Lightning strike prediction system and lightning strike prediction method

Abstract

To provide a lightning strike prediction system capable of improving lightning strike prediction accuracy in a wide lightning generation area (region).SOLUTION: A lightning strike prediction system 1 includes an electrostatic field value acquisition unit 10 for acquiring atmospheric electrostatic field values at a plurality of measurement points Pn arranged in a plurality of areas An, an information acquisition unit 20 for acquiring environmental information in a predetermined range around the measurement points Pn, and a lightning strike probability prediction unit 30 for predicting a lightning strike probability. The area An is set as a predetermined range including one measurement point Pn. The lightning strike probability prediction unit 30 predicts a lightning strike probability for each area An based on the electrostatic field value and the environmental information, and notifies an area An in which the predicted lightning strike probability is equal to or greater than a predetermined threshold value. The lightning strike probability prediction unit 30 predicts a lightning strike probability based on a prediction algorithm generated by using, as learning data, the amount of change in the electrostatic field value and the environmental information in a predetermined period.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a lightning strike prediction system and a lightning strike prediction method, and more particularly to a lightning strike prediction system and a lightning strike prediction method for predicting areas where lightning is likely to strike based on electrostatic field values ​​in the atmosphere. [Background technology]

[0002] Conventionally, lightning strikes have been predicted by predicting the movement of lightning based on the history of lightning strikes and meteorological information, and issuing an alarm, but in such cases there has been a problem in that it is not possible to predict the first lightning strike. Therefore, a lightning strike warning device installed at a specified installation location is used to acquire electrostatic field values ​​and lightning activity (weather information data), and based on the acquired electrostatic field values ​​and lightning activity, the possibility of a lightning strike is predicted and an alarm is issued (for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-018254 Summary of the Invention [Problem to be solved by the invention]

[0004] The conventional technology disclosed in Patent Document 1 predicts lightning strikes at a single location based on the electrostatic field value measurement results and lightning activity at that location. However, the conventional technology disclosed in Patent Document 1 has a problem in that the range of lightning strike predictions is limited, reducing the accuracy of the lightning strike predictions. Furthermore, the conventional technology disclosed in Patent Document 1 predicts the movement of lightning based on the history of lightning strikes and meteorological information, and issues an alert or the like. However, the conventional technology disclosed in Patent Document 1 predicts lightning strikes based on the electrostatic field value and lightning activity at a single location, which has the problem of making false or overlooked predictions.

[0005] Therefore, an object of the present invention is to provide a lightning strike prediction system and a lightning strike prediction method that can improve the accuracy of lightning strike prediction in a wide-area lightning strike area (region). [Means for solving the problem]

[0006] (1) The lightning strike prediction system of the present invention, which is provided to solve the above-mentioned problems, comprises an electrostatic field value acquisition unit that acquires atmospheric electrostatic field values ​​at multiple measurement points located in multiple regions, an information acquisition unit that acquires environmental information within a predetermined range centered on the measurement points, and a lightning strike probability prediction unit that predicts the probability of lightning strikes, wherein the regions are set as predetermined ranges that include one of the measurement points, and the lightning strike probability prediction unit predicts the lightning strike probability for each region based on the electrostatic field value and the environmental information, and notifies the regions where the predicted lightning strike probability is equal to or greater than a predetermined threshold.

[0007] The lightning strike prediction system of the present invention can predict the probability of lightning strikes based on atmospheric electrostatic field values ​​at multiple measurement points located in multiple regions and environmental information for a predetermined range centered on the measurement points, thereby making it possible to predict lightning strikes based on meteorological information for a wide area. Therefore, the lightning strike prediction system of the present invention can make highly accurate lightning strike predictions. Here, the environmental information can include various types of environmental information, such as information about the weather at each measurement point and each region, topographical information, and physical information (altitude, atmospheric pressure, humidity, and temperature).

[0008] Furthermore, the lightning strike prediction system of the present invention predicts the probability of lightning strikes for multiple regions and notifies regions where the predicted lightning strike probability is above a predetermined threshold, preventing unnecessary notifications. Therefore, the lightning strike prediction system of the present invention can prevent notifications from becoming cumbersome and lightning strike predictions from ending up as false alarms. Various notification methods can be used for notifications, such as issuing an alarm or sending notifications via email. Examples of recipients of notifications include various businesses and facilities that are highly susceptible to lightning strikes (e.g., semiconductor factories, automobile factories, steel mills with blast furnaces, schools, etc.).

[0009] (2) In the lightning strike prediction system of the present invention described above, the lightning strike probability prediction unit may be characterized in that it predicts the lightning strike probability based on a prediction algorithm generated using learning data based on the amount of change in the electrostatic field value and the environmental information over a predetermined period of time.

[0010] The lightning prediction system of the present invention can predict the probability of lightning strikes (also called AI prediction) based on a prediction algorithm generated as learning data from electrostatic field values ​​and environmental information, and can therefore make appropriate lightning strike probability predictions according to the environmental characteristics of each region. As a result, the lightning prediction system of the present invention can reduce false positives and missed lightning strike predictions and provide highly accurate lightning strike predictions.

[0011] Furthermore, the lightning strike prediction system of the present invention uses the amount of change in the electrostatic field value and environmental information over a predetermined time period as training data, thereby reducing the influence of the location of the measurement point (e.g., elevation differences). Specifically, the lightning strike prediction system of the present invention uses the amount of change over a predetermined time period, rather than the absolute values ​​of the electrostatic field value and environmental information (e.g., atmospheric pressure), as training data, thereby reducing the influence of, for example, elevation differences at the measurement point. Here, the training data may include, for example, the amount of change per predetermined time for atmospheric pressure, electrostatic field value, humidity, temperature, etc. Furthermore, measurements of the electrostatic field value and environmental information (e.g., atmospheric pressure, humidity, temperature) at the measurement point may be performed, for example, at 10-second intervals, and the training data may be sampled, for example, by sampling the amount of change per minute for items that change significantly over time, such as electrostatic field values, or by sampling the amount of change per 10 minutes for items that change relatively slowly over time, such as environmental information. Note that the measurement interval and the interval at which training data is acquired may be appropriately changed depending on the environment of the measurement point, the measurement equipment, etc. Here, various prediction algorithms, such as XGBoost and LightGBM, may be used. Furthermore, by configuring the lightning strike prediction system of the present invention as described above in (2), it is possible to prevent unnecessary notifications (for example, issuing alarms).

[0012] (3) The lightning strike prediction system of the present invention described above may be characterized in that it designates an area where the predicted lightning strike probability is equal to or greater than a predetermined threshold as a candidate area, and notifies the candidate area on the condition that the electrostatic field value in the candidate area is higher than the electrostatic field values ​​in other candidate areas adjacent to the candidate area.

[0013] The lightning prediction system of the present invention is designed to notify candidate areas where the probability of lightning strikes is equal to or greater than a threshold and where the electrostatic field value in the candidate area is higher than the electrostatic field values ​​in other adjacent candidate areas. Therefore, the lightning prediction system of the present invention can not only notify areas with a high probability of lightning strikes, but also areas with a high probability of lightning strikes and high surrounding electrostatic field values. This allows the lightning prediction system of the present invention to make more accurate lightning strike predictions and further reduce false positives and missed lightning strike predictions.

[0014] (4) In the above-described lightning strike prediction system of the present invention, the lightning strike probability prediction unit may be characterized in that it uses the electrostatic field value of 1 kV / m or more as a threshold value for predicting the lightning strike probability.

[0015] The lightning strike prediction system of the present invention uses electrostatic field values ​​of 1 kV / m or more to predict the probability of a lightning strike, thereby reducing the influence of noise and enabling the system to make highly accurate lightning strike predictions.

[0016] Here, it is generally believed that the scale of a thundercloud often forms within a radius of approximately 10 km. Meanwhile, the acquisition range of electrostatic field values ​​at the electrostatic field value acquisition unit (electric field sensor, etc.) at the measurement point is set to a radius of approximately 5 km. Therefore, in consideration of the scale of a thundercloud and the acquisition range of electrostatic field values, it is believed that if electrostatic field value measurement points are placed at intervals of 10 km or less, it will be possible to comprehensively predict lightning strikes in the surrounding area.

[0017] (5) Therefore, the lightning strike prediction system of the present invention described above may be characterized in that the plurality of measurement points are located within 10 km of each other adjacent measurement points.

[0018] If the lightning prediction system of the present invention is configured as described above in (5), it can efficiently predict lightning strikes with a minimum number of measurement points (electrostatic field value acquisition units). This allows the lightning prediction system of the present invention to perform accurate lightning strike predictions while reducing costs. Here, the measurement points may be arranged, for example, in a grid pattern. This allows the lightning prediction system of the present invention to perform more comprehensive and accurate lightning strike predictions for each region.

[0019] (6) The lightning strike prediction system of the present invention may be characterized in that the environmental information includes at least one of altitude, atmospheric pressure, temperature, and humidity.

[0020] By configuring the lightning strike prediction system of the present invention as described above in (6), the influence of differences in topography and installation environment in each region can be reduced, thereby enabling more accurate lightning strike prediction. Among the environmental information, altitude and atmospheric pressure are considered to have a strong influence on the measured electrostatic field value. Therefore, it is desirable to include altitude and atmospheric pressure in the environmental information.

[0021] (7) In the lightning prediction system of the present invention described above, the electrostatic field value acquired by the electrostatic field value acquisition unit may be corrected based on the altitude at the measurement point so that it becomes a measurement value at a predetermined altitude.

[0022] By configuring the lightning strike prediction system of the present invention as described above in (7), the influence of the difference in altitude at the installation location of the electrostatic field value acquisition unit can be reduced. This allows the lightning strike prediction system of the present invention to perform lightning strike prediction with high accuracy. Furthermore, by configuring the lightning strike prediction system of the present invention as described above in (7), it is possible to perform calculation processing using electrostatic field values ​​after correcting for the difference in altitude, thereby reducing the load associated with the calculation processing.

[0023] (8) The lightning strike prediction system of the present invention described above may be characterized in that, in a non-measured area where the electrostatic field value acquisition unit is not located, an estimated electric field value and predicted environmental information are calculated based on the electrostatic field value and the environmental information in at least one area adjacent to the non-measured area, and a lightning strike probability prediction is made in the non-measured area based on the estimated electric field value and the predicted environmental information.

[0024] By configuring the lightning strike prediction system of the present invention as described above in (8), it is possible to predict the probability of lightning strikes even in unmeasured areas where no electrostatic field value acquisition units (electric field sensors, etc.) are installed. As a result, the lightning strike prediction system of the present invention can efficiently utilize the installed electrostatic field value acquisition units, and can therefore predict the probability of lightning strikes even in areas where it is difficult to install electrostatic field value acquisition units (due to cost, topography, etc.). Furthermore, the lightning strike prediction system of the present invention is able to predict the probability of lightning strikes in unmeasured areas while minimizing the number of installed electrostatic field value acquisition units, which is expected to reduce costs.

[0025] (9) The lightning strike prediction system of the present invention described above may be characterized in that the learning data includes a first electrostatic field value and a second electrostatic field value, the first electrostatic field value is the amount of change in the electrostatic field value per predetermined time, and the second electrostatic field value is the sum of the first electrostatic field value and the absolute value of the electrostatic field value obtained per predetermined time.

[0026] By configuring the lightning strike prediction system of the present invention as described above in (9), it is possible to increase the amount of learning data (explanatory variables) related to electrostatic field values ​​that are considered to be highly dependent, thereby improving the accuracy of lightning strike prediction.

[0027] (10) The lightning strike prediction system of the present invention described above may be characterized in that it estimates the gradient state of each electrostatic field value based on each electrostatic field value acquired at a plurality of the measurement points, and predicts the lightning strike probability in the lightning strike probability prediction unit based on the gradient state.

[0028] By configuring the lightning strike prediction system of the present invention as described above in (10), it is possible to predict the probability of a lightning strike even in areas where no measurement points have been established. Therefore, the lightning strike prediction system of the present invention can predict the probability of a lightning strike even in locations where it is difficult to install an electrostatic field value acquisition unit (electrostatic field value acquisition device). Furthermore, the lightning strike prediction system of the present invention can reduce the number of electrostatic field value acquisition units installed, thereby expected to have a cost-saving effect. Here, the gradient state of each electrostatic field value can be estimated, for example, based on the gradient (electrostatic field value gradient) of the line (electrostatic field value) when electrostatic field values ​​acquired in adjacent areas are connected in a straight line. The gradient of the electrostatic field value can be either a two-dimensional (2D) gradient or a three-dimensional (3D) gradient.

[0029] Here, for example, if a person or the like passes near an electrostatic field value acquisition unit (such as an electric field sensor) at a measurement point, there is a concern that the electrostatic field value acquired by the electrostatic field value acquisition unit may be affected by static electricity, etc. In such a case, the electrostatic field value acquired by the electrostatic field value acquisition unit may indicate a high value, raising a concern that a lightning strike prediction may be erroneously notified. Therefore, as a result of extensive research, the present inventors have come to the conclusion that, taking into consideration lightning strike prediction information (e.g., nowcast) from the Japan Meteorological Agency, etc., even if the electrostatic field value indicates a high value, if the lightning activity (lightning strike risk) in the lightning strike prediction information from the Japan Meteorological Agency, etc. is low, the predicted lightning strike probability should be discarded.

[0030] (11) Therefore, in order to solve the above problem, the lightning strike prediction system of the present invention may be characterized in that the environmental information includes lightning strike prediction information in either or both of the weather information from the Japan Meteorological Agency and private weather information, and the lightning strike probability prediction unit discards the predicted lightning strike probability on the condition that the predicted lightning strike probability is equal to or greater than a predetermined threshold and the risk level in the lightning strike prediction information is below a predetermined threshold.

[0031] The lightning strike prediction system of the present invention, configured as described above in (11), can discard lightning strike probabilities predicted based on electrostatic field values ​​due to erroneous detection, thereby improving the accuracy of lightning strike predictions. In the lightning strike prediction system of the present invention, even if the lightning strike probability predicted by the lightning strike probability prediction unit is equal to or greater than a predetermined threshold, the predicted lightning strike probability is discarded if the risk level in the weather information (e.g., nowcast) from the Japan Meteorological Agency or private weather information is below a predetermined threshold. In other words, even if the predicted lightning strike probability is high, the predicted lightning strike probability is discarded if the risk level of the lightning strike forecast information in either or both of the weather information from the Japan Meteorological Agency and private weather information is low. This allows the lightning strike prediction system of the present invention to improve the reliability of lightning strike predictions by using both actual measurement data, such as electrostatic field values, at measurement points and weather information based on observation data from the Japan Meteorological Agency, etc.

[0032] (12) The above-mentioned lightning strike prediction system of the present invention may be characterized in that the lightning strike probability prediction unit is located on a server, and the electrostatic field value acquisition unit and the information acquisition unit are connected to the lightning strike probability prediction unit via a network.

[0033] By configuring the lightning strike prediction system of the present invention as described above in (12), it is possible to perform calculations related to lightning strike probability prediction, which involves a heavy load, on a server. This reduces the load on the lightning strike prediction system, thereby simplifying the system itself and reducing costs. The server may be, for example, a cloud server. Note that the server is not limited to a cloud server, and various types of servers may be used.

[0034] (13) The lightning prediction method of the present invention, which is provided to solve the above-mentioned problems, uses the lightning prediction system and includes an electrostatic field value acquisition step in which the electrostatic field value acquisition unit acquires the electrostatic field values ​​in the plurality of regions including the measurement point; an information acquisition step in which the information acquisition unit acquires the environmental information in the plurality of regions; a region designation step in which one of the plurality of regions is designated as a candidate region; a lightning probability prediction step in which the lightning probability prediction unit predicts the lightning probability based on the electrostatic field values ​​and the environmental information; a high electrostatic field area existence determination step for determining whether or not there is a high electrostatic field area in the plurality of areas where the electrostatic field value is equal to or greater than a predetermined threshold and has an electrostatic field value higher than that of the designated candidate area, provided that the lightning strike probability is equal to or greater than the predetermined threshold in the lightning strike probability determination step; and an alarm issuance step for issuing an alarm to the candidate area, provided that no high electrostatic field area exists in the high electrostatic field area existence determination step.

[0035] The lightning prediction method of the present invention, configured as described above in (13), can accurately predict lightning strikes in a designated candidate area in the area designation step. In the high electrostatic field area presence determination step, the lightning prediction method of the present invention determines whether a high electrostatic field area exists that exhibits a higher electrostatic field value than a designated candidate area, and if no high electrostatic field area exists, issues an alarm for the candidate area in the alarm issuance step. That is, the lightning prediction method of the present invention does not uniformly issue an alarm for areas with electrostatic field values ​​equal to or greater than a threshold, but issues an alarm only if no area exhibits a higher electrostatic field value than the designated candidate area, thereby preventing unnecessary alarms from being issued. Therefore, the lightning prediction method of the present invention can prevent cumbersome lightning strike warnings from being issued and can perform accurate lightning strike predictions. [Effects of the Invention]

[0036] According to the present invention, it is possible to provide a lightning strike prediction system and a lightning strike prediction method that can improve the accuracy of lightning strike prediction in a wide-area lightning strike area (region). [Brief explanation of the drawings]

[0037] [Figure 1] 1 is a block diagram of a lightning strike prediction system according to the present invention. [Figure 2] 1 is a map illustrating an example of an area set in the lightning strike prediction system of the present invention. [Figure 3] 1 is a schematic explanatory diagram showing areas and measurement points set in the lightning strike prediction system of the present invention. [Figure 4] 1 is a data table showing an example of learning data used in the lightning strike prediction system of the present invention. [Figure 5] FIG. 2 is a flow chart showing an example of the operation of the lightning strike prediction system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0038] A lightning strike prediction system 1 according to one embodiment of the present invention will be described in detail below with reference to the drawings. Note that each drawing is a schematic representation for ease of understanding, and may differ from the actual shape, size, and arrangement of components. Also, Figure 3 is an explanatory diagram in which the map in Figure 2 has been removed for ease of understanding, and the area and measurement points are shown.

[0039] As shown in FIG. 1, the lightning strike prediction system 1 includes an electrostatic field value acquisition unit 10, an information acquisition unit 20, and a lightning strike probability prediction unit 30. As shown in FIGS. 2 and 3, the lightning strike prediction system 1 is configured to predict lightning strikes in multiple areas A1 to An (n is any number). In the following description, when there is no need to particularly distinguish between the multiple areas, they may be referred to as area An. Furthermore, an electrostatic field value acquisition unit 10 is disposed in the center of each of the multiple areas An, forming multiple measurement points P1 to Pn (n is any alphabetical letter). When there is no need to particularly distinguish between the multiple measurement points P1 to Pn, they may be referred to as measurement point Pn.

[0040] As shown in FIG. 1, the electrostatic field value acquiring unit 10 is configured as, for example, a surface potential sensor (charge detector). The electrostatic field value acquiring unit 10 can acquire (measure) electrostatic field values ​​(also referred to as electric field values) in the atmosphere by measuring the potential (charge) within a predetermined region. As shown in FIGS. 2 and 3, the electrostatic field value acquiring unit 10 is disposed at the center (measurement point Pn) of a plurality of circular regions An. In other words, a plurality of circular regions An are formed with the respective measurement points Pn at which the electrostatic field value acquiring unit 10 is disposed as centers. Here, the region An is a range in which the electrostatic field value can be acquired (detected) by the electrostatic field value acquiring unit 10, and in this embodiment, the region An is set to a range with a radius of 5 km, as will be described in detail later. In other words, the electrostatic field value acquiring unit 10 acquires electrostatic field values ​​in the atmosphere at a plurality of measurement points Pn disposed within the plurality of regions An.

[0041] In this embodiment, an electrostatic field value acquiring unit 10 is disposed at each of five measurement points P1 to P5, and areas A1 to A5 are set around each of the measurement points P1 to P5. If such a configuration is permitted, it is desirable to arrange the multiple measurement points Pn in a grid pattern. This allows the multiple electrostatic field value acquiring units 10 to comprehensively acquire electrostatic field values ​​in each area An. It is also desirable that the multiple measurement points Pn be disposed within 10 km of each other adjacent measurement points Pn. This is because, for example, the scale of a thundercloud 2 often forms within a range of approximately 10 km. That is, since the measurement range of the electrostatic field value acquired by the electrostatic field value acquiring unit 10 (one area An) has a diameter of approximately 10 km, one electrostatic field value acquiring unit 10 can cover lightning strike predictions in one area An. Since the acquisition of electrostatic field values ​​is easily obstructed by an overhead obstruction such as a roof, it is desirable to dispose the electrostatic field value acquiring unit 10 in a location free of obstructions above.

[0042] As shown in FIG. 1 , the information acquisition unit 20 can acquire environmental information within a predetermined range centered on a plurality of measurement points Pn. Here, the environmental information can include various types of information related to the environment, such as information on the weather at each measurement point or each region, topographical information, and physical information. Specifically, the environmental information includes at least one of altitude, air pressure, temperature, and humidity. In this embodiment, altitude, air pressure, temperature, and humidity are used as the environmental information. The information acquisition unit 20 is configured as, for example, an altimeter, barometer, thermometer, hygrometer, or sensors capable of measuring these. Note that the altitude is determined when the electrostatic field value acquisition unit 10 is initially installed, and therefore an altimeter does not need to be installed at all times.

[0043] Here, among the environmental information, altitude and atmospheric pressure are considered to have a strong influence on the electrostatic field value. After altitude and atmospheric pressure, temperature and humidity are considered to have a weaker influence on the electrostatic field value. Therefore, in this embodiment, a correction process is performed on the electrostatic field value acquired by the electrostatic field value acquisition unit 10 based on the altitude at the measurement point Pn so that the value is a measurement value at a predetermined altitude. The altitude at the measurement point Pn is determined by the topography and the installation position of the electrostatic field value acquisition unit 10, and the correction coefficient for the electrostatic field value can be determined in advance based on the amount of change in the electrostatic field value depending on the altitude. Therefore, the electrostatic field value acquired by the electrostatic field value acquisition unit 10 can be corrected based on the correction coefficient determined in advance.

[0044] In this embodiment, the environmental information includes lightning strike forecast information from either or both of the weather information from the Japan Meteorological Agency and private weather information. The information acquisition unit 20 can acquire the weather information from the Japan Meteorological Agency and private weather information via appropriate means (e.g., a network, etc.). Here, the weather information from the Japan Meteorological Agency can be, for example, lightning strike forecast information such as lightning activity levels (levels L1 to L4) in nowcast. The lightning activity levels are defined as follows: level L4 is "intense lightning: numerous lightning strikes are occurring"; level L3 is "moderately intense lightning: lightning strikes are occurring"; level L2 is "thunder: lightning is visible and thunder is heard. The possibility of lightning strikes is increasing"; and level L1 is "lightning possible: no lightning is currently occurring, but there is a possibility of lightning strikes in the future." In addition to the weather information from the Japan Meteorological Agency, private weather information, etc. can also be used as environmental information.

[0045] 1, in this embodiment, the lightning strike probability prediction unit 30 is arranged on a server 31 (cloud server 31). The lightning strike probability prediction unit 30 is connected to the electrostatic field value acquisition unit 10 and the information acquisition unit 20 via a network (not shown). Therefore, the electrostatic field value acquired by the electrostatic field value acquisition unit 10 and the environmental information acquired by the information acquisition unit 20 are transmitted to the lightning strike probability prediction unit 30 via the network.

[0046] The lightning strike probability prediction unit 30 can calculate (predict) the probability of a lightning strike through calculations performed by the server 31. The lightning strike probability prediction unit 30 can predict the probability of a lightning strike for each region An based on the electrostatic field value and environmental information. The lightning strike probability prediction unit 30 can also notify regions An where the predicted lightning strike probability is equal to or greater than a predetermined threshold (e.g., 50%). The notification can be performed by an appropriate notification unit (not shown). The notification can also be sent to businesses (e.g., semiconductor factories, automobile factories, steel mills with blast furnaces, etc.) or facilities (schools, hospitals), for example, via communication means such as issuing an alarm or email. The lightning strike probability prediction unit 30 uses an electrostatic field value of 1 kV / m or greater as a threshold for predicting the probability of a lightning strike. This is because an accurate electrostatic field value cannot be obtained when the electrostatic field value is less than 1 kV / m due to environmental influences at the measurement point, noise, etc.

[0047] In this embodiment, the lightning strike probability prediction unit 30 predicts the lightning strike probability (also referred to as AI prediction) based on a prediction algorithm generated using the amount of change in the electrostatic field value and the environmental information over a predetermined period as learning data (see FIG. 4). Here, various algorithms such as XGBoost and LightGBM can be used as the prediction algorithm.

[0048] As shown in FIG. 4, the training data consists of the amount of change in atmospheric pressure, first electrostatic field value (first electric field value), second electrostatic field value (second electric field value), humidity, and temperature over a predetermined period of time. In other words, the training data is based on the amount of change in environmental information over a predetermined period of time. This is because, if absolute values ​​for each target item (explanatory variable) were used as training data, differences in atmospheric pressure due to differences in elevation between locations would be included, making accurate predictions difficult. Specifically, at high altitudes, the original atmospheric pressure is low, so the absolute value of atmospheric pressure would approximate the atmospheric pressure at which thunderclouds 2 and lightning strikes occur, raising concerns about misjudgment of lightning strike predictions.

[0049] Therefore, in this embodiment, the current measurement data (absolute values) of atmospheric pressure, humidity, and temperature acquired (measured) every 10 seconds are compared with the previously measured data (absolute values) of atmospheric pressure, humidity, and temperature acquired 10 minutes ago, and the difference (amount of change) between the original measurement data and the previously measured data is sampled as learning data. Also, learning data related to electrostatic field values, which have a large influence, is sampled at shorter intervals (every minute) than atmospheric pressure, humidity, and temperature. Specifically, the first electrostatic field value is sampled as the amount of change between the current electrostatic field value (current electric field absolute value) acquired every 10 seconds and the previously measured electrostatic field value (previous electric field absolute value) acquired 1 minute ago.

[0050] Furthermore, since the electrostatic field value has a large influence on lightning strike probability prediction, in this embodiment, in order to improve prediction accuracy, the number of explanatory variables is increased and a second electrostatic field value is also included in the training data. Here, the second electrostatic field value is the sum of the first electrostatic field value (amount of change) and the absolute value (electrostatic field absolute value) of the electrostatic field value acquired every predetermined time (every 10 seconds in this embodiment).

[0051] Here, as shown in FIG. 3, the lightning probability prediction unit 30 determines an area A3 where the predicted lightning probability is equal to or greater than a predetermined threshold (e.g., 50%) as a candidate area A3, and desirably notifies the candidate area A3 on the condition that the electrostatic field value (e.g., 10 kV / m) in the candidate area A3 is higher than the electrostatic field value (e.g., 5 kV / m) of another candidate area A1 (area A1) adjacent to the candidate area A3. In other words, when there are multiple candidate areas A1 and A3 where the lightning probability exceeds a predetermined threshold, the lightning probability prediction unit 30 desirably notifies the candidate area A3 with the higher electrostatic field value among the multiple adjacent candidate areas A1 and A3. This enables the lightning probability prediction unit 30 to perform more accurate lightning predictions with fewer false positives and missed strikes.

[0052] Furthermore, it is desirable for the lightning strike probability prediction unit 30 to discard the predicted lightning strike probability on the condition that the predicted lightning strike probability is equal to or greater than a predetermined threshold (e.g., 50%) and the risk level in the lightning strike forecast information from the Japan Meteorological Agency or the like is below a predetermined threshold (e.g., lightning activity level L2). This allows the lightning strike probability prediction unit 30 to suppress the influence of erroneous detection of the electrostatic field value even when the electrostatic field value acquisition unit 10 is affected by static electricity or the like (e.g., when a person passes near the electrostatic field value acquisition unit 10). This allows the lightning strike probability prediction unit 30 to make accurate lightning strike probability predictions.

[0053] 3, the lightning strike prediction system 1 can also predict the probability of a lightning strike in a non-measured area a (also referred to as a blank area a) where no electrostatic field value acquisition unit 10 is located. Specifically, in the non-measured area a where no electrostatic field value acquisition unit 10 is located, the lightning strike probability prediction unit 30 calculates an estimated electric field value and predicted environmental information based on the electrostatic field value and environmental information in at least one area A3 (measurement point P3) adjacent to the non-measured area a, and can predict the probability of a lightning strike in the non-measured area a based on the estimated electric field value and the predicted environmental information.

[0054] Furthermore, in this embodiment, when calculating the estimated electric field value and predicted environmental information in the non-measurement area a, the probability of a lightning strike can be predicted based on the gradient of each electrostatic field value between multiple measurement points Pn. Specifically, the gradient of each electrostatic field value is estimated based on each electrostatic field value acquired at multiple measurement points Pn, and the lightning strike probability can be predicted in the lightning strike probability prediction unit 30 based on the gradient. Here, the gradient of the electrostatic field value can be, for example, a two-dimensional (2D) gradient (e.g., the gradient of a line or plane) of each electrostatic field value at measurement points P3 and P4, or a three-dimensional (3D) gradient that includes information other than the electrostatic field value, such as topography and environmental information.

[0055] The above is one embodiment of the lightning strike prediction system 1 of the present invention. Next, the effects achieved by the lightning strike prediction system 1 of the present invention will be described below.

[0056] <Action and effect> The above-described lightning strike prediction system 1 has the following characteristic configurations (a) to (l). Therefore, the lightning strike prediction system 1 of the present invention can achieve the following unique effects that cannot be achieved by conventional techniques.

[0057] (a) The lightning strike prediction system 1 of the present invention comprises an electrostatic field value acquisition unit 10 that acquires atmospheric electrostatic field values ​​at a plurality of measurement points Pn arranged in a plurality of regions An, an information acquisition unit 20 that acquires environmental information within a predetermined range centered on the measurement point Pn, and a lightning strike probability prediction unit 30 that predicts the probability of a lightning strike, wherein each region An is set as a predetermined range including one measurement point Pn, and the lightning strike probability prediction unit 30 predicts the lightning strike probability for each region An based on the electrostatic field value and the environmental information, and notifies the region An where the predicted lightning strike probability is equal to or greater than a predetermined threshold.

[0058] The lightning strike prediction system 1 of the present invention can predict the probability of lightning strikes based on atmospheric electrostatic field values ​​at multiple measurement points Pn located in multiple regions An and environmental information for a predetermined range centered on the measurement points Pn, thereby making it possible to predict lightning strikes based on meteorological information for a wide range of regions An. As a result, the lightning strike prediction system 1 of the present invention can make highly accurate lightning strike predictions. Here, the environmental information can include information about the weather at each measurement point Pn and each region An, as well as various types of environmental information such as topographical information and physical information.

[0059] Furthermore, the lightning strike prediction system 1 of the present invention predicts the probability of a lightning strike for each of multiple regions An, and notifies regions An where the predicted lightning strike probability is equal to or greater than a predetermined threshold, preventing unnecessary notifications. Therefore, the lightning strike prediction system 1 of the present invention can prevent notifications from becoming cumbersome and lightning strike predictions from ending up as false alarms. Various notification methods can be used for notifications, such as issuing an alarm or sending notifications via email. Examples of recipients of notifications include various businesses and facilities that are highly susceptible to lightning strikes (e.g., semiconductor factories, automobile factories, steel mills with blast furnaces, schools, etc.).

[0060] (b) In the lightning strike prediction system 1 of the present invention, the lightning strike probability prediction unit 30 is characterized in that it predicts the lightning strike probability based on a prediction algorithm generated using the amount of change in the electrostatic field value and the environmental information over a predetermined period as learning data.

[0061] The lightning strike prediction system 1 of the present invention can predict the probability of lightning strikes (also called AI prediction) based on a prediction algorithm generated as learning data from electrostatic field values ​​and environmental information, and can therefore make appropriate lightning strike probability predictions according to the environmental characteristics of each region. As a result, the lightning strike prediction system of the present invention can reduce false positives and missed lightning strike predictions and provide highly accurate lightning strike predictions.

[0062] Furthermore, the lightning strike prediction system 1 of the present invention uses the amount of change in the electrostatic field value and environmental information over a predetermined time period as training data, thereby reducing the influence of the location of the measurement point (e.g., elevation differences). Specifically, the lightning strike prediction system 1 of the present invention uses the amount of change over a predetermined time period, rather than the absolute values ​​of the electrostatic field value and environmental information (e.g., atmospheric pressure), as training data. This reduces the influence of, for example, elevation differences at the measurement point. Here, the training data may include, for example, the amount of change per predetermined time for atmospheric pressure, electrostatic field value, humidity, temperature, etc. Furthermore, measurements of the electrostatic field value and environmental information (e.g., atmospheric pressure, humidity, temperature) at the measurement point may be performed, for example, at 10-second intervals. As training data, for example, the amount of change per minute for items that change significantly over time, such as electrostatic field values, and the amount of change per 10 minutes for items that change relatively slowly over time, such as environmental information. Note that the measurement interval and the interval at which training data is acquired may be appropriately changed depending on the environment of the measurement point, the measurement equipment, etc. Here, various prediction algorithms, such as XGBoost and LightGBM, may be used. Furthermore, by configuring the lightning strike prediction system of the present invention as described above in (b), it is possible to prevent unnecessary notifications (for example, issuing alarms).

[0063] (c) The lightning strike prediction system 1 of the present invention is characterized in that it designates an area An where the predicted lightning strike probability is equal to or greater than a predetermined threshold as a candidate area An, and notifies the candidate area An on the condition that the electrostatic field value in the candidate area An is higher than the electrostatic field value in other candidate areas An adjacent to the candidate area An.

[0064] The lightning prediction system 1 of the present invention is configured to send notifications to candidate areas An where the probability of lightning strikes is equal to or greater than a threshold and where the electrostatic field value in the candidate area An is higher than the electrostatic field values ​​in other adjacent candidate areas An. Therefore, the lightning prediction system 1 of the present invention can send notifications not only to areas An where the probability of lightning strikes is high, but also to areas An where the probability of lightning strikes is high and the surrounding electrostatic field values ​​are high. This enables the lightning prediction system 1 of the present invention to make more accurate lightning predictions and further reduce false or missed lightning predictions.

[0065] (d) In the lightning strike prediction system 1 of the present invention, the lightning strike probability prediction unit 30 uses the electrostatic field value of 1 kV / m or more as a threshold value for predicting the lightning strike probability.

[0066] The lightning strike prediction system 1 of the present invention uses an electrostatic field value of 1 kV / m or more to predict the probability of a lightning strike, thereby reducing the influence of noise. Therefore, the lightning strike prediction system 1 of the present invention can make lightning strike predictions with high accuracy.

[0067] Here, the scale of a thundercloud 2 is generally considered to be within a range of approximately 10 km. Meanwhile, the electrostatic field value acquisition range of the electrostatic field value acquisition unit 10 (electric field sensor, etc.) at measurement point Pn is considered to be approximately a 5 km radius. Therefore, in consideration of the scale of a thundercloud 2 and the electrostatic field value acquisition range, if the electrostatic field value measurement points Pn are arranged at intervals of 10 km or less, it is possible to comprehensively predict lightning strikes in the surrounding area An.

[0068] (e) Therefore, the lightning strike prediction system 1 of the present invention is characterized in that a plurality of measurement points Pn are located within 10 km of each other adjacent measurement points Pn.

[0069] If the lightning prediction system 1 of the present invention is configured as described above in (e), it can efficiently predict lightning strikes with a minimum number of measurement points Pn (electrostatic field value acquisition units 10). This allows the lightning prediction system 1 of the present invention to perform accurate lightning strike predictions while reducing costs. Here, the measurement points Pn may be arranged, for example, in a grid pattern. This allows the lightning prediction system 1 of the present invention to perform more comprehensive and accurate lightning strike predictions for each area An.

[0070] (f) The lightning strike prediction system 1 of the present invention described above is characterized in that the environmental information includes at least one of altitude, atmospheric pressure, temperature, and humidity.

[0071] By configuring the lightning strike prediction system 1 of the present invention as described above in (f), the influence of differences in topography and installation environment in each region An can be reduced, thereby enabling more accurate lightning strike prediction. Among the environmental information, altitude and atmospheric pressure are considered to have a strong influence on the measured electrostatic field value. Therefore, it is desirable to include altitude and atmospheric pressure in the environmental information.

[0072] (g) In the lightning strike prediction system 1 of the present invention, the electrostatic field value acquired by the electrostatic field value acquisition unit 10 is corrected based on the altitude at the measurement point Pn so that it becomes a measurement value at a predetermined altitude.

[0073] By configuring the lightning strike prediction system 1 of the present invention as described in (g) above, it is possible to reduce the influence of differences in altitude at the installation location of the electrostatic field value acquisition unit 10. This allows the lightning strike prediction system 1 of the present invention to perform lightning strike predictions with high accuracy. Furthermore, by configuring the lightning strike prediction system 1 of the present invention as described in (g) above, it is possible to perform calculation processing using electrostatic field values ​​after correction for differences in altitude, thereby reducing the load associated with the calculation processing.

[0074] (h) The lightning strike prediction system 1 of the present invention is characterized in that, in a non-measured area a where an electrostatic field value acquisition unit 10 is not located, an estimated electric field value and predicted environmental information are calculated based on the electrostatic field value and the environmental information in at least one area An adjacent to the non-measured area a, and a lightning strike probability prediction is made in the non-measured area a based on the estimated electric field value and the predicted environmental information.

[0075] By configuring the lightning strike prediction system 1 of the present invention as described in (h) above, it is possible to predict the probability of lightning strikes even in non-measured areas a where no electrostatic field value acquisition units 10 (electric field sensors, etc.) are installed. As a result, the lightning strike prediction system 1 of the present invention can efficiently utilize the installed electrostatic field value acquisition units 10, and can therefore predict the probability of lightning strikes even in areas An where it is difficult to install electrostatic field value acquisition units 10 (due to cost, topography, etc.). Furthermore, the lightning strike prediction system 1 of the present invention can predict the probability of lightning strikes in non-measured areas a while minimizing the number of installed electrostatic field value acquisition units 10, which is expected to reduce costs.

[0076] (i) The lightning strike prediction system 1 of the present invention is characterized in that the learning data includes a first electrostatic field value and a second electrostatic field value, the first electrostatic field value is the amount of change in the electrostatic field value per predetermined time, and the second electrostatic field value is the sum of the first electrostatic field value and the absolute value of the electrostatic field value obtained per predetermined time.

[0077] By configuring the lightning strike prediction system 1 of the present invention as described above in (i), it is possible to increase the amount of learning data (explanatory variables) relating to electrostatic field values ​​that are considered to be highly dependent, thereby improving the accuracy of lightning strike prediction.

[0078] (j) The lightning strike prediction system 1 of the present invention described above is characterized in that it estimates the gradient state of each electrostatic field value based on each electrostatic field value acquired at a plurality of measurement points Pn, and predicts the lightning strike probability in the lightning strike probability prediction unit 30 based on the gradient state.

[0079] By configuring the lightning strike prediction system 1 of the present invention as described in (j) above, it is possible to predict the probability of a lightning strike even in an area An where no measurement point Pn has been set. Therefore, the lightning strike prediction system 1 of the present invention can predict the probability of a lightning strike even in a location where it is difficult to install an electrostatic field value acquisition unit 10 (electrostatic field value acquisition device). Furthermore, the lightning strike prediction system 1 of the present invention can reduce the number of electrostatic field value acquisition units 10 installed, which is expected to result in cost reduction. Here, the gradient state of each electrostatic field value can be estimated based on, for example, the gradient (electrostatic field value gradient) of a line (electrostatic field value) when electrostatic field values ​​acquired in adjacent areas An are connected in a straight line. The gradient of the electrostatic field value can be either a two-dimensional (2D) gradient or a three-dimensional (3D) gradient.

[0080] Here, for example, if a person or the like passes near the electrostatic field value acquisition unit 10 (electric field sensor or the like) at measurement point Pn, there is a concern that the electrostatic field value acquired by the electrostatic field value acquisition unit 10 may be affected by static electricity or the like. In such a case, the electrostatic field value acquired by the electrostatic field value acquisition unit 10 may indicate a high value, raising the concern that a lightning strike prediction may be erroneously notified. Therefore, the lightning strike prediction system 1 of the present invention takes into consideration lightning strike prediction information (e.g., nowcast) from the Japan Meteorological Agency, etc., and discards the predicted lightning strike probability if the lightning strike prediction information from the Japan Meteorological Agency or the like indicates a low lightning activity level (lightning strike risk level), even if the electrostatic field value indicates a high value.

[0081] (k) Specifically, the lightning strike prediction system 1 of the present invention is characterized in that the environmental information includes lightning strike prediction information in either or both of meteorological information from the Japan Meteorological Agency and private meteorological information, and the lightning strike probability prediction unit 30 discards the predicted lightning strike probability on the condition that the predicted lightning strike probability is equal to or greater than a predetermined threshold and the risk level in the lightning strike prediction information is below a predetermined threshold.

[0082] By configuring the lightning strike prediction system 1 of the present invention as described in (k) above, it is possible to discard lightning strike probabilities predicted based on electrostatic field values ​​due to erroneous detection, thereby improving the accuracy of lightning strike predictions. In the lightning strike prediction system 1 of the present invention, even if the lightning strike probability predicted by the lightning strike probability prediction unit 30 is equal to or greater than a predetermined threshold, the predicted lightning strike probability is discarded if the risk level in the weather information (e.g., nowcast) from the Japan Meteorological Agency or private weather information is below a predetermined threshold. In other words, even if the predicted lightning strike probability is high, the predicted lightning strike probability is discarded if the risk level of the lightning strike forecast information in either or both of the weather information from the Japan Meteorological Agency and private weather information is low. This allows the lightning strike prediction system 1 of the present invention to improve the reliability of lightning strike predictions by using both actual measurement data, such as electrostatic field values, at measurement point Pn and weather information based on observation data from the Japan Meteorological Agency and the like.

[0083] (l) The lightning strike prediction system 1 of the present invention is characterized in that the lightning strike probability prediction unit 30 is placed on a server 31, and the electrostatic field value acquisition unit 10 and the information acquisition unit 20 are connected to the lightning strike probability prediction unit 30 via a network.

[0084] By configuring the lightning strike prediction system 1 of the present invention as described above in (l), calculations related to lightning strike probability prediction, which impose a heavy load, can be performed on the server 31. This reduces the load on the lightning strike prediction system 1, thereby simplifying the system itself and reducing costs. Here, it is preferable to use a cloud server, for example, as the server 31. Note that the server 31 is not limited to a cloud server, and various types of servers can be used. Note that the server 31 can be located in various locations.

[0085] The above are the effects of the lightning strike prediction system 1 according to one embodiment of the present invention. Next, a lightning strike prediction method using the lightning strike prediction system 1 of the present invention will be described in detail. In the following description, it is assumed that area A1 is designated as a candidate area among multiple areas An.

[0086] <Lightning strike prediction method> Fig. 5 is a flow diagram of an embodiment of a lightning prediction method of the present invention. The lightning prediction method uses the lightning prediction system 1 described above. As shown in Fig. 5, when lightning prediction starts, electrostatic field values ​​are acquired by the electrostatic field value acquisition unit 10 at multiple measurement points Pn (areas An) (electrostatic field value acquisition step S1). In addition to the electrostatic field value acquisition step S1, environmental information (barometric pressure, humidity, temperature) is acquired by the information acquisition unit 20 (information acquisition step S2). The electrostatic field value acquisition step S1 and the information acquisition step S2 are processed in parallel or one after the other. Next, one area A1 of the multiple areas An is designated as a candidate area A1 (area designation step S3).

[0087] Next, a lightning strike prediction (AI prediction) is performed based on the electrostatic field value and environmental information at the measurement point P1 (candidate area A1). That is, the lightning strike probability prediction unit 30 predicts the lightning strike probability based on the electrostatic field value and environmental information (lightning strike probability prediction step S4). In parallel with the lightning strike probability prediction step S4, the lightning strike probability is also predicted for candidate areas An surrounding the candidate area A1.

[0088] Next, it is determined whether or not the probability (predicted value) of lightning strikes in the candidate area A1 predicted in the lightning strike probability prediction step S4 is equal to or greater than a predetermined threshold (lightning strike probability determination step S5).

[0089] In the lightning strike probability determination step S5, provided that the lightning strike probability is equal to or greater than a predetermined threshold (e.g., 50%), it is determined whether or not there are high electrostatic field areas in the multiple areas An where the electrostatic field value is equal to or greater than a predetermined threshold (e.g., 1 kV / m) and has a higher electrostatic field value than the designated candidate area A1 (high electrostatic field area existence determination step S6).

[0090] If a high electrostatic field area exists in the high electrostatic field area determination step S6, an alarm (notification) is not issued in the candidate area A1 (step S7). Here, when step S7 is executed, the series of processes ends. Note that when the series of processes ends, the series of processes are repeated from the beginning as necessary.

[0091] If the lightning strike probability is less than a predetermined threshold (for example, 50%) in the lightning strike probability determination step S5, the process proceeds to step S7, and the series of processes ends without issuing an alarm (notification) in the candidate area A1.

[0092] Furthermore, if no high electrostatic field region exists in the high electrostatic field region existence determination step S6, an alarm is issued for candidate region A1 (alarm issuance step S10). That is, if no high electrostatic field region exists in the high electrostatic field region existence determination step S6, candidate region A1 exhibits the highest electrostatic field value among the surrounding candidate regions An, and the lightning strike probability is also above a threshold, so it is determined that there is a high possibility of a lightning strike. Therefore, an alarm is issued for candidate region A1. Note that, when alarm issuance step S10 is completed, the series of processes ends. Note that, when the series of processes ends, the series of processes are repeated from the beginning as necessary.

[0093] That is, the lightning strike prediction method of the present invention can be configured as follows (m), and these configurations can provide the following operational effects.

[0094] (m) The lightning strike prediction method of the present invention uses the above-described lightning strike prediction system 1, and includes an electrostatic field value acquisition step S1 in which the electrostatic field value acquisition unit 10 acquires electrostatic field values ​​in a plurality of areas An including the measurement point Pn, an information acquisition step S2 in which the information acquisition unit 20 acquires environmental information in the plurality of areas An, an area designation step S3 in which one area A1 of the plurality of areas An is designated as a candidate area A1, a lightning strike probability prediction step S4 in which the lightning strike probability prediction unit 30 predicts the lightning strike probability based on the electrostatic field value and the environmental information, and a lightning strike probability prediction step S5 in which the lightning strike probability in the candidate area A1 predicted in the lightning strike probability prediction step S5 is calculated. is equal to or greater than a predetermined threshold in the lightning probability determination step S5; a high electrostatic field area existence determination step S6, on the condition that the lightning probability is equal to or greater than the predetermined threshold in the lightning probability determination step S5, of determining whether or not there exists a high electrostatic field area in a plurality of areas An where the electrostatic field value is equal to or greater than the predetermined threshold and has an electrostatic field value higher than that of a designated candidate area A1; and an alarm issuance step S10, on the condition that no high electrostatic field area exists in the high electrostatic field area existence determination step S6.

[0095] The lightning prediction method of the present invention, configured as described above in (m), can accurately predict lightning strikes in a designated candidate area A1 in the area designation step S3. The lightning prediction method of the present invention determines whether a high electrostatic field area exists that exhibits a higher electrostatic field value than the designated candidate area A1 in the high electrostatic field area existence determination step S6, and if no high electrostatic field area exists, issues an alarm for the candidate area A1 in the alarm issuance step S10. That is, the lightning prediction method of the present invention does not uniformly issue an alarm for all areas An whose electrostatic field value is equal to or greater than a threshold value, but issues an alarm only if there is no area An whose electrostatic field value is higher than the designated candidate area A1, thereby preventing unnecessary alarms from being issued. Therefore, the lightning prediction method of the present invention can prevent cumbersome lightning strike warnings from being issued and can perform accurate lightning strike predictions.

[0096] The above is the configuration and effects of the lightning strike prediction system 1 and lightning strike prediction method according to one embodiment of the present invention. However, the lightning strike prediction system 1 and lightning strike prediction method of the present invention are not limited to the above-described embodiment and can be modified in various ways within the scope of the present invention. For example, the lightning strike prediction system 1 may be configured as described in (a) above and can be formed in various shapes and sizes. Furthermore, the lightning strike prediction system 1 of the present invention may not include some or all of the components described in (b) to (l) above, or may include some or all of the components described in (b) to (l) above and other components. Furthermore, the lightning strike prediction method of the present invention may be configured as described in (m) above and can use various means or change the order within the scope of the invention.

[0097] In this embodiment, a surface potential sensor or a charge detector is used as an example of the electrostatic field value acquisition unit 10, but the present invention is not limited to this, and various types of sensors can be used for the electrostatic field value acquisition unit 10. Furthermore, the electrostatic field value acquisition unit 10 can be placed in various locations where the electric field is not blocked. Furthermore, in this embodiment, the environmental information includes altitude, atmospheric pressure, humidity, and temperature, but the environmental information can be one or more of these pieces of information. Furthermore, the environmental information is not limited to altitude, atmospheric pressure, humidity, and temperature, and can include various types of information that may contribute to lightning strike prediction.

[0098] In this embodiment, the lightning strike probability prediction unit 30 is provided on the cloud server 31, but the lightning strike probability prediction unit 30 need not be provided on a server 31 such as the cloud server 31, and may instead be incorporated into the main body of the lightning strike prediction system 1. Furthermore, when the lightning strike probability prediction unit 30 is provided on the server 31, various types of servers can be used as the server 31, not just a cloud server. Furthermore, the server 31 can be an internal server, not just an external server. Furthermore, the server 31 can be located in various locations.

[0099] In this embodiment, notification is sent to an area An where the lightning strike probability is equal to or greater than a predetermined threshold, but the threshold is not limited to the example given and can be set to various values ​​taking into consideration environmental information, the topography of the measurement point Pn, etc. Also, in this embodiment, the explanation has been given mainly on area A1 (measurement point P1) among multiple areas An, but lightning strike probability can also be predicted for other areas A2 to An in the same way as for area A1.

[0100] In this embodiment, the lightning strike probability prediction unit 30 predicts the lightning strike probability based on a prediction algorithm generated using the amount of change in the electrostatic field value and environmental information over a predetermined period as training data. However, this is not limited to this; the lightning strike probability can also be predicted based on the actual amount of change without using a prediction algorithm (AI prediction). When using a prediction algorithm, various algorithms can be used, in addition to the exemplified XGBoost and LightGBM. Furthermore, the training data can be not only related to the atmospheric pressure, first electrostatic field value, second electrostatic field value, humidity, and temperature, but also various types of environmental information. Furthermore, in this embodiment, the training data is the amount of change in various types of environmental information. However, the training data may not only be based on the amount of change, but may also be composed of one or more combinations of absolute values ​​and amounts of change. Furthermore, in this embodiment, the training data includes the second electrostatic field value (amount of change + absolute value). However, the second electrostatic field value may be included as appropriate, or the second electrostatic field value may not be included. Furthermore, the training data may include, as an explanatory variable, the amount of change in environmental information plus the absolute value, such as the second electrostatic field value. Furthermore, the intervals at which the electrostatic field values ​​and environmental information are acquired and the intervals at which the amount of change is acquired can be set to various intervals depending on the characteristics of each piece of data.

[0101] In this embodiment, the lightning strike probability prediction unit 30 uses an electrostatic field value of 1 kV / m or more as a threshold for predicting the lightning strike probability, but the threshold can be set to various thresholds within a range that does not affect the lightning strike probability prediction, taking into account the performance of the electrostatic field value acquisition unit 10, etc. Furthermore, in this embodiment, an example is shown in which multiple measurement points Pn are located within 10 km of each other adjacent measurement points Pn, but the present invention is not limited to this, and the measurement points Pn can be located within various ranges. Furthermore, correction of the electrostatic field values ​​and environmental information at the measurement points Pn based on altitude can be performed as needed, or correction based on altitude can be omitted.

[0102] In this embodiment, the environmental information includes meteorological information (nowcast) from the Japan Meteorological Agency, but instead of or in addition to this, various types of meteorological information from government agencies and private companies can be used. Also, in this embodiment, lightning strike forecast information is used to predict the probability of lightning strikes, but the present invention can use not only the lightning strike forecast information itself but also various types of information that may be related to lightning strikes to predict the probability of lightning strikes.

[0103] Furthermore, in this embodiment, the estimated electrostatic field value and predicted environmental information in the non-measured area a are calculated based on the gradient state of each electrostatic field value acquired at a plurality of measurement points Pn, and the probability of lightning strikes in the non-measured area a is predicted from the calculated estimated electrostatic field value and predicted environmental information, but this is not limited to this, and the present invention can also be used in areas An other than the non-measured area a based on the above-mentioned gradient state. Furthermore, the above-mentioned gradient state may be not only a planar two-dimensional (2D) gradient but also a stereoscopic three-dimensional (3D) gradient.

[0104] The lightning strike prediction method in this embodiment is configured as described above in (n), but the lightning strike prediction system 1 used is not limited to the above embodiment and can be modified in various ways as described above. Furthermore, in this embodiment, area A1 is specified in area designation step S3, but any area A1 to An can be designated in area designation step S3, not just area A1. Furthermore, notifications in the lightning strike prediction method and lightning strike prediction system 1 can be made not only by issuing an alarm but also by various means such as email, communication, or telephone to businesses, facilities, etc. Furthermore, notifications can also be made to, for example, an alarm device installed in area An.

[0105] The above are various embodiments and modifications of the lightning strike prediction system and lightning strike prediction method according to the present invention. However, the present invention is not limited to the above-described embodiments and modifications, and it will be readily apparent to those skilled in the art that other embodiments are possible within the scope of the claims and the teachings and spirit of the present invention. [Industrial Applicability]

[0106] The lightning strike prediction system and method of the present invention can be used to predict lightning strikes in various regions, and can be used to notify and warn of lightning strike predictions to various businesses and facilities, such as semiconductor factories, automobile factories, steel mills with blast furnaces, schools, hospitals, and transportation facilities. [Explanation of symbols]

[0107] 1: Lightning strike prediction system 2: Thundercloud 10: Electrostatic field value acquisition unit 20: Information acquisition department 30: Lightning strike probability prediction section 31: Server, Cloud Server A1~An: Areas, candidate areas P1~Pn: Measurement points

Claims

1. an electrostatic field value acquisition unit that acquires atmospheric electrostatic field values ​​at a plurality of measurement points arranged in a plurality of regions; an information acquisition unit that acquires environmental information within a predetermined range centered on the measurement point; a lightning strike probability prediction unit that predicts the probability of a lightning strike; and the area is set as a predetermined range including one of the measurement points, The lightning strike probability prediction unit predicts the lightning strike probability for each region based on the electrostatic field value and the environmental information, and notifies the region where the predicted lightning strike probability is above a predetermined threshold.

2. The lightning strike prediction system of claim 1, wherein the lightning strike probability prediction unit predicts the lightning strike probability based on a prediction algorithm generated using learning data of the amount of change in the electrostatic field value and the environmental information over a predetermined period of time.

3. 3. The lightning strike prediction system of claim 1, wherein the area where the predicted lightning strike probability is equal to or greater than a predetermined threshold is designated as a candidate area, and notification is sent to the candidate area on the condition that the electrostatic field value in the candidate area is higher than the electrostatic field values ​​in other candidate areas adjacent to the candidate area.

Images