JP2023030962A - Information processing apparatus, control method of information processing apparatus, control program of information processing apparatus, and delivery system - Google Patents

Abstract

To provide an efficient delivery system.SOLUTION: An information processing apparatus in a delivery system related to delivery of gas to a consumer includes: an acquisition unit which acquires information on the amount of gas used by a consumer; a prediction unit which estimates future gas consumption from past gas consumption; a group generation unit which generates a group, on the basis of the estimated consumption, including an essential consumer of the highest gas delivery priority and a quasi-essential consumer of low gas delivery priority; and a delivery order setting unit which sets a gas delivery order in the group. The group generation unit generates a group for a delivery date on the basis of a first change-over date on which the consumer becomes an essential consumer and a second change-over date on which the consumer becomes a quasi-essential consumer.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information processing device, a control method for the information processing device, a control program for the information processing device, and a delivery system.

Conventionally, there has been proposed a delivery plan generation system that automatically creates a comprehensive delivery plan including delivery times for consumables such as LP gas cylinders and efficient delivery routes (for example, Patent Document 1).

JP 2019-219783 A

According to an embodiment of the present invention, an information processing apparatus related to gas delivery to a consumer includes an acquisition unit that acquires information on the amount of gas used by the customer; A group that includes a forecasting unit that calculates predicted usage and a group that includes essential customers with the highest gas delivery priority and semi-essential customers with lower gas delivery priority than essential customers based on the forecasted usage. and a delivery order setting unit for setting the order of gas delivery in the group. A group for the delivery date is generated based on the second conversion date of the consumer.

In the information processing apparatus according to one embodiment of the present invention, the group generation unit includes the essential customers and the semi-essential customers when the number of essential customers on the delivery date is equal to or less than the upper limit number of deliveries. Groups may be created.

In the information processing apparatus according to one embodiment of the present invention, the prediction unit performs machine learning using linear regression or nonlinear regression based on at least a data set regarding gas usage over a predetermined period in the past for each customer. A prediction model may be generated by and based on the prediction model, a predicted usage over a predetermined period in the future may be calculated for each customer.

In an information processing apparatus according to an embodiment of the present invention, the prediction unit includes the gas usage amount on the first day in the past as an objective variable, the gas usage amount over a predetermined period before the first day, and at least the past A prediction model may be generated by machine learning using information about the weather on the first day as an explanatory variable.

In an information processing apparatus according to an embodiment of the present invention, the prediction unit uses a past usage amount and a predicted usage amount predicted by a prediction model to calculate a predicted usage amount for a predetermined future period. , the group generation unit calculates the forecast accuracy of the forecast model based on the past forecast usage, which is the forecast usage at the past point in time, calculated using the forecast model, and the actual gas usage amount at the past point in time calculating the information, calculating the first probability that the remaining amount of gas is equal to or less than the first threshold value based on the prediction accuracy information and the predicted future usage amount, and calculating the first probability that the remaining amount of gas is equal to or less than the first reference value for the first time; may be the first conversion date.

In the information processing apparatus according to an embodiment of the present invention, in the group generation unit, the first conversion date may be calculated closer to the prediction execution date as the first reference value decreases.

In the information processing apparatus according to an embodiment of the present invention, the group generation unit further calculates a second probability that the remaining amount of gas is equal to or less than the second threshold value, and the second probability is the second reference value or more first. The second conversion date may be calculated closer to the prediction date as the second reference value becomes smaller.

In the information processing apparatus according to one embodiment of the present invention, the acquisition unit acquires information about the amount of gas used at a first interval, and the prediction unit calculates the amount of continuous gas used by one consumer. If the interval at which the information on gas consumption is acquired includes a section of the first interval or more and less than the second interval, a predetermined algorithm is used to complement the information on the gas usage amount in the section of the first interval or more and less than the second interval. , may generate a data set of gas usage over a period of time in the past.

In the information processing apparatus according to one embodiment of the present invention, the prediction unit predicts the gas A plurality of other consumers whose usage trends are similar to the gas usage trends of one consumer are extracted, and a data set obtained by linearly combining the information on the gas usage of the plurality of other consumers is stored as a predetermined past A data set of gas usage over a period of time may be used to calculate future projected usage of a customer.

In the information processing apparatus according to one embodiment of the present invention, the prediction unit determines that there is a section where the interval at which the information on the continuous gas usage amount of one consumer is acquired is equal to or greater than a second interval, and one consumer when the third quartile of the daily usage of the consumer is greater than the largest third quartile of the daily usage of a plurality of other consumers A data set relating to gas usage over a predetermined period of time in the past may be generated at the third quartile of daily usage for a customer.

In the information processing device according to one embodiment of the present invention, the prediction unit uses a prediction model as linear regression, support vector regression, random forest regression, gradient boosting regression. (Gradient Boosting Regression) or an algorithm using a Gradient Boosting Decision Tree.

In the information processing apparatus according to one embodiment of the present invention, the group generation unit determines at least the delivery areas of the essential customers and the semi-essential customers in one group under the constraint condition regarding the delivery capacity of the delivery vehicles for gas cylinders. is within a predetermined range, semi-essential consumers to be included in one group may be extracted.

In the information processing apparatus according to one embodiment of the present invention, the group generation unit at least groups the essential consumers under a constraint condition related to the delivery capacity of the delivery vehicles for the gas cylinders. A semi-essential customer located in the delivery area of the group may be extracted as a semi-essential customer to be included in one group.

In the information processing apparatus according to one embodiment of the present invention, the delivery order set by the delivery order setting unit is for one group generated by the group generation unit only for the essential consumers included in one group. , a delivery order that tours the essential customers by the Traveling Salesman Problem (TSP).

In the information processing apparatus according to one embodiment of the present invention, the delivery order set by the delivery order setting unit is included in one group generated by the group generation unit under at least a constraint condition regarding the working hours of the delivery person. It may include a delivery order set based on a first mathematical model that maximizes the number of gas cylinders to be delivered to semi-essential customers included in one group after always delivering to essential customers included in one group.

In the information processing apparatus according to one embodiment of the present invention, the delivery order set by the delivery order setting unit is included in one group generated by the group generation unit under the constraint condition regarding the working hours of the delivery person. When the delivery order for delivery to semi-essential customers cannot be set, the delivery order may be set based on a second mathematical model that minimizes the delivery time for delivery to essential customers included in one group.

In the information processing apparatus according to one embodiment of the present invention, the delivery order set by the delivery order setting unit is included in one group generated by the group generation unit under at least a constraint condition regarding the working hours of the delivery person. The delivery order set based on the third mathematical model that minimizes the delivery distance among the delivery orders that do not deliver the gas cylinders to the semi-essential customers included in one group after always delivering to the essential customers included in one group. may contain

In the information processing apparatus according to one embodiment of the present invention, the delivery order set by the delivery order setting unit is such that the delivery order based on the first to third mathematical models cannot be set under the constraint conditions regarding the working hours of the delivery staff. In this case, the gas cylinders will not be delivered to the semi-essential customers included in one group, but will be delivered to the essential customers included in one group within the prescribed upper limit of working hours of the delivery staff, regardless of the constraints. It may include a delivery order set to:

In the information processing apparatus according to one embodiment of the present invention, the delivery order set by the delivery order setting unit is such that the delivery order based on the first to third mathematical models cannot be set under the constraint conditions regarding the working hours of the delivery staff. In this case, the gas cylinders will not be delivered to the semi-essential customers included in one group regardless of the constraints, and will be delivered within the prescribed upper limit of working hours of the delivery staff among the essential customers included in one group. May include order of delivery to essential customers that maximizes the number of gas containers served.

In the information processing apparatus according to one embodiment of the present invention, the delivery order setting unit determines that when delivery to the essential consumers included in the group generated by the group generation unit exceeds the predetermined upper limit of working hours of the delivery person, , the essential customers may be deleted from the group in descending order of the first probability, and a delivery order that allows delivery within a predetermined upper limit of working hours may be set.

In the information processing apparatus according to one embodiment of the present invention, the delivery order set by the delivery order setting unit is the delivery time specified by the essential customers and semi-essential customers included in the group generated by the group generation unit. may include a delivery order set by the Time Constrained Traveling Salesman Problem (TSP-TW) according to .

In the information processing apparatus according to one embodiment of the present invention, the delivery order setting unit determines whether delivery is possible within the upper limit of working hours of the delivery person according to the delivery order set by the time-constrained traveling salesman problem. , among the semi-essential customers included in one group, the semi-essential customers contributing to reduction of the delivery time may be updated to the delivery order by thinning out the semi-essential customers.

In the information processing apparatus according to an embodiment of the present invention, when information about one gas container is associated with a plurality of consumers, the group generation unit generates predicted usage amounts predicted for the plurality of consumers. You may calculate a 1st threshold value by adding.

The information processing apparatus according to one embodiment of the present invention may further include a communication unit that transmits information regarding the delivery order set by the delivery order setting unit to the communication terminal of the delivery person.

A control method for an information processing apparatus according to an embodiment of the present invention includes steps in which the information processing apparatus obtains information about the amount of gas used by a customer; and a step of generating a group including essential customers with the highest delivery priority of gas and semi-essential customers with a lower delivery priority than the essential customers based on the predicted usage; and the grouping step is performed on a first conversion date when the customer becomes an essential customer and a second conversion date when the customer becomes a semi-essential customer. Based on this, a group is generated for the delivery date.

A control program for an information processing apparatus according to an embodiment of the present invention provides the information processing apparatus with a function of acquiring information on the amount of gas used by a consumer, and a predicted amount of gas used in the future based on the amount of gas used in the past. and a function to generate a group that includes essential customers with the highest gas delivery priority and semi-essential customers with a lower gas delivery priority than essential customers based on the predicted usage. , the function of setting the gas delivery order in the group, and the function of creating the group by realizing the first conversion date when the customer becomes an essential customer, and the second conversion date when the customer becomes a semi-essential customer. Create a group for the delivery date based on the date.

A delivery system for gas delivery to a consumer according to an embodiment of the present invention includes an acquisition unit that acquires information about the amount of gas used by the customer, and a prediction of future gas from the past gas usage. Based on the prediction unit that calculates the usage amount and the predicted usage amount, a group is generated that includes essential customers with the highest gas delivery priority and semi-essential customers with a lower gas delivery priority than the essential customers. and a delivery order setting unit for setting the delivery order of gas in the group, the group creation unit determines the first conversion date when the customer becomes the essential customer and the semi-essential demand when the customer becomes the semi-essential demand. A group is generated for the delivery date based on the second conversion date that becomes the home.

1 is a schematic diagram of a delivery system configuration according to one embodiment of the present invention; FIG. It is an example of a customer information table stored in the delivery system according to one embodiment of the present invention. 1 is information stored in a delivery system according to an embodiment of the present invention, where (a) is an example of a delivery person information table and (b) is an example of a delivery vehicle information table. BRIEF DESCRIPTION OF THE DRAWINGS It is the schematic explaining the outline|summary of the delivery system which concerns on one Embodiment of this invention. It is a schematic diagram explaining an algorithm in the delivery system according to one embodiment of the present invention. It is an example of a delivery list generated by the delivery system according to one embodiment of the present invention. (a) and (b) are schematic diagrams showing essential customers and semi-essential customers appearing as days pass in the delivery system according to the embodiment of the present invention. It is a flow chart of prediction model generation processing concerning one embodiment of the present invention. (a), (b) is a figure which shows an example of the information memorize|stored in the delivery system which concerns on one Embodiment of this invention. (a), (b) is the schematic explaining the complementary process in the delivery system which concerns on one Embodiment of this invention. It is a schematic diagram explaining complement processing in the delivery system according to one embodiment of the present invention. It is a schematic diagram explaining complement processing in the delivery system according to one embodiment of the present invention. It is a figure which shows an example of the weather information memorize|stored in the delivery system which concerns on one Embodiment of this invention. It is a figure which shows an example of the prediction model information memorize|stored in the delivery system which concerns on one Embodiment of this invention. It is a schematic diagram explaining an outline of prediction processing in a delivery system concerning one embodiment of the present invention. It is an example of the flowchart concerning the prediction process in the delivery system which concerns on one Embodiment of this invention. 4(a) and 4(b) are schematic diagrams plotting a risk function indicating the probability that the remaining amount of gas is equal to or less than a predetermined threshold for one consumer in a delivery system according to an embodiment of the present invention; be. 4(a) and 4(b) are schematic diagrams plotting a risk function indicating the probability that the remaining amount of gas is equal to or less than a predetermined threshold for one consumer in a delivery system according to an embodiment of the present invention; be. (a) to (c) are schematic diagrams for explaining grouping in the delivery system according to one embodiment of the present invention. It is a schematic diagram explaining grouping in the delivery system according to one embodiment of the present invention. (a) and (b) are schematic diagrams illustrating grouping in the delivery system according to one embodiment of the present invention. It is a schematic diagram explaining grouping in the delivery system according to one embodiment of the present invention. FIG. 4 is a diagram illustrating an overview of delivery route processing in the delivery system according to one embodiment of the present invention; It is an example of a table related to delivery routes in one embodiment of the present invention. 6 is an example of a flowchart relating to delivery order setting processing in the delivery system according to one embodiment of the present invention. 6 is an example of a flowchart relating to delivery order setting processing in the delivery system according to one embodiment of the present invention. It is a hardware configuration example of a computer capable of realizing the server 100 according to one embodiment of the present invention.

An embodiment of the invention (also referred to as the present invention) according to the present disclosure will be described below with reference to the drawings. The drawings are only examples, and the present invention is not limited to those shown in the drawings. For example, the configuration diagram of the delivery system, communication terminal and information processing device (server), meter, communication device, number of gas containers, data set (table), and flow chart shown in the figure are examples, and the present invention is limited to these. not to be

<System configuration>
FIG. 1 is a diagram showing a configuration example of a delivery system according to one embodiment of the present invention. The distribution system 800 may be a system related to gas distribution to consumers. For example, the distribution system 800 may be a system for efficiently distributing gas in a gas supply service in which gas containers are replaced according to the amount of gas used by consumers. Efficient delivery of gas may be aimed at, for example, minimizing the replacement frequency and the time required for replacement per customer and maximizing the amount of gas delivered per delivery person. For example, if there is a large amount of remaining gas and the gas container does not need to be replaced even after visiting the customer, the frequency of replacement will increase, resulting in poor efficiency. In addition, depending on the order of delivery to each customer, it takes time to move between customers, which is inefficient. According to one embodiment of the present invention, a delivery system 800 that accurately predicts the amount of remaining gas at a consumer and appropriately sets the order of gas delivery to each consumer for efficient delivery of gas cylinders. can be realized.

The delivery system 800 includes a server (arithmetic processing server) 100, a management server 210, a business DB (database) 220, an external DB 230, a communication device 310 installed in the meter 320 of the gas containers 330A and 330B, and a communication terminal 510 owned by the delivery person 500 .

The meter 320 is a so-called "smart meter" that includes the communication device 310, is connected to the gas containers 330A and 330B, and may measure the amount of gas used by the consumer. The communication device 310 may transmit to the management server 210 information about gas, such as the meter reading of the meter 320 and information about replacement of the gas container. Note that the communication device 310 may transmit the meter reading of the meter 320 to the management server 210 at predetermined intervals. The communication device 310 may transmit the meter reading data to the management server 210 at a set time, for example, once a day. In addition, the communication device 310 may transmit, as information about gas, information about anomalies and alarms, such as gas leaks, out of gas, and stoppage of supply from gas containers, to the management server 210 via the network 600 . Communication device 310 may be connected to an input/output I/F outside meter 320 and provided outside meter 320 as a separate communication device. That is, communication device 310 may be retrofitted to meter 320 . Alternatively, the communication device 310 may be built into the meter 320 when it is manufactured and integrated with the meter 320 . Alternatively, communication device 310 may be incorporated within meter 320, for example, as a communication board.

The management server 210 may collect information from meters 320 installed in homes, businesses, facilities, etc., and may remotely control the meters 320 via the network 600 . That is, the management server 210 processes information of each meter 320 transmitted from the communication device 310 and functions as an IoT-PF (platform) that transfers necessary data to a meter administrator (not shown). good. The management server 210 may include a meter reading data DB 212 that stores (stores) meter reading data of each meter. In addition, prediction model DB 213 is mentioned later.

The communication method in the network 600 between the management server 210 and the communication device 310 is, for example, LTE, LTE-Advanced, fourth generation communication (4G), fifth generation communication (5G), sixth generation communication (6G) or later. communication method, CDMA, or the like. Also, for example, it is a wireless communication system for IoT such as Category M, Category M1, and NB-IoT (Narrow Band IoT), and may be a communication system that extends LTE. Note that the communication method is not limited to these examples.

Currently, although smart meters are becoming more popular, there are consumers who do not have a communication device and require regular meter reading by the meter reader 440 . In this specification, hereinafter, a customer having a meter 320 equipped with a communication device 310 is referred to as an "installed customer", and a customer having a meter 420 without a communication device 310 is referred to as a "non-installed customer". . As for the amount of gas used by non-installed consumers, the meter reader 440 checks the meter readings of the meters 420 connected to the gas containers 430A and 430B every other month, for example, and sends the meter readings to the management server 210 via the communication terminal 441, for example. may be sent. When there is no particular need to distinguish between installed customers and non-installed customers, they may be collectively referred to as "customers". In addition, in FIG. 1, one installed customer and one non-installed customer are shown as detached houses, but the customers are not limited to this, and may be collective housing, corporations, etc. There may be installed customers and non-installed customers.

The company DB 220 is a database of companies that provide gas delivery services, and may include a consumer information DB 221 , a delivery person information DB 222 , and a delivery vehicle information DB 223 . FIG. 2 shows an example of consumer information stored in the consumer information DB 221. As shown in FIG. The consumer information table TB1 is information for identifying each consumer, and includes a consumer ID (Identifier) (an example of an identifier) that uniquely identifies each consumer, a meter ID, a delivery destination ID, and capacity information. , delivery designation information and the like may be associated with each other and stored. Any type of identifier can be used as long as it is information that uniquely identifies an assigned object.

Referring to the consumer information table TB1, for the consumer with the consumer ID "CS0001", the meter with the meter ID "mt0001" is installed, and the area identified by the delivery destination ID "LC0001" is the delivery destination of the gas cylinder. I understand. Note that the delivery destination information that stores the delivery destination ID and the area (address) in association with each other may be stored as a database (not shown), and the delivery destination address may be determined from the delivery destination ID. The capacity information may be information indicating the total capacity of gas containers supplied to the consumer. In the case of the consumer with the consumer ID "CS0001", the capacity of the gas container is "50L", the number of gas containers is "2", and the total capacity of the gas containers is 100L. Also, the delivery designation information may be information on conditions specified by each consumer regarding delivery. For example, the delivery designation information may indicate that the consumer with the consumer ID “CS0001” has designated the delivery of the gas cylinder “after 16:00 from Monday to Thursday”.

Further, referring to the consumer information table TB1, the consumer with the consumer ID "CS0005" is a non-installed consumer and does not need to be associated with a meter ID. Further, consumer IDs "CS0015" and "CS0016" are associated with the same delivery destination ID "LC0040". This may indicate a case where a plurality of consumers exist at one address, such as collective housing such as condominiums and apartments, or a building in which a plurality of corporations reside. In this case, the capacity information may indicate the total capacity of the gas containers installed at one delivery destination ID.

Note that the table TB1 in FIG. 2 is an example, and is not limited to this. For example, the capacity information is not limited to the illustrated format as long as it is information that can determine the total capacity of the gas container. Also, the consumer information may be stored in a plurality of tables. Further, the consumer information table TB1 contains the identifier of the communication device 310, the date of installation of the meters 320 and 420, the information on the gas distribution service provider (the service provider used by the consumer), the contact information of the consumer and other information. Data may be stored.

FIG. 3( a ) shows an example of delivery member information stored in the delivery member information DB 222 . The delivery member information table TB2 is information for identifying each delivery member, and stores a delivery member ID that uniquely identifies each delivery member in association with working hours, working days, assigned area ID, and the like. good. Referring to the delivery person information table TB2, the delivery person with the delivery person ID “DP0001” works from 9:00 to 17:00 and is in charge of the area identified by the area ID “AR0100”. It can be seen that it is a delivery area. Note that a table (not shown) that associates the responsible area ID with the above-described delivery destination ID may be further stored. Note that if the delivery person does not have a responsible area, the responsible area ID may not be stored. In addition, the delivery staff information table TB2 may further store the weight of the delivery staff and the upper limit of overtime hours (for example, the working hours are until 17:00, but overtime work is possible until 20:00, etc.).

FIG. 3B shows an example of delivery vehicle information stored in the delivery vehicle information DB 223. As shown in FIG. The delivery vehicle information table TB3 is information for identifying each delivery vehicle, and may store a delivery vehicle ID that uniquely identifies each delivery vehicle in association with the vehicle type, maximum loading capacity, and the like. Referring to the delivery vehicle information table TB3, it can be seen that the delivery vehicle with the delivery vehicle ID "TR0001" has a city vehicle type of "large size" and a maximum loading capacity of "20000 kg". Further, the delivery vehicle information table TB3 may further store information on the maximum loading capacity, information on the next use schedule, such as the scheduled date of use, time, delivery person, and the like. Note that the delivery vehicle information table TB3 may store, for each business, the delivery vehicles owned by each business.

The server (arithmetic processing server) 100 is a server that executes various arithmetic processes in the delivery system 800 , and may include a communication section 110 , a prediction section 120 , a group generation section 130 and a delivery order setting section 140 .

Here, an outline of the delivery system 800 according to one embodiment of the present invention will be described. FIG. 4 is a schematic diagram illustrating an overview of the delivery system 800. As shown in FIG. The delivery system 800 includes a “remaining amount prediction” of gas for each consumer, a “delivery list setting” for setting gas delivery destinations based on the predicted remaining gas amount, and a delivery route (delivery route) for the set delivery destinations. Gas delivery may be performed efficiently through the three steps of "delivery route setting" for setting the order).

Each step will be briefly explained. In the "remaining amount prediction" step 10, first, the past meter reading data of the consumer is complemented by a predetermined complementing algorithm as necessary. Next, based on a data set of past meter reading data of consumers and past weather information, a prediction model for predicting future gas usage is set by machine learning. Then, using the set prediction model and the meter reading data, prediction data regarding future gas usage is calculated.

In the "delivery list setting" step 20, the future gas remaining amount is predicted from the forecast data of the future gas usage calculated in the remaining amount forecasting step 10 and the information on the gas container capacity of the customer. Then, the degree of risk of running out of gas is calculated from the estimated remaining amount of gas, and a delivery list is set for consumers with a high degree of risk as consumers requiring gas delivery.

In the "delivery route setting" step 30, a delivery order is set for efficient delivery to the customers included in the delivery list. In setting the order of delivery, the delivery time zone desired by the customer, the travel time between customers, the time required to replace the gas container, the working hours of the delivery staff, and the like may be taken into consideration. Therefore, the possession time for delivery in the set delivery order may be output in the delivery route setting step. Elements taken into consideration in setting the delivery order are not limited to the above examples as long as they can make delivery more efficient.

Among the above three steps, the "remaining amount prediction" and "delivery list setting" steps are executed for each of the large number of consumers. Further, to reduce the possibility of running out of gas, the above three steps are preferably performed on a daily basis, with gas meter readings read and sent to the management server 210 daily. That is, the delivery system 800 according to one embodiment of the present invention may be a system that performs arithmetic processing using a large amount of data about many customers.

<Algorithm in delivery system>
Details will be described later, but the algorithms used in the above three steps will be briefly described with reference to FIG. The prediction unit 120 may execute remaining capacity prediction step 10 . The prediction unit 120 may acquire meter reading data from the meter reading data DB 212 and complement the meter reading data with a predetermined complementing algorithm 11 as necessary. After that, using a predetermined supervised machine learning method using past meter reading data or a predetermined approximation method, a prediction model 12 that predicts the future gas usage of each customer is generated. may be generated. The predetermined machine learning method is, for example, Linear Regression, Support Vector Regression (SVR), Random Forest regression, or Gradient Boosting Regression (GBT). , k-Nearest Neighbor (kNN), Gradient Boosting Decision Tree (GBDT), and the like. Also, a TQ (Third Quartile) model may be used as a predetermined approximation method. The prediction unit 120 may use the generated prediction model to calculate the amount of gas used for several days in the future for each customer.

The group generation unit 130 may execute the delivery list setting step 20 . The group generation unit 130 may calculate the predicted value of the remaining amount of gas from the future gas usage amount predicted by the prediction unit 120 .

Here, the consumers may be categorized into three categories of "essential delivery customers", "semi-essential delivery customers", and "non-delivery customers" according to the priority of gas delivery. A “must-delivery consumer” may refer to a consumer with the highest gas delivery priority. For example, a "must-delivery customer" may refer to a customer that requires gas delivery on a given delivery date. Specifically, the "essential delivery consumer" means, for example, that the probability that the predicted value of the remaining amount of gas is equal to or less than a first threshold value (first probability described later) is equal to or greater than a predetermined reference value (first reference value). You can point to a certain customer. In addition, the "semi-essential delivery customer" may refer to a customer whose gas delivery priority is lower than that of the essential customer but whose gas delivery priority is higher than that of the non-delivery customer. For example, a “semi-essential delivery customer” is a customer who will not run out of gas even if no delivery is made on a certain delivery date, but who can become an essential customer within a few days (near future) after that delivery date. You can point Specifically, the "semi-essential delivery consumer" is, for example, the probability that the first probability is less than the first reference value and the predicted value of the remaining amount of gas is equal to or less than the second threshold (second probability) is equal to or greater than a predetermined reference value (second reference value). In the following description, the first threshold is a value corresponding to 5% or less of the gas container, and the second threshold is a value corresponding to 7% or less of the gas container. However, the values of the first threshold and the second threshold are not limited to these, and may be arbitrary values. In addition, the “customer not targeted for delivery” may refer to a customer whose gas delivery priority is lower than that of the semi-essential delivery customer. For example, a “customer not subject to delivery” is a consumer who does not run out of gas even if delivery is not made on the delivery date and has a low possibility of running out of gas for several days (near future) after the delivery date. You can point Specifically, a “customer not targeted for delivery” may refer to, for example, a consumer who has a sufficient remaining amount of gas and a low possibility of running out of gas on the prediction execution date. Hereinafter, the essential delivery customer is also referred to as the "essential customer", and the semi-essential delivery customer is also referred to as the "semi-essential customer".

The group generation unit 130 trip-splits essential customers and semi-essential customers to be delivered on the delivery date. A trip may refer to a single delivery starting from the distribution center where the gas container is loaded, going around the destination, and returning to the distribution center. In other words, the group generation unit 130 may group the consumers to be delivered in one delivery by a predetermined optimization problem method. The predetermined optimization problem method may be, for example, a greedy method. Note that the group generation unit 130 may first divide trips only for the essential consumers. That is, the group generation unit 130 may calculate the number of delivery vehicles (trucks) required for delivery to the essential consumers, and calculate the trucks to load the gas cylinders of each essential consumer. In addition, the group generation unit 130 may output the information of the consumers divided into trips as, for example, a delivery list. FIG. 6 shows an example of a delivery list. As shown in the figure, the delivery list TB10 may be a table in which information on consumers to whom delivery is to be made is associated with each trip. The figure is only an example, and the format of the delivery list is as long as it is possible to identify which trip the customer to be delivered is included in and whether it is a semi-essential customer or an essential customer. is not limited to this.

The delivery order setting unit 140 may set the order of delivery to consumers when delivering according to the delivery list generated by the group generation unit 130 . That is, the delivery order setting unit 140 may set an efficient delivery order by using a predetermined optimization problem method, for example, by solving a mathematical model. The method of the predetermined optimization problem may be, for example, an integer programming problem (integer program: IP), a traveling salesman problem (TSP), or the like.

<Functional configuration>
Returning to FIG. 1, the function of each functional unit of the server 100 will be briefly described. The server 100 is a server that executes the above-described arithmetic processing in the delivery system 800 , and may include a communication section 110 , a prediction section 120 , a group generation section 130 and a delivery order setting section 140 . Note that FIG. 1 shows the prediction unit 120 , the group generation unit 130 and the delivery order setting unit 140 as functional units in one server 100 . However, each functional unit may be realized by a separate server or engine. Also, the server 100 may be, for example, a distributed server system that cooperates by communicating via a network, or may be a so-called cloud server. In other words, the server 100 is not limited to a physical server, and may include a software virtual server.

The communication unit 110 may acquire information about the amount of gas used by the consumer from the management server 210 . The information about the amount of gas used by the customer is information indicating the amount of gas used by each customer transmitted from the communication device 310 or the communication terminal 441 of the meter reader 440, and stored in the meter reading data DB 212. It can be information.

The prediction unit 120 may be a functional unit that executes the remaining capacity prediction step 10 described above. That is, the prediction unit 120 may calculate the predicted future gas usage from the past gas usage.

The group generator 130 may be a functional unit that executes the distribution list setting step 20 described above. Based on the predicted usage amount calculated by the prediction unit 120, the group generation unit 130 divides the essential customer with the highest gas delivery priority and the semi-essential customer with a lower gas delivery priority than the essential customer. It may be possible to generate a group containing The group generation unit 130 determines that the demander becomes the essential consumer with the highest delivery priority of the gas cylinder, with the first probability that the predicted value of the remaining amount of gas is equal to or less than the first threshold value is equal to or greater than the first reference value. A first conversion date may be calculated. Further, the group generation unit 130 determines that the first probability that the remaining amount of gas is equal to or less than the first threshold is less than the first reference value and the second probability that the remaining amount of gas is greater than the first threshold for the plurality of consumers. A second conversion date may be further calculated in which the second probability of being equal to or less than the threshold becomes equal to or greater than the second reference value and the customer becomes a semi-essential customer having the second highest delivery priority of gas cylinders. Note that the first reference value and the second reference value may be different values or may be the same value.

Further, the group generation unit 130 may generate a group for the delivery date based on the first conversion date when the customer becomes the essential customer and the second conversion date when the customer becomes the semi-essential customer. That is, the group generation unit 130 may group a plurality of essential consumers on the delivery date, which are extracted based on the first conversion date of each consumer, based on a predetermined condition regarding delivery. . Predetermined conditions for delivery will be described later. Further, when the number of essential consumers on the delivery date is equal to or less than the upper limit number of deliveries, the group generation unit 130 may further group semi-essential customers extracted based on the second conversion date. . The delivery date for calculating the first conversion date and the second conversion date may be the next delivery date, or may be one week, one month, or the like after the next delivery. That is, the delivery date is not particularly limited.

The delivery order setting unit 140 may be a functional unit that executes the delivery route setting step 30 described above. That is, the delivery order setting unit 140 may set the delivery order of the gas containers in each group grouped by the group generation unit 130 .

Here, with reference to FIG. 7, the advantage of generating a group including not only essential customers but also semi-essential customers will be described. According to one embodiment of the present invention, as described above, the group generation unit 130 extracts essential consumers and semi-essential consumers from a plurality of consumers according to the predicted value of the remaining amount of gas. FIG. 7 is a schematic diagram showing essential customers and semi-essential customers appearing as the number of days elapses. In FIG. 7, the upper part shows the case of delivery only to essential customers, and the lower part shows the case of delivering not only essential customers but also semi-essential customers.

First, the case of delivering only to essential consumers (upper stage) will be described. On the first day, all consumers are non-delivery consumers and no delivery is performed (MD1). On the second day, although semi-essential customers are found as a result of the prediction, there are no essential customers, so delivery is not performed (MD2). On the third day, semi-essential customers newly appeared, and some percentage of the semi-essential customers on the second day became essential customers (MD3a), so delivery is performed only to essential customers ( MD3b). At this time, since the percentage of essential consumers is small, the delivery volume is small and the burden on the delivery person is small. However, on the 4th day, new semi-essential customers arise, and many of the semi-essential customers who were not delivered on the 3rd day become essential customers (MD4a), and the delivery volume is reduced compared to the 3rd day. increased (MD4b). In other words, the burden on the delivery person varies depending on the delivery date, and if there are many essential customers, there is a risk that the delivery will not be completed within working hours. In other words, if the delivery is made only to essential consumers, the efficiency is lowered.

On the other hand, a case (lower stage) in which delivery is made not only to essential customers but also to some semi-essential customers will be explained. As in the case of the upper row, on the first day, all customers are non-delivery customers and no delivery is performed (PD1). On the second day, semi-essential customers arise as a result of the prediction (PD2a), and although they are semi-essential customers, delivery is made to some of them (PD2b). As a result, the number of essential customers decreases on the third day (PD3a) and delivery is no longer essential compared to the case of delivering only to essential customers. At this time, however, some of the semi-essential customers may also be delivered (PD3b). As a result, the number of customers who become essential customers on the fourth day can be reduced (PD4a), and the number of essential customers to be delivered on the next delivery day can be reduced (PD4b).

According to one embodiment of the present invention, essential customers and semi-essential customers are extracted according to future gas forecast values. By delivering gas not only to essential customers but also to semi-essential customers, it is possible to level the delivery burden of delivery personnel and to provide an efficient delivery system.

<Prediction model generation processing>
Here, the process of generating a prediction model according to one embodiment of the present invention will be described with reference to the flowchart of FIG. Note that the flowchart of FIG. 8 may be executed for each consumer. First, the prediction unit 120 may determine whether or not one customer is an installed customer who has installed a smart meter (step S10). In the case of an installation customer, the meter reading of meter 320 is obtained at predetermined first intervals and transmitted from communication device 310 to management server 210 . The predetermined first interval may be, for example, one day. That is, the meter reading value may be transmitted from the communication device 310 to the management server 210 at a predetermined time every day.

FIG. 9 shows an example of information about the amount of gas used, which is transmitted from the communication device 310 to the management server 210. As shown in FIG. For example, the communication device 310 may transmit information about the meter ID, the date and time of the meter reading, and the guideline value of the meter to the management server 210 . FIG. 9A is an example of the data table TB4 when the meter reading value is transmitted from the communication device 310 to the management server 210. FIG. Further, information indicating whether or not the gas container has been replaced may be transmitted by a flag of, for example, "1" if replaced and "0" if not replaced. In the management server 210, as shown in FIG. 9B, a meter reading data table TB5 for each meter may be stored in the meter reading data DB 212 based on the received meter reading values.

If the one customer is an installed customer (YES in step S10), the prediction unit 120 refers to the meter reading data over a predetermined period in the past, and information about the continuous gas usage amount of the one customer is acquired. It may be determined whether or not there is a section whose interval is equal to or greater than the above-described first interval (step S11). Thereby, it may be determined whether or not there is a period during which the meter reading data did not reach the management server 210 due to some trouble and the meter reading data could not be acquired.

Note that if there is no interval longer than the first interval in the generation of the prediction model (NO in step S11), supplementation of meter reading data is unnecessary, and the generation of the prediction model may proceed. If there is an interval longer than or equal to the first interval (YES in step S11), the prediction unit 120 determines that the interval at which the information on the continuous gas usage amount of one consumer is acquired is the second interval longer than the first interval. It may be determined whether or not there is an interval equal to or greater than the interval (step S12). Note that the second interval may be, for example, 80 days, but is not limited to this. If there is a section longer than the second interval (YES in step S12), the process may proceed to step S17. If the meter reading data cannot be acquired for the second interval or more, the customer may be treated as a non-installed customer instead of an installed customer. Therefore, in the case of YES in step S12, the processing for the case where it is determined that the customer is not an installation customer in step S10 may be performed.

When there is no section longer than the second interval, that is, when the interval at which the information on the continuous gas usage amount of one consumer is acquired is the first interval or more and less than the second interval (in step S12 NO), the prediction unit 120 complements the information about the gas usage in the interval between the first interval and the second interval using a predetermined algorithm, and generates a data set about the gas usage over the past predetermined period. Good (step S13).

Algorithms used for complementation will be described with reference to FIGS. For example, for a consumer j, the cumulative usage x _i (i∈[t, t+N-1]) is missing for N days from a certain date t, and the daily gas usage {μ ^j _t-1 ,μ ^j _t ,...,μ ^j _t+N-1 } cannot be calculated.

<Linear interpolation (interpolation)>
FIG. 10A is a diagram for explaining interpolation interpolation among linear interpolations. First, calculate the difference D(t, t+N-1) of the cumulative usage during the deficit period, and divide the difference D(t, t+N-1) by the number of deficit days N to calculate the average daily usage during the deficit period. Calculate the quantity μ1. Furthermore, it is possible to calculate the cumulative usage amount for the missing day iε[t, t+N−1] using the calculated average daily usage amount μ1 for the missing period. That is, the cumulative usage amount on missing days is as follows.

<Linear interpolation (extrapolation)>
FIG. 10B is a diagram for explaining extrapolation among linear interpolations. In the case of extrapolation, calculate the average daily usage μ2 of the complement target consumer, and use the calculated average daily usage μ2 for the missing period to accumulate missing days i ∈ [t, t + N-1] It is sufficient to calculate the usage amount. That is, the cumulative usage amount on missing days is as follows. Note that the third quartile may be used instead of the average daily usage.

<Supplementation of similar consumers (interpolation)>
FIG. 11 is a diagram for explaining a case where similar consumers whose meter reading data tends to be similar to a deficient consumer are extracted from a plurality of consumers and used for complementation. First, for a consumer i to be complemented, a similarity function L is used to extract K installation consumers with a high degree of similarity to obtain a set N(K, L). Then, the daily usage amount for i∈[t−1, t+N−1] during the deficit period is calculated from the set N(K, L) using the following formula.

In addition, in order to ensure the monotonicity of the cumulative usage amount, the daily usage amount may be corrected by the following formula.

Then, using the calculated daily usage amount μ ^* _i for the missing period, the cumulative usage amount for the missing day iε[t, t+N-1] may be calculated by the following formula.

<Supplementation of similar consumers (extrapolation)>
Although not shown, in the case of extrapolation supplementation by similar consumers, in the case of interpolation described above, the daily usage is not corrected, and µ _i in Equation (3) is used to calculate the missing day i ∈ [t, t+N-1] may be calculated using the following formula:

<Cycle interpolation (interpolation)>
FIG. 12 is a diagram for explaining periodic supplementation in which missing data is supplemented using the periodicity of meter reading data of one consumer. For missing days i ∈ [t-1, t+N-1] of daily usage, let Λ _i be the set of dates with the same days of the week. At this time, the daily usage amount for missing day i can be expressed by the following formula.

In order to ensure the monotonicity of the cumulative usage, the daily usage is corrected by the following formula.

Then, using the calculated daily usage amount μ ^* _i for the missing period, the cumulative usage amount for the missing day iε[t, t+N-1] may be calculated by the following formula.

<Cycle interpolation (extrapolation)>
Although not shown, in the case of extrapolation by periodic interpolation using the periodicity of the meter reading data of one consumer, in the case of interpolation described above, the daily usage amount is not corrected, and μ in Equation (7) Using _i , the cumulative usage for missing day i∈[t, t+N−1] may be calculated by the following formula.

The prediction unit 120 performs prediction for predicting future gas usage by machine learning using linear regression or nonlinear regression based on the data set of the past predetermined period supplemented as necessary as described above. A model may be created (step S14).

According to one embodiment of the present invention, a predictive model of future gas usage is generated by machine learning using a data set of meter reading data for a predetermined period in the past. Therefore, since the lack of past meter reading data affects the accuracy of the prediction model, if there is a lack of meter reading data, the missing portion is complemented to generate a data set. Thereby, the accuracy of the prediction model can be improved.

<Prediction model creation>
Generating a predictive model according to one embodiment of the present invention will now be described. The prediction unit 120 generates a prediction model by machine learning using the amount of gas used on the first day in the past as an objective variable and the amount of gas used over a predetermined period before the first day and weather information as explanatory variables. You can For example, the forecast execution date is t, x _i is the daily usage amount on the i day, K is the period of input to the model, S is the number of samples, and for weather data, the maximum temperature, minimum temperature, daytime weather conditions, nighttime be w _i ^hight , w _i ^low , w _i ^daytime , w _i ^night respectively. At this time, the explanatory variable matrix X and the objective variable vector Y used for learning in the prediction model for one consumer can be expressed by the following equations.

The prediction unit 120 may calculate a prediction model f that satisfies [Y]=f[X] from the explanatory variables and objective variables described above. Information stored in the weather information DB 231 of the external DB 230 may be used as the weather information. FIG. 13 shows an example of the weather information table TB6. The weather information table TB6 is stored, for example, by city, and may store dates, daytime and nighttime weather, maximum temperature, average temperature, minimum temperature, and the like. Note that the prediction unit 120 inputs category information such as “sunny”, “clear”, “cloudy”, “rainy”, “heavy rain”, and “fog” for the explanatory variables of formula (11). , may be converted to a binary vector. Note that the prediction unit 120 may convert the weather information into a one-hot vector when the weather information is "clear", "cloudy", or "rainy", and into a zero vector in other cases.

According to one embodiment of the present invention, a prediction model is generated for each customer according to the past usage of the customer itself. Therefore, it is possible to generate a highly accurate prediction model that conforms to the usage trend of the consumer.

In addition, since the predictive model is generated using explanatory variables including weather information that affects gas usage, it is possible to generate a highly accurate predictive model.

In addition, the prediction unit 120 uses the prediction model as Linear Regression, Support Vector Regression, Random Forest regression, Gradient Boosting Regression, or Gradient Boosting Regression. It may be generated by an existing algorithm using a decision tree (Gradient Boosting Decision Tree: GBDT) or the like. These methods can generate a prediction model with high accuracy.

Returning to the flow chart of FIG. The prediction model generated by the prediction unit 120 in step S14 may be stored, for example, in the prediction model DB 213 of the management server 210 (step S15). Then, as shown in FIG. 14, the prediction model DB 213 may store a prediction model table TB7 that associates and stores a prediction model corresponding to each consumer, that is, each meter. The prediction model table TB7 may store information related to the prediction model generated for each consumer in association with the consumer ID. Note that since the prediction model is generated on a daily basis, information regarding the generation date may also be stored. When predicting the future gas usage for one consumer, the prediction unit 120 may refer to the prediction model table TB7, call the prediction model of the consumer, and perform prediction processing (step S16).

<Customer without installation>
According to an embodiment of the present invention, it may be possible to predict the future gas usage even for customers who do not have a smart meter installed. This will be further described with reference to the flowchart of FIG.

The prediction unit 120 may proceed to step S17 when it is determined in step S10 that one consumer is a non-installed consumer without a smart meter. It should be noted that even if meter reading values are not acquired continuously for the second interval or longer (YES in step S12) for the installation consumer described above, the process may proceed to step S17. In step S17, the prediction unit 120 determines that the third quartile of past daily usage of one consumer is the third quartile of past daily usage of a plurality of other consumers. It may be determined whether it is greater than the largest of the numbers. By this determination, it is possible to extract non-installed customers having tendencies that are significantly different from those of other customers' meter reading data. A customer whose past daily usage amount in the 3rd quartile is greater than the largest of the past daily usage amounts in the 3rd quartile of multiple other customers , for example, a coin laundry, a restaurant, or the like, which corresponds to a consumer who consumes a large amount of gas compared to the average amount of gas used by all consumers.

If the third quartile of the past daily usage of one consumer is greater than the largest third quartile of the past daily usage of a plurality of other consumers If so (YES in step S17), the prediction unit 120 may linearly interpolate the daily usage of the one consumer with the third quartile of past daily usage (step S18). In addition, the prediction unit 120 may set a TQ model in which the predicted daily usage of the consumer is the third quartile of past daily usage as the prediction model (step S19). The prediction model generated by the prediction unit 120 in step S19 may be stored in the prediction model DB 213 (step S15).

Since the 3rd quartile is a relatively large value in the daily gas usage, the kNN interpolation algorithm described later extracts similar installed customers for customers with high gas usage. becomes unnecessary. Therefore, the prediction accuracy can be improved more than using kNN interpolation, which will be described later. That is, the possibility of running out of gas can be reduced.

In addition, the third quartile of past daily usage of one consumer is higher than the largest third quartile of past daily usage of a plurality of other consumers. If determined to be smaller (NO in step S17), the process may proceed to step S22. In step S22, the prediction unit 120 may interpolate the data by a k-Nearest Neighbor (kNN) algorithm. Using the kNN algorithm, gas consumption for a certain period of time can be determined from two consecutive meter readings of customers with low meter reading frequency. Therefore, it is possible to extract K installed consumers who are close to the amount of gas used for a certain period of time, and perform demand forecasting and supplementation for consumers with low meter reading frequency by linearly combining them.

That is, the prediction unit 120 extracts a plurality of other consumers whose gas usage trends are similar to the gas usage trends of one consumer, and uses information about the gas usage of the plurality of other consumers. , the data may be complemented (step S22). Further, a data set obtained by linearly combining the information on the gas usage of the plurality of other consumers is used as a data set on the gas usage over a predetermined period in the past to predict the future usage of one consumer. You may set the prediction model which calculates (step S23). Thus, according to an embodiment of the present invention, even non-installation consumers can supplement data and generate a prediction model. Note that the prediction model generated by the prediction unit 120 in step S23 may be stored in the prediction model DB 213 (step S15).

In addition, in step S14, for example, if the prediction model cannot be generated due to learning failure or the like, the process proceeds to step S19, and the TQ model may be set as the prediction model. Also, a phenomenon may occur in which the set prediction model does not work well, for example, depending on the model generated in steps S19 and S23, the prediction accuracy may not be obtained. In this case, as the prediction model, an overall average model may be set in which the average usage amount of all consumers is used as a prediction value.

<Prediction processing>
Next, the prediction processing (step S16) in one embodiment of the present invention will be described. FIG. 15 is a schematic diagram explaining an overview of the prediction process. Moreover, FIG. 16 is an example of the flowchart regarding a prediction process.

The prediction unit 120 may call the prediction model of the consumer who executes prediction from the prediction model DB 213 (step T20). Next, the prediction unit 120 may use the called prediction model to calculate the predicted usage for L days from the prediction execution date (step T21). Note that the prediction unit 120 may use the past usage amount and the predicted usage amount predicted by the prediction model to calculate the predicted usage amount over a predetermined future period.

Description will be made with reference to FIG. The forecast model is f, the forecast execution date is t, the daily gas usage (actual value) is x _t , the daily gas usage (predicted value) is x ^* _t , the period to be input to the forecast model (predetermined past Assuming that the period of time is K and the period of prediction is L, the daily gas usage amount (predicted value) on the prediction execution date t is expressed by the following formula.

That is, in FIG. 15, the daily gas usage amount (predicted value) on the prediction execution date “2/2” is the most recent K days “…, 1/29, 1/30, 1/31, 2/1”. It is calculated by inputting the daily gas usage (actual value) of

However, the daily gas usage (predicted value) on the day t+1, one day after the prediction execution date t, is given by the following formula.

That is, in FIG. 15, the daily unit gas usage (predicted value) of "2/3" one day after the prediction execution date "2/2" is the most recent K days "..., 1/29, 1/30, It is calculated by inputting the daily gas usage amount (actual value) of "1/31, 2/1" and the daily gas usage amount (predicted value) of "2/2" into the prediction model.

Furthermore, the daily gas consumption (predicted value) on the day t+2, two days after the prediction execution date t, is expressed by the following formula.

That is, in FIG. 15, the daily gas consumption (predicted value) of "2/4" two days after the prediction execution date "2/2" is the most recent K days "..., 1/29, 1/30, By inputting the daily gas usage amount (actual value) of 1/31, 2/1 and the daily gas usage amount (predicted value) of 2/2, 2/3 into the prediction model Calculated.

As described above, the numerical values input to the prediction model include prediction values, so the extraction of essential customers and semi-essential customers may depend on the accuracy of the forecast model. Therefore, according to one embodiment of the present invention, the probability that the remaining amount of gas falls below the threshold (first threshold, second threshold) (hereinafter also referred to as "risk of running out of gas") ), and based on the risk of running out of gas, essential customers and semi-essential customers may be extracted.

A detailed description will be given below. The group generation unit 130 calculates the prediction accuracy of the prediction model based on the past predicted usage amount, which is the predicted usage amount at the past point in time, calculated using the prediction model, and the actual gas usage amount at the past point in time. Information may be calculated. That is, the group generation unit 130 may obtain the accuracy of the prediction model from the difference between the output value obtained by inputting the past data set into the prediction model and the actual measurement value.

First, assuming that the predicted distribution f _i of gas usage X _i on day i follows a normal distribution, and the variance of the point estimation model uses the predicted value of the learning data and the unbiased variance of the correct value, the following is represented by the formula

If the predicted distribution g _n of the total gas usage X for 1 to n days independently follows the same distribution, g _n follows the following normal distribution due to the reproducibility of the normal distribution. In addition, when they do not independently follow the same distribution, the display of the parameters of g _n will be different.

Based on the above assumptions, for consumer h∈H, if the current remaining gas amount is z and the remaining gas amount at gas container α% is z _α , the remaining gas amount at 24:00 after n days is A risk function that returns a probability of less than α% can be defined by the following equation.

Using the risk function described above, the group generation unit 130 may calculate the probability that the remaining amount of gas will be equal to or less than a predetermined threshold, taking into consideration the prediction accuracy information (step T22). Further, the group generation unit 130 may calculate the date when the customer becomes an essential customer or a semi-essential customer based on the calculated probability (step T23).

Steps T22 and T23 will be described with reference to the drawings. FIG. 17(a) is a schematic diagram plotting a risk function indicating the probability that the remaining amount of gas for one consumer will be equal to or less than a predetermined threshold, with the horizontal axis representing the number of days from the predicted date. Group generation unit 130 calculates a first probability that the remaining amount of gas is equal to or less than a first threshold value, which is determined as an essential consumer, and determines the date on which the first probability is equal to or greater than a first reference value for the first time. Good as conversion date. In FIG. 17(a), the solid line represents the transition of the first probability when the first threshold value for extracting essential consumers is set to a value equivalent to 5% of the gas container. Here, when the predetermined reference value is 0.3, the day when the first probability becomes 0.3 or more for the first time is "t12 days later" from the predicted date. Therefore, one consumer is determined to be an essential consumer "after t12 days" from the predicted date.

In addition, the group generation unit 130 further calculates a second probability that the remaining amount of gas is equal to or less than the second threshold value, which is determined as a semi-essential consumer, and the second probability first becomes equal to or greater than the second reference value. day may be the second conversion date. The dashed line in FIG. 17(a) represents the transition of the second probability when the second threshold value for extracting semi-essential consumers is set to a value equivalent to 7% of the gas container. Here, if the second reference value is set to 0.3 like the first reference value, the day when the second probability first becomes 0.3 or more is "t11 days later" from the predicted date. Therefore, one consumer is determined to be a semi-essential consumer "after t11 days" from the forecast date.

For example, if two consumers have the same predicted remaining amount of gas as a result of prediction using their respective prediction models, the consumer whose prediction accuracy is poorer can be quickly extracted as an essential consumer, is preferable from the viewpoint of preventing running out of gas. According to one embodiment of the present invention, a risk function that takes into account the prediction accuracy is introduced, and the higher the risk (the higher the probability of running out of gas) is output, the more the customer for whom the prediction model with the lower accuracy is set. be. Therefore, the possibility of running out of gas is reduced, and more efficient gas delivery becomes possible.

Note that the first conversion date may be calculated closer to the prediction execution date as the predetermined reference value becomes smaller. This will be described with reference to FIG. 17(b). In FIG. 17(b), when the first reference value is set from "0.3" to "0.2", the day when the first probability first becomes 0.2 or more is "t22 days later" from the prediction date. Therefore, it is faster than when the first reference value is "0.3". Accordingly, the date on which one customer is determined to be an essential customer is also earlier than when the first reference value is "0.3". In this way, it is possible to set the date to be extracted as an essential customer according to the setting of the first reference value. This is suitable for the following cases.

FIG. 18 is a schematic diagram plotting risk functions for consumers different from the consumers in FIG. 17, with the horizontal axis representing the number of days from the prediction date. In FIG. 18, the solid line indicates the transition of the first probability when the first threshold value for extracting the essential consumer is set to a value equivalent to 5% of the gas container, and the dashed line indicates the second probability for extracting the semi-essential consumer. It shows the transition of the second probability when the threshold is set to a value equivalent to 7% of the gas container. Here, in FIG. 18, N1 to N3 mean delivery dates of gas containers. In FIG. 18A, when the first and second reference values are "0.3", the customer is not determined to be a semi-essential customer or an essential customer on delivery dates N1 and N2, and gas is not delivered. . “t31 days after the forecast date”, the second probability is less than 0.3, it is determined to be a semi-essential customer, and “t32 days later” from the forecast date, the first probability is less than 0.3, and it is an essential customer. be judged. Therefore, at the time of the next delivery date N3 after "t32 days" from the predicted date, the remaining amount of gas will be less than 5% of the gas container, and the possibility of running out of gas is high.

For the above phenomenon, when the first and second thresholds are set to "0.15", as shown in FIG. Therefore, it is earlier than "t31 days later" when the first and second thresholds are "0.3". Similarly, the day when the customer is determined to be an essential consumer is "t42 days" from the predicted date, which is earlier than "t32 days" when the first and second thresholds are "0.3". As a result, at the delivery date N2, the consumer is determined as an essential consumer and can be extracted as a gas delivery destination.

That is, according to one embodiment of the present invention, it is possible to quickly extract a customer who is likely to run out of gas and deliver gas. Therefore, it is possible to reduce the possibility of running out of gas and provide a delivery system with good usability.

Returning to the flow chart of FIG. The group generation unit 130 may determine whether or not the processing of steps T20 to T23 has been performed for all consumers (step T24). If not executed for all consumers (NO in step T24), steps T20 to T23 may be executed for the next consumer. If all consumers have been processed (YES in step T24), the trip splitting process may be executed for the extracted essential and semi-essential customers (step T25).

<Trip division processing>
The group generation unit 130 may group essential customers and semi-essential customers (trip division) based on predetermined conditions regarding delivery. Each algorithm for trip division will be described with reference to FIGS.

<Trip division algorithm>
The trip splitting algorithm may aim to roughly create efficient trips when considering delivering only to essential customers. This will be explained using FIG. If there is a mandatory consumer group CON10 as shown in FIG. 19(a), a trip is created by gathering together the mandatory consumer groups with close delivery destinations, such as trips TR1 and TR2 as shown in FIG. 19(b). is preferred. On the other hand, trips TR3 and TR4 as shown in FIG. 19(c) are inefficient because the movement route between essential consumers is long. As a trip division algorithm, an improved version of the NF (Next-Fit) method, which is an existing bin-packing problem algorithm, is used, and an optimization problem method targeting all essential consumers is used to achieve the optimal solution. Alternatively, an approximate solution may be obtained.

<Greedy Algorithm>
The greedy algorithm may aim to incorporate, as much as possible, high-risk semi-essential customers into trips of essential customers divided by the above-described algorithm. The group generation unit 130 sorts the semi-essential customers in descending order of risk, and, for each trip of each essential customer, sorts the semi-essential customers in descending order of risk, and calculates the load weight of the delivery vehicle (truck) and the With the loading space as a constraint, it may be calculated whether or not the delivery vehicle can be loaded.

<Delivery Area Minimization Algorithm>
FIG. 20 is a schematic diagram illustrating an overview of the delivery area minimization algorithm. The delivery area minimization algorithm may be an algorithm aiming at delivery to some semi-essential customers while aiming at efficient delivery by making the delivery area as small as possible. That is, the group generation unit 130 selects a plurality of essential customers and semi-essential customers in one group under at least a constraint condition related to the delivery capacity of gas cylinder delivery vehicles as a predetermined condition for extracting semi-essential customers. Semi-essential customers to be included in one group may be extracted so that the distribution area of the essential customers is within a predetermined range.

For example, as shown in FIG. 20, consider a case where a customer group CON11 includes an essential customer C1 and semi-essential customers P1 to P6. Note that in the figure, the darker color indicates the consumer with a higher risk of running out of gas as described above. When simply delivering to a customer with a high degree of risk, delivery to an area including essential customer C1 and semi-essential customers P1 and P6 can be considered. Long and inefficient. According to the delivery area minimization algorithm, a rectangular area that partially includes semi-essential customers while including the essential customer C1 is defined as the delivery area AR10, and gas is delivered to the customers within the delivery area AR10.

The delivery area AR10 may be calculated by solving a mathematical optimization model that formulates constraints. For example, the optimization model is based on the formulation of minimizing the sum of the short side and long side of the delivery area AR10 with the constraint conditions of delivery to the essential customer, and the loading weight and loading space of the delivery vehicle. can be set. The data on semi-essential customers and essential customers can be read from the processing results of steps T20 to T24 described above. Also, the loading weight and loading space of the delivery vehicle (truck) can be read from the delivery vehicle information table TB3. Furthermore, the weight and number of each gas container of each consumer and the location information of each consumer can be read from the consumer information table TB1. The group generation unit 130 may obtain an optimal solution or an approximate solution from the set optimization model.

According to one embodiment of the present invention, a rectangular area including essential customers is set as a delivery area, and quasi-essential customers in the delivery area are extracted. can be realized.

<(multi-day) delivery area minimization algorithm>
A (multi-day) delivery area minimization algorithm may consider the number of required customers on a delivery day and aim to even out the delivery area across delivery days. Description will be made with reference to FIG. FIG. 21(a) assumes that a delivery area AR41 including essential customers is set as the delivery area for the first delivery date in the consumer group CON12. After that, the semi-essential customer P1 who was not delivered on the first delivery date becomes the essential customer C1, and in order to include other essential customers, the delivery area AR42 is set as the delivery area on the second delivery date. Become. In this case, the delivery area AR42 on the second delivery date becomes large.

On the other hand, according to the multi-day delivery area minimization algorithm, as shown in FIG. A delivery area AR43 including essential consumer P1 is set and gas is delivered. Therefore, a delivery area AR52 smaller than the delivery area AR42 is set as the delivery area for the second delivery date. By minimizing the delivery area for multiple days in this way, it is possible to level the delivery area over the entire delivery days and provide an efficient delivery system.

<Acquisition Risk Maximization Algorithm>
FIG. 22 is a schematic diagram illustrating an overview of the acquisition risk maximization algorithm. The acquisition risk maximization algorithm may be an algorithm aiming at shortening the movement route between essential customers and delivering if there is a semi-essential customer on the way between essential customers. That is, the group generating unit 130, as a predetermined condition, at least, when the group generating unit groups the essential consumers under the constraint condition regarding the delivery capacity of the delivery vehicle of the gas cylinder, The located semi-essential customers may be extracted as semi-essential customers to be included in one group.

For example, as shown in FIG. 22, consider a case where essential customers C1 to C4 and semi-essential customers P1 to P5 exist in the customer group CON13. Considering delivery to essential consumers C1 to C4, elliptical delivery areas AR20 to AR23 can be set. At that time, the semi-essential customers P2, P3, and P4, which are included in the delivery areas AR20 to AR22 and located on the movement route between the essential customers, may be set as delivery destinations. However, semi-essential customer P3 may be excluded from the delivery destination due to constraints on the loading weight and loading space of the delivery vehicle. That is, in the customer group CON13, the semi-essential customers P1 and P5 outside the delivery area and the semi-essential customer P3, which is within the delivery area, do not have to be included in the delivery destination due to the limitations of the delivery vehicles. .

It is preferable that the semi-essential customers included in the delivery areas AR20 to AR23 (AR24) are customers who have a high degree of risk and are highly likely to run out of gas. Therefore, by making sure to deliver to the essential customers, and by limiting the loading weight and loading space of the delivery vehicle, and maximizing the risk of the customers included in each of the delivery areas AR20 to AR23, , may set the optimization model. The group generation unit 130 may obtain an optimal solution or an approximate solution from the set optimization model.

According to one embodiment of the present invention, semi-essential customers within the delivery area of the essential customer are extracted, so an efficient delivery system can be realized.

<Delivery route settings>
Next, delivery route setting processing according to one embodiment of the present invention will be described. FIG. 23 shows an outline of the delivery route setting process. By the delivery route setting process, the trips TP10 and TP11 calculated by the delivery list setting process shown in FIG. 23(a) are input, and as shown in FIG. Information regarding appropriate delivery routes to semi-essential customers may be output. FIG. 24 shows an example of the table TB20 regarding delivery routes. As shown, information about delivery routes may include information about delivery times to each customer to be delivered.

As shown in FIG. 5, the delivery route setting step 30 is performed by an algorithm based on an integer program (IP) or an algorithm based on a traveling salesman problem (TSP). , a delivery route within the group generated by the group generation unit 130 may be set.

First, an algorithm based on the integer programming problem will be described with reference to FIG. FIG. 25 is a flow chart of delivery route setting processing by an algorithm based on an integer programming problem.

First, a delivery list related to consumers to be delivered, generated by the group generation unit 130, is acquired. The delivery list may be information about customers to be delivered on the next delivery date, for example the next day. As noted above, the distribution list may include information on essential and semi-essential customers included in each trip. The group generation unit 130 can refer to the customer information table TB1 from the customer information included in the delivery list to obtain information about the location (delivery location) of the gas cylinder of each customer.

First, for one group generated by the group generation unit 130, the delivery order setting unit 140 targets only the essential customers included in one group, and selects the essential customers by the known traveling salesman problem (TSP). You may set the delivery order to circulate (step P10). The Traveling Salesman Problem is, given a set of points and a travel cost (distance, time, etc.) between each two points, in a circuit that visits all points exactly once and returns to the starting point. , is a type of combinatorial optimization problem that seeks the route with the lowest total travel cost. As a result, a rough solution (approximate solution) of the delivery order can be obtained, and input data for other algorithms can be obtained.

<Maximize the number of delivery cylinders>
If a solution based on the traveling salesman problem is found in step P10, the process proceeds to step P11, and the delivery order setting unit 140 may obtain a delivery order that maximizes the number of delivery cylinders. That is, the delivery order setting unit 140 always delivers to the essential consumers included in one group generated by the group generation unit 130 under at least the constraint conditions related to the working hours of the delivery staff, and then to the one group. The delivery order may be set based on a first mathematical model that maximizes the number of gas containers delivered to the involved semi-essential customers. As a constraint, a condition regarding delivery time specified by the consumer may be added.

The delivery order setting unit 140 may, for example, acquire from a predetermined application a function that gives the travel time between consumers included in one trip or between a distribution center and a consumer. Furthermore, the delivery order setting unit 140 may acquire the delivery staff's working hours, the customer's specified delivery time, and the like from the consumer information DB 221 and the delivery staff information DB 222 .

As a first mathematical model, the delivery order setting unit 140 is designed to maximize the number of gas cylinders to be exchanged (delivered). A time-related condition may be formulated as a constraint and an optimal or approximate solution may be obtained by a mixed integer optimization problem approach.

<Maximize the number of delivery cylinders only for essential customers>
The step P15 of maximizing the number of delivery cylinders only for essential consumers will be explained. The case where step P15 is applied means that, in steps P10, P13, and P14, the restrictions on the working hours of the delivery staff (for example, the working hours of the delivery staff included in the delivery staff information table TB2 in FIG. 3) are satisfied. It may be the case that no solution is obtained. In this case, the purpose of step P15 may be to deliver to the essential consumers as much as possible within the upper limit of the overtime hours. That is, when it is impossible to set a delivery order that satisfies the constraints on the working hours of the delivery staff, the delivery order setting unit 140 delivers the gas cylinders to the semi-essential consumers included in one group regardless of the constraints. The delivery order may be set to the essential customers that maximize the number of gas cylinders to be delivered within the predetermined upper limit of working hours of the delivery staff, among the essential customers that are not delivered and are included in one group.

<Minimization of working hours>
The algorithm for minimizing working hours is used when finding a delivery route to a customer included in a trip, as described above, when a solution that satisfies the constraints on the working hours of delivery workers cannot be obtained (overtime work occurs). may be executed). Therefore, minimizing working hours can be considered synonymous with minimizing the overtime hours of delivery personnel. In order to minimize working hours, it is sufficient to find a solution to a mathematical model aimed at minimizing the time it takes to go to a delivery destination.

For example, the delivery order setting unit 140 may obtain a delivery order that minimizes working hours for only essential customers (step P13). That is, if the delivery order setting unit 140 cannot set the delivery order for delivery to the semi-essential customers included in one group generated by the group generation unit 130 under the constraint conditions regarding the working hours of the delivery staff, The delivery order may be set based on a second mathematical model that minimizes the delivery time to the essential customers included in the group of .

Alternatively, the delivery order setting unit 140 may obtain a delivery order that minimizes working hours in addition to semi-essential customers (step P12).

If the optimum solution cannot be obtained by steps P13 and P14, even if the above-described maximization of the number of delivery cylinders is executed again for only the essential customers among the customers included in the delivery list, good. In other words, when the delivery order cannot be set due to constraints on the working hours of the delivery staff, the delivery order setting unit 140 delivers the gas cylinders to the semi-essential consumers included in one group regardless of the constraints. Instead, the delivery order may be set so as to deliver to the essential consumers included in one group within the predetermined upper limit of working hours of the delivery person.

<Thinning out only essential TSP output>
If the overtime work limit is exceeded even if only the essential customers are delivered, this algorithm will continue to deliver the low-risk demand among the essential customers until the time to complete the delivery is earlier than the overtime work maximum time. A house may be deleted from a trip. In other words, the delivery order setting unit 140 determines that if the delivery to the essential consumers included in the group generated by the group generation unit 130 exceeds the predetermined upper limit of working hours of the delivery person, the above-described first probabilities are determined in descending order. , the essential customer may be deleted from the group, and a delivery order that allows delivery within a predetermined upper limit of working hours may be set. In addition, the predetermined upper limit of working hours of the delivery person may refer to the upper limit of hours that the delivery worker can work as overtime.

Note that the flowchart shown in FIG. 25 is an example, and the order and position of each algorithm may be changed.

Next, a delivery route setting process using an algorithm based on the traveling salesman problem will be described with reference to FIG. In the flowchart of FIG. 26, the same reference numerals are assigned to the same items as in FIG. 25, and the description thereof is omitted.

<TSP with time constraint (TSP-TW)>
The time-constrained TSP algorithm (step P20) refers to a problem in which the arrival time of each point is constrained in TSP. That is, the delivery order setting unit 140 solves the time-constrained traveling salesman problem (TSP-TW ) may be set. An existing method such as ant colony optimization (ACO) may be used as an approximate solution method for TSP with time constraints. As a result, it becomes possible to correspond to the specification of the delivery time of each customer in the delivery route setting.

<Edit delivery list>
In the delivery list correction algorithm (step P21), in the case of the optimal solution by the time-constrained TSP algorithm, if the delivery worker's prescribed upper limit of working hours is exceeded, it is considered which consumer should be removed from the delivery list. can be a problem. That is, the delivery order setting unit 140 selects the semi-essential customers included in one group when delivery is impossible within the upper limit of working hours of the delivery staff depending on the delivery order set by the time-constrained TSP. , the delivery order may be updated by thinning out the semi-essential customers that contribute to the reduction of the delivery time.

The delivery order setting unit 140 may, for example, calculate the time that can be saved when semi-essential customers are reduced one by one from the delivery list. Note that the time that can be reduced may include the travel time between consumers and the work time required to replace the gas container. Alternatively, the delivery order setting unit 140 may calculate the time that can be saved when the number of customers to be delivered consecutively is reduced by two, which is output as the optimal solution of the time-constrained TSP from the delivery list. good.

As described above, according to one embodiment of the present invention, a delivery route to a customer to be delivered is calculated in consideration of the delivery staff's working hours. Therefore, for the gas supply service provider, it is possible to provide a cost effective and efficient delivery service.

Information about the delivery route may be provided to the delivery person. That is, the communication unit 110 may transmit information regarding the delivery order set by the delivery order setting unit 140 to the communication terminal 510 of the delivery person 500 . In the communication terminal 510 of the delivery person 500, the delivery order may be superimposed and displayed on a map, for example. The information about the delivery order includes information about the delivery time, and may be displayed on the communication terminal 510 on a map or as text information. As a result, the delivery staff 500 can carry out the delivery work while confirming the order of delivery.

<Hardware configuration>
A hardware configuration of the server 100 will be described. FIG. 22 is a hardware configuration example of a computer that can implement the server 100 in this embodiment. The server 100 includes a processor 101 , a storage 102 , a memory 103 , an input/output interface (input/output I/F) 104 and a communication interface (communication I/F) 105 . Each component is interconnected via a bus B. FIG. The server 100 implements the functions and methods described in this embodiment through cooperation of these components. For example, when each functional unit of the server 100 is implemented by software, it is implemented by the processor 101 executing instructions included in a program read from the storage 102 to the memory 103 . That is, the server 100 according to this embodiment functions as the communication unit 110 , the prediction unit 120 , the group generation unit 130 and the delivery order setting unit 140 by executing the program read into the memory 103 by the processor 101 .

The processor 101 is, for example, a central processing unit (CPU), MPU (Micro Processing Unit), GPU (Graphics Processing Unit), microprocessor, processor core, multiprocessor, ASIC (Application- Specific Integrated Circuit), FPGA (Field Programmable Gate Array), etc., and are realized by logic circuits (hardware) and dedicated circuits formed in integrated circuits (IC (Integrated Circuit) chip, LSI (Large Scale Integration)), etc. you can The server 100 preferably has a processor 101 with high computing power for processing the large amount of data described above.

The communication I/F 105 is implemented as hardware such as a network adapter, communication software, or a combination thereof, and transmits and receives various data to and from an external device. The communication may be performed by wire or wirelessly, and any communication protocol may be used as long as mutual communication can be performed.

The input/output I/F 104 includes an input device for inputting various operations to the server 100 and an output device for outputting processing results processed by the server 100 . The input device includes, for example, hardware keys such as a touch panel, a touch display, and a keyboard, a pointing device such as a mouse, a camera (operation input via images), and a microphone (operation input by voice). The output device outputs the processing result processed by the processor 101 . The output device includes, for example, a touch panel, a speaker, and the like.

Although the present invention has been described with reference to the drawings and examples, it should be noted that various variations and modifications will be readily apparent to those skilled in the art based on this disclosure. Therefore, it should be noted that these variations and modifications are included in the scope of the present invention. For example, the functions included in each component, each step, etc. can be rearranged so as not to be logically inconsistent, and multiple components, steps, etc. can be combined into one or divided. is. Further, the configurations shown in the above embodiments may be combined as appropriate. For example, each component described as being included in the server 100 may be distributed and implemented by a plurality of servers.

In addition, generally, in an apartment complex or a building, a plurality of gas containers are collectively installed in the entire apartment complex or building, and gas is shared with each consumer. That is, multiple consumers share multiple gas cylinders. A meter is installed for each consumer, and the consumption of each consumer is read. Therefore, when a gas cylinder is shared, a prediction model for the amount of gas used by each customer is calculated, and the predicted amount of gas used by each customer is added up to predict the remaining amount of gas in multiple shared gas cylinders. you can That is, when information about one gas container is associated with a plurality of consumers, the group generation unit 130 sums the predicted usage amounts predicted for the plurality of consumers, and calculates the above-described first threshold value. can be calculated. As a result, even when a gas container is shared by a plurality of consumers, it is possible to calculate an appropriate replacement timing.

Information on roads may also be used in setting the delivery order. For example, whether the road is two lane or one lane, road construction, parking lot locations, etc., can affect the time it takes to deliver a gas canister. Therefore, it affects the determination of whether or not the delivery staff's working hours are exceeded. Therefore, information about roads may be formulated as a constraint when setting the delivery order.

The program of each embodiment of the present disclosure may be provided in a state stored in a storage medium readable by the information processing device. The storage medium can store the program in a "non-temporary tangible medium". Programs include, for example, software programs and information processing device programs. When each functional unit of the server 100 as an information processing device is realized by software, the server 100 executes a program loaded on the memory by the processor, so that the communication unit 110, the prediction unit 120, the group generation unit 130, and the It functions as the delivery order setting unit 140 .

The storage medium may, where appropriate, be one or more semiconductor-based or other integrated circuits (ICs) (e.g., field programmable gate arrays (FPGAs), application specific ICs (ASICs), etc.); Disk drive (HDD), hybrid hard drive (HHD), optical disk, optical disk drive (ODD), magneto-optical disk, magneto-optical drive, floppy diskette, floppy disk drive (FDD), magnetic tape, solid state drive (SSD), RAM drive, secure digital card or drive, any other suitable storage medium, or any suitable combination of two or more thereof. Storage media may, where appropriate, be volatile, nonvolatile, or a combination of volatile and nonvolatile.

Also, the program of the present disclosure may be provided to server 100 via any transmission medium (communication network, broadcast wave, etc.) capable of transmitting the program.

Embodiments of the present disclosure may also be implemented in the form of a data signal embedded in a carrier wave in which the program is embodied by electronic transmission.

Note that the program of the present disclosure may be implemented using, for example, script languages such as JavaScript (registered trademark) and Python, C language, Go language, Swift, Koltin, Java (registered trademark), and the like.

100 server (arithmetic processing server)
110 communication unit 120 prediction unit (remaining amount prediction engine)
130 group generator (delivery list setting engine)
140 delivery order setting unit (delivery route setting engine)
210 Management server 212 Meter reading data DB (database)
213 Prediction Model DB
220 Business DB
221 Customer information DB
222 Delivery person information DB
223 Delivery vehicle information DB
230 External database
231 weather information database
232 Road Information DB
310 Communication Device 320, 420 Meter 330A, 330B, 430A, 430B Gas Container 440 Meter Reader 441 Communication Terminal 500 Delivery Person 510 Communication Terminal 800 Delivery System

Claims (27)

An information processing device in a distribution system for gas distribution to consumers,
an acquisition unit that acquires information about the amount of gas used by the consumer;
a prediction unit that calculates a predicted future gas usage from past gas usage;
a group generation unit capable of generating a group including essential customers with the highest gas delivery priority and semi-essential customers with a lower gas delivery priority than the essential customers, based on the predicted usage;
a delivery order setting unit that sets the delivery order of gas in the group;
with
The group generation unit
generating the group for the delivery date based on a first conversion date on which the customer becomes the essential customer and a second conversion date on which the customer becomes the semi-essential customer;
Information processing equipment. The group generation unit
generating the group including the essential customers and including the semi-essential customers when the number of the essential customers on the delivery date is equal to or less than the maximum number of deliveries;
The information processing device according to claim 1 . The prediction unit generates a prediction model by machine learning using linear regression or nonlinear regression based on at least a data set related to gas usage over a predetermined period in the past for each consumer, and based on the prediction model to calculate the predicted usage over a predetermined period in the future for each of the consumers;
The information processing apparatus according to claim 1 or 2. The prediction unit uses the amount of gas used on the first day in the past as an objective variable, the amount of gas used over a predetermined period prior to the first day in the past, and information about the weather on at least the first day in the past. generating the predictive model by the machine learning as an explanatory variable;
The information processing apparatus according to claim 3. The prediction unit uses the past usage amount and the predicted usage amount predicted by the prediction model to calculate the predicted usage amount for a predetermined future period,
The group generation unit calculates the prediction model based on the past predicted usage amount, which is the predicted usage amount at the past point in time, and the actual gas usage amount at the past point in time, calculated using the prediction model. calculating prediction accuracy information, calculating a first probability that the remaining amount of gas is equal to or less than a first threshold value based on the prediction accuracy information and the predicted future usage amount, and calculating the first probability that the remaining amount of gas is equal to or less than a first threshold The first conversion date is the day when the reference value or more is reached,
The information processing apparatus according to claim 3 or 4. In the group generation unit, the first conversion date is calculated closer to the prediction execution date as the first reference value becomes smaller.
The information processing device according to claim 5 . The group generation unit further calculates a second probability that the remaining amount of gas is equal to or less than a second threshold value, sets a date that the second probability is equal to or greater than a second reference value for the first time as the second conversion date, and The second conversion date is calculated closer to the date of execution of the prediction as the second reference value decreases,
The information processing apparatus according to claim 5 or 6. The acquisition unit acquires information about the usage amount of the gas at a first interval,
When the interval at which the information on the continuous gas usage amount of one consumer is acquired has an interval equal to or greater than the first interval and less than the second interval, the prediction unit uses a predetermined algorithm to determine the first Complementing the information about the amount of gas usage in an interval of one interval or more and less than the second interval to generate a data set about the amount of gas usage over the past predetermined period;
The information processing device according to any one of claims 3 to 7. When the interval at which the information on the continuous gas usage amount of one consumer is acquired is equal to or greater than the second interval, the prediction unit detects that the gas usage trend of the one consumer is A data set obtained by extracting a plurality of other consumers similar to gas usage trends and linearly combining information on the gas usage of the plurality of other consumers, is obtained as data on the gas usage over the past predetermined period. using it as a set to calculate the future predicted usage of the one consumer;
The information processing apparatus according to claim 8 . The prediction unit predicts that there is a section in which the interval at which the information on the continuous gas usage amount of one consumer is acquired is equal to or greater than the second interval, and the daily usage amount of the one consumer is If the third quartile is greater than the largest third quartile of the daily usage of a plurality of other consumers, the amount of gas of the one consumer over a past predetermined period generating a usage data set at the third quartile of daily usage of the one consumer;
The information processing apparatus according to claim 8 or 9. The prediction unit uses the prediction model as Linear Regression, Support Vector Regression, Random Forest regression, Gradient Boosting Regression, or Gradient Boosting Determination. Generated by an algorithm using a tree (Gradient Boosting Decision Tree),
The information processing device according to any one of claims 3 to 10. The group generating unit generates the one gas container so that the delivery areas of the essential customers and the semi-essential customers in one group are within a predetermined range under at least a constraint condition related to the delivery capacity of the delivery vehicles for the gas cylinders. Extracting the semi-essential consumers to be included in the group of
The information processing apparatus according to any one of claims 1 to 11. The group generation unit, at least, when the group generation unit groups the essential customers under a constraint condition related to the delivery capacity of the delivery vehicle for the gas cylinder, the semi-essential customers located in the delivery area of one group. extracting the customer as a semi-essential customer to be included in the one group;
The information processing apparatus according to any one of claims 1 to 12. The delivery order set by the delivery order setting unit is
For one group generated by the group generation unit, only the essential customers included in the one group are targeted, and a delivery order that tours the essential customers by the traveling salesman problem (TSP) is included,
The information processing apparatus according to any one of claims 1 to 13. The delivery order set by the delivery order setting unit is
Under the constraint condition at least regarding the working hours of the delivery staff, delivery is always made to essential customers included in one group generated by the group creation unit, and then delivery to semi-essential customers included in the one group. including a delivery order set based on a first mathematical model that maximizes the number of gas containers that
The information processing apparatus according to any one of claims 1 to 14. The delivery order set by the delivery order setting unit is
Under the constraint on the working hours of the delivery staff, if the delivery order for delivery to the semi-essential customers included in the one group generated by the group generation unit cannot be set, the essential customers included in the one group Including a delivery order set based on a second mathematical model that minimizes the delivery time to deliver to
The information processing apparatus according to any one of claims 1 to 15. The delivery order set by the delivery order setting unit is
Under the constraint condition at least regarding the working hours of the delivery staff, the gas is delivered to the essential consumers included in the one group generated by the group generation unit without fail, and then to the semi-essential consumers included in the one group. Among the delivery orders that do not deliver containers, including the delivery order set based on the third mathematical model that minimizes the delivery distance,
The information processing apparatus according to any one of claims 1 to 16. The delivery order set by the delivery order setting unit is
gas to the semi-essential consumers included in the one group regardless of the constraint conditions when the delivery order based on the first to third mathematical models cannot be set due to the constraint conditions regarding the working hours of the delivery staff; Including a delivery order set so as not to deliver the container but to deliver it to the essential consumers included in the one group within a predetermined upper limit of working hours of the delivery person,
The information processing apparatus according to claim 17. The delivery order set by the delivery order setting unit is
gas to the semi-essential consumers included in the one group regardless of the constraint conditions when the delivery order based on the first to third mathematical models cannot be set due to the constraint conditions regarding the working hours of the delivery staff; including an order of delivery to essential consumers who do not deliver containers and maximize the number of gas containers to be delivered within a predetermined upper limit of working hours of the delivery staff, among the essential consumers included in the one group;
The information processing apparatus according to claim 17. The delivery order setting unit, when delivering to the required consumers included in the group generated by the group generation unit, exceeds a predetermined upper limit of working hours of the delivery person, the group delete the essential consumer from the above, and set a delivery order that can be delivered within the predetermined upper limit of working hours;
The information processing device according to claim 5 . The delivery order set by the delivery order setting unit is
including a delivery order set by a time-constrained traveling salesman problem (TSP-TW) according to delivery times specified by essential customers and semi-essential customers included in the group generated by the group generation unit; ,
The information processing apparatus according to any one of claims 1 to 20. According to the delivery order set by the time-constrained traveling salesman problem, the delivery order setting unit selects the semi-essential demands included in the one group when delivery is impossible within the upper limit of working hours of the delivery person. Update the delivery order by thinning out the semi-essential customers that contribute to the reduction of delivery time,
The information processing apparatus according to claim 21. When information about one gas container is associated with a plurality of consumers, the group generation unit calculates the first threshold by summing the predicted usage amounts predicted for the plurality of consumers. do,
The information processing device according to claim 5 . further comprising a communication unit that transmits information about the delivery order set by the delivery order setting unit to a communication terminal of a delivery person;
The information processing device according to any one of claims 1 to 23. An information processing device in a distribution system related to gas distribution to consumers,
obtaining information about the amount of gas used by the consumer;
calculating a predicted future gas usage from past gas usage;
generating a group including essential customers with the highest gas delivery priority and semi-essential customers with a lower gas delivery priority than the essential customers based on the predicted usage;
setting the order of gas delivery in the group;
and run
The step of generating the group generates the group on the delivery date based on a first conversion date when the customer becomes the essential customer and a second conversion date when the customer becomes the semi-essential customer. , a control method for an information processing device; a function of acquiring information on the amount of gas used by the consumer to an information processing device in a distribution system related to gas delivery to the consumer;
A function that calculates the predicted future gas usage from the past gas usage,
a function of generating a group including essential customers with the highest gas delivery priority and semi-essential customers with a lower gas delivery priority than the essential customers, based on the predicted usage;
the ability to set the order of gas delivery in the group;
to realize
The group generating function generates the group on a delivery date based on a first conversion date when a customer becomes the essential customer and a second conversion date when a customer becomes the semi-essential customer. , a control program for an information processing device. A delivery system for delivering gas to consumers,
an acquisition unit that acquires information about the amount of gas used by the consumer;
a prediction unit that calculates a predicted future gas usage from past gas usage;
Based on the predicted usage amount, a group can be generated for the consumers that includes essential consumers with the highest gas delivery priority and semi-essential consumers with a lower gas delivery priority than the essential consumers. a group generator;
a delivery order setting unit that sets the delivery order of gas in the group;
with
The group generation unit generates the group on a delivery date based on a first conversion date when a customer becomes the essential customer and a second conversion date when a customer becomes the semi-essential customer. system.

Images