent via email to the administrator's smartphone or other portable device.

[0030] As described above, in Embodiment 1, time-series data of ambient temperature is used as input data, and time-series data of internal equipment temperature is used as label data. A trained model, which has been trained by machine learning based on the input data and the label data, is used to output the internal equipment temperature under normal conditions from the time-series data of ambient temperature. Therefore, we can provide a temperature monitoring system, machine learning device, and temperature monitoring device that can accurately predict the extremely complex and changing temperature inside the facility.

[0031] For example, if an annual inspection determines that no abnormal overheating has occurred, time-series data of ambient temperature and internal equipment temperature prior to that point can be acquired and used for machine learning. This allows for the construction of a highly accurate temperature monitoring system, machine learning device, and temperature monitoring device that take into account complex internal equipment temperature changes and are tailored to individual operating environments, without requiring manual intervention.

[0032] Furthermore, for example, if the daily fluctuation cycle of the load current value of equipment within the power distribution facility does not change significantly, accurate predictions of the internal temperature under normal conditions can be obtained by inputting only time-series data of ambient temperature and internal temperature. In this case, the measurement system can be constructed with only two temperature sensors, ambient temperature and internal temperature, resulting in very low implementation costs. Moreover, if meteorological time-series data from the Japan Meteorological Agency or other sources is used for ambient temperature, ambient temperature sensors become unnecessary, making implementation even more cost-effective.

[0033] Furthermore, for example, if an overheating incident occurs in power distribution equipment, a power outage is necessary to carry out restoration work. However, since the operation of buildings and factories connected to the power distribution equipment system also stops, it is not possible to immediately restore power. By using the temperature monitoring system, machine learning device, and temperature monitoring device of this embodiment, abnormal overheating can be detected in the precursor stage, and time can be secured before a malfunction occurs, allowing for planned responses such as restoration during the next annual inspection, thereby preventing damage caused by sudden power outages.

[0034] Furthermore, unlike the device described in Patent Document 1, this device does not use exhaust and intake air temperatures, which are affected by the internal temperature of the equipment. Therefore, it can prevent the inability to detect abnormalities, such as the inability to capture the effects of overheating as in the device described in Patent Document 1.

[0035] Furthermore, this method is not limited to electrical panels equipped with air intake ports, as in the device described in Patent Document 1, but can also be applied to electrical panels without air intake ports.

[0036] Embodiment 2. In Embodiment 1, time-series data of ambient temperature is used as input data, and time-series data of internal equipment temperature is used as label data. A trained model, which has been trained by machine learning based on the input data and the label data, is used to output the internal equipment temperature under normal conditions from the time-series data of ambient temperature. In Embodiment 2, time-series data of ambient temperature and time-series data of factors other than ambient temperature that affect the temperature inside the equipment are used as input data, and time-series data of the temperature inside the equipment are used as label data. A trained model, which has been trained by machine learning based on the input data and the label data, is used to output the normal temperature inside the equipment from the time-series data of ambient temperature and time-series data of factors other than ambient temperature that affect the temperature inside the equipment. Here, time-series data of factors that affect the temperature inside the facility, other than time-series data of ambient temperature, refers to time-series data that affect the temperature inside the facility, such as load current of equipment inside the facility, humidity, sunlight intensity, wind direction, wind speed, and rainfall. In particular, the load current of equipment within the facility, such as the main circuit current of a switchboard in a power distribution system, affects the temperature inside the facility. The time-series data of factors that affect the temperature inside the equipment, other than the time-series data of the outside temperature, may be one or more of the aforementioned load current, humidity, solar irradiance, wind direction, wind speed, and rainfall.

[0037] Figure 13 is a block diagram showing the overall configuration of the temperature monitoring system according to Embodiment 2. As shown in Figure 13, the temperature monitoring system according to Embodiment 2 comprises a machine learning device 200A and a temperature monitoring device 500A having a temperature prediction unit 300A and an abnormal overheating determination unit 400A.

[0038] Figure 14 is a block diagram showing the configuration of the machine learning device according to Embodiment 2. As shown in Figure 14, the machine learning device 200A includes a first input data acquisition unit 2100A, a first training data construction unit 230A, a first training model generation unit 240A, and a first data storage unit 250A. The first input data acquisition unit 2100A includes an input data acquisition unit 210A and a label data acquisition unit 220A. The input data acquisition unit 210A acquires input data A1, which is time-series data of the outside air temperature, and input data D1, which is time-series data of factors other than the outside air temperature that affect the temperature inside the equipment. The label data acquisition unit 220A acquires label data B1, which is time-series data of the temperature inside the equipment.

[0039] The first training data construction unit 230A constructs training data (training dataset) based on time-series data A1a of the outside temperature acquired by the input data acquisition unit 210A and time-series data D1a of factors other than outside temperature that affect the temperature inside the equipment, and time-series data B1a of the temperature inside the equipment acquired by the label data acquisition unit 220A. The training data constructs training data (training dataset) using time-series data A1a of the outside temperature for the period from training time T1 to a predetermined past time as input data, time-series data D1a of factors other than outside temperature that affect the temperature inside the equipment for the period from training time T1 to a predetermined past time as input data, and the temperature inside the equipment at training time T1 B1a as label data.

[0040] The first learning model generation unit 240A performs machine learning based on the learning data constructed by the first learning data construction unit 230A to construct a trained model M2 capable of predicting the temperature inside the equipment. For machine learning in Embodiment 2, the same methods as those used for machine learning in Embodiment 1 can be used. The first data storage unit 250A stores the trained model M2 constructed by the first learning model generation unit 240A.

[0041] Figure 15 is a flowchart showing the operation of the machine learning device according to Embodiment 2. As shown in Figure 15, in step S501, the input data acquisition unit 210A acquires input data, which is time-series data A1 of outside temperature. In step S502, the input data acquisition unit 210A acquires input data, which is time-series data D1 of factors other than the outside air temperature that affect the temperature inside the equipment. In step S503, the label data acquisition unit 220A acquires label data, which is time-series data B1 of the temperature inside the equipment. In step S504, the first training data construction unit 230A constructs training data (training dataset) based on the time series data A1 of ambient temperature acquired by the input data acquisition unit 210A, the time series data D1 of factors other than ambient temperature that affect the temperature inside the equipment, and the time series data B1 of the temperature inside the equipment acquired by the label data acquisition unit 220A. The training data constructs training data using the time series data A1a of ambient temperature for the period from training time T1 to a predetermined past time as input data, the time series data D1 of factors other than ambient temperature that affect the temperature inside the equipment for the period from training time T1 to a predetermined past time as input data, and the temperature inside the equipment at training time T1 as label data. In step S505, the first learning model generation unit 240A performs machine learning based on the training data constructed in step S504. In step S506, the first learning model generation unit 240A determines whether or not to terminate the machine learning process. If machine learning is completed in step S506, the first learning model generation unit 240A saves the trained model M2 constructed by machine learning to the first data storage unit 250A. In step S507, the trained model M2 constructed by the first learning model generation unit 240A is transmitted to the temperature prediction unit 300A of the temperature monitoring device 500A.

[0042] Figure 16 is a block diagram showing the configuration of the temperature prediction unit according to Embodiment 2. As shown in Figure 16, the temperature prediction unit 300A includes a second input data acquisition unit 310A, a second input data construction unit 320A, a second temperature inference unit 330A, and a second data storage unit 340A.

[0043] Next, the operation of the temperature prediction unit in Embodiment 2 will be described based on the flowchart in Figure 17. As shown in Figure 17, in step S601, the second input data acquisition unit 310A acquires input data, which is time-series data A2 of outside temperature. In step S602, the second input data acquisition unit 310A acquires input data D2, which is time-series data of factors other than the outside air temperature that affect the temperature inside the equipment. In step S603, the second input data construction unit 320A constructs inference data using time series data A2a of outside air temperature for the period from the prediction time to a predetermined past time, based on time series data A2 of outside air temperature acquired by the second input data acquisition unit 310A. In addition, the second input data construction unit 320A constructs inference data using time series data D2a of factors other than outside air temperature that affect the temperature inside the equipment for the period from the prediction time to a predetermined past time, based on time series data D2 of factors other than outside air temperature that affect the temperature inside the equipment from the second input data acquisition unit 310A. In step S604, the second temperature inference unit 330A inputs the time series data A2a and time series data D2a, constructed by the second input data construction unit 320A, to the trained model M2 constructed by the machine learning device 200A. In step S605, the second temperature inference unit 330A performs calculations based on the trained model M2 to obtain the predicted result value C3 (prediction target date and time and temperature prediction value) of the normal temperature inside the equipment at the prediction time, and stores it in the second data storage unit 340A. In step S606, the predicted value C3 of the normal internal temperature of the equipment, which has been stored in the second data storage unit 340A, is transmitted to the abnormal overheating determination unit 400A.

[0044] Figure 18 is a block diagram showing the configuration of the abnormal overheating detection unit according to Embodiment 2. As shown in Figure 18, the abnormal overheating determination unit 400A includes a third sensor data acquisition unit 410A, a third prediction data acquisition unit 420A, a third difference calculation unit 430A, a third data storage unit 440A, a third determination unit 450A, and a third display unit 460A.

[0045] Next, the operation of the abnormal overheating detection unit in Embodiment 2 will be described based on the flowchart in Figure 19. As shown in Figure 19, in step S710, the third sensor data acquisition unit 410A acquires time-series data (measurement date and time and temperature) of the actual temperature inside the equipment from the first temperature sensor 10 which measures the temperature inside the equipment as shown in Figure 2. In step S720, the third prediction data acquisition unit 420A acquires the predicted result value C3 (target date and time and predicted temperature) of the normal temperature inside the equipment from the temperature prediction unit 300A. In step S730, the third difference calculation unit 430A combines the time-series data of the actual temperature inside the equipment acquired by the third sensor data acquisition unit 410A with the time-series data of the predicted temperature inside the equipment acquired by the third prediction data acquisition unit 420A, calculates the difference between the actual temperature value and the predicted temperature value, and stores it in the third data storage unit 440A. In step S740, the third determination unit 450A determines whether the difference between the measured temperature value and the predicted temperature value calculated by the third difference calculation unit 430A exceeds a preset threshold. In step S750, the third display unit 460A displays that if the difference between the measured temperature and the predicted temperature exceeds a preset threshold, it indicates that the temperature difference indicates abnormal overheating. In this case, it may be set to sound an alarm. Alternatively, for example, if remote monitoring is being performed, it may be displayed as an alert on the monitoring software, or an alarm may be sent via email to the administrator's smartphone or other portable device.

[0046] As described above, in Embodiment 2, in equipment where the internal temperature changes due to the influence of the outside air temperature, a trained model is used, which has been trained by machine learning using time-series data of outside air temperature and time-series data of factors other than outside air temperature that affect the internal temperature of the equipment as input data, and time-series data of the internal temperature of the equipment as label data, to output the normal internal temperature of the equipment from the time-series data of outside air temperature and time-series data of factors other than outside air temperature that affect the internal temperature of the equipment. Therefore, we can provide a temperature monitoring system, machine learning device, and temperature monitoring device that can more accurately predict the highly complex and ever-changing temperature inside the facility.

[0047] In other words, by adding time-series data of at least one of the following as input values—load current, humidity, illuminance, wind speed, wind direction, or rainfall—the time-series changes of these parameters can be taken into account, further improving prediction accuracy.

[0048] Furthermore, the temperature monitoring device of the second embodiment can also obtain predicted values ​​of the temperature inside the equipment by inputting arbitrary time-series data. For example, it can simulate whether there is a margin of safety for the allowable temperature of the equipment inside the equipment, or whether there is a risk of condensation occurring inside the equipment due to changes in outside temperature, etc., by inputting temperature data and humidity data. This can be used to improve the equipment environment, such as by adding heaters inside the equipment.

[0049] Embodiment 3. In Embodiment 1, a first temperature sensor 10 is placed in the equipment 100 to measure the temperature inside the equipment. Then, using the time-series data of the ambient temperature as input data and the time-series data of the temperature inside the equipment measured by the first temperature sensor 10 as label data, a trained model trained by machine learning is used to output the normal temperature inside the equipment from the time-series data of the ambient temperature. In Embodiment 3, instead of the first temperature sensor 10 that measures the temperature inside the equipment, an infrared camera 30 is placed in the equipment 100 as shown in Figure 20. The infrared camera 30 is used to photograph equipment inside the equipment 100, such as energized parts, and the temperature of the equipment obtained is taken as the temperature inside the equipment. The time-series data of ambient temperature is used as input data, and the time-series data of the temperature inside the equipment measured by the infrared camera 30 is used as label data. A trained model trained by machine learning is used to output the normal temperature inside the equipment from the time-series data of ambient temperature.

[0050] By using the temperature data of equipment within the facility acquired by the infrared camera 30 as label data, it is possible to identify not only whether or not abnormal overheating has occurred within the facility, but also the location where the overheating occurred. In other words, by taking images so that multiple pieces of equipment within the facility are within the field of view of the infrared camera 30, the temperature of equipment over a wide area can be acquired at once, and costs can be reduced when monitoring overheating of multiple pieces of equipment within the facility.

[0051] The infrared camera 30 suffers from reduced measurement accuracy due to factors such as the emissivity, surface properties, reflectivity, and measurement distance of the equipment within the facility. In this embodiment, as long as fluctuations in the measured temperature can be understood, even if the fluctuation scale or the measured temperature value shifts, a learning model can be constructed accordingly, thus preventing any hindrance to detecting abnormal overheating of equipment within the facility.

[0052] However, instead of using the infrared camera 30, the temperature data obtained by directly measuring the temperature of equipment within the facility using a sensor such as a thermistor or thermocouple as the first temperature sensor 10 may also be used as label data. Thermistors, thermocouples, etc., can measure the temperature of equipment inside the facility more accurately without being affected by the emissivity or surface properties of the equipment inside the facility, unlike infrared cameras 30. Therefore, by directly measuring the temperature of equipment inside the facility using thermistors, thermocouples, etc., it is possible to obtain predicted values ​​that are closer to the actual temperature. This allows for more accurate detection of abnormal overheating of equipment inside the facility, as well as determining whether the temperature inside the facility exceeds the allowable temperature of the equipment, and whether there are any risks due to fluctuations in ambient temperature, etc., through simulation.

[0053] As described above, according to Embodiment 3, the temperature inside the facility is the temperature of the equipment inside the facility as captured by the infrared camera. By capturing images so that multiple pieces of equipment inside the facility are within the field of view of the infrared camera, the temperatures of equipment inside a wide area can be acquired at once, and overheating of multiple pieces of equipment inside the facility can be monitored.

[0054] Alternatively, since the internal temperature of the equipment is temperature data obtained by directly measuring the temperature of the equipment inside the equipment using a sensor, it is possible to predict the temperature of the equipment inside the equipment more accurately.

[0055] The machine learning device 200, temperature prediction unit 300, and abnormal overheating detection unit 400 described in Embodiment 1, and the machine learning device 200A, temperature prediction unit 300A, and abnormal overheating detection unit 400A described in Embodiment 2, consist of a processor 1000 and a storage device 1010, as shown in Figure 21 as an example of the hardware. The storage device 1010 includes a volatile storage device such as random access memory (not shown) and a non-volatile auxiliary storage device such as flash memory. Alternatively, a hard disk may be provided as an auxiliary storage device instead of flash memory. The processor 1000 executes a program input from the storage device 1010. In this case, the program is input to the processor 1000 from the auxiliary storage device via the volatile storage device. The processor 1000 may also output data such as calculation results to the volatile storage device of the storage device 1010, or it may save the data to the auxiliary storage device via the volatile storage device.

[0056] While this disclosure describes various exemplary embodiments and examples, the various features, aspects, and functions described in one or more embodiments are not limited to the application of a particular embodiment, but can be applied individually or in various combinations to the embodiments. Accordingly, countless variations not illustrated are conceivable within the scope of the art disclosed in this specification. These include, for example, modifying, adding or omitting at least one component, or even extracting at least one component and combining it with components of other embodiments.

[0057] The various aspects of this disclosure are summarized below as an appendix.

[0058] (Note 1) A first input data acquisition unit acquires learning data including time-series data of the outside temperature of the equipment and time-series data of the internal temperature of the equipment. A first learning model generation unit generates a trained model for inferring the internal temperature of the equipment under normal conditions from time-series data of the external temperature of the equipment using the aforementioned training data, A second input data acquisition unit acquires time-series data of the outside temperature of the aforementioned equipment, A second temperature inference unit that uses the trained model to output the normal internal temperature of the equipment from time-series data of the outside temperature of the equipment, A temperature monitoring system equipped with the following features. (Note 2) The aforementioned training data further includes, in addition to time-series data of the ambient temperature of the equipment, time-series data of factors that affect the temperature inside the equipment. The first learning model generation unit generates the trained model for inferring the normal internal temperature of the equipment using the training data, The second input data acquisition unit acquires time-series data of factors that affect the internal temperature of the equipment, in addition to the time-series data of the outside temperature of the equipment. The temperature monitoring system described in Appendix 1, wherein the second temperature inference unit uses the trained model to output the normal internal temperature of the equipment from time-series data of the ambient temperature of the equipment and time-series data of factors other than the ambient temperature of the equipment that affect the internal temperature of the equipment. (Note 3) The temperature monitoring system described in Appendix 2 further comprises a first learning data construction unit that constructs the learning data using time-series data of the outside temperature of the equipment during a period from the learning point to a predetermined past point in time, and time-series data of factors that affect the internal temperature of the equipment during the same period, in addition to the time-series data of the outside temperature of the equipment during the same period, as input data, and the internal temperature of the equipment at the learning point as label data. (Note 4) The system further comprises a second input data construction unit that constructs inference data using as input data time-series data of the outside temperature of the equipment during a period from the prediction point to a predetermined past point in time, and time-series data of factors that affect the internal temperature of the equipment during the same period, in addition to the time-series data of the outside temperature of the equipment during the same period. The temperature monitoring device described in Appendix 2, wherein the second temperature inference unit uses the trained model to output a predicted value of the normal internal temperature of the equipment at the prediction time from the inference data. (Note 5) The temperature monitoring system described in any one of the appendices 2 to 4, wherein factors affecting the internal temperature of the equipment include the load current of the equipment within the equipment. (Note 6) A temperature monitoring system according to any one of the appendices 1 to 5, comprising an abnormal overheating determination unit that detects an abnormality in the equipment based on the difference between a predicted value of the equipment's internal temperature under normal conditions output by the second temperature inference unit and an actual measured value of the equipment's internal temperature. (Note 7) The temperature monitoring system described in any one of the appendices 1 to 6, wherein the internal temperature of the equipment is the internal temperature of the power distribution equipment. (Note 8) The temperature monitoring system described in any one of the appendices 1 to 7, wherein the internal temperature of the equipment is the temperature of the equipment inside the equipment as captured by an infrared camera. (Note 9) The temperature monitoring system described in any one of the appendices 1 to 7, wherein the internal temperature of the equipment is the temperature measured by a sensor at the equipment inside the equipment. (Note 10) A first input data acquisition unit acquires learning data including time-series data of the outside temperature of the equipment and time-series data of the internal temperature of the equipment. A first learning model generation unit generates a trained model for inferring the internal temperature of the equipment under normal conditions from time-series data of the external temperature of the equipment using the aforementioned training data, A machine learning device equipped with the following features. (Note 11) The aforementioned training data further includes, in addition to time-series data of the ambient temperature of the equipment, time-series data of factors that affect the temperature inside the equipment. The machine learning apparatus described in Appendix 10, wherein the first learning model generation unit generates a trained model for inferring the normal internal temperature of the equipment from time-series data of the outside temperature of the equipment and time-series data of factors other than the outside temperature of the equipment that affect the internal temperature of the equipment, using the training data. (Note 12) The machine learning apparatus described in Appendix 11, further comprising a first training data construction unit that constructs training data using time-series data of the outside temperature of the equipment during a period from the training time to a predetermined past time, and time-series data of factors that affect the internal temperature of the equipment during the same period, in addition to the time-series data of the outside temperature of the equipment during the same period, as input data, and the internal temperature of the equipment at the training time as label data. (Note 13) A second input data acquisition unit acquires time-series data of the outside temperature of the equipment, A second temperature inference unit outputs the normal internal temperature of the equipment from the time-series data of the outside temperature of the equipment, using a trained model for inferring the normal internal temperature of the equipment from the time-series data of the outside temperature of the equipment. A temperature monitoring device equipped with the following features. (Note 14) The second input data acquisition unit acquires time-series data of factors that affect the internal temperature of the equipment, in addition to the time-series data of the outside temperature of the equipment. The temperature monitoring device described in Appendix 13, wherein the second temperature inference unit uses the trained model to output the normal internal temperature of the equipment from time-series data of the ambient temperature of the equipment and time-series data of factors other than the ambient temperature of the equipment that affect the internal temperature of the equipment. (Note 15) The system further comprises a second input data construction unit that constructs inference data using as input data time-series data of the outside temperature of the equipment during a period from the prediction point to a predetermined past point in time, and time-series data of factors that affect the internal temperature of the equipment during the same period, in addition to the time-series data of the outside temperature of the equipment during the same period. The temperature monitoring device described in Appendix 14, wherein the second temperature inference unit uses the trained model to output a predicted value of the normal internal temperature of the equipment at the prediction time from the inference data. (Note 16) A third prediction data acquisition unit acquires a predicted value of the internal temperature of the equipment under normal conditions, A third sensor data acquisition unit acquires the actual measured temperature inside the equipment, A temperature monitoring device according to any one of the appendices 13 to 15, further comprising: a third determination unit that determines an abnormality in the equipment based on the difference between a predicted value of the internal temperature of the equipment under normal conditions and a measured value of the internal temperature of the equipment. [Explanation of symbols]

[0059] 10 First temperature sensor, 20 Second temperature sensor, 30 Infrared camera, 100 Equipment, 200 Machine learning device, 210 Input data acquisition unit, 220 Label data acquisition unit, 230 First training data construction unit, 240 First training model generation unit, 250 First data storage unit, 300 Temperature prediction unit, 310 Second input data acquisition unit, 320 Second input data construction unit, 330 Second temperature inference unit, 340 Second data storage unit, 400 Abnormal overheating determination unit, 410 Third sensor data acquisition unit, 420 Third prediction data acquisition unit, 430 Third difference calculation unit, 440 Third data storage unit, 450 Third determination unit, 460 Third display unit, 500 Temperature monitoring device, 200A Machine learning device, 210A Input data acquisition unit, 220A Label data acquisition unit, 230A First training data construction unit, 240A First learning model generation unit, 250A First data storage unit, 300A Temperature prediction unit, 310A Second input data acquisition unit, 320A Second input data construction unit, 330A Second temperature inference unit, 340A Second data storage unit, 400A Abnormal overheating detection unit, 410A Third sensor data acquisition unit, 420A Third prediction data acquisition unit, 430A Third difference calculation unit, 440A Third data storage unit, 450A Third determination unit, 460A Third display unit, 500A Temperature monitoring device, 1000 Processor, 1010 Storage device, 2100 First input data acquisition unit, 2100A First input data acquisition unit.

Claims

1. A first input data acquisition unit acquires learning data including time-series data of the outside temperature of the equipment and time-series data of the internal temperature of the equipment. A first learning model generation unit generates a trained model for inferring the internal temperature of the equipment under normal conditions from time-series data of the external temperature of the equipment using the aforementioned training data, A second input data acquisition unit acquires time-series data of the outside temperature of the aforementioned equipment, A second temperature inference unit that uses the trained model to output the normal internal temperature of the equipment from time-series data of the outside temperature of the equipment, A temperature monitoring system equipped with the following features.

2. The aforementioned training data further includes, in addition to time-series data of the ambient temperature of the equipment, time-series data of factors that affect the temperature inside the equipment. The first learning model generation unit generates the trained model for inferring the normal internal temperature of the equipment using the training data, The second input data acquisition unit acquires time-series data of factors that affect the internal temperature of the equipment, in addition to the time-series data of the outside temperature of the equipment. The temperature monitoring system according to claim 1, wherein the second temperature inference unit uses the trained model to output the normal internal temperature of the equipment from time-series data of the ambient temperature of the equipment and time-series data of factors other than the ambient temperature of the equipment that affect the internal temperature of the equipment.

3. The temperature monitoring system according to claim 2, further comprising a first learning data construction unit that constructs the learning data using time-series data of the outside temperature of the equipment during a period from the learning point to a predetermined past point in time, and time-series data of factors that affect the internal temperature of the equipment during the period in addition to the time-series data of the outside temperature of the equipment during the period as input data, and the internal temperature of the equipment at the learning point as label data.

4. The system further comprises a second input data construction unit that constructs inference data using as input data time-series data of the outside temperature of the equipment during a period from the prediction point to a predetermined past point in time, and time-series data of factors that affect the internal temperature of the equipment during the same period, in addition to the time-series data of the outside temperature of the equipment during the same period. The temperature monitoring device according to claim 2, wherein the second temperature inference unit uses the trained model to output a predicted value of the normal internal temperature of the equipment at the prediction time from the inference data.

5. The temperature monitoring system according to any one of claims 2 to 4, wherein the factors affecting the temperature inside the equipment include the load current of the equipment inside the equipment.

6. A temperature monitoring system according to any one of claims 1 to 4, further comprising an abnormal overheating determination unit that detects an abnormality in the equipment based on the difference between a predicted value of the equipment's internal temperature under normal conditions output by the second temperature inference unit and an actual measured value of the equipment's internal temperature.

7. The temperature monitoring system according to any one of claims 1 to 4, wherein the internal temperature of the equipment is the internal temperature of the power distribution equipment.

8. The temperature monitoring system according to any one of claims 1 to 4, wherein the internal temperature of the equipment is the temperature of the equipment inside the equipment as captured by an infrared camera.

9. The temperature monitoring system according to any one of claims 1 to 4, wherein the internal temperature of the equipment is the temperature measured by a sensor at the temperature of the equipment inside the equipment.

10. A first input data acquisition unit acquires learning data including time-series data of the outside temperature of the equipment and time-series data of the internal temperature of the equipment. A first learning model generation unit generates a trained model for inferring the internal temperature of the equipment under normal conditions from time-series data of the external temperature of the equipment using the aforementioned training data, A machine learning device equipped with the following features.

11. The aforementioned training data further includes, in addition to time-series data of the ambient temperature of the equipment, time-series data of factors that affect the temperature inside the equipment. The machine learning apparatus according to claim 10, wherein the first learning model generation unit generates a trained model for inferring the normal internal temperature of the equipment from time-series data of the outside temperature of the equipment and time-series data of factors other than the outside temperature of the equipment that affect the internal temperature of the equipment, using the training data.

12. The machine learning apparatus according to claim 11, further comprising a first training data construction unit that constructs training data using time-series data of the outside temperature of the equipment during a period from the training time to a predetermined past time, and time-series data of factors that affect the internal temperature of the equipment during the period, in addition to the time-series data of the outside temperature of the equipment during the period, as input data, and the internal temperature of the equipment at the training time as label data.

13. A second input data acquisition unit acquires time-series data of the outside temperature of the equipment, A second temperature inference unit outputs the normal internal temperature of the equipment from the time-series data of the outside temperature of the equipment, using a trained model for inferring the normal internal temperature of the equipment from the time-series data of the outside temperature of the equipment. A temperature monitoring device equipped with the following features.

14. The second input data acquisition unit acquires time-series data of factors that affect the internal temperature of the equipment, in addition to the time-series data of the outside temperature of the equipment. The temperature monitoring device according to claim 13, wherein the second temperature inference unit uses the trained model to output the normal internal temperature of the equipment from time-series data of the ambient temperature of the equipment and time-series data of factors other than the ambient temperature of the equipment that affect the internal temperature of the equipment.

15. The system further comprises a second input data construction unit that constructs inference data using as input data time-series data of the outside temperature of the equipment during a period from the prediction point to a predetermined past point in time, and time-series data of factors that affect the internal temperature of the equipment during the same period, in addition to the time-series data of the outside temperature of the equipment during the same period. The temperature monitoring device according to claim 14, wherein the second temperature inference unit uses the trained model to output a predicted value of the normal internal temperature of the equipment at the prediction time from the inference data.

16. A third prediction data acquisition unit acquires a predicted value of the internal temperature of the equipment under normal conditions, A third sensor data acquisition unit acquires the actual temperature inside the equipment of the aforementioned equipment, A temperature monitoring device according to any one of claims 13 to 15, further comprising: a third determination unit that determines an abnormality in the equipment based on the difference between a predicted value of the internal temperature of the equipment under normal conditions and a measured value of the internal temperature of the equipment.

Images