Device for estimating exploration well candidates for geothermal power generation, method for estimating exploration well candidates for geothermal power generation, and program

Abstract

To find out a new geothermal development area by correctly estimating the prospects of a pilot well candidate area including an area with a small amount of surface survey data.SOLUTION: An exploration well candidate position estimation system 1 for geothermal power generation using machine learning includes a model generation device 2 that generates a learned model for estimating an objective variable from an explanatory variable using teacher data including the explanatory variable including an ore deposit distribution, a quarternary volcanic distribution, and a hypocenter distribution, and the objective variable being an exploration promising place and an exploration unpromising place, and an estimation device 3 that estimates an exploration candidate position for geothermal power generation using the learned model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an apparatus for estimating candidate sites for exploratory wells for geothermal power generation, a method for estimating candidate sites for exploratory wells for geothermal power generation, and a program therefor. [Background technology]

[0002] In order to promote measures against global warming and ensure energy security, it is important to further strengthen the use of renewable energy. Geothermal power generation in particular is capable of generating electricity more stably than solar or wind power, and is expected to have a high potential as a base power source.

[0003] Conventionally, when selecting potential drilling sites for geothermal power generation, regions estimated to have large amounts of hydrothermal resources using volumetric methods or the like have been selected as promising sites. For example, a tool has been disclosed for selecting such drilling sites that displays basic information on the potential and characteristics of geothermal resources (particularly hydrothermal resources) across Japan on a nationwide scale (see, for example, Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] "Report on the FY2013 Detailed Survey and Analysis of Geothermal Power Generation Introduction Potential," June 2014, EX Urban Research Institute Ltd., National Institute of Advanced Industrial Science and Technology, Asia Air Survey Co., Ltd. Summary of the Invention [Problem to be solved by the invention]

[0005] In the conventional method of selecting potential drilling sites, it is necessary to estimate promising areas from temperature logging data of a small number of wells and hot springs scattered widely across the country. This results in areas with low estimation accuracy, which is an issue. Specifically, in the current geothermal survey, 2Surface surveys are conducted within a certain survey area using geological, geophysical, and geochemical methods, and the results are comprehensively interpreted before drilling sites are selected. However, such detailed data is only available in a limited number of areas, making it difficult to perform a uniform nationwide assessment. While several reports on nationwide assessment studies, including Non-Patent Document 1, have been published, they all spatially interpolate temperature logging data from a small number of wells and hot springs scattered throughout the country to create an underground temperature structure model and estimate the national geothermal resource volume using the volumetric method. Areas estimated to have abundant geothermal resources using these methods are then considered promising areas for exploration. However, in reality, there are areas that are not properly evaluated due to the lack of well and hot spring data. As a result, potentially promising candidate sites for exploration well drilling are overlooked, and less desirable areas are sometimes evaluated as promising.

[0006] The present invention was made in light of these circumstances, and its purpose is to accurately estimate the promising potential of candidate sites for geothermal power generation exploratory wells, including areas with little surface survey data, and to discover new geothermal development areas. [Means for solving the problem]

[0007] In order to solve the above problem, one embodiment of the present invention is an apparatus that uses machine learning to estimate candidate locations for geothermal power plant exploration wells, and includes: a model generation device that generates a trained model that estimates a target variable from the explanatory variables using training data consisting of explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake source distribution, and a target variable that is a promising area for exploration and an unpromising area for exploration; and an estimation device that uses the trained model to estimate candidate locations for geothermal power plant exploration wells.

[0008] Another aspect of the present invention is a method for using machine learning to estimate candidate sites for geothermal power plant exploration wells, the method including: a model generation step of generating a trained model that estimates a target variable from the explanatory variables using training data including explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake source distribution, and a target variable indicating promising and unpromising areas for exploration; and a step of estimating candidate sites for geothermal power plant exploration wells using the trained model.

[0009] Another aspect of the present invention is a program for estimating potential sites for geothermal power generation exploration wells using machine learning, the program causing a computer to execute a method including: a model generation step of generating a trained model that estimates a target variable from the explanatory variables using training data including explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake source distribution, and a target variable indicating promising and unpromising areas for exploration; and a step of estimating potential sites for geothermal power generation exploration wells using the trained model.

[0010] Any combination of the above components, and any transformation of the present invention into an apparatus, method, system, recording medium, computer program, etc., are also valid aspects of the present invention. [Effects of the Invention]

[0011] According to the present invention, it is possible to correctly estimate the potential of potential exploration well sites, including areas with little surface survey data, and to discover new geothermal development areas. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a functional block diagram of a prospective well site estimation system 1 according to a first embodiment. [Figure 2] 10 is a flowchart showing the flow of processing in a method for estimating a potential exploratory well site according to a second embodiment. [Figure 3] The explanatory variables in Verification 2 were data created using a geographic information system (GIS). (A) shows the distribution of ore deposits, (B) shows the distribution of Quaternary volcanoes, and (C) shows the D90 data created from the distribution of earthquake centers. [Figure 4] This is data created using a geographic information system (GIS) for the preferred explanatory variables in Verification 2. (D) shows the data for topography quantity, and (E) shows the data for topography type. [Figure 5] This is a diagram showing the distribution of promising and non-promising areas for exploratory wells across Japan, using the trained model from Verification 2. DETAILED DESCRIPTION OF THE INVENTION

[0013] The present invention will be described below based on preferred embodiments with reference to the drawings. In the embodiments and modifications, identical or equivalent components and members are designated by the same reference numerals, and redundant descriptions will be omitted where appropriate. The dimensions of the components in the drawings are enlarged or reduced as appropriate for ease of understanding. Some components that are not important for explaining the embodiments are omitted from the drawings. Terms including ordinal numbers such as "first" and "second" are used to describe various components, but these terms are used only to distinguish one component from another and do not limit the components.

[0014] Because geothermal resources exist underground in places that cannot be seen through the naked eye, it is difficult to accurately estimate the location of geothermal resources that are useful for power generation. In conventional geothermal development surveys, in areas where abundant geothermal resources are estimated using the volumetric method, the geothermal resources are evaluated using multiple exploratory wells, and detailed studies are conducted on the feasibility of power generation, etc. In this case, drilling exploratory wells costs several hundred million yen per well. However, especially in areas where no previous surveys have been conducted and there is little data, the probability that an exploratory well will encounter a geothermal resource is said to be less than 25%. The development risks resulting from these inaccurate geothermal resource estimation techniques are one of the barriers to promoting geothermal development.

[0015] In response to this, the inventors have discovered that it is possible to use machine learning to accurately estimate potential locations for exploratory wells, based on a variety of geoscience information collected uniformly across the country, without relying on conventional well and hot spring data.

[0016] [First embodiment] FIG. 1 is a functional block diagram of a prospective well site estimation system 1 according to a first embodiment. The prospective well site estimation system 1 shown in FIG. 1 is an apparatus that estimates prospective well sites for geothermal power generation using machine learning, and includes a model generation device 2 and an estimation device 3. The model generation device 2 and the estimation device 3 can communicate with each other via a communication line 4. Another device (not shown) may be communicatively connected to the communication line 4. Here, the other device may be a server of a geographic information system (GIS) that provides explanatory variables for the trained model of the present invention.

[0017] The communication line 4 may be a communication module configured to be able to communicate with other devices via wire or wirelessly, and may be, for example, 4G ( 4th Generation) or 5G (5 th The communication module may be compatible with a mobile communication standard such as IEEE 802.11 generation. The communication module may be compatible with a wired or wireless communication standard. The communication module is not limited to these and may be compatible with various communication standards.

[0018] In the exploration well candidate site estimation system 1 shown in Fig. 1, the model generating device 2 and the estimation device 3 are separate devices, but the exploration well candidate site estimation system 1 may also be configured such that the model generating device 2 and the estimation device 3 are integrated. In this specification, the explanation will be focused on the system in which the model generating device 2 and the estimation device 3 are separate devices, as shown in Fig. 1.

[0019] The model generating device 2 includes a storage unit 21, an input unit 22, a model generating unit 23, an output unit 24, and a communication unit 25.

[0020] The estimation device 3 includes a storage unit 31, an input unit 32, an estimation unit 33, an output unit 34, and a communication unit 35.

[0021] The model generating device 2 and the estimation device 3 are realized as software by a processor such as a CPU (Central Processing Unit) executing a program stored in a storage device having a non-volatile recording medium (non-transitory recording medium) and a memory. The program may be recorded on a computer-readable recording medium. Examples of computer-readable recording media include portable media such as flexible disks, magneto-optical disks, ROMs (Read Only Memory), and CD-ROMs (Compact Disk Read Only Memory), and non-transitory recording media such as hard disks or solid-state drives (SSDs) built into computer systems.

[0022] The model generating device 2 and the estimation device 3 may be realized using hardware (accelerator) including an electronic circuit (electronic circuit or circuitry) using, for example, an LSI (Large Scale Integrated circuit), an ASIC (Application Specific Integrated circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array).

[0023] First, the model generating device 2 will be described. The storage unit 21 stores in advance training data used in machine learning and a machine learning computer program. The training data may be input to the model generation device 2 by the input unit 22, or may be input to the model generation device 2 via the communication line 4.

[0024] Examples of the input unit 22 include devices such as a mouse, a touch panel, and a keyboard. The input unit 22 may have a plurality of types of devices.

[0025] The output unit 24 is, for example, a display device, such as a liquid crystal display or an organic EL display, and is used to visually confirm input data and the degree of generation of a trained model.

[0026] The model generation unit 23 generates a trained model that estimates the dependent variable from the explanatory variables using training data consisting of the explanatory variables and the dependent variable.

[0027] The estimation unit 33 estimates potential exploration sites for geothermal power generation using the trained model generated by the model generation unit 23.

[0028] The explanatory variables of the training data include distribution of mineral deposits, distribution of Quaternary volcanoes, and distribution of earthquake centers.

[0029] First, let us explain the distribution of mineral deposits. When high-temperature hydrothermal fluids rise to the surface due to their high vapor pressure, they react with the rocks in the path, reducing the solubility of minerals, causing them to crystallize and form mineral deposits. Therefore, mineral deposits suggest the presence of the high-temperature hydrothermal fluids necessary for geothermal power generation underground, and are thought to have a strong correlation with promising geothermal power exploration areas. Specifically, the distribution of mineral deposits can be calculated based on the distance from the target prospective exploration area to the nearest mineral deposit. Thus, the inventors have newly discovered that mineral deposit distribution data, which is completely different from the previously used well and hot spring data, can be used to estimate potential geothermal power exploration well locations.

[0030] Next, we will explain the distribution of Quaternary volcanoes. Quaternary volcanoes are those active between approximately 2.6 million years ago and the present. The more recent or active a volcano is, the more likely it is that a magma chamber, which contributes to the formation of a geothermal system as a heat source, is located shallow underground. Therefore, the distribution of Quaternary volcanoes is thought to correlate strongly with promising geothermal exploration sites. Furthermore, areas surrounding volcanoes with too recent activity may be unsuitable for geothermal power generation due to insufficient hydrothermal system development, making this classification important. Specifically, the distribution of Quaternary volcanoes can be determined based on the following volcanic classifications: the Quaternary volcanoes closest to the target prospecting site are a volcano group; the Quaternary volcanoes closest to the target prospecting site are a caldera; or the Quaternary volcanoes closest to the target prospecting site are neither a volcano group nor a caldera. In this way, the inventors have newly discovered that Quaternary volcano distribution data, which is completely different from the well and hot spring data that has been used conventionally, can be used to estimate candidate locations for exploratory wells for geothermal power generation.

[0031] Next, we will explain the hypocenter distribution. The hypocenter distribution may be calculated based on hypocenter distribution data, such as microearthquakes detected by highly sensitive vibration sensors installed underground. This hypocenter distribution data can provide information on fracture systems that regulate underground hot water flow and are likely to form geothermal reservoirs.

[0032] Alternatively, the hypocenter distribution may be constructed based on the lowest depth limit of the seismogenic layer, D90 (defined as the depth at which 90% of the total earthquakes occurring in the shallow crust occur). Most natural earthquakes in the shallow crust occur shallower than the brittle-ductile transition (BDT). Based on past measurements, the BDT temperature in granitic rocks is known to be approximately 380°C, so D90 indicates the subsurface thermal structure. The shallower the D90, the higher the expected temperature at shallower depths, and therefore it is thought to correlate with potential exploration well locations.

[0033] In this way, the inventors have newly discovered that earthquake source distribution data, which is completely different from the well and hot spring data that has been used conventionally, can be used to estimate candidate locations for exploratory wells for geothermal power generation.

[0034] It is preferable to standardize explanatory variables consisting of continuous values ​​before using them as explanatory variables. When using multiple explanatory variables whose values ​​differ greatly due to differences in units, etc., standardizing each variable before using them as explanatory variables can reduce the excessive influence of the differences in values. One method of standardization is to calculate the mean and standard deviation of each variable and scale them so that the mean is 0 (zero) and the standard deviation is 1.

[0035] The objective variable of the training data is a binary variable of promising areas for exploration and non-promising areas for exploration.

[0036] As a specific algorithm for generating a trained model that estimates a target variable from explanatory variables, any suitable algorithm may be used, such as a random forest, a decision tree, SVC (Support Vector Classification), XGBoost, etc. According to the study by the present inventor, in this embodiment, XGBoost achieved the highest accuracy rate.

[0037] As described above, potential locations for geothermal power plant exploration wells can be estimated by using the distribution of mineral deposits, Quaternary volcanoes, and earthquake source distributions. Note that this is not disclosed in the prior art, nor is it common general knowledge that there is a correlation between the two.

[0038] The explanatory variables of the training data may further include topographical quantities, which may be calculated based on valley density, drainage basin area, slope, etc., which are highly correlated with potential well sites.

[0039] By including topographical quantities in the explanatory variables, it is possible to estimate the potential locations of exploratory wells with even greater accuracy.

[0040] The explanatory variables of the training data may further include terrain types, such as volcanic areas, plains, and ocean areas, which are highly correlated with potential exploration well sites.

[0041] By including the terrain type in the explanatory variables, it is possible to estimate the potential locations of exploratory wells with even greater accuracy.

[0042] The target variable, promising exploration areas, may be production wells of existing geothermal power plants or wells or deep drilling hot spring data locations distributed within a certain distance therefrom, and non-promising exploration areas may be wells or deep drilling hot spring data locations other than promising exploration areas.

[0043] By identifying promising and unpromising areas for exploration in this manner, it is possible to estimate candidate sites for exploration wells with even greater accuracy.

[0044] The trained model generated by the model generation unit 23 is sent to the estimation device 3 via the communication line 4, for example, and stored in the storage unit 31 of the estimation device 3.

[0045] Next, the estimation device 3 will be described. The hardware configurations of the memory unit 31, input unit 32, output unit 34, and communication unit 35 are the same as those of the memory unit 21, input unit 22, output unit 24, and communication unit 25 of the above-mentioned model generation device 2. By inputting data for a location where it is desired to estimate whether the location is a promising area for prospecting or not, using the input data and the trained model, the estimation unit 33 estimates whether the location is a promising area for prospecting or not.

[0046] [Second embodiment] 2 is a flowchart showing the process flow of a method for estimating potential exploration well sites according to the second embodiment. This method is a method for estimating potential exploration well sites for geothermal power generation using machine learning, and includes a model generation step S1 (model generation device 2) and an estimation step S2 (estimation device 3).

[0047] In this method, in a model generation step S1, a trained model is generated that estimates the objective variables from the explanatory variables using training data consisting of explanatory variables including the distribution of ore deposits, the distribution of Quaternary volcanoes, and the distribution of earthquake epicenters, and objective variables representing areas with good prospecting potential and areas without good prospecting potential. The distribution of ore deposits, the distribution of Quaternary volcanoes, and the distribution of earthquake epicenters are as described in the first embodiment.

[0048] In this method, in the estimation step S2, candidate drilling sites for geothermal power generation are estimated using the trained model generated in the model generation step S1.

[0049] According to this embodiment, it is possible to correctly estimate the potential of prospective exploration well sites, including areas with little surface survey data, and to find new geothermal development areas.

[0050] The explanatory variables in this method may further include topographical quantities. By including topographical quantities in the explanatory variables, it is possible to estimate the potential locations of the exploratory wells with even higher accuracy.

[0051] The explanatory variables in this method may further include terrain type. By including terrain type in the explanatory variables, it is possible to estimate the potential locations of exploratory wells with even higher accuracy.

[0052] The terrain quantities and terrain types suitable as explanatory variables are also as explained in the first embodiment.

[0053] In this method, promising areas for exploration may be production wells of existing geothermal power plants or wells within a certain distance from them, or locations of data on deep drilling hot springs, while non-promising areas for exploration may be locations of wells or deep drilling hot springs other than promising areas for exploration. By identifying promising and non-promising areas for exploration in this way, it is possible to estimate candidate locations for exploration wells with even greater accuracy.

[0054] [Third embodiment] The third embodiment is a program for estimating potential locations for exploratory wells for geothermal power generation using machine learning, and causes a computer to execute a method including a model generation step S1 and an estimation step S2.

[0055] In the model generation step S1, this program generates a trained model that estimates the objective variable from the explanatory variables using training data consisting of explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake center distribution, and objective variables including promising and unpromising areas for exploration.

[0056] In estimation step S2, this program estimates potential exploration sites for geothermal power generation using the trained model generated in model generation step S1.

[0057] According to this embodiment, a method for correctly estimating the potential of potential exploration well sites, including areas with little surface survey data, and finding new geothermal development areas can be implemented in software as a computer program.

[0058] The explanatory variables in this program may further include topographical quantities. By including topographical quantities in the explanatory variables, it is possible to estimate the potential locations of exploratory wells with even greater accuracy.

[0059] The explanatory variables in this program may further include terrain type. By including terrain type in the explanatory variables, it is possible to estimate the potential locations of exploratory wells with even greater accuracy.

[0060] Each of the explanatory variables in the model generation step S1 of this program is the same as that described in the second embodiment and the first embodiment.

[0061] In this program, promising areas for exploration may be production wells of existing geothermal power plants or wells located within a certain distance from them, or locations of data on deep drilling hot springs, while non-promising areas for exploration may be locations of wells other than promising areas for exploration or locations of data on deep drilling hot springs. By identifying promising and non-promising areas for exploration in this way, it is possible to estimate candidate locations for exploration wells with even greater accuracy.

[0062] [Verification 1] The inventors conducted experiments to verify the effectiveness of the present invention. The explanatory variables were ore deposit distribution, Quaternary volcano distribution, earthquake source distribution, topographical volume, and topographical type. The objective variables were existing geothermal power plant production wells and their surrounding areas (within a 3 km radius) as promising areas for exploration, and other areas as non-prospective areas for exploration. XGBoost was used as the trained model generation algorithm, and cross-validation was performed by dividing 70% of all samples for training and 30% for testing.

[0063] As a result of the verification, Successful prediction results and correct data: 81 Prediction results are successful, but the correct data is unsuccessful: 11 Prediction results failed, correct answer data successful: 8 Prediction results are incorrect and the correct data is incorrect: 168 As a result, when the areas with promising prospecting are considered positive and the areas without promising prospecting are considered negative, Accuracy rate (percentage of correct predictions out of all predictions): 0.9291 Precision (the percentage of predicted positive results that were actually correct): 0.8804 Recall (the percentage of predicted positive results among the actual positive results): 0.9101 F-measure (harmonic mean of precision and recall): 0.8950 The above results were obtained, demonstrating that the present invention can estimate promising areas for exploration with high accuracy.

[0064] [Verification 2] Figure 3 shows data created using GIS regarding the distribution of mineral deposits, Quaternary volcanoes, and epicenters throughout Japan. Figure 3(A) shows the locations of mineral deposits within Japan. Figure 3(B) shows the locations of Quaternary volcanoes within Japan. Figure 3(C) shows D90 data created from the epicenter distribution. D90 was created from data on earthquakes that have occurred in the shallow crust (less than 25 km shallow) since 2003, and is used to calculate the distribution of earthquakes within the shallow crust (less than 25 km shallow) within the 100 km range. 2 This indicates the depth within the range where 90% of the total earthquakes occur, counting from the shallowest part.

[0065] In Verification 2, a trained model was generated using the preferred explanatory variables of terrain quantity and terrain type. Figure 4 shows the terrain quantity and terrain type data created using GIS. Figure 4(D) shows the slope angle (slope amount) of the earth's surface as data indicating terrain quantity. Figure 4(E) shows data representing terrain type, dividing Japan into volcanic areas, coastal areas, and others.

[0066] A trained model using the XGBoost algorithm was generated from the data in Figures 3(A) to (C) and Figures 4(D) to (E). Figure 5 shows the results of using this trained model to distinguish between areas across Japan that are promising for exploration and other areas that are not.

[0067] The present invention has been described above based on the embodiments. These embodiments are merely examples, and it will be understood by those skilled in the art that various modifications are possible in the combination of each component and each treatment process, and that such modifications are also within the scope of the present invention.

[0068] When understanding the abstract technical ideas of the embodiments and modifications, the technical ideas should not be interpreted as being limited to the contents of the embodiments and modifications. The above-described embodiments and modifications are merely illustrative examples, and many design modifications, such as changes, additions, and deletions of components, are possible. In the embodiments, the contents in which such design modifications are possible are emphasized by adding the notation "embodiment." However, design modifications are also permitted even in contents without such notation. [Industrial Applicability]

[0069] The present invention can be widely used in the design and development of geothermal power generation systems, the construction of digital geothermal databases, and consulting for geothermal developers. [Explanation of symbols]

[0070] 1. Exploration well candidate site estimation system 2. Model generator, 3...estimation device, 4. Communication lines, 21...Storage section, 22··Input section, 23··Model generation part, 24··Output section, 25··Communications Department, 31...Storage section, 32··input section, 33... Estimation Department, 34··Output section, 35··Communications Department, S1··Model generation step, S2··Estimation step.

Claims

1. An apparatus for estimating potential locations for geothermal power plant exploration wells using machine learning, a model generation device that generates a trained model using training data consisting of explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake source distribution, and objective variables including promising areas for exploration and non-prospective areas for exploration, to estimate the objective variables from the explanatory variables; an estimation device that estimates candidate drilling sites for geothermal power generation using the trained model; An apparatus comprising:

2. the model generation device includes a storage unit, an input unit, a model generation unit, an output unit, and a communication unit; the estimation device includes a storage unit, an input unit, an estimation unit, an output unit, and a communication unit; a communication unit of the model generation device and a communication unit of the estimation device are capable of communicating with each other via a communication line; the storage unit of the model generation device stores in advance teacher data to be used in machine learning and a computer program for machine learning; the model generation unit generates a trained model that estimates the objective variables from the explanatory variables using training data consisting of explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake source distribution, and objective variables that are promising areas for exploration and areas not promising for exploration; The device according to claim 1 , wherein the estimation unit estimates potential exploration sites for geothermal power generation using the trained model.

3. 3. The apparatus according to claim 1, wherein the explanatory variables further include topographical quantities.

4. 3. The apparatus according to claim 1, wherein the explanatory variables further include a terrain type.

5. The prospective exploration area is a production well of an existing geothermal power plant or a location of wells or deep drilling hot spring data distributed within a certain distance therefrom, 3. The apparatus according to claim 1, wherein the non-prospective exploration areas are data locations of wells or deep drilling hot springs other than the prospective exploration areas.

6. A method for estimating potential locations for geothermal power plant exploration wells using machine learning, comprising: a model generation step of generating a trained model using training data consisting of explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake source distribution, and objective variables including prospective exploration areas and non-prospective exploration areas, to estimate the objective variables from the explanatory variables; a step of estimating candidate sites for geothermal power generation exploration using the trained model; A method comprising:

7. The method of claim 6 , wherein the explanatory variables further include topographical quantities.

8. The method of claim 6 , wherein the explanatory variables further include terrain type.

9. The prospective exploration area is a production well of an existing geothermal power plant or a location of wells or deep drilling hot spring data distributed within a certain distance therefrom, The method according to any one of claims 6 to 8, wherein the non-prospective exploration areas are wells or deep drilling hot spring data locations other than the prospective exploration areas.

10. A program that uses machine learning to estimate candidate locations for geothermal power plant exploration wells, a model generation step of generating a trained model using training data consisting of explanatory variables including ore deposit distribution, Quaternary volcano distribution, and earthquake source distribution, and objective variables including prospective exploration areas and non-prospective exploration areas, to estimate the objective variables from the explanatory variables; a step of estimating candidate sites for geothermal power generation exploration using the trained model; A program that causes a computer to execute a method including the steps of:

11. 11. The program according to claim 10, wherein the explanatory variables further include topographical quantities.

12. 11. The program according to claim 10, wherein the explanatory variables further include a terrain type.

13. The prospective exploration area is a production well of an existing geothermal power plant or a location of wells or deep drilling hot spring data distributed within a certain distance therefrom, 13. The program according to claim 10, wherein the non-prospective exploration areas are locations of wells or deep drilling hot spring data other than the prospective exploration areas.

Images