Quietness evaluation method, prediction model generation method, quietness evaluation device, and program

Abstract

To provide a technique that allows for more easily evaluating calmness of target waters in detail.SOLUTION: Calmness evaluation method according to an embodiment of the present invention comprises a first step (S531) of acquiring information on waves in first waters, and a second step (S534, S535) of outputting information on calmness of second waters based on the information on waves in the first waters using a prediction model built by machine learning in advance.SELECTED DRAWING: Figure 11

Description

[Technical Field]

[0001] The present invention relates to a quietness evaluation method, a prediction model generation method, a quietness evaluation device, and a program. [Background technology]

[0002] In port construction work and port management, it is important to understand the availability of work vessels and commercial vessels entering and leaving the port. In order to understand the availability, it is necessary to evaluate the wave height distribution within the port in response to incoming waves, which usually requires wave transformation analysis using a physical model.

[0003] Cited document 1 discloses that wave forecast data for time N is estimated by interpolating data based on wind direction and wind speed data contained in atmospheric analysis data in which the first estimated weather forecast value is corrected with observed values, and this is set in a wave prediction program as the initial condition value for predicting waves at time N+α, thereby predicting waves at time N+α.

[0004] Furthermore, cited document 2 discloses that a prediction model is constructed by machine learning using GMDH on observed values of wave information, such as at least one of the observed significant wave height and significant wave period, and forecast values for a location corresponding to a desired location using a global wave model. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2010-54460 [Patent Document 2] Japanese Patent Application Publication No. 2020-134315 Summary of the Invention [Problem to be solved by the invention]

[0006] To evaluate the operating rate of a port, it is necessary to estimate the detailed wave height distribution in the target sea area, and then perform a statistical analysis of the obtained estimates to estimate the calmness of the target sea area. Conventional wave transformation analysis requires time and effort not only for the calculation itself, but also for creating input data (initial topography data, expected wave conditions, etc.) and processing output data, as well as for dealing with errors that occur during the analysis. For this reason, a simple evaluation method that does not rely on detailed numerical analysis is needed.

[0007] Port utilization rates are also used as an indicator of the efficiency of work by ships entering and leaving the port. For example, it is necessary to accurately estimate the calmness of extremely small sea areas, such as the area (sea area) within a port where construction work ships anchor while working, or the area (sea area) where commercial ships dock to load and unload cargo. For this reason, there is a need for an evaluation method that can estimate the calmness of a target sea area in more detail and accurately.

[0008] For example, the technology in Patent Document 1 uses a mechanical calculation model (SWAN) as a wave transformation analysis model, which requires time and effort to create input data, process output data, and deal with errors that occur during analysis in order to obtain predicted values at a desired location and time.Furthermore, the technology in Patent Document 2 obtains predicted results for locations where observation data exists, so it is not suitable for estimating detailed wave height distribution within ports and harbors where no observation data exists.

[0009] The present invention has been made in view of the above-mentioned problems, and aims to provide a technique for more simply and precisely evaluating the calmness of a target sea area. [Means for solving the problem]

[0010] In order to solve the above problem, a calmness evaluation method according to one embodiment of the present invention is implemented by an information processing device that performs a first step of acquiring information about waves in a first sea area, and a second step of outputting information about the calmness of a second sea area using a predictive model that has been constructed in advance by machine learning based on the information about waves in the first sea area.

[0011] According to the above configuration, information about waves in a first sea area is acquired, and information about the calmness of a second sea area is output using a prediction model previously constructed by machine learning based on the information about waves in the first sea area. This significantly reduces the time required to obtain information about calmness compared to, for example, calculating information about the calmness of the second sea area using wave transformation analysis.

[0012] The information regarding waves obtained in the first step is wave conditions including at least wave height, period, and wave direction, and in the second step, information regarding calmness including a wave height contour diagram, which is wave height distribution information for the second sea area predicted by the prediction model, may be output.

[0013] According to the above configuration, the wave height distribution in the second sea area is predicted based on the wave height, period, and wave direction in the first sea area, thereby making it possible to predict the calmness in the second sea area in detail.

[0014] In the second step, the wave height contour map may be further divided into grid intervals of 5 m to 100 m, and converted into wave height data for each grid obtained from the wave height contour map.

[0015] According to the above configuration, the wave height contour map is further converted into wave height data for each grid, which makes it possible to check only the wave height distribution information for any location in the second sea area based on the output wave height contour map.

[0016] In the first step, a plurality of wave conditions are acquired, and in the second step, the occurrence ratio of waves below a predetermined limit wave height in the second sea area is calculated based on wave height data corresponding to a plurality of wave height distribution information predicted for each of the plurality of wave conditions, and the calculated ratio is output as information relating to the calmness.

[0017] According to the above configuration, the occurrence rate of waves equal to or less than a predetermined critical wave height in the second sea area is calculated based on wave height data corresponding to a plurality of pieces of wave height distribution information predicted for a plurality of wave conditions, thereby enabling more accurate prediction of the efficiency of vessel work in the second sea area.

[0018] The wave conditions may further include at least one of wind speed, wind direction, and tidal current.

[0019] According to the above configuration, at least one of wind speed, wind direction, and tidal current is further included in the wave conditions, which allows for more accurate prediction of wave height distribution.

[0020] In the first step, information regarding the topographical structures of the second sea area may further be input as an explanatory variable, and in the second step, information regarding the calmness of the second sea area may be output using a predictive model previously constructed by machine learning based on information regarding waves in the first sea area and information regarding the topographical structures of the second sea area.

[0021] According to the above configuration, by inputting information about the topographical structures in the second sea area as explanatory variables, it is possible to take into account the influence of the topographical structures in the target sea area on wave height, thereby enabling more accurate prediction of wave height distribution.

[0022] The method may further include a third step of acquiring training data including multiple pairs of information on waves in the first sea area and information on the calmness of the second sea area obtained by performing wave transformation analysis on the information on waves, and a fourth step of training a prediction model by referring to the training data, in which the information on waves in the first sea area is used as input data and the information on the calmness of the second sea area is used as output data.

[0023] According to the above configuration, training data including a plurality of pairs of information on waves in a first sea area and information on calmness in a second sea area obtained by performing wave transformation analysis on the information on waves is acquired, and a prediction model in which the information on waves in the first sea area is used as input data and the information on calmness in the second sea area is used as output data is trained with reference to the training data. A prediction model can be constructed through this training.

[0024] The information regarding waves obtained as the training data is wave conditions including at least wave height, period, and wave direction, and the information regarding calmness obtained as the training data may include a wave height contour diagram, which is the wave height distribution in the second sea area obtained by performing the wave deformation analysis.

[0025] According to the above configuration, the information on waves acquired as training data is wave conditions including at least wave height, period, and wave direction, and the information on calmness acquired as training data includes a wave height contour map, which is the wave height distribution in the second sea area obtained by performing wave transformation analysis. This makes it possible to perform learning suitable for, for example, evaluating the calmness of the second sea area in detail.

[0026] The second sea area is divided into a plurality of divided sea areas, and the method further comprises a fifth step of acquiring analytical wave height values and measured wave height values for each divided sea area based on an analytical wave height contour map, which is a wave height contour map for the second sea area obtained by performing wave deformation analysis under specified wave conditions for the first sea area, and a measured wave height contour map, which is a wave height contour map for the second sea area actually measured under the same wave conditions as the specified wave conditions, and a sixth step of calculating a correction value for each divided sea area to correct the analytical wave height value to the measured wave height value, and in the fourth step, information about the calmness including a corrected wave height contour map, which is a wave height contour map obtained by correction using the correction value, may be re-learned as the training data.

[0027] According to the above configuration, a correction value for correcting the analytical wave height value to the measured wave height value is calculated for each divided sea area, and information on calmness including the corrected wave height contour map obtained by correcting with the correction value is re-learned as training data, thereby enabling the construction of a prediction model that enables more accurate predictions.

[0028] In the wave transformation analysis, information regarding the topographical structures of the second sea area may further be used, and in the third step, training data may be acquired that includes multiple sets of information regarding the waves of the first sea area and information regarding the calmness of the second sea area obtained by performing wave transformation analysis on the information regarding the waves and the information regarding the topographical structures of the second sea area.

[0029] According to the above configuration, in the wave transformation analysis, information on the topography and structures of the second sea area is further used to generate training data, which makes it possible to acquire more appropriate training data during learning.

[0030] A method for generating a predictive model according to one embodiment of the present invention is carried out by an information processing device, which performs the following steps: acquiring training data including multiple pairs of information relating to wave conditions in a first sea area and information relating to the calmness of a second sea area; and learning a predictive model by referring to the training data, in which information relating to wave conditions in the first sea area is input data and information relating to the calmness of the second sea area is output data.

[0031] According to the above configuration, training data including multiple pairs of information on wave conditions in a first sea area and information on the calmness of a second sea area is acquired, and a prediction model in which the information on wave conditions in the first sea area is used as input data and the information on the calmness of the second sea area is used as output data is trained with reference to the training data. This makes it possible to train a prediction model that can significantly reduce the time required to output information on calmness, compared to when, for example, information on the calmness of the second sea area is calculated using wave transformation analysis.

[0032] A calmness evaluation device according to one embodiment of the present invention comprises an information acquisition unit that acquires information regarding waves in a first sea area, and an information output unit that outputs information regarding the calmness of a second sea area using a predictive model that has been constructed in advance by machine learning based on the information regarding waves in the first sea area.

[0033] According to the above configuration, information about waves in a first sea area is acquired, and information about the calmness of a second sea area is output using a prediction model previously constructed by machine learning based on the information about waves in the first sea area. This significantly reduces the time required to output information about calmness compared to, for example, calculating information about the calmness of the second sea area using wave transformation analysis.

[0034] A program according to one embodiment of the present invention is a program for causing a computer to function as a quietness evaluation device, and causes the computer to function as the information acquisition unit and the information output unit.

[0035] According to the above configuration, information about waves in a first sea area is acquired, and information about the calmness of a second sea area is output using a prediction model previously constructed by machine learning based on the information about waves in the first sea area. This significantly reduces the time required to output information about calmness compared to, for example, calculating information about the calmness of the second sea area using wave transformation analysis. [Effects of the Invention]

[0036] According to one aspect of the present invention, the calmness of a target sea area can be evaluated more simply and in detail. [Brief explanation of the drawings]

[0037] [Figure 1] 1 is a block diagram showing an example of the configuration of a learning device according to a first embodiment; [Figure 2] FIG. 1 is a diagram illustrating a first sea area and a second sea area. [Figure 3] 1 is a diagram showing an example of wave height distribution in a sea area 72. FIG. [Figure 4] FIG. 10 is a diagram illustrating conversion from a wave height contour map to wave height data. [Figure 5] FIG. 10 is a diagram showing an example of wave height data. [Figure 6] 10 is a flowchart illustrating an example of a teacher data generation process performed by a learning device. [Figure 7] 10 is a flowchart illustrating an example of a learning process performed by a learning device. [Figure 8] FIG. 10 is a diagram illustrating an area related to correction of wave height data. [Figure 9] FIG. 10 is a block diagram showing an example of the configuration of a learning and prediction device according to a second embodiment. [Figure 10] 10 is a flowchart illustrating an example of a quietness evaluation process performed by the learning and prediction device. [Figure 11] 10 is a flowchart illustrating details of a wave height distribution prediction process performed by the learning and prediction device. [Figure 12]10 is a flowchart illustrating details of a pulse height statistical analysis process performed by the learning and prediction device. [Figure 13] FIG. 10 is a diagram illustrating the occurrence rate of wave height. [Figure 14] FIG. 10 is a diagram illustrating the occurrence rate of wave height. [Figure 15] FIG. 10 is a block diagram showing an example of the configuration of a prediction device according to a third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0038] [Embodiment 1] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Fig. 1 is a block diagram showing an example of the configuration of a learning device 1 according to the first embodiment.

[0039] As shown in the figure, the learning device 1 includes a control unit 10, a storage unit 20, a communication unit 30, an input unit 40, and an output unit 50. The learning device 1 is configured by, for example, a general-purpose computer.

[0040] The control unit 10 includes an acquisition unit 11, a teacher data generation unit 12, and a learning unit 13.

[0041] The acquisition unit 11 acquires information about waves in the sea area 71, which is the first sea area. For example, the acquisition unit 11 acquires wave conditions for the sea area 71 as information about waves in the sea area 71. The wave conditions are, for example, the wave height, period, and wave direction of the sea area 71. The wave conditions for the sea area 71 may further include at least one of wind speed, wind direction, and tidal current.

[0042] The acquisition unit 11 also acquires information (e.g., terrain structure conditions) related to the topography and structures of the sea area 72 as the second sea area. The terrain structure conditions of the sea area 72 are, for example, information indicating a topography contour map of the seabed of the sea area 72, and the position, arrangement, size, height, etc. of the topography and structures constructed in the sea area 72.

[0043] The topographical structures include offshore breakwaters, breakwaters, depressions and other irregularities on the seabed, etc. Note that the topographical structure conditions for the sea area 72 may not be acquired.

[0044] The acquisition unit 11 further acquires the wave height distribution of the sea area 72. An example of the wave height distribution is a wave height contour map of the sea area 72. The wave height distribution acquired here is information related to the calmness of the sea area 72. The wave height distribution of the sea area 72 is obtained, for example, by performing a wave transformation analysis that takes into account the topographical and structural conditions of the sea area 72 with respect to the wave conditions of the sea area 71. Note that if the topographical and structural conditions of the sea area 72 have not been acquired, the wave height distribution of the sea area 72 may be obtained by performing a wave transformation analysis with respect to the wave conditions of the sea area 71.

[0045] Wave transformation analysis is a mechanical model that analyzes how waves propagate under given wave conditions, and is a computational process that solves equations of motion and continuity equations. For example, by performing wave transformation analysis for offshore wave conditions, it is possible to virtually determine the wave height distribution in a harbor when offshore waves propagate to the coast. Wave transformation analysis may be performed, for example, by a computer or by desk calculation.

[0046] FIG. 2 is a diagram illustrating a sea area 71 (first sea area) and a sea area 72 (second sea area). The sea area 71 is, for example, a larger sea area than the sea area 72, but this does not limit the present embodiment. The sea area 72 is a sea area for which calmness should be evaluated, such as a sea area where a ship is anchored. The sea area 72 is, for example, a sea area within a harbor. In this example, the sea area 72 is the sea area within the bay of a port (Yokohama Port). The sea area 72 may also be a predetermined sea area close to the coast, or may be a sea area of an offshore anchorage outside the harbor. The sea area 71 is, for example, a sea area located offshore from the sea area 72, but this does not limit the present embodiment. The sea area 71 may, for example, be a sea area that includes the entire coastal area of the Japanese archipelago. The sea area 72 may be included within the sea area 71 as part of the sea area 71.

[0047] FIG. 3 is a diagram showing an example of wave height distribution in a sea area 72. In this example, the wave height distribution in the sea area 72 is shown as a wave height contour map. A wave height contour map is a diagram that displays the wave height in a sea area so that it can be visually recognized, for example. In this example, waves ranging from 0 m to 1.3 m in height are shown color-coded. Note that in this figure, the color coding in the wave height contour map is expressed by shades of gray.

[0048] The wave height contour map of the sea area 72 may be further converted into wave height data, as will be described later.

[0049] FIG. 4 is a diagram illustrating the conversion of a wave height contour map into wave height data for each grid. For example, a grid mesh is overlaid on the wave height contour map, and numerical values are read from the contour shades in the grid sections and converted into wave height data, thereby converting the wave height contour map into wave height data. The wave height contour map for the sea area 72 is divided into a predetermined grid interval (hereinafter referred to as "meshing" where appropriate). In other words, the sea area 72 is meshed on the wave height contour map for the sea area 72. Each meshed section is a rectangular area with one side measuring 5 m to 100 m.

[0050] In the example of Figure 4, one section (grid) of the meshed sea area 72 is shown as a dotted rectangle. Although Figure 4 shows only one section as an example, many sections of the same size will be set in the sea area 72. Here, data indicating the wave height corresponding to each section is wave height data. In the example of Figure 4, if the coordinates (X, Y) of the lower left corner in the figure, which are the representative coordinates of the section, are expressed as (nkm, mKm), the coordinates of the lower left corner of the section shown by the dotted rectangle are expressed as 2.400 km in the X-axis direction (horizontal direction in the figure) and 1.700 km in the Y-axis direction (vertical direction in the figure). The wave height (H) of the section in question is identified as 1.05 m from the color of the wave height contour map. In addition, when the size of a section is large and multiple colors can be confirmed on the wave height contour map within the same section (when multiple wave heights can be confirmed within the same section), it is desirable to identify the wave height of the section from the average color of the section.

[0051] In this way, wave height data is generated by identifying the wave height of each section of the meshed wave height contour map. Figure 5 is a diagram showing an example of wave height data. For example, as shown in Figure 5, the wave height data is generated as each of the representative coordinates (X, Y) of each section and the corresponding wave height (H). Note that the representative coordinate of each section may be one of the same four corners of each grid (lower left, upper left, lower right, or upper right corner), or may be set to the center point of the section. When one of the four corners is used, it is not limited to the lower left coordinate (X·Y), and may also be the upper right coordinate or the lower right coordinate.

[0052] The teacher data generating unit 12 generates teacher data linking the wave conditions of the sea area 71 with the wave height contour map of the sea area 72. The teacher data generating unit 12 may generate teacher data linking the topographical and structural conditions of the sea area 72 with the wave height contour map of the sea area 72 in addition to the wave conditions of the sea area 71.

[0053] The learning unit 13 uses teacher data to train a prediction model. That is, the learning unit 13 trains a prediction model in which information about waves in the sea area 71 (for example, wave conditions) is used as input data, and information about the calmness of the sea area 72 (for example, a wave height contour map) is used as output data. The learning unit 13 may also train a prediction model in which information about waves in the sea area 71 (for example, wave conditions) and topographical and structural conditions in the sea area 72 are used as input data, and information about the calmness of the sea area 72 (for example, a wave height contour map) is used as output data.

[0054] The memory unit 20 stores training data TD. The training data TD consists of a set of wave conditions WC and a wave height contour map WF linked to the wave conditions, and multiple sets of training data are stored. Here, multiple sets of training data will be collectively referred to as training data TD.

[0055] It should be noted that the wave height data corresponding to the wave height contour diagram included in the training data may be corrected during learning by the learning unit 13. The details of the correction of the wave height data will be described later.

[0056] Here, an example of training data consisting of a set of wave conditions for sea area 71 and a wave height contour map for sea area 72 has been described, but the present invention is not limited to this. In short, it is sufficient to generate training data that links information about the waves for sea area 71 with information about the calmness of sea area 72.

[0057] The storage unit 20 further stores model parameters MP. The model parameters MP are, for example, coefficients included in a prediction model having a CNN (Convolutional Neural Network) structure, and are updated by machine learning. However, this does not limit the present embodiment, and the model parameters MP may be, for example, parameters for defining a linear or nonlinear regression model. These machine learning operations are performed by the learning unit 13. Note that the storage unit 20 does not have to be built into the learning device 1, and may be configured to be externally connected to the learning device 1 via a communication unit 30, which will be described later, for example.

[0058] In the above description, neural networks such as CNN and regression models are used as examples of prediction models, but this does not limit the present embodiment, and a recurrent neural network (RNN) or the like may also be used as the prediction model. Furthermore, non-neural network models such as random forests and support vector machines may also be used as the prediction model.

[0059] The communication unit 30 has a function of connecting to, for example, a local area network, a wide area network, etc. For example, it is also possible to obtain teacher data via the communication unit 30 from an external server, etc.

[0060] The input unit 40 is configured by, for example, a keyboard, an OCR scanner, etc. For example, teacher data, topographical and structural conditions, etc. are input via the input unit 40.

[0061] The output unit 50 is configured by, for example, a display, a speaker, a printer, etc., and displays the results of various processes performed by the device on a screen or outputs them as sound or figures.

[0062] In this way, the learning device 1 is configured to learn the model parameters MP of a prediction model that takes the wave conditions of a first sea area (e.g., sea area 71) as input and outputs a wave height contour map as the wave height distribution of a second sea area (e.g., sea area 72).

[0063] Next, an example of the teacher data generation process performed by the learning device 1 will be described with reference to the flowchart of FIG.

[0064] 6, in step S11, the acquisition unit 11 acquires information about waves in the sea area 71. In this example, in step S11, a plurality of wave conditions including the wave height, period, and wave direction in the sea area 71 are acquired as information about waves. The wave conditions may further include at least one of wind speed, wind direction, and tidal current.

[0065] In step S12, the acquisition unit 11 acquires topographical and structural conditions. The topographical and structural conditions are acquired as data obtained from, for example, a topographical contour map of the seabed of the sea area 72, and information indicating the layout, position, size, height, etc. of topographical and structural structures such as offshore breakwaters and breakwaters constructed in the sea area 72. This data may be, for example, the height of the seabed or structures corresponding to each section of the wave height data, and data in which the distance between the sea surface and the seabed or the sea surface and the structures is associated with the coordinates of each section.

[0066] For example, the topography of the seabed (such as depressions and other irregularities on the seabed surface) affects the height of waves that appear on the water surface within a port. Furthermore, structures within a port (such as breakwaters) also affect the height of waves that appear on the water surface within a port. Therefore, if the topographical and structural conditions are taken into account in wave transformation analysis, the wave height distribution within a port can be calculated more accurately.

[0067] Alternatively, if a new breakwater is planned to be constructed in a port, the geographical structure conditions including the constructed breakwater may be acquired. In this way, the wave height distribution in a port with a breakwater that does not actually exist yet can be virtually determined by wave transformation analysis. Alternatively, the conditions (location, position, size, height, etc.) of the geographical structure to be constructed to achieve the desired wave height distribution in the port can be determined.

[0068] In step S13, the teacher data generation unit 12 performs a wave deformation analysis using the wave conditions for the sea area 71 acquired in step S11 and the terrain and structure conditions for the sea area 72 acquired in step S12. Note that the processing of step S12 may not be executed. In other words, the wave deformation analysis may be performed without acquiring the terrain and structure conditions.

[0069] In step S14, the teacher data generating unit 12 generates information relating to the calmness of the sea area 72, which is derived from the results of the wave transformation analysis performed in step S13. That is, information relating to a plurality of calmness levels of the sea area 72 obtained by performing wave transformation analysis on each of the pieces of information relating to the plurality of waves of the sea area 71 acquired in step S11 is generated by the teacher data generating unit 12. In this example, in step S14, for example, a plurality of wave height contour maps described with reference to Fig. 3 are generated.

[0070] It should be noted that, for example, if it is possible to obtain a wave height distribution in the sea area 72 corresponding to predetermined wave conditions in the sea area 71 from publicly available weather information, etc., it is not necessary to perform wave transformation analysis in step S13. In this case, in step S14, it is sufficient to obtain the wave height distribution in the sea area 72 obtained from publicly available weather information, etc.

[0071] In step S15, the teacher data generation unit 12 generates teacher data that links the plurality of pieces of information about waves (wave conditions) acquired in the processing of step S11 with the plurality of pieces of information about calmness (e.g., wave height contour maps) generated in the processing of step S14. In this way, teacher data TD is generated that includes a plurality of pairs of wave conditions WC for the sea area 71 and wave height contour maps WF corresponding to those wave conditions. Teacher data may also be generated that includes a plurality of pairs of wave height contour maps WF corresponding to the wave conditions WC for the sea area 71 and the topographical and structural conditions for the sea area 72. The generated teacher data TD is stored in the storage unit 20.

[0072] The wave conditions used as training data may be obtained from publicly available data via the communication unit 30. For example, wave conditions derived from Grid Point Value (GPV) data published by the Japan Meteorological Agency may be obtained. Similarly, if publicly available information is available about the wave height contour map used as training data, that information may be used.

[0073] Next, an example of the learning process performed by the learning device 1 will be described with reference to the flowchart of FIG.

[0074] In step S31, the learning unit 13 acquires training data including multiple pairs of wave conditions WC for the sea area 71 and wave height contour maps WF corresponding to those wave conditions, and corrects wave height data WD corresponding to the wave height contour maps WF included in the training data TD. At this time, the learning unit 13 reads and acquires the training data TD from the storage unit 20, and corrects wave height data for a predetermined region in the sea area 72, which is wave height data obtained by converting the wave height contour maps, using a correction coefficient.

[0075] Wave transformation analysis is a calculation using a predetermined physical model, and the wave height contour map (analyzed wave height contour map) obtained as a result of the wave transformation analysis as the wave height distribution in the sea area 72 may differ from the wave height contour map (measured wave height contour map) as the wave height distribution actually observed in the sea area 72. Correction of wave height data is performed to bring the wave height of the analytical wave height contour map obtained as the analysis result closer to the wave height of the measured wave height contour map obtained as the actual measurement result. For this reason, the wave height value of the analysis result for the sea area 72 (analyzed wave height value) and the actually measured wave height value (measured wave height value) are obtained.

[0076] For example, one or more predetermined areas (divisions of sea areas) related to the correction of wave height data are set within the sea area 72. The predetermined areas are set, for example, by the learning unit 13. FIG. 8 is a diagram illustrating the areas related to the correction of wave height data. These areas may include, for example, multiple sections as described with reference to FIG. 4, and are areas 1 to 5 represented by circled numbers 1 to 5 in the wave height contour diagram shown in FIG. 8. The sizes of the areas may be the same or different. In the example of FIG. 8, five areas are set, but more areas may be set. Alternatively, only one area may be set.

[0077] Next, the peak value of the analysis result (analysis peak value) is compared with the actually measured peak value (measured peak value) for each of the above-mentioned regions. If multiple peak values exist in one region, the average values of these peak values may be compared. The comparison between the analysis peak value and the measured peak value is performed, for example, by the learning unit 13. Then, for example, the learning unit 13 calculates a correction value A = measured peak value / analysis peak value for each region. For example, correction values A1 to A5 are calculated as correction values to be applied to regions 1 to 5, respectively.

[0078] In step S31, the learning unit 13 corrects the wave height contour map by multiplying the wave height data obtained by converting the wave height contour map WF in the training data by a correction value. At this time, the learning unit 13 determines whether the wave height data belongs to any of regions 1 to 5 based on the representative coordinates (X, Y) of the section of the wave height data, and identifies the wave height data belonging to regions 1 to 5. The learning unit 13 then multiplies the wave height (H) of the wave height data belonging to region 1 by the correction value A1, multiplies the wave height (H) of the wave height data belonging to region 2 by the correction value A2, multiplies the wave height (H) of the wave height data belonging to region 3 by the correction value A3, ..., multiplies the wave height (H) of the wave height data belonging to region 5 by the correction value A5. A wave height contour map (corrected wave height contour map) is then created from the wave height data multiplied by the correction value. Note that the processing of step S31 may be omitted.

[0079] In step S32, the learning unit 13 performs machine learning using the teacher data corrected in the processing of step S31 (teacher data TD consisting of wave conditions WC and corrected wave height contour map WF). That is, the learning unit 13 trains a prediction model using information about waves in the sea area 71 as input data and information about the calmness of the sea area 72 as output data. Note that if the correction processing in step S31 is not performed, in step S32, machine learning is performed using the teacher data TD read out from the storage unit 20 as is. Alternatively, machine learning may be performed once using the teacher data TD read out from the storage unit 20 as is, and then re-learning may be performed using the corrected teacher data. In this way, the correlation between the wave conditions WC and the wave height contour map WF is machine-learned. That is, the prediction model is trained by updating the model parameters of the prediction model so that the difference between the information about the calmness of the sea area 72 output by the prediction model and the information about calmness included in the teacher data becomes smaller.

[0080] As described above, the training data TD may include multiple sets of wave height contour maps WF corresponding to the wave conditions WC of the sea area 71 and the topographical and structural conditions of the sea area 72. In this case, the correlation between the wave conditions WC and the topographical and structural conditions of the sea area 72 and the wave height contour maps WF is machine-learned. In other words, the training data TD may include the topographical and structural conditions of the sea area 72 to train a prediction model, or may not include the topographical and structural conditions of the sea area 72 to train a prediction model.

[0081] In step S33, the learning unit 13 updates the model parameters. At this time, the learning unit 13 stores in the storage unit 20 the model parameters learned in the process of step S32.

[0082] In the above example, in step S31, the wave height is corrected using correction values A1 to A5 corresponding to regions 1 to 5, but more detailed corrections may be performed. For example, different correction values may be applied depending on whether the wave height of the wave conditions is equal to or greater than a preset threshold value.

[0083] For example, the analytical wave height value corresponding to a wave condition with a wave height less than a predetermined threshold is compared with the measured wave height value, and a correction value B01 to be applied to the wave height data for region 1 is calculated. Similarly, the analytical wave height value corresponding to a wave condition with a wave height equal to or greater than a threshold is compared with the measured wave height value, and a correction value B11 to be applied to the wave height data for region 1 is calculated. In this manner, correction values B01, B11, B02, B12, B03, B13, ..., B05, B15 to be applied when the wave height of the wave condition is less than the threshold and when the wave height is equal to or greater than the threshold are calculated for regions 1 to 5. Note that in this example, the wave height of the wave condition is divided into two by one threshold, and two correction values are calculated for each region. However, the wave height of the wave condition may be divided into three by two thresholds, and three correction values may be calculated for each region. Furthermore, correction values may be calculated by classifying wave heights by three or more thresholds.

[0084] Alternatively, multiple correction values may be calculated for each region by comparing the period or direction of the wave conditions with one or more thresholds. Additionally, multiple correction values may be calculated for each region by comparing the wave height, period, and / or a combination of wave heights of the wave conditions with one or more thresholds.

[0085] The learning process is carried out in this manner. Once the learning process is completed using the training data TD, which contains a sufficient number of pairs of wave conditions WC and wave height contour maps WF, predictions can be made using a prediction model. For example, the training data TD includes multiple wave conditions WC, but it becomes possible to predict the wave height distribution in the sea area 72 corresponding to wave conditions that do not fall under any of these wave conditions WC.

[0086] For example, consider the case of wave conditions consisting of wave height, period, and direction. Furthermore, suppose that, in order to evaluate the calmness of sea area 72, it is necessary to identify the wave height distribution in sea area 72 corresponding to wave conditions of 20 wave heights, 20 periods, and 20 wave directions in sea area 71. In this case, conventional methods would require performing wave transformation analysis on 8,000 (= 20 × 20 × 20) wave conditions to determine the wave height distribution. In contrast, with the present invention, if only 1 / 5 to 1 / 10 of the analysis results are available, the wave height distribution can be predicted using a prediction model obtained as a result of machine learning. In other words, it becomes possible to quickly predict how the wave height distribution in sea area 72 will change in response to any wave condition in sea area 71.

[0087] [Embodiment 2] Next, a learning and prediction device according to another embodiment of the present invention will be described with reference to Fig. 9. Fig. 9 is a block diagram showing an example of the configuration of a learning and prediction device 101 according to embodiment 2. In addition to the functions of the learning device 1 described with reference to Fig. 1, the learning and prediction device 101 in Fig. 9 is capable of predicting wave height distribution using a prediction model and performing wave height statistical analysis.

[0088] 9, the learning and prediction device 101 includes a control unit 110, a storage unit 20, a communication unit 30, an input unit 40, and an output unit 50. The learning and prediction device 101 is configured, for example, by a general-purpose computer. In the following description, functional blocks having the same functions as those in FIG. 1 are designated by the same reference numerals as those in FIG. 1, and detailed description thereof will be omitted.

[0089] In the control unit 110, the acquisition unit 11, the teacher data generation unit 12, and the learning unit 13 are functional blocks similar to the acquisition unit 11, the teacher data generation unit 12, and the learning unit 13 in Fig. 1. Note that the acquisition unit 11 may also be referred to as the first acquisition unit 11 to distinguish it from the second acquisition unit 114 described later.

[0090] The control unit 110 further includes a second acquisition unit 114, a prediction unit 115, and an information output unit .

[0091] The second acquisition unit 114 acquires information about waves to be used for prediction by the prediction model obtained as a result of machine learning. The information about waves acquired by the second acquisition unit may be, for example, wave conditions in the sea area 71 obtained from publicly available weather information, or wave conditions in the sea area 71 estimated from wind speed and wind direction. Alternatively, it may be wave conditions actually measured in the sea area 71.

[0092] 1, the wave conditions include, for example, the wave height, period, and wave direction of the sea area 71. The wave conditions of the sea area 71 may further include at least one of the wind speed, wind direction, and tidal current.

[0093] Furthermore, the second acquisition unit 114 acquires the latest topographical and structural conditions of the sea area 72 as explanatory variables used in predictions by the prediction model. For example, because the topography of the seabed can change due to tidal currents, more accurate predictions can be made by using newer topographical and structural conditions as explanatory variables than those used when generating the training data.

[0094] Alternatively, if a new breakwater is planned to be constructed in a port, the topographical and structural conditions including the constructed breakwater may be acquired as explanatory variables. In this way, the prediction model can be used to predict the wave height distribution in a port with a breakwater that does not actually exist yet.

[0095] Alternatively, other conditions that are different from the topographical structure conditions and that may affect the wave height distribution in the sea area 72 may be acquired as explanatory variables. Note that the second acquisition unit 114 may not acquire explanatory variables.

[0096] The prediction unit 115 outputs information about the calmness of the sea area 72 using the wave conditions acquired by the second acquisition unit 114, or the wave conditions and topographical structure conditions acquired by the second acquisition unit 114. For example, the wave height distribution in the sea area 72 is predicted and output as information about the calmness. As an example, the prediction unit 115 performs a calculation process on the wave conditions according to a prediction model, and predicts a wave height contour map for the sea area 72 corresponding to the wave conditions of the sea area 71. In this calculation process, the model parameters MP stored in the storage unit 20 are used. That is, in predicting the wave height distribution by the prediction unit 115, a prediction model defined by the model parameters MP updated by the learning process described above with reference to FIG. 7 is used.

[0097] The information output unit 116 outputs information about the calmness of the sea area 72 based on information about waves in the sea area 71 using the prediction model previously constructed by machine learning. That is, the information output unit 116 outputs information representing the prediction result of the prediction unit 115 as information about the calmness of the sea area 72. The information output unit 116 outputs, for example, a wave height contour map of the sea area 72 obtained as the prediction result. Furthermore, as described above with reference to FIG. 4, the information output unit 116 generates wave height data consisting of representative coordinates (X, Y) and wave heights (H) corresponding to each region (section) when the wave height contour map of the sea area 72 is meshed into rectangular regions with sides of 5 to 100 meters. The information output unit 116 then outputs the wave height data for each region.

[0098] Furthermore, after predictions of multiple wave height distributions corresponding to multiple wave conditions have been made, the information output unit 116 statistically analyzes the wave height data for the sea area 72 and calculates the occurrence frequency of waves with wave heights equal to or less than a predetermined wave height in the sea area 72. For example, the wave height data of wave height distributions corresponding to each of several hundred different wave conditions is statistically analyzed. As a result of the statistical analysis, the information output unit 116 calculates the occurrence frequency of waves with each wave height in the sea area 72. The calculated occurrence frequency of waves with wave heights equal to or less than a critical wave height is used as information regarding the calmness of the sea area 72.

[0099] For example, wave height values from 0 m to 10 m are classified into ranges every 50 cm, and the occurrence frequency of wave height values belonging to each classified range is calculated by the information output unit 116. Furthermore, the information output unit 116 outputs the occurrence frequency of waves with a wave height equal to or less than a predetermined wave height as the calmness of the sea area 72. Here, the predetermined wave height is, for example, the maximum wave height at which a ship anchored in a port can load and unload cargo, carry out construction work, and the like, and is called the critical wave height.

[0100] The calmness index is expressed, for example, by classifying the occurrence rate of wave heights below the critical wave height into five levels. If the occurrence rate of wave heights below the critical wave height is 95% or more, the calmness index is expressed as 5, if it is 90% or more but less than 95%, the calmness index is expressed as 4, if it is 85% or more but less than 90%, the calmness index is expressed as 3, etc. The calmness index may also be output in daily, half-day, or three-hourly units.

[0101] In this way, the learning and prediction device 101 is configured to learn the model parameters MP of the prediction model, and to use the learned model parameters to generate a wave height contour map for the sea area 72 from the wave conditions in the sea area 71 and predict the wave height distribution. The learning and prediction device 101 is further configured to output the degree of calmness for the sea area 72, which is derived from the predicted wave height distribution.

[0102] Next, an example of the quietness evaluation process performed by the learning and prediction device 101 in FIG. 9 will be described with reference to the flowchart in FIG.

[0103] In step S51, the learning and prediction device 101 executes a teacher data generation process. As a result, teacher data TD is generated, and the generated teacher data TD is stored in the storage unit 20. This process is similar to the process described with reference to FIG. 6, and therefore a detailed description thereof will be omitted.

[0104] In step S52, the learning and prediction device 101 executes a learning process, thereby updating the model parameters MP in the storage unit 20. This process is similar to the process described with reference to Fig. 7, and therefore a detailed description thereof will be omitted.

[0105] In step S53, the learning and prediction device 101 executes a wave height distribution prediction process, thereby predicting the wave height distribution in the sea area 72. Details of this process will be described later with reference to FIG.

[0106] In step S54, the learning and prediction device 101 executes a wave height statistical analysis process. This executes a statistical analysis of the wave height data for the sea area 72, and calculates the occurrence frequency of waves with a wave height equal to or less than the critical wave height in the sea area 72, which is output as a calmness index. Details of this process will be described later with reference to FIG. 12.

[0107] In this manner, the quietness evaluation process is carried out.

[0108] Next, the details of the wave height distribution prediction process in step S53 in FIG. 10 will be described with reference to the flowchart in FIG.

[0109] In step S531, the second acquisition unit 114 acquires information about waves in the sea area 71. In this example, in step S531, wave conditions including wave height, period, and wave direction in the sea area 71 are acquired. The wave conditions may further include at least one of wind speed, wind direction, and tidal current. In step S532, the second acquisition unit 114 acquires explanatory variables. As a result, for example, the latest topographical and structural conditions for the sea area 72 are acquired. Note that the processing of step S532 is executed when, in the teacher data generation processing of step S51, teacher data including multiple pairs of wave height contour maps WF corresponding to the wave conditions WC for the sea area 71 and the topographical and structural conditions for the sea area 72 is generated, and when, in the learning processing of step S52, correlations between the wave conditions WC and the topographical and structural conditions for the sea area 72 and the wave height contour maps WF are machine-learned. In other words, if the prediction model used in the wave height distribution prediction processing of step S53 is a prediction model learned without including the topographical and structural conditions in the teacher data, the topographical and structural conditions do not need to be acquired as explanatory variables.

[0110] In step S533, the prediction unit 115 performs calculation processing using the prediction model. As a result, calculation processing according to the prediction model is performed using the wave conditions acquired in step S531 and the explanatory variables acquired in step S532. In this calculation processing, the model parameters MP stored in the storage unit 20 and updated by the learning processing in step S52 are used.

[0111] As described above, the process of step S532 may not be executed. That is, in step S533, calculation processing may be performed using the prediction model without acquiring explanatory variables. In this case, the prediction unit 115 performs calculation processing according to the prediction model for the wave conditions acquired in step S531.

[0112] In step S534, the prediction unit 115 outputs information about the calmness of the sea area 72 from the result of the calculation process using the prediction model performed in step S533. In this example, wave height distribution information for the sea area 72 is output in step S534. As a result, the prediction unit 115 generates and outputs a wave height contour diagram as wave height distribution information for the sea area 72. In step S535, the information output unit 116 divides the wave height contour map into grid intervals of 5 m to 100 m, and converts the data into wave height data for each grid obtained from the wave height contour map. The wave height data is generated as representative coordinates (X, Y) of each section with a side length of 5 m to 100 m and the corresponding wave height (H), as shown in Fig. 5, for example.

[0113] In this way, the wave height distribution prediction process is executed.

[0114] Next, the pulse height statistical analysis process in step S54 of FIG. 10 will be described in detail with reference to the flowchart of FIG.

[0115] In step S541, the information output unit 116 statistically analyzes the wave height data. At this time, the information output unit 116 statistically analyzes, for example, a plurality of wave height distributions predicted corresponding to each of a plurality of wave conditions in the sea area 71. As a result, the occurrence frequency of wave heights in the sea area 72 is calculated.

[0116] 13 and 14 are diagrams explaining the occurrence of wave heights. FIG. 13 is a table explaining the occurrence of each wave height in sea area 71, and FIG. 14 is a table explaining the occurrence of each wave height in sea area 72. In the tables of FIGS. 13 and 14, the leftmost column indicates the range of wave heights. In this example, wave heights from 0 m to 10 m are categorized in 20 ranges, each in 50 cm increments, and wave heights over 10 m are categorized into one range.

[0117] In the tables of Figures 13 and 14, the top row indicates the wave direction. N, E, S, and W represent north, east, south, and west, respectively, with NE representing northeast, NNE representing north-northeast, ENE representing east-northeast, etc. Each wave direction is indicated in the table. The cells corresponding to each wave direction and wave height indicate the number of times waves of that wave height occurred in that wave direction. The bottom two rows of the table show the total number of times and percentage of waves that occurred for each wave direction. The rightmost two columns of the table show the total number of times and percentage of waves that occurred for each wave height.

[0118] In the process of step S541, for example, the occurrence frequency of each wave height is calculated, which is shown as a ratio in the rightmost column of Fig. 14. The occurrence frequency of a wave height may also be referred to as the occurrence ratio of the wave height.

[0119] In step S542, the information output unit 116 calculates the occurrence frequency of wave heights equal to or less than a critical wave height. The critical wave height is, for example, the maximum wave height at which loading and unloading of cargo, construction work, etc., on a ship anchored in a port can be carried out without any problems, and is assumed to be determined in advance.

[0120] For example, if the critical wave height is 1 m, then according to the table in FIG. 14, the occurrence rate of wave heights equal to or less than the critical wave height in sea area 72 is 88.91+10.92=99.83%.

[0121] In step S543, the information output unit 116 outputs the calmness of the sea area 72. The calmness is expressed, for example, by classifying into five levels the occurrence frequency of wave heights equal to or less than the critical wave height, as described above.

[0122] In this way, the pulse height statistical analysis process is carried out.

[0123] In this way, by predicting the wave height distribution in sea area 72 and calculating the calmness, it is possible to predict the degree to which ship operations can be carried out without hindrance in the port in sea area 72. For example, if a wave forecast for the next month in sea area 71 can be obtained as weather forecast information, it is possible to identify the wave conditions for the next month from the obtained wave forecast and predict the wave height distribution and calmness in sea area 72 corresponding to the identified wave conditions. In other words, it is possible to predict the degree to which ship operations can be carried out without hindrance in the port for the next month. By planning ship operations based on such predictions, it is possible to carry out the operations more efficiently. It is also possible to predict the wave height distribution and calmness in the sea area 72 corresponding to specified wave conditions on a specific day within the next month, for example.

[0124] It is also possible to identify past wave conditions in sea area 71 and predict the wave height distribution and wave calmness in sea area 72 that correspond to the identified wave conditions. For example, by identifying wave conditions in a year with similar weather conditions (water temperature, ocean currents, frequency of typhoons, etc.) to this year and predicting the wave height distribution and wave calmness in sea area 72 that correspond to the identified wave conditions, it is possible to roughly predict the wave calmness for the entire year.

[0125] [Embodiment 3] Next, a prediction device according to another embodiment of the present invention will be described with reference to Fig. 15. Fig. 15 is a block diagram showing an example of the configuration of a prediction device 201 (information processing device) according to a third embodiment. The prediction device 201 in Fig. 15 is a version of the learning and prediction device 101 shown in Fig. 9, with the functions of the learning device 1 described with reference to Fig. 1 removed, and is capable of predicting wave height distribution and performing wave height statistical analysis using a prediction model. In other words, the prediction device 201 in Fig. 15 does not perform learning processing, but instead predicts wave height distribution (wave height distribution prediction processing) and calculates calmness (wave height statistical analysis processing) using model parameters updated through machine learning by another device (e.g., the learning device 1 in Fig. 1 or the learning and prediction device in Fig. 9).

[0126] 15, the prediction device 201 includes a control unit 210, a storage unit 220, a communication unit 30, an input unit 40, and an output unit 50. The prediction device 201 is configured by, for example, a general-purpose computer.

[0127] Functional blocks having the same functions as those described with reference to FIGS. 9 to 14 are designated by the same reference numerals, and detailed description thereof will be omitted.

[0128] The model parameters MP stored in the storage unit 220 are model parameters MP that have already been updated by machine learning. The storage unit 220 may be realized, for example, by copying the contents stored in the storage unit 20. Also, in FIG. 15, the storage unit 220 stores the teacher data TD, but the teacher data TD may not be stored.

[0129] [Software implementation example] The learning device 1, learning and prediction device 101, and prediction device 201 may each be realized by a logic circuit (hardware) formed on an integrated circuit (IC chip) or the like, or by software. That is, each functional block shown in Figures 1, 9, and 15 may be realized by hardware or software.

[0130] In the latter case, each of the learning device 1, the learning and prediction device 101, and the prediction device 201 includes a computer that executes instructions of a program, which is software that realizes each function. This computer includes, for example, one or more processors and a computer-readable recording medium that stores the program.

[0131] The object of the present invention is achieved by the computer having the processor read and execute the program from the recording medium. The processor may be, for example, a CPU (Central Processing Unit). The recording medium may be a "non-transitory tangible medium," such as a ROM (Read Only Memory), tape, disk, card, semiconductor memory, or programmable logic circuit. The computer may further include a RAM (Random Access Memory) for storing the program. The program may be supplied to the computer via any transmission medium capable of transmitting the program (such as a communication network or broadcast waves). The present invention may also be realized in the form of a data signal embedded in a carrier wave, in which the program is embodied by electronic transmission.

[0132] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0133] 1 Learning device 10,110,210 Control unit 11 Acquisition Department 12 Teacher data generation unit 13 Learning Department 20,220 storage unit 30 Communications Department 40 Input section 50 Output section 101 Learning and Prediction Device 114 Second Acquisition Department 115 Prediction Department 116 Information output section 201 Prediction Device

Claims

1. A first step of acquiring information about waves in a first sea area; a second step of outputting information about the calmness of a second sea area using a prediction model previously constructed by machine learning based on information about waves in the first sea area; The information processing device executes The information about waves acquired in the first step is wave conditions including at least wave height, period, and wave direction; In the second step, information on calmness including a wave height contour diagram, which is wave height distribution information for the second sea area predicted by the prediction model, is output. Calmness assessment method.

2. In the second step, the wave height contour map is further divided into grid intervals of 5 m to 100 m, and converted into wave height data for each grid obtained from the wave height contour map. A method for evaluating quietness according to claim 1.

3. In the first step, a plurality of wave conditions are acquired, and in the second step, the occurrence ratio of waves having a predetermined limit wave height or less in the second sea area is calculated based on wave height data corresponding to a plurality of pieces of wave height distribution information predicted in accordance with the plurality of wave conditions, and the calculated ratio is output as information regarding the calmness. The method for evaluating quietness according to claim 2.

4. The wave conditions further include at least one of wind speed, wind direction, and tidal current. The method for evaluating tranquility according to any one of claims 1 to 3.

5. In the first step, information about topographical features of the second sea area is further input as an explanatory variable, In the second step, information about the calmness of the second sea area is output by a prediction model previously constructed by machine learning based on information about waves in the first sea area and information about topography and structures in the second sea area. The method for evaluating tranquility according to any one of claims 1 to 4.

6. a third step of acquiring teacher data including a plurality of pairs of information on waves in the first sea area and information on calmness in the second sea area obtained by performing wave transformation analysis on the information on waves; a fourth step of learning a prediction model with information about waves in the first sea area as input data and information about calmness in the second sea area as output data, by referring to the training data; The method for evaluating tranquility according to claim 1 , further comprising:

7. The information about waves acquired as the training data is wave conditions including at least wave height, period, and wave direction, The information about calmness acquired as the training data includes a wave height contour diagram, which is a wave height distribution in the second sea area obtained by performing the wave transformation analysis. The method for evaluating quietness according to claim 6.

8. The second sea area is divided into a plurality of sub-sea areas, a fifth step of acquiring analytical wave height values and measured wave height values for each of the divided sea areas based on an analytical wave height contour map, which is a wave height contour map for the second sea area obtained by performing a wave transformation analysis on predetermined wave conditions for the first sea area, and a measured wave height contour map, which is a wave height contour map for the second sea area actually measured under the same wave conditions as the predetermined wave conditions; and a sixth step of calculating a correction value for correcting the analytical wave height value to the measured wave height value for each of the divided sea areas, In the fourth step, the information on the calmness including the corrected wave height contour diagram, which is a wave height contour diagram corrected by the correction value, is re-learned as the training data. The method for evaluating quietness according to claim 7.

9. In the third step, information about topographical features of the second sea area is further acquired as the training data, In the fourth step, a prediction model is trained with reference to the training data, the prediction model having information on waves and information on topographical structures in the second sea area as input data and information on calmness in the second sea area as output data. The method for evaluating tranquility according to any one of claims 6 to 8.

10. acquiring training data including a plurality of pairs of information on waves in a first sea area and information on calmness in a second sea area; a step of learning a prediction model with information about waves in a first sea area as input data and information about calmness in a second sea area as output data, by referring to the training data; The information processing device executes The information about waves acquired as the training data is wave conditions including at least wave height, period, and wave direction, The information on calmness acquired as the training data includes a wave height contour diagram, which is a wave height distribution in the second sea area obtained by performing a wave transformation analysis on the wave conditions. Predictive model generation method.

11. an information acquisition unit that acquires information about waves in a first sea area; an information output unit that outputs information about the calmness of the second sea area using a prediction model that has been constructed in advance by machine learning based on information about waves in the first sea area; Equipped with The information acquisition unit acquires wave conditions including at least wave height, period, and wave direction as information about the waves, The information output unit outputs, as the information about the calmness, information about the calmness including a wave height contour diagram which is wave height distribution information in the second sea area predicted by the prediction model. A quietness evaluation device characterized by:

12. A program for causing a computer to function as the quietness evaluation device according to claim 11, the program causing a computer to function as the information acquisition unit and the information output unit.

Images