Creation method, creation system, and creation program

Abstract

To optimize transportation plans when transporting materials from multiple sources to individual storage units for multiple users. [Solution] In this disclosure, a transportation plan is created in which the amount of goods used by each user in the storage compartments owned by each of the multiple users is predicted using a machine learning model, and the goods are transported from multiple sources to each storage compartment using predetermined means of transport. In the transportation plan, it is determined when, from which source, and using which means of transport the goods to each storage compartment.

Description

Technical Field

[0001] The present disclosure relates to a creation method, a creation system, and a creation program.

Background Art

[0002] Patent Document 1 discloses a delivery plan control device that plans the delivery by the vehicle in consideration of the business requirements necessary for delivering a load from a base serving as a delivery base to a delivery destination.

[0003] The delivery plan control device of Patent Document 1 includes a reception unit that receives an order for delivery of a load, a creation unit that creates a delivery pattern to a delivery destination based on the order content received by the reception unit and a first business requirement selected from the business requirements, and a vehicle allocation unit that allocates the vehicle to the delivery pattern created by the creation unit and performs vehicle allocation based on a second business requirement selected from the business requirements.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Here, when transporting a consumable (e.g., liquefied natural gas) from a plurality of sources (e.g., liquefied natural gas bases) to each storage unit (e.g., tank) of a plurality of users (e.g., customers), it is conceivable for a seller to create a transportation plan. In Patent Document 1, the seller receives an order from a user and creates a transportation plan based on the order content. Therefore, when the assumed usage amount at the time of the user's order is significantly different from the actual usage amount thereafter and there is a possibility of exceeding the upper and lower limits of the tank, it is necessary to change the transportation plan, which increases the business load of the seller.

[0006] This disclosure aims to optimize transportation plans when transporting goods from multiple sources to the respective storage units of multiple users. [Means for solving the problem]

[0007] The first embodiment involves creating a transportation plan that transports items stored in storage compartments owned by multiple users to each storage compartment from multiple sources using predetermined means of transport, based on predicted values ​​obtained using a machine learning model, and determining when, from which source, and using which means of transport to each storage compartment.

[0008] According to the first embodiment, a transportation plan is created to transport items stored in storage compartments owned by multiple users from multiple sources to each storage compartment using predetermined means of transport, based on predicted values ​​obtained using a machine learning model. In the transportation plan, it is determined when, from which source, and by which means of transport items will be transported to each storage compartment. In this way, since the transportation plan is created based on predicted values ​​obtained using a machine learning model, the transportation plan can be optimized.

[0009] In the second embodiment, as in the first embodiment, the amount of usage by each user for each unit period is predicted using a machine learning model, and the machine learning model is performed using actual data showing the actual amount of usage by the user for each unit period and schedule data showing the schedule for the use of the product.

[0010] Therefore, it is possible to improve the accuracy of users' predictions regarding the amount of material they use. As a result, transportation plans can be optimized. In other words, transportation plans can be created that match the actual usage of the customer.

[0011] In the third embodiment, the transportation plan is created such that the objective function value obtained by weighting the following elements 1 to 3 is minimized in the first embodiment. Element 1: The amount of material to be stored in each storage section is within the upper and lower limits, taking into account the prediction error. Element 2: The transport distance from the source to each storage unit is minimized. Element 3: The maximum number of operational modes of transport is minimized.

[0012] According to the third embodiment, the transportation plan is created such that the objective function value obtained by weighting the elements 1 to 3 described above is minimized, thereby optimizing the transportation plan.

[0013] In the fourth embodiment, based on the transportation plan described in the first embodiment, a transportation date is proposed to the user.

[0014] According to the fourth aspect, since the transportation date is proposed based on the optimized transportation plan, the user can be offered an optimized transportation date.

[0015] The fifth embodiment includes a processor, which creates a transportation plan for transporting items stored in storage units owned by multiple users from multiple sources to each storage unit using predetermined means of transport, based on predicted values ​​obtained using a machine learning model, and in the transportation plan, it determines when, from which source, and using which means of transport to transport items to each storage unit.

[0016] According to the fifth embodiment, the same effects and advantages as those of the first embodiment are achieved.

[0017] The sixth aspect involves providing a computer with a machine learning model to predict the amount of usage by each user of the items stored in the storage compartments owned by each of the multiple users, based on the predicted values ​​obtained. This is a program for creating a transportation plan that involves transporting goods from multiple sources to each storage unit using predetermined means of transport, and for executing a process in the transportation plan to determine when, from which source, and by which means of transport goods should be transported to each storage unit.

[0018] According to the sixth aspect, the same effects and advantages as those of the first aspect are achieved.

Advantages of the Invention

[0019] Since the present disclosure has the above configuration, when transporting articles from a plurality of origins to the respective storage parts of a plurality of users, the transportation plan can be optimized.

Brief Description of the Drawings

[0020] [Figure 1] It is a schematic diagram showing an example of the transportation system according to the present embodiment. [Figure 2] It is a block diagram showing an example of the customer terminal, the seller terminal, and the transporter terminal according to the present embodiment. [Figure 3] It is a block diagram showing an example of the management system according to the present embodiment. [Figure 4] It is a block diagram showing an example of the functional configuration of the processor in the management system according to the present embodiment. [Figure 5] It is a conceptual diagram of the accuracy index according to the present embodiment. [Figure 6] It is a graph showing the prediction and actual results of the LNG usage amount for a specific customer. [Figure 7] It is a graph showing the evaluation results regarding the optimization of the transportation plan in the conventional aspect. [Figure 8] It is a graph showing the evaluation results regarding the optimization of the transportation plan in the aspect of the present embodiment.

Modes for Carrying Out the Invention

[0021] Hereinafter, an example of an embodiment according to the present invention will be described based on the drawings.

[0022] <Transportation System 10> The transportation system 10 according to the present embodiment will be described. FIG. 1 is a schematic diagram showing an example of the transportation system 10 according to the present embodiment.

[0023] The transportation system 10 shown in Figure 1 is a system for transporting liquefied natural gas 24 (hereinafter referred to as LNG 24). In the transportation system 10, the seller 14 predicts the amount of LNG 24 that the customer 12 will use, and based on the predicted amount of use, the seller 14 creates a transportation plan and proposes the amount of LNG 24 to be transported and the transport dates from the seller 14 to the customer 12.

[0024] In this embodiment, there are multiple customers 12, and the seller 14 proposes to each of the customers 12 the amount and date of LNG 24 to be transported, tailored to each customer 12. When a customer 12 accepts this proposal, the seller 14 receives an order from the customer 12 for the transport of LNG 24, and the transporter 15 transports the LNG 24 from the seller 14 to the customer 12.

[0025] Specifically, the transportation system 10 includes a management system 11, a tank 22, user equipment 25, a customer terminal 30, a seller terminal 40, a transporter terminal 50, an LNG terminal 49, and a transport vehicle 59, as shown in Figure 1. The following describes each part of the transportation system 10.

[0026] Customer 12 (e.g., a manufacturing company) is an entity that uses LNG24 and is an example of a user. Seller 14 (e.g., a sales company) is an entity that sells LNG24 to Customer 12. Transporter 15 (e.g., a transport company) is an entity that transports LNG24 from Seller 14 to Customer 12.

[0027] LNG is an example of a material used and is a transported item. However, the transported items in the transport system 10 are not limited to LNG 24; for example, they could include petroleum products such as gasoline and kerosene, pharmaceuticals, food products, and various other items (including fluids and solids).

[0028] <Tank 22 and Equipment 25> Tank 22 is an example of a storage unit. Tank 22 is a facility owned by each of the plurality of customers 12 and is a storage facility for storing LNG 24. A detection sensor 23 for detecting the storage capacity is installed in Tank 22. The detection sensor 23 is constituted by, for example, a sensor that detects the level (liquid level value) of the liquid surface in Tank 22. This detection information is transmitted to the management system 11. Tank 22 has an allowable range for allowing the storage of LNG 24, and the upper and lower limits of the storage capacity are set. LNG 24 is stored within this range of the upper and lower limits.

[0029] The usage facility 25 is a facility that uses LNG 24 in each of the plurality of customers 12. As the usage facility 25, for example, a facility that generates fuel gas from the LNG 24 supplied from Tank 22 and a facility that uses that fuel gas (for example, a manufacturing factory that manufactures products using the fuel gas) are provided.

[0030] <LNG Base 49 and Transport Vehicle 59> LNG Base 49 is an example of a shipping origin. LNG Base 49 is a facility owned by the seller 14 and is equipped with a storage facility 48 (LNG tank) for storing the LNG 24 sold to the customer 12. The seller 14 has, for example, a plurality of LNG Bases 49.

[0031] Transport Vehicle 59 is an example of a means of transportation. Transport Vehicle 59 is a means of transportation owned by the transporter 15, and for example, a truck is used. Transport Vehicle 59 transports LNG 24 from the storage facility 48 of LNG Base 49 to the tank 22 of the customer 12. The transport volume of Transport Vehicle 59 is, for example, predetermined.

[0032] <Customer Terminal 30, Seller Terminal 40, and Transporter Terminal 50> Customer Terminal 30 is a terminal used by each of the plurality of customers 12. Seller Terminal 40 is a terminal used by the seller 14. Transporter Terminal 50 is a terminal used by the transporter 15. Customer Terminal 30, Seller Terminal 40, and Transporter Terminal 50 are similarly configured, and the same reference numerals are given to the common components.

[0033] The customer terminal 30, the seller terminal 40, and the transporter terminal 50 are equipped with computer functions and, as shown in Figure 2, include a processor 31, memory 32, storage 33, input unit 34, communication unit 35, and display unit 36.

[0034] For example, a general-purpose processor such as a CPU (Central Processing Unit) can be used as the processor 31. Alternatively, the processor 31 may be a dedicated processor composed of circuits specifically designed to perform a particular task. Furthermore, the processor 31 is not limited to a single processor, but may consist of multiple processors physically separated from each other.

[0035] Storage 33 stores various programs and various data. Specifically, storage 33 is implemented by recording devices such as HDDs (Hard Disk Drives), SSDs (Solid State Drives), and flash memory.

[0036] Memory 32 is a workspace for the processor 31 to execute various programs, and it temporarily stores various programs or data when the processor 31 is executing processing. The processor 31 reads various programs from storage 33 into memory 32 and executes the programs using memory 32 as a workspace.

[0037] The input unit 34 is a component into which various information and instructions are input by the user. Specifically, the input unit 34 is composed of, for example, a pointing device such as a mouse and input keys such as a keyboard.

[0038] The input unit 34 is not limited to a pointing device and input keys; it may also consist of a touch panel or the like, as long as it can accept various types of information and instructions.

[0039] The communication unit 35 is a connection unit (communication interface) for communicating with other devices (for example, the management system 11). Specifically, the communication unit 35 communicates with other devices through a communication line using at least one of wired and wireless connections.

[0040] The display unit 36 ​​notifies the user of the information to be presented by displaying the information to be presented to the user. This display unit 36 ​​is composed of, for example, a liquid crystal display and an organic EL (Electro Luminescence) display.

[0041] Furthermore, the aforementioned user is Customer 12 on Customer Terminal 30, Seller 14 on Seller Terminal 40, and Transporter 15 on Transporter Terminal 50.

[0042] In the customer terminal 30, the information that customer 12 inputs through the input unit 34 includes customer information about customer 12. Examples of customer information include schedule data showing whether or not the equipment 25 will be operated for each unit period. Examples of instructions that customer 12 inputs through the input unit 34 in the customer terminal 30 include an instruction to accept a proposal from seller 14. This instruction corresponds to an instruction to order the transport of LNG 24.

[0043] In the seller terminal 40, an instruction that the seller 14 inputs through the input unit 34 could be, for example, an instruction to send a transport request to the transporter 15 based on the transport plan. In the transporter terminal 50, an instruction that the transporter 15 inputs through the input unit 34 could be, for example, an instruction to accept the transport plan from the seller 14.

[0044] In the customer terminal 30, the information displayed on the display unit 36 ​​may include information on the LNG 24 transport volume and transport date proposed by the seller 14 to the customer 12. In the transporter terminal 50, the information displayed on the display unit 36 ​​may include information on the LNG 24 transport volume and transport date in the transport plan created by the seller 14. In the seller terminal 40, the information displayed on the display unit 36 ​​may include, for example, information on the transport plan, information on whether the proposal to the customer 12 has been accepted, and information on whether the transport request to the transporter 15 has been accepted.

[0045] <Management System 11> Management system 11 is a system that manages the transportation of LNG24. Management system 11 predicts the amount of LNG24 used by customer 12 and creates a transportation plan based on this prediction. Then, management system 11 executes the transportation of LNG24 based on the transportation plan.

[0046] The management system 11 is managed by the seller 14, who is the managing entity for the transportation of LNG 24. The management system 11 is managed and operated by the seller 14 through the seller terminal 40. Therefore, in this embodiment, the seller 14 predicts the amount of LNG 24 used by the customer 12 using a machine learning model (specifically, learning model 63B), and creates a transportation plan based on the predicted value.

[0047] Specifically, the management system 11 has computer-like functions and includes a processor 61, memory 62, storage 63, and a communication unit 65, as shown in Figure 3.

[0048] For example, a general-purpose processor such as a CPU (Central Processing Unit) can be used as the processor 61. Alternatively, the processor 61 may be a dedicated processor composed of circuits specifically designed to perform a particular task. Furthermore, the processor 61 is not limited to a single processor, but may consist of multiple processors physically separated from each other.

[0049] Storage 63 stores various programs, including the management program 63A and the learning model 63B (see Figure 4), as well as various data. Specifically, storage 63 is implemented by recording devices such as HDDs (Hard Disk Drives), SSDs (Solid State Drives), and flash memory.

[0050] Memory 62 is a workspace for the processor 61 to execute various programs, and it temporarily stores various programs or data when the processor 61 is executing processing. The processor 61 reads various programs, including the management program 63A and the learning model 63B, from the storage 63 into memory 62, and executes the programs using memory 62 as a workspace.

[0051] The communication unit 65 is a connection unit for communicating with other devices (for example, the customer terminal 30, the seller terminal 40, and the transporter terminal 50). Specifically, the communication unit 65 communicates with other devices through a communication line using at least one of wired and wireless communication.

[0052] In the management system 11, the processor 61 implements various functions by executing the management program 63A and the learning model 63B. The functional configuration realized through the cooperation of the processor 61 as a hardware resource and the management program 63A and the learning model 63B as software resources will be described below. Figure 4 is a block diagram showing the functional configuration of the processor 61.

[0053] As shown in Figure 4, in the management system 11, the processor 61 functions as an acquisition unit 61A, a prediction unit 61B, and a creation unit 61C by executing the management program 63A and the learning model 63B.

[0054] The acquisition unit 61A acquires information on the liquid level of the tank 22 detected by the detection sensor 23. The acquisition unit 61A also acquires customer information entered by the customer 12 through the customer terminal 30. The acquisition unit 61A acquires various instructions entered in the customer terminal 30, the seller terminal 40, and the transporter terminal 50, respectively.

[0055] The prediction unit 61B predicts the amount of LNG 24 used by the customer 12 per unit period using the learning model 63B. The unit period is, for example, one day. In this embodiment, the prediction unit 61B predicts the amount of LNG used per unit period (e.g., one day) for a predetermined period (e.g., one month) in the future, for example, two months ahead, for each predetermined period (e.g., one month). The learning model 63B is an example of a machine learning model.

[0056] The learning model 63B performs machine learning using actual usage data from customer 12, schedule data showing the schedule for LNG 24 usage, and other data that affects usage (hereinafter referred to as influence data). In other words, the actual usage data, schedule data, and influence data are used as features in the machine learning.

[0057] Examples of performance data include data showing the actual usage by customer 12 for each unit period. Specifically, examples of performance data include daily performance data, performance data for the same month of the previous year, performance data for the previous month of the same year, performance data for the same season of the previous year, performance data for the same day of the same month of the previous year, or data obtained by transforming these variables. The performance data for the same month of the previous year shows the actual daily usage for the same month of the previous year for the day on which the usage forecast is being made. The performance data for the same month of the same year shows the actual daily usage for the same month of the same year for the day on which the usage forecast is being made. The performance data for the same season of the previous year shows the actual daily usage for the same season of the previous year for the day on which the usage forecast is being made. The performance data for the same day of the same month of the previous year shows the actual usage for the same day of the same month of the previous year for the day on which the usage forecast is being made.

[0058] For example, the actual usage data is calculated based on the daily difference in liquid level values ​​acquired by the acquisition unit 61A. If there is additional transport of LNG 24 to tank 22, the usage is calculated taking into account the amount of transport (additional amount). In the above example, the actual usage data for LNG 24 was calculated based on liquid level values, but this is not the only method. For example, the actual usage data could be the actual amount of LNG 24 used by customer 12 at the equipment 25, and the method for obtaining the actual usage data is not limited to a specific method.

[0059] Examples of schedule data include data showing the year, month, day, day of the week, and public holidays for the days on which usage is to be predicted; operational schedule data showing the schedule for each unit period of whether or not the equipment 25 is operated; and data obtained by transforming these into variables. Specifically, operational schedule data may include data showing the operating days and non-operating days for which the equipment 25 is not operated for the customer 12. This operational schedule data can be included, for example, in the customer information acquired by the acquisition unit 61A.

[0060] The operational schedule data may include data indicating whether the days are before or after non-operating days when the equipment 25 is not in operation. On days before and after non-operating days, the amount of LNG 24 used by the equipment 25 may decrease. Therefore, identifying the days before and after non-operating days can improve the accuracy of usage forecasts.

[0061] Examples of impact data include temperature, weather, economic indicators, or data obtained by transforming these into variables. Note that impact data does not necessarily have to be used as features in machine learning.

[0062] While there are no specific learning models (algorithms) used in machine learning, it is possible to use models such as "LightGBM" and "Prophet."

[0063] A learning model 63B is prepared for each user. In other words, in this embodiment, the learning model 63B for each user is used to predict usage for each user.

[0064] In this embodiment, the learning model 63B is evaluated using a predetermined accuracy index (hereinafter referred to as TGSC) as the evaluation function for the learning model 63B. As the TGSC, an accuracy index calculated based on the acceptable range in the tank 22 for each customer 12 is used.

[0065] Therefore, in this embodiment, the amount of LNG 24 used per customer 12 is predicted using a learning model 63B that has been evaluated using TGSC calculated based on the tolerance range in the tank 22 for each customer 12.

[0066] The evaluation function is a function (index) used to quantify and measure (evaluate) the performance (accuracy) of the learning model 63B. The acceptable range is an example of the upper and lower limits.

[0067] TGSC is an index shown in Figure 5 and can be calculated using the following formulas (A), (B), and (C).

[0068]

number

[0069] Here, in equation (A) above, y1[t]: tank tolerance range, tank_upper[t]: tank upper limit, tank_lower[t]: tank lower limit, and d[t]: gas filling amount per cycle.

[0070]

number

[0071] Here, in equation (B) above, y2[t]: prediction accuracy, t[t]: mean prediction error, A: safety factor, s[t]: standard deviation of error, and l: dispatch date proposal period [days].

[0072]

number

[0073] In equation (C) above, a TGSC value closer to 0 indicates higher prediction accuracy relative to the tank size. A TGSC value below a certain level can be used to evaluate high accuracy.

[0074] In this way, the amount of LNG 24 used by each of the 12 customers per unit period is predicted using the learning model 63B.

[0075] The creation unit 61C creates a transportation plan based on the predicted values ​​predicted by the prediction unit 61B, for transporting LNG from multiple LNG terminals 49 to each of the customer 12's tanks 22 using transport vehicles 59. The transportation plan determines when, from which LNG terminal 49, and using which transport vehicle 59 to deliver to each of the customer 12. In other words, it determines the combination of transportation timing, source, and means of transportation.

[0076] In this embodiment, the creation unit 61C creates a transport plan such that the objective function value obtained by weighting the following elements 1 to 3 is minimized. Element 1: The capacity of LNG 24 will be within the upper and lower limits of each tank 22, taking into account prediction errors. Element 2: The transport distance from LNG terminal 49 to each tank 22 is minimized. Element 3: The maximum number of operational transport vehicles (59) is minimized.

[0077] Specifically, the following objective function and constraint equations can be used.

[0078]

number

[0079] The first term of the objective function shown in equation (1) above represents the transportation cost for transporting LNG from LNG terminal 49 to customer 12 (tank 22). The second term represents the maximum number of operating vehicles (daily) per transporter 15. The second term aims to reduce the number of transport vehicles 59 to be secured by reducing the number of operating vehicles on peak days. The third term onward represents the amount of tank liquid level violation, defined as the amount that violates the upper and lower limits. Tank liquid level violations can be calculated by selecting whether or not to consider prediction errors for the combination of {upper limit, lower limit} × {most recent period, next period}. It is also possible to specify only the most recent period. In the case of {without considering prediction errors}, a penalty value for the liquid level is calculated based on the predicted LNG usage, and in the case of {considering prediction errors}, a penalty value for the liquid level is calculated considering upward and downward deviations from the predicted LNG usage. Upward and downward deviations from the predicted usage are defined as shown in equations (2) and (3) below. Note that while formula (1) above performs the calculation in two stages, it is not limited to this. Formula (1) may also perform the calculation using only the first or second stage.

[0080]

number

[0081] Here, w is the correction factor, A is the safety factor, t_1[t] is the mean prediction error (working days), t_2[t] is the mean prediction error (non-working days), s_1[t] is the error standard deviation (working days), s_2[t] is the error standard deviation (non-working days), N_1 is the proposed dispatch date period (working days), and N_2 is the proposed dispatch date period (non-working days). The objective is to minimize the weighted average of the objective function value, which is calculated based on transportation costs, the maximum number of vehicles in operation, and the amount of tank liquid level violations.

[0082] (constraint expression) • All combinations of customer-garage, customer-vehicle type, and customer-base are transportable. • There is a limit to the number of transports each transport vehicle (59) can make. The number of operating vehicles per day, per base, and per vehicle type must not exceed the upper limit. The following equation (4) satisfies the requirement for daily continuity of the tank liquid level value.

[0083]

number

[0084] In addition, while equation (4) above calculates the amount of LNG24 used from the liquid level in the tank, as mentioned above, this is not the only way to determine the amount of LNG24 used. For example, the amount of LNG24 used can be determined using the actual amount of LNG24 used by customer 12 at the equipment 25, and the method for determining the amount of LNG24 used is not limited to a specific method.

[0085] <Suggested delivery date> The seller 14 confirms the transportation plan created by the management system 11 through the display unit 36 ​​of the seller terminal 40. The seller inputs an instruction to propose a transportation date to the customer 12 through the input unit 34 of the seller terminal 40. As a result, the management system 11 proposes a transportation date to the customer 12 through the display unit 36 ​​of the customer terminal 30. Alternatively, the management system 11 may automatically propose a transportation date to the customer 12 after creating the transportation plan.

[0086] <Effects and Effects According to This Embodiment> According to this embodiment, the amount of LNG 24 used by customer 12 for each unit period is predicted using a learning model 63B. In this embodiment, machine learning is performed on the learning model 63B using actual data showing the actual amount of LNG 24 used by customer 12 and schedule data showing the schedule for LNG 24 use. Therefore, the accuracy of predicting the amount of LNG 24 used by customer 12 can be improved.

[0087] Furthermore, according to this embodiment, since usage is predicted for each of the 12 customers using a learning model 63B for each customer 12, the accuracy of predicting usage for each of the 12 customers can be improved.

[0088] Furthermore, according to this embodiment, since the learning model 63B is evaluated using an accuracy index calculated based on the upper and lower limits of the tank 22 for each customer 12, the learning model 63B can be appropriately evaluated even if the capacity of the tank 22 differs for each customer 12.

[0089] Furthermore, according to this embodiment, the seller 14 predicts the amount of LNG 24 that the customer 12 will use, and based on this prediction, the seller 14 can propose to the customer 12 the amount of LNG 24 to sell.

[0090] According to this embodiment, the management system 11 creates a transportation plan in which the amount of LNG 24 used by each of the customers 12 is predicted using a learning model 63B, and transports it from multiple LNG terminals 49 to each of the customers' tanks 22 using transport vehicles 59. In this way, since the transportation plan is created based on predicted values ​​predicted using the learning model 63B, the transportation plan can be optimized. That is, a transportation plan that matches the actual usage of the customers can be created. Note that "optimization of the transportation plan" includes being able to create a transportation plan that matches the actual usage of the customers and leveling out the number of transport vehicles in operation.

[0091] As described above, in this embodiment, the learning model 63B performs machine learning using actual data, schedule data, and impact data. Therefore, the accuracy of predicting the amount of LNG used by LNG24 users can be improved, and as a result, the transportation plan can be optimized.

[0092] Furthermore, according to this embodiment, the transportation plan is created such that the objective function value obtained by weighting the aforementioned elements 1 to 3 is minimized, thus optimizing the transportation plan.

[0093] Furthermore, according to this embodiment, since the transportation date is proposed based on the optimized transportation plan, the customer 12 can be offered an optimized transportation date.

[0094] <Evaluation of improvements in prediction accuracy> Figure 6 shows a graph comparing the predicted LNG24 usage per unit period by the customer 12, as predicted by the aforementioned management system 11, with the actual LNG24 usage. As shown in Figure 6, it was confirmed that the predicted values ​​can be matched with the trend of actual LNG24 usage.

[0095] <Evaluation of the optimization of transportation plans> We evaluated whether the transportation plan was optimized when it was created using the aforementioned management system 11 versus when it was created using the conventional method. This optimization was evaluated based on whether the number of operational transport vehicles 59 could be leveled out, that is, whether the maximum number of operational transport vehicles 59 could be reduced. In the conventional configuration, a transport plan is created based on the transport volume ordered by the customer 12. As a result, as shown in Figure 7 (conventional configuration) and Figure 8 (configuration of this embodiment), it was confirmed that the number of operational transport vehicles 59 is leveled out according to the configuration of this embodiment in which the transport plan is created by the management system 11.

[0096] <Modified representation of the main body of the transportation system 10> In the transportation system 10, as described above, there was a transporter 15 that acted as the transporting entity for transporting LNG 24 from the seller 14 to the customer 12. However, the transportation system 10 is not limited to this configuration. In the transportation system 10, the seller 14 may function as the transporting entity. That is, the seller 14 itself may transport LNG 24 to the customer 12 using the transport vehicle 59.

[0097] The present invention is not limited to the embodiments described above, and various modifications, changes, and improvements are possible without departing from the spirit of the invention. For example, the modified forms shown above may be combined in any way.

[0098] In this embodiment, machine learning is performed using performance data and schedule data in the learning model 63B. In this embodiment, in addition to performance data and schedule data, machine learning is performed using at least one of change data, date data, usage data, and day of the week data, but is not limited to this. The learning model 63B can perform machine learning using various types of data.

[0099] <Note> [Aspect 1] Based on predicted values ​​obtained using a machine learning model, the amount of usage by each user of the items stored in the storage compartments owned by each of the multiple users is predicted. A transportation plan is created to transport goods from multiple sources to each storage unit using predetermined means of transport, with a predetermined transport volume. In the aforementioned transportation plan, it is determined when, from which source, and by what means of transport each storage unit will be transported. How to create a transportation plan. [Aspect 2] The amount of usage per unit period by each of the aforementioned users is predicted using a machine learning model. The machine learning model is executed using performance data showing the actual amount of usage by the user for each unit period, and schedule data showing the schedule for the use of the product. A method for creating a transportation plan as described in Embodiment 1. [Aspect 3] The transportation plan is created such that the objective function value obtained by weighting the following elements 1 to 3 is minimized. A method for creating a transportation plan as described in Embodiment 1 or 2. Element 1: The amount of material to be stored in each storage section is within the upper and lower limits, taking into account the prediction error. Element 2: The transport distance from the source to each storage unit is minimized. Element 3: The maximum number of operational modes of transport is minimized. [Aspect 4] Based on the aforementioned transportation plan, we propose a transportation date to the user. A method for creating a transportation plan as described in any one of the three embodiments. [Aspect 5] Equipped with a processor, The aforementioned processor, Based on predicted values ​​obtained using a machine learning model, the amount of usage by each user of the items stored in the storage compartments owned by each of the multiple users is predicted. A transportation plan is created to transport goods from multiple sources to each storage unit using predetermined means of transport, with a predetermined transport volume. In the aforementioned transportation plan, it is determined when, from which source, and by what means of transport each storage unit will be transported. Creation system. [Aspect 6] For computers, Based on predicted values ​​obtained using a machine learning model, the amount of usage by each user of the items stored in the storage compartments owned by each of the multiple users is predicted. A transportation plan is created to transport goods from multiple sources to each storage unit using predetermined means of transport, with a predetermined transport volume. In the aforementioned transportation plan, it is determined when, from which source, and by what means of transport each storage unit will be transported. A program created to execute a process. [Explanation of symbols]

[0100] 10 Transportation Systems 11 Management Systems 12 customers 14 Sellers 15 Transporter 22 tanks 23 Detection Sensors 24. Liquefied natural gas 25 Equipment used 30 Customer terminals 31 processors 32 memory 33 Storage 34 Input section 35 Communications Department 36 Display section 40 Seller terminal 48 Accommodation facilities 49 Base 50 Transporter terminals 59 Transport vehicles 61 processors 61A Acquisition Department 61B Prediction Unit 61C Creation Section 62 memory 63 storage 63A Management Program 63B Learning Model 65 Communications Department

Claims

1. Based on predicted values ​​obtained using a machine learning model, the amount of usage by each user of the items stored in the storage compartments owned by each of the multiple users is predicted. A transportation plan is created to transport goods from multiple sources to each storage unit using predetermined means of transport, with a predetermined transport volume. In the aforementioned transportation plan, it is determined when, from which source, and by what means of transport each storage unit will be transported. How to create a transportation plan.

2. The amount of usage per unit period by each of the aforementioned users is predicted using a machine learning model. The machine learning model is executed using performance data showing the actual amount of usage by the user for each unit period, and schedule data showing the schedule for the use of the product. A method for creating a transportation plan as described in claim 1.

3. The transportation plan is created such that the objective function value obtained by weighting the following elements 1 to 3 is minimized. The method of preparation described in claim 1. Element 1: The amount of material to be stored in each storage section is within the upper and lower limits, taking into account the prediction error. Element 2: The transport distance from the source to each storage unit is minimized. Element 3: The maximum number of available means of transport is minimized. A method for creating a transportation plan as described in claim 1.

4. Based on the aforementioned transportation plan, we propose a transportation date to the user. A method for creating a transportation plan as described in claim 1.

5. Equipped with a processor, The aforementioned processor, Based on predicted values ​​obtained using a machine learning model, the amount of usage by each user of the items stored in the storage compartments owned by each of the multiple users is predicted. A transportation plan is created to transport goods from multiple sources to each storage unit using predetermined means of transport, with a predetermined transport volume. In the aforementioned transportation plan, it is determined when, from which source, and by what means of transport each storage unit will be transported. Creation system.

6. For computers, Based on predicted values ​​obtained using a machine learning model, the amount of usage by each user of the items stored in the storage compartments owned by each of the multiple users is predicted. A transportation plan is created to transport goods from multiple sources to each storage unit using predetermined means of transport, with a predetermined transport volume. In the aforementioned transportation plan, it is determined when, from which source, and by what means of transport each storage unit will be transported. A program created to execute a process.

Images