JPH08217343A - Elevator group management device - Google Patents

Abstract

(57) [Summary] [Objective] A neural network is applied to obtain a group management device for an elevator that operates a plurality of vehicles as a group. [Structure] Control means (3) for a plurality of cars (1), operation control means (4) for controlling a group of cars (1), and call assignment means for deciding the assigned car (1) for a hall call (6) is also provided, and traffic flow discriminating means (8) that collects the generated traffic data and discriminates the traffic flow by applying a neural network is provided, and based on the traffic data of the traffic flow discriminating means (8). In the memory (13) for storing the traffic data, the teacher data and the information of the neural network, a timer (14) for learning the neural network by the teacher data corresponding to the neural network and connecting to the operation control means (4) is provided. Is provided to identify traffic flow
The driving pattern for the traffic flow is selected by the output of (8). [Effect] Different traffic flows are discriminated and the management function for driving multiple cars is improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a group management device for an elevator that manages and operates a plurality of cars as a group, and more particularly to a group management device that discriminates a traffic flow in a building by applying a neural network.

[0002]

2. Description of the Related Art A group management device for controlling a plurality of elevators can realize arithmetic processing of a large amount of information and sophisticated control by adopting a microcomputer.
Against this background, fuzzy logic and neural networks, which are one of the artificial intelligence technologies, have been applied to elevator devices. As a conventional elevator group management device,
For example, as disclosed in Japanese Patent Laid-Open No. 3-124676, there is an elevator group management device including a partial model part of a neural network and a plurality of membership functions.

Further, as disclosed in, for example, Japanese Unexamined Patent Publication No. 3-279178, there is an elevator group management device which constitutes a neural network which outputs a congestion degree in a car with a traffic state as an input. Further, for example, Japanese Patent Laid-Open No. 4-324
As shown in Japanese Patent Laid-Open No. 72, there is an elevator group management device that selects a most suitable neural network by providing a time zone or a traffic pattern.

Further, for example, as disclosed in Japanese Unexamined Patent Publication No. 6-263346, there is an elevator group management device for determining a traffic flow in a building by applying a neural network and re-learning if there is an unlearned traffic flow. is there. In addition, the traffic flow that occurs in the building is a concept that represents the traffic state, the number of passengers,
It is described by a passenger occurrence interval, an OD (Origin and Destination) matrix, and the like.

[0005]

The conventional elevator group management device as described above has the following problems. That is, in the elevator group management device composed of the partial model part of the neural network and the plurality of membership functions, although the neural network is applied to the hall call allocation control, the traffic flow generated in the building The effect is uncertain.

In addition, in the elevator group management device which constitutes a neural network which outputs the congestion degree in the car with the traffic condition as an input, although the neural network is applied to predict the congestion degree in the car, The question remains how to control the elevator from the predicted data. Also, in the elevator group management device that sets the time zone or traffic pattern and selects the most suitable neural network, only the neural network is selected and used for hall call assignment control. It is unclear how to distinguish the different traffic flows that occur.

In addition, in an elevator group management device that determines a traffic flow in a building by applying a neural network and re-learns if there is an unlearned traffic flow, a concrete example of the elevator after the determination of the traffic flow is performed. Control is unknown.

In each of the conventional elevator group management devices described above, it is not clear about the following matters. That is, the learning of the neural network usually requires a huge amount of calculation time, which affects the control of other elevators. However, such a problem has not been solved. Further, teacher data is required in the learning process of the neural network, but the procedure for extracting that data is unclear.

Further, when real-time control such as an elevator is required, the output from the neural network may not appear clearly if the learning of the neural network is insufficient. However, it is unclear how to improve it so that it does not affect the elevator even when the output from the neural network does not appear clearly. Further, when various information is input from the outside of the elevator, it is unclear about the priority relationship between the information and the control information from the neural network.

The present invention has been made in order to solve such a problem, and it is an elevator that efficiently manages and operates a plurality of cars as a group by applying a neural network to determine the traffic flow. The purpose is to obtain a group management device.

[0011]

In an elevator group management system according to the present invention, a control means for controlling the operation of each of a plurality of cars, and an operation control means for managing and controlling all the cars as a group. , Call allocation means for deciding an allocated car based on a predetermined evaluation method for calls generated at the halls corresponding to the cars, and traffic data for collecting traffic data generated in the building where the cars are installed A collection means, a traffic flow discrimination means for discriminating a traffic flow obtained from the traffic data collection means by applying a neural network, and a neural network learning based on the traffic data collected from the traffic flow discrimination means Teacher data extraction means for extracting the teacher data necessary for the operation, learning means for learning the neural network based on the teacher data, and operation control means. A timer that continuously measures time, a memory that stores at least traffic data, teacher data, and information that forms a neural network, and driving that selects a driving pattern corresponding to a traffic flow based on the output from the traffic flow determination means. Pattern selecting means is provided.

Further, in the elevator group management device according to the present invention, the elevator group management device which manages and operates a plurality of cars as a group collects traffic data generated in a building in which the cars are installed. A data collection means, a traffic flow discrimination means for discriminating a traffic flow obtained from the traffic data collection means by applying a neural network, and a neural network based on the traffic data collected from the traffic flow discrimination means. Teacher data extracting means for extracting the teacher data necessary for learning, learning means for learning a neural network based on the teacher data, and traffic flow discriminating means are the first priority computing means, and the learning means is the third.
There is provided task management means for managing the operation of each computing means in accordance with the order in which the computing means of priority is used and the other computing means of the traffic flow discrimination means and the learning means are designated as second priority computing means.

Further, in the elevator group management system according to the present invention, the elevator group management system which manages and operates a plurality of cars as a group collects traffic data generated in the building in which the car is installed. A data collection means, a traffic flow discrimination means for discriminating a traffic flow obtained from the traffic data collection means by applying a neural network, and a neural network based on the traffic data collected from the traffic flow discrimination means. Teacher data extraction means for extracting the teacher data necessary for learning, learning means for learning a neural network based on the teacher data, and a driving pattern corresponding to the traffic flow based on the output from the traffic flow determination means And a memory for storing a plurality of neural nets for discriminating the traffic flow for each driving pattern It is provided.

Further, in the elevator group management device according to the present invention, the elevator group management device which manages and operates a plurality of cars as a group collects traffic data generated in the building in which the car is installed. A data collection means, a traffic flow discrimination means for discriminating a traffic flow obtained by the traffic data collection means by applying a neural network, and a peak of traffic volume from the traffic data collected by the traffic data collection means. A peak time detecting means for detecting a time zone, a teacher data extracting means for extracting continuous traffic data a predetermined time before the peak time detected by the peak time detecting means as teacher data, and a neural network based on the teacher data. Learning means for learning the net is provided.

Further, in the elevator group management device according to the present invention, the first group of the traffic data generated in the building in which the cars are installed is provided to the elevator group management device which manages and operates a plurality of cars as one group. A traffic data collecting means for collecting at a time interval, and a second traffic time shorter than the first time interval for the traffic flow obtained from the traffic data collecting means.
Traffic flow discriminating means for discriminating by applying a neural network at time intervals, and teacher data extracting means for extracting teacher data necessary for learning the neural network based on the traffic data collected by the traffic data collecting means. And learning means for learning the neural network based on the teacher data.

Further, in the elevator group management device according to the present invention, the elevator group management device which manages and operates a plurality of cars as a group collects traffic data generated in a building where the cars are installed. A data collection means, a traffic flow discrimination means for discriminating a traffic flow obtained from the traffic data collection means by applying a neural network, and a neural network learning based on the traffic data collected by the traffic data collection means. Teacher data extraction means for extracting the teacher data necessary for the training, learning means for learning the neural network based on the teacher data, and a driving pattern corresponding to the traffic flow based on the output from the traffic flow determination means. The operation pattern selection means, the service status measurement means for measuring the margin for the service status of each car in the building, and the The control of each car in the driving pattern and the degree based on selected by the operation pattern selecting means and the operation control means for performing by applying fuzzy logic is provided.

In addition, in the elevator group management device according to the present invention, the elevator group management device that manages and operates a plurality of cars as a group collects traffic data generated in the building where the cars are installed. A data collection means, a traffic flow discrimination means for discriminating a traffic flow obtained from the traffic data collection means by applying a neural network, and if the output result from the neural network does not clearly appear, it is impossible to discriminate or identify. The number of times when the neural network is discriminated for each collection day based on the traffic data collected by the traffic data collection means over a predetermined period by the filtering means that outputs as either impossible or unidentifiable Teacher data extraction means for extracting traffic data on the day when the number of times is the smallest as teacher data, And learning means for learning the neural network based on chromatography data, and operation pattern selecting means for selecting a driving pattern corresponding to the traffic flow is provided on the basis of the output from the traffic flow determination unit.

Further, in the elevator group management system according to the present invention, the elevator group management system, which manages and operates a plurality of cars as a group, collects traffic data generated in a building where the cars are installed. A data collection means, a traffic flow discrimination means for discriminating a traffic flow obtained from the traffic data collection means by applying a neural network, and a neural network learning based on the traffic data collected by the traffic data collection means. Teacher data extraction means for extracting the teacher data necessary for the training, learning means for learning the neural network based on the teacher data, and a driving pattern corresponding to the traffic flow based on the output from the traffic flow determination means. Operation pattern selection means, external communication means that enables connection between the group management device and external equipment, and neural network identification And monitoring means are provided for displaying at least one of status and learning status to an external device.

Further, in the elevator group management device according to the present invention, the elevator group management device which manages and operates a plurality of cars as one group collects traffic data generated in the building in which the car is installed. A data flow collecting means, a traffic flow determining means for applying a neural network to a traffic flow obtained from the traffic data collecting means to make a determination, and a neural network based on the traffic data collected by the traffic data collecting means. Teacher data extraction means for extracting teacher data necessary for learning, learning means for learning a neural network based on teacher data, reservation information input means for inputting reservation information from the outside, and reservation information input means The memory that stores the reservation information input from the And the operation pattern selecting means for selecting the operation pattern is provided with priority information.

[0020]

In the elevator group management device configured as described above, when the neural network is applied to determine the traffic flow of the elevator, the elevator operation control suitable for the traffic flow is based on the output of the neural network. Is done.

In addition, in the elevator group management device configured as described above, when the traffic flow of the elevator is discriminated by applying the neural network, the task management unit prioritizes the computations related to the neural network. Based on the output of the neural network, operation control of the elevator that is suitable for the traffic flow is performed.

In addition, in the elevator group management device configured as described above, when a neural network is applied to determine the traffic flow of the elevator, a plurality of neural networks that determine the traffic flow for each operation pattern are provided. Then, based on the outputs of these neural networks, the operation control of the elevator adapted to the traffic flow is performed.

In addition, in the elevator group management device configured as described above, when the traffic flow of the elevator is discriminated by applying the neural network, the continuous traffic before the predetermined time from the time zone when the traffic peaks. Teacher data is extracted from the data, a neural network is learned based on the teacher data, and operation control of the elevator that is suitable for the traffic flow is performed.

In addition, in the elevator group management device configured as described above, when the traffic flow of the elevator is discriminated by applying the neural network, the discrimination calculation cycle and the learning calculation cycle are separately performed. Thus, the operation control of the elevator suitable for the traffic flow is performed.

In addition, in the elevator group management device configured as described above, when the traffic flow of the elevator is discriminated by applying the neural network, the fuzzy theory is applied based on the discrimination result of the neural network. Control of multiple elevators is performed to control the operation of elevators that are suitable for traffic flow.

In addition, in the elevator group management device configured as described above, when the traffic flow of the elevator is discriminated by applying the neural network, the traffic data on the day when the undecidable or unidentifiable is reduced is given priority. The neural network is learned by using this as teacher data, and the elevator operation control suitable for the traffic flow is performed.

Further, in the elevator group management device constructed as described above, the traffic flow of the elevator is discriminated by applying the neural network, and the discrimination result and the learning result are displayed on the external device by the monitoring means.

In addition, in the elevator group management device configured as described above, when the reservation information is input when the traffic flow of the elevator is determined by applying the neural network, the reservation is given priority and the operation pattern is set. Is selected, and operation control of the elevator suitable for the traffic flow is performed.

[0029]

【Example】

Example 1. 1 to 5 are views showing an embodiment of the present invention, FIG. 1 is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added, and FIG. 2 is an elevator group management device of FIG. FIG. 3 is a configuration diagram of the neural network of FIG. 3, FIG. 3 is a flowchart showing the operation of the elevator group management device of FIG. 1, FIG. 4 is an explanatory diagram of the zone division state in the building in the elevator group management device of FIG. 1, and FIG. It is explanatory drawing of the storage state of the traffic data in the elevator group management apparatus of No. 1.

In the figure, (1) is not shown, but a plurality of elevator cars are installed in each car, and (2) a car (1) is installed in a landing (not shown) located in the hoistway. Landing button, (3) control means for controlling the operation of each car (1), (4) operation control means for managing and controlling all the cars (1) as a group, (5) control means Each car communication means for performing optical communication between (3) and the operation control means (4), (6) is connected to the operation control means (4), and a response car is generated each time a call is made by the landing button (2). It is a call assigning means for selecting (1).

(7) is connected to the operation control means (4) to collect traffic information such as load of passengers getting on and off the elevator, and (8) is connected to the traffic data collection means (7). Traffic flow discriminating means for discriminating against the traffic flow collected by applying a neural network, (9) is connected to the traffic data collecting means (7), and traffic collected by the traffic data collecting means (7) This is a teacher data extraction means for extracting teacher data necessary for learning the neural network (10) from the data.

Reference numeral (11) is a learning means connected to the traffic flow discrimination means (8) and the teacher data extraction means (9) for learning the neural network (10), and (12) is a driving control means (4). Driving pattern selection means connected to the traffic flow determination means (8) and selecting a driving pattern such as driving at work by the output of the traffic flow determination means (8), (13) is traffic data collection means (7), traffic flow It is a memory that is connected to the discriminating means (8), the teacher data extracting means (9) and the learning means (11) to configure the collected data, the teacher data and the neural network (10).

Reference numeral (14) is a timer which is connected to the operation control means (4) and measures time. (15) is operation control means (4), each car communication means (5), call allocation means (6), traffic data collection means (7), traffic flow discrimination means (8), teacher data extraction means (9),
Learning means (11), operation pattern selecting means (12), memory (13)
And a group management device composed of a timer (14).

The structure of the neural network (10) used in the elevator group management device configured as described above will be described with reference to FIG. That is, the neural network (10) has four layers of the input layer, the intermediate layer 1, the intermediate layer 2 and the output layer.
It consists of three layers. The number of intermediate layers may be arbitrary, but two are prepared in the embodiments of FIGS. To the neural network (10), the data on the number of passengers collected from the traffic data collecting means (7) is used as input data, and the characteristic mode of the traffic flow to be selected at that time is used as output data.

The characteristic mode appearing in this output corresponds to, for example, an early morning mode, a work mode, a normal mode. Further, each layer such as the intermediate layer 1 has a predetermined number of neurons, and the input layer and the output layer have the same number of neurons as the input data and the output data, respectively.

Next, the operation of the group management device (14) will be described with reference to the flow chart shown in FIG. I.e. step
At (ST31), traffic data is collected by the traffic data collection means (7). Next, in step (ST32), if the first predetermined time has not elapsed, the process returns to step (ST31), and if it has elapsed, the process proceeds to step (ST33). This traffic data is constantly collected, and is statistically processed, for example, every first predetermined time such as every minute. Then, in step (ST33), the traffic flow is discriminated in real time based on the collected traffic data.

Then, the process proceeds to step (ST34)
Elevator operation control is performed based on the traffic flow discrimination result of (ST33). Then, the process proceeds to step (ST35), and returns to step (ST34) if the second predetermined time has not elapsed, and proceeds to step (ST36) if the second predetermined time has elapsed. In addition, the second said here
The predetermined time is a time interval for learning the neural network (10), and is set such as 2 weeks. Then, the neural network (10) is learned in step (ST36).

In the group management device (15), in order to collect traffic data to be input to the neural network (10),
A procedure for dividing the inside of the building into zones will be described with reference to FIG.
That is, it was mentioned above that the data input to the neural network (10) is passenger data, but input data is prepared for each hall placed in the hoistway where the car (1) is provided, that is, for each floor. In that case, a huge amount of memory and calculation time are required, which is not practical. Therefore, multiple floors with similar traffic flow will be collected in the same zone.

By doing so, the memory is not consumed too much and the accuracy of the neural network (10) is not deteriorated. FIG. 4 shows B1 to 15th floor 16
Assuming a building with floors, Fig. 4 (a) shows, for example, to determine the traffic flow during office work in the office building.
This is a case where the B1 floor is set as the first zone, the first floor is set as the second zone, and the second to fifteenth floors are set as the third zone. In the case of going to work in the morning, most of the passengers head to the B1 floor or the main floor, that is, the floor from the first floor to the upper floor, so such a setting is sufficient.

Further, in FIG. 4B, in order to determine the traffic flow during lunch, if the 15th floor is the dining room floor, the 2nd to 14th floors are the third
Zone, 15th floor is set as the 4th zone. With this setting, the number of input data in each case is shown in FIG.
(a) 2 (boarding / alighting) x 2 (UP / DOWN) x 3
(Zone) = 12, and similarly in FIG. 4 (b), 2 (ride /
It is enough to get off the vehicle × 2 (UP / DOWN) × 4 (zone) = 16.

Next, a method of storing traffic data in the group management device (15) will be described with reference to FIG. That is, the data collected and collected here is stored in the memory (13). Then, as described in FIG. 4, the building is divided into a predetermined number of zones and data is collected. The traffic data is roughly divided into data for applying the neural network (10) to discriminate every first predetermined time in real time and data for learning the neural network (10) every second predetermined time. Two
Different types of data are needed.

Therefore, first, an area for storing data for a predetermined number of days including data for the current day is prepared every day. Then, on each day, for example, the data is statistically collected every 5 minutes, and the collected traffic data is the loading / unloading load for each zone and each direction. Based on the data stored in this way,
For example, when determining the traffic flow, data around the current time of the day is used. Besides, neural network (10)
The data collected in the past few days will be used when learning. It is not always necessary to store the traffic data for one day, and only the traffic data for a specific time zone such as a busy hour may be stored.

In this way, when the traffic flow of the elevator is determined by applying the neural network (10), the operation control of the elevator suitable for the traffic flow is performed based on the output of the neural network (10). Therefore, it becomes possible to distinguish various traffic flows that occur in the building, and obtain a group management device for elevators that efficiently manages and operates a plurality of cars as a group by applying neural networks to distinguish traffic flows. Things will come.

Example 2. 6 to 8 are views showing another embodiment of the present invention. FIG. 6 is a block diagram showing a configuration of an elevator group management device with an elevator configuration added, and FIG. 7 is an elevator group management of FIG. FIG. 8 is a time chart for explaining the calculation sequence of the device, and FIG. 8 is a flowchart showing the operation of the elevator group management device in FIG. In the figure, the same reference numerals as those in FIGS. 1 to 5 indicate corresponding parts, and (16) is a traffic data collecting means (7), a traffic flow discriminating means (8), a teacher data extracting means (9), a learning means (11). And task management means connected to the memory (13) to control the order of operations of the plurality of operation means.

The calculation sequence of each calculation means in the elevator group management device configured as described above will be described with reference to the time chart of FIG. That is, each task is divided into three, and the traffic discriminating means (8) is set as the first priority, the learning means (11) is set as the third priority, and the traffic data collecting means (7) and other computing means are set as the second priority. . In addition, traffic discrimination means (8)
Determines the traffic flow in real time, and if the determination result is not output in a short time, the control is delayed, so it is set as the first priority.

The learning means (11) is a neural network (1
Since the learning of 0) requires a huge amount of time, it is not necessary to finish the learning in a short time, for example, by performing it at night when traffic is low, and thus the third priority is set. In this way, priority is given to each computing means so that the elevator control is not affected.

Next, the operation of the group management device (15) will be described with reference to the flow chart shown in FIG. I.e. step
If the BUSY1 flag of the first priority task is turned on in (ST81), the process proceeds to END. If not, the process proceeds to step (ST82) to turn on the BUSY1 flag. Then step (S
T83) and calculate the first priority task, and then step (S
Proceed to (T84) to turn off the BUSY1 flag, and then step (ST8
Go to 5).

BUS of the second priority task in step (ST85)
If the Y2 flag is on, the operation proceeds to END, and if not, the operation proceeds to step (ST86) to turn on the BUSY2 flag. Then, the process proceeds to step (ST87) to calculate the second priority task, proceeds to step (ST88) to turn off the BUSY2 flag, and proceeds to step (ST89) to calculate the third priority task.

As described above, when the traffic flow of the elevator is discriminated by applying the neural network (10), the arithmetic operations relating to the neural network (10) are prioritized and the output of the neural network (10) is used as a basis. Elevator operation control suitable for traffic flow is performed. Therefore, with respect to the discrimination of the neural network (10) and the execution of learning, the calculation is performed in accordance with the respective priorities, so that the related computer resources can be effectively used. This makes it possible to obtain an elevator group management device that efficiently manages and operates a plurality of cars as a group.

Example 3. FIG. 9 is also a diagram showing another embodiment of the present invention, and FIG. 9 is a diagram showing a neural network for discriminating the traffic flow in the elevator group management device, which is similar to the embodiment shown in FIGS. An elevator group management device is configured in. In FIG. 9, a plurality of neural nets (10) for discriminating the traffic flow for each driving pattern are set. In the embodiment shown in FIG. 9, neural networks are set for three traffic flows when the office building is at work, at lunch, and at work. In addition to these, it is possible to set a neural network (10) for discriminating the traffic flow even when traffic is concentrated on a specific floor due to a meeting or a banquet, for example.

As described above, when the traffic flow of the elevator is discriminated by applying the neural network (10), the neural network (10) for discriminating the traffic flow for each driving pattern is set, and the neural network (10) is set. Based on the output of, the operation control of the elevator that is suitable for the traffic flow is performed. Therefore, the teacher data required in the learning process of the neural network (10) can be extracted from the traffic data in which the characteristic mode is likely to appear, and the accuracy of the neural network (10) can be improved. This makes it possible to obtain an elevator group management device that efficiently manages and operates a plurality of cars as a group.

Example 4. 10 and 11 are also diagrams showing another embodiment of the present invention. FIG. 10 is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added, and FIG. 11 is an elevator group management of FIG. It is a flow chart which shows operation of a device. In the figure, the same reference numerals as those in FIGS. 1 to 5 indicate corresponding parts, and (17) is connected to the traffic data collecting means (7), the teacher data extracting means (9) and the memory (13) to collect traffic data. It is a peak time detecting means for detecting a time zone in which the traffic volume peaks in each time zone from the traffic data collected by the means (7).

A processing procedure in the elevator group management device configured as described above will be described with reference to the flowchart shown in FIG. That is, in step (ST111), the time zone i and the maximum value MAX of traffic volume are initialized. The time zone mentioned here is set, for example, every 5 minutes. Next, the procedure proceeds to step (ST112), sets the traffic volume of time zone i as T (i), and the traffic volume T is set at step (ST113).
If (i) is greater than the maximum value MAX, the process proceeds to step (ST114), and if traffic volume T (i) is less than the maximum value MAX, the process proceeds to step (ST115).

Then, at step (ST114), the traffic volume T (i) and time zone i are stored, and at step (ST115), if the storage is not completed for all time zones i, the time zone i is updated at step (ST116). Then, the process returns to step (ST112), and if the storage is completed, the process proceeds to step (ST117). In step (ST117), traffic data for a predetermined time before the time zone tu in which the traffic volume is maximum is extracted as teacher data.

As described above, when the traffic flow of the elevator is determined by applying the neural network (10), the traffic data of a predetermined time before the peak time zone of each traffic flow is set as the teacher data. Therefore, the teacher data required in the learning process of the neural network (10) can be extracted from the traffic data in which the characteristic mode is likely to appear, and the accuracy of the neural network (10) can be improved. This makes it possible to obtain an elevator group management device that efficiently manages and operates a plurality of cars as a group.

Example 5. FIG. 12 is also a diagram showing another embodiment of the present invention, and FIG. 12 is a time chart for explaining the calculation sequence of the elevator group management device, and the elevator group management is the same as the example of FIGS. 1 to 5. The device is configured. As mentioned above, in the elevator group management device,
A means to determine the traffic flow by applying the neural network (10) in real time, and a neural network (1
The means for learning 0) is roughly divided into two means.

Explaining the calculation cycle of these two means,
First, in the traffic flow determination means, for example, every 1 minute, the traffic data for the past 5 minutes is used to determine the current traffic flow. The reason for making the determination every minute is to follow the traffic flow that changes with time. The learning means statistically collects traffic data every five minutes, for example, and learns the neural network (10) every predetermined period.

As described above, when the traffic flow of the elevator is discriminated by applying the neural network (10), the neural network is divided into a discrimination calculation cycle and a learning calculation cycle.
Based on the output of (10), the operation control of the elevator suitable for the traffic flow is performed. Therefore, it is possible to obtain a group management device for elevators that can follow changes in traffic flow occurring in a building and efficiently manage and operate a plurality of cars as a group.

Example 6. 13 to 15 are also diagrams showing another embodiment of the present invention. FIG. 13 is a block diagram showing a configuration of an elevator group management device with an elevator configuration added, and FIG. 14 is an elevator group management of FIG. FIG. 15 is a graph showing a membership function representing a fuzzy amount used in the calculation of the device, and FIG. 15 is a flowchart showing the operation of the elevator group management device in FIG. In the figure,
The same reference numerals as those in FIG. 5 indicate the corresponding parts, and (18) is the operation control means.
It is a service status measuring means that is connected to (4) and measures the margin of the elevator for the traffic in the building at the current time.

A procedure for controlling the elevators by applying the fuzzy theory in the elevator group management device configured as described above will be described. That is, after the traffic flow is discriminated using the neural network (10), it is necessary to control the operation of the elevator according to the traffic flow. For example, if there is an output from the neural network (10) that the current time should be driving at work, the driving pattern selecting means (12) selects driving at work.

At this time, usually, a plurality of elevators are paid to the main floor of getting on and off, but fuzzy logic is applied to the payout of the elevators.
The fuzzy rule used here is described in the IF-THEN format, and the condition part, that is, the condition including the ambiguous expression is described in the IF part, and the execution procedure is described in the execution part, that is, the THEN part. It was done.

Plural kinds of fuzzy rules described in this way are prepared. As a description example of this fuzzy rule, for example, the following rules can be considered. That is, [rule n] IF ((the ratio of the time when the car (1) is present on the main floor is low) and (the number of departures from the main floor without passengers is small) and (main floor (The ratio of the time when the car (1) is waiting for the door to close on the upper floors is high.) THEN (Increase the number of elevators to be distributed to the main floor by one)

The membership function used in this rule and representing the fuzzy amount has the form shown in FIG. FIG. 14 (a) is a membership function representing the first condition in the condition part of rule n, and FIG. 14 (b) is a membership function representing the third condition. The rules described here are for increasing the number of payouts for the elevator, but conversely, it is also possible to write rules for reducing the number of payouts and controlling the timing of payouts.

Next, the operation of the group management device (15) will be described with reference to the flowchart shown in FIG. That is, in step (ST151), the rule counter i and the maximum fuzzy amount MAX are initialized. Then, the process proceeds to step (ST152) to calculate the fuzzy amount of the conditional part of the rule i and set it as F (i). Then, the process proceeds to step (ST153), and if F (i) is larger than the maximum value MAX, the process proceeds to step (ST154) to update the maximum value MAX. Also, in step (ST153), F
If (i) is smaller than the maximum value MAX, step (ST15
Go to 5).

If the counter i has not been finished for all rules in step (ST155), step (ST1
The counter i is updated in 56) and the process returns to step (ST152). If the counters i have been completed for all the rules, the process proceeds to step (ST157) to execute the execution unit of the rule mu. The elevator is controlled according to such a procedure, and the number of payouts of the elevator and the payout timing of the elevator are controlled.

As described above, when the traffic flow of the elevator is discriminated by applying the neural network (10), fuzzy theory is applied based on the discrimination result of the neural network (10) to control a plurality of elevators. To do. That is, for example, a fuzzy rule is described in the IF-THEN format,
Further, the condition section, that is, the IF section describes the condition including the ambiguous expression, and the execution section, that is, the THEN section describes the execution procedure. According to such a rule, control is performed to increase or decrease the number of elevators to be paid out or to adjust the payout timing. Therefore, it is possible to obtain an elevator group management device that efficiently manages and operates a plurality of cars as a group.

Example 7. 16 and 17 are diagrams showing another embodiment of the present invention. FIG. 16 is a configuration diagram of a neural network of an elevator group management device, and FIG. 17 is a flowchart showing the operation of the elevator group management device of FIG. Thus, an elevator group management device is configured as in the embodiment shown in FIGS. In the figure, (19) is the above-mentioned FIG.
Filtering means at the output of the neural network (10)
This is a neural network with (20) added.

In the elevator group control device configured as described above, two modes, i.e., unidentifiable and unidentifiable modes, are added to the data output by the filtering means (20) in addition to the n characteristic modes. . This filtering means (20) is provided with a threshold value in advance, and if the output from any neuron does not exceed the threshold value with respect to the output from each neuron, it is output as unidentifiable, and the output from multiple neurons When exceeds the threshold, it is output as unspecified. As a result, if the neural network (19) is not well constructed and the output is unclear, it is possible to avoid wrong control.

Next, the operation of the group management apparatus corresponding to the embodiment shown in FIGS. 16 and 17 will be described with reference to the flow chart shown in FIG. Here, a means for extracting the teacher data necessary for learning the neural network (19) will be described. The teacher data extracting means (9) is executed every predetermined period, for example, every two weeks, like the learning means (11). That is, in step (ST171), the day i and the minimum value MIN are initialized, and the process proceeds to step (ST172), in which the neural network (19) discriminates the traffic data on the i-th day.

Next, the procedure proceeds to step (ST173), the number of times Un (i) at which unidentifiable or unidentifiable is output is counted, and the procedure proceeds to step (ST174) to count Un more than the minimum value MIN.
If (i) is small, proceed to step (ST175) and set the minimum value M
If the number of times Un (i) is larger than IN, step (ST17
Go to 6). Then, in step (ST175), the minimum value MIN is updated, and in step (ST176), if all days i have ended, proceed to step (ST178), and if all days i have not ended, step (ST177). Go to and update day i (ST17
Return to 2).

In step (ST178), the traffic data on the day cn, which has the smallest number of times the unidentifiable or unidentifiable is output, is extracted. Thereby, teacher data can be extracted from the traffic data on the day when the traffic flow clearly appears, and a highly accurate neural network (19) can be constructed.

As described above, when the traffic flow of the elevator is discriminated by applying the neural network (19), the teacher data is set by giving priority to the traffic data on the day when the discrimination or unidentification is reduced. As a result, even if the output from the neural network (19) does not appear clearly, it is possible to control the elevator without affecting it. Therefore, it is possible to obtain an elevator group management device that efficiently manages and operates a plurality of cars as a group.

Example 8. 18 and 19 are also diagrams showing another embodiment of the present invention. FIG. 18 is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added, and FIG. 19 is an elevator group management of FIG. It is a figure which shows the example of a display screen of the monitoring means in an apparatus. In the figure, the same reference numerals as those in FIGS.
Is an external communication means that is connected to the operation control means (4) and communicates between the group management device (15) and external equipment, and (22) is an external communication means
The monitoring means is connected to (21) and displays the learning progress and the discrimination result of the neural network (10). The monitoring means (22) can be realized by, for example, a laptop type personal computer generally used as a maintenance tool for elevators or a tool for changing the operation mode.

In the elevator group management device configured as described above, for example, a screen output as shown in FIG. 19 can be considered. That is, the right half of FIG. 19 shows the configuration of the neural network (10) as an illustration. In this case, the thickness of the line is changed or the density of the neuron is changed according to the strength of the degree of connection between neurons and the weight of the neuron. Further, the graph in the upper left part of FIG. 19 shows the fluctuation of passengers for each time zone, the middle part of the left part of FIG. 19 shows the output level of the characteristic mode determined by the neural network (10), and the lower left part of FIG. Change of each is displayed.

As described above, when the traffic flow of the elevator is discriminated by applying the neural network (10), the discrimination result and the learning result are externally displayed by the monitoring means (22), so that the group management situation can be easily performed. Can be monitored and appropriate countermeasures can be taken to optimize the group management situation. Therefore, it is possible to obtain an elevator group management device that efficiently manages and operates a plurality of cars as a group.

Example 9. FIG. 20 is also a diagram showing another embodiment of the present invention, and is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added. In the figure, the same reference numerals as those in FIGS. 1 to 5 indicate the corresponding parts, and (21) is an external communication means that is connected to the operation control means (4) and communicates between the group management device (15) and an external device, 23) is an external communication means (21)
A laptop type personal computer is used as a reservation information input means that can be connected to the elevator to input various reservation information to the elevator.

In the elevator group management device configured as described above, the operation pattern can be automatically selected by the neural network (10) using the reservation information input means (23). Then, some elevator managers may want to select an operation pattern according to their preference. In such a case, the driving pattern is selected by the driving pattern selecting means (12) according to the input time so as to give priority to the reservation information over the determination result of the neural network (10). The reservation information input means (23) can also be configured by an input from, for example, a CRT type monitoring panel or a building management system.

As described above, when the traffic flow of the elevator is determined by applying the neural network (10), the reservation information is input by the reservation information input means (23). In this case, the reservation information is prioritized and the elevator is given priority. Are managed in groups. For this reason, it becomes a user-friendly group management device capable of selecting an operation pattern in accordance with a reservation determined by the elevator administrator. Therefore, it is possible to obtain an elevator group management device that efficiently manages and operates a plurality of cars as a group.

[0079]

As described above, according to the present invention, the control means for controlling the operation of a plurality of cars, the operation control means for managing and controlling the operation of all the cars as a group, and the call generated at the hall are described. A call assigning means for determining an assigned car based on a predetermined evaluation method is provided, and a neural network is applied to the traffic flow obtained from the traffic data collecting means for collecting the traffic data generated in the building to make a discrimination. A traffic flow discriminating means is provided, and learning means for learning the neural network based on the teacher data necessary for learning the neural network based on the traffic data collected from the traffic flow discriminating means is provided, and the driving control means A timer connected to the and a memory for storing at least traffic data, teacher data, and information forming a neural network,
Further, there is provided driving pattern selection means for selecting a driving pattern corresponding to the traffic flow based on the output from the traffic flow determination means.

As a result, when the traffic flow of the elevator is determined by applying the neural network, the elevator operation control suitable for the traffic traffic is performed based on the output of the neural network. Therefore, it becomes possible to distinguish various traffic flows that occur in a building, and to apply the neural network to distinguish the traffic flows, to efficiently manage the management function of the elevator group management device that manages and operates multiple cars as a group. Has the effect of

As described above, according to the present invention, the group management device for elevators that manages and operates a plurality of cars as a group is used for the traffic flow obtained from the traffic data collecting means for collecting the traffic data generated in the building. A traffic flow discriminating means for discriminating by applying a neural network to the neural network, and the neural network based on the teacher data necessary for learning the neural network based on the traffic data collected from the traffic flow discriminating means. And a traffic flow discriminating means as a first-priority computing means,
The learning means is the third-priority computing means, and the traffic flow determining means and the other computing means of the learning means are the second-priority computing means. is there.

Thus, when the traffic flow of the elevator is discriminated by applying the neural network, the task management means gives priority to the operation related to the neural network and adapts to the traffic flow based on the output of the neural network. The operation control of the elevator is performed. Therefore, with respect to the discrimination of the neural network and the execution of learning, the calculation is performed according to the respective priorities. Therefore, related computer resources can be effectively used, and the management function of the elevator group management device that manages and operates a plurality of cars as a group is effective.

As described above, according to the present invention, the group management device for the elevators, which manages and operates a plurality of cars as a group, has the traffic flow obtained from the traffic data collecting means for collecting the traffic data generated in the building. A traffic flow discriminating means for discriminating by applying a neural network to the neural network, and the neural network based on the teacher data necessary for learning the neural network based on the traffic data collected from the traffic flow discriminating means. Further, a learning means for learning is provided, a driving pattern selecting means for selecting a driving pattern corresponding to the traffic flow based on the output from the traffic flow judging means, and a plurality of neural nets for judging the traffic flow for each driving pattern. A memory for storing is provided.

Thus, when the traffic flow of the elevator is discriminated by applying the neural net, a plurality of neural nets for discriminating the traffic flow for each driving pattern are provided, and based on the outputs of these neural nets. Elevator operation control suitable for traffic flow is performed. Therefore, the teacher data required in the learning process of the neural network can be extracted from the traffic data in which the characteristic mode is likely to appear, and the accuracy of the neural network can be improved. Accordingly, there is an effect that the management function of the elevator group management device that manages and operates a plurality of cars as a group is made efficient.

As described above, the present invention provides a group management device for elevators, which manages and operates a plurality of cars as a group, with a traffic flow obtained from traffic data collecting means for collecting traffic data generated in a building. A traffic flow discriminating means for discriminating by applying a neural net is provided, and a peak time detecting means for detecting a peak time zone of traffic volume from the traffic data collected by the traffic data collecting means is provided. The teaching data extracting means extracts the continuous traffic data of a predetermined time before the peak time detected by the peak detecting means as the teacher data, and the learning means for learning the neural network based on the teacher data.

As a result, when the traffic flow of the elevator is discriminated by applying the neural network, the teacher data is extracted from the continuous traffic data a predetermined time before the peak traffic time, and the teacher data is extracted. Based on this, a neural network is learned to control the operation of the elevator that is suitable for the traffic flow. Therefore, the teacher data required in the learning process of the neural network can be extracted from the traffic data in which the characteristic mode is likely to appear, and the accuracy of the neural network can be improved. Accordingly, there is an effect that the management function of the elevator group management device that manages and operates a plurality of cars as a group is made efficient.

As described above, according to the present invention, the elevator group control device for controlling and operating a plurality of cars as a group collects the traffic data generated in the building at the first time interval. A traffic flow discriminating means for discriminating the traffic flow obtained by the means by applying a neural network at a second time interval shorter than the first time interval is provided, and the traffic collected by the traffic data collecting means. The teaching data extracting means for extracting the teaching data necessary for learning the neural network based on the data and the learning means for learning the neural network based on the teaching data are provided.

As a result, when the traffic flow of the elevator is discriminated by applying the neural network, the discrimination operation cycle and the learning operation cycle are performed separately, and the elevator operation control suitable for the traffic flow is performed. Be seen. Therefore, it is possible to easily follow changes in the traffic flow occurring in the building, and there is an effect that streamlines the management function of the elevator group management device that manages and operates a plurality of cars as a group.

As described above, the present invention provides a group management device for elevators, which manages and operates a plurality of cars as a group, to a traffic flow obtained from traffic data collecting means for collecting traffic data generated in a building. A traffic flow discriminating means for discriminating by applying a neural network is provided, and teacher data extraction for extracting the teacher data necessary for learning the neural network based on the traffic data collected by the traffic data collecting means A learning means for learning the neural network based on the means and the teacher data is provided,
Further, a driving pattern selecting means for selecting a driving pattern corresponding to the traffic flow based on the output from the traffic flow judging means and a service status measuring means for measuring a margin for the service status of each car in the building are provided. The operation control means is provided for controlling each car in the operation pattern selected by the operation pattern selection means based on the fuzzy theory.

Accordingly, when the traffic flow of the elevator is discriminated by applying the neural network, the fuzzy theory is applied based on the discrimination result of the neural network to control a plurality of elevators, and the traffic flow is controlled. Operation control of the adapted elevator is performed. That is, the fuzzy rule is used to increase or decrease the number of elevators to be paid, or to adjust the payout timing. Therefore, there is an effect that the management function of the elevator group management device that manages and operates a plurality of cars as a group is made efficient.

As described above, according to the present invention, the group management device for elevators, which manages and operates a plurality of cars as a group, has a traffic flow obtained from the traffic data collecting means for collecting the traffic data generated in the building. A traffic flow discriminating means for discriminating by applying a neural network is provided, and if the output result from the neural net does not clearly appear, filtering means and traffic data collection that output as undecidable or unidentifiable Based on the traffic data collected by the means for a predetermined period, the neural network is discriminated for each collection day, and the traffic data on the day when the number of times of unidentifiable or unidentifiable is outputted is extracted as teacher data. A learner who has a teacher data extraction means and learns a neural network based on the teacher data. And it is provided with a operation pattern selecting means for selecting a driving pattern corresponding to the traffic flow on the basis of the output from the traffic flow determination unit.

Thus, when the traffic flow of the elevator is discriminated by applying the neural network, the traffic data on the day when the number of undecidable or unidentifiable is reduced is given priority, and the neural network is learned by using this as the teacher data. However, the operation control of the elevator suitable for the traffic flow is performed. Therefore, even if the output from the neural network does not appear clearly, it is possible to control without affecting the elevator. Accordingly, there is an effect that the management function of the elevator group management device that manages and operates a plurality of cars as a group is made efficient.

As described above, according to the present invention, an elevator group management device that manages and operates a plurality of cars as a group, and a traffic flow obtained from traffic data collecting means for collecting traffic data generated in a building. A traffic flow discriminating means for discriminating by applying a neural network is provided, and teacher data extraction for extracting the teacher data necessary for learning the neural network based on the traffic data collected by the traffic data collecting means A learning means for learning the neural network based on the means and the teacher data is provided,
Further, a driving pattern selecting means for selecting a driving pattern corresponding to the traffic flow based on the output from the traffic flow judging means and an external communication means for enabling connection between the group management device and the external device are provided, and the neural network discrimination is performed. A monitoring means for displaying at least one of the situation and the learning situation on an external device is provided.

Accordingly, the traffic flow of the elevator is discriminated by applying the neural network, and the discrimination result and the learning result are displayed on the external device by the monitoring means. Therefore, the group management status can be easily monitored, and appropriate countermeasures for optimizing the group management status can be taken. Accordingly, there is an effect that the management function of the elevator group management device that manages and operates a plurality of cars as a group is made efficient.

Further, as described above, the present invention provides a group management device for elevators, which manages and operates a plurality of cars as a group, to a traffic flow obtained from traffic data collecting means for collecting traffic data generated in a building. A traffic flow discriminating means for discriminating by applying a neural network is provided, and teacher data extraction for extracting the teacher data necessary for learning the neural network based on the traffic data collected by the traffic data collecting means A learning means for learning the neural network based on the means and the teacher data is provided,
Further, a memory for storing reservation information input from the reservation information input means for inputting reservation information from the outside and an operation for selecting the operation pattern by prioritizing the reservation information over the output from the traffic flow judging means by inputting the reservation information The pattern selecting means is provided.

Accordingly, when the traffic flow of the elevator is discriminated by applying the neural network, when the reservation information is input, the operation pattern of the elevator is selected with priority to the reservation, and the operation control of the elevator suitable for the traffic flow is performed. Is done. Therefore, the user-friendly group management device can select the operation pattern according to the reservation determined by the elevator administrator. Accordingly, there is an effect that the management function of the elevator group management device that manages and operates a plurality of cars as a group is made efficient.

[Brief description of drawings]

FIG. 1 is a diagram showing a first embodiment of the present invention, and is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added.

FIG. 2 is a configuration diagram of a neural network of the elevator group management device in FIG.

FIG. 3 is a flowchart showing the operation of the elevator group management device in FIG.

4 is an explanatory diagram of a zone division state in a building in the elevator group management device in FIG. 1. FIG.

5 is an explanatory diagram of a storage state of traffic data in the elevator group management device in FIG. 1. FIG.

FIG. 6 is a diagram showing a second embodiment of the present invention, and is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added.

7 is a time chart for explaining the calculation sequence of the elevator group management device in FIG.

8 is a flowchart showing the operation of the elevator group management device in FIG.

FIG. 9 is a view showing a third embodiment of the present invention, and is a view showing a neural network for discriminating traffic flow in the elevator group management device.

FIG. 10 is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added according to a fourth embodiment of the present invention.

11 is a flowchart showing the operation of the elevator group management device in FIG.

FIG. 12 is a diagram showing a fifth embodiment of the present invention and is a time chart for explaining the calculation sequence of the elevator group management device.

FIG. 13 is a diagram showing a sixth embodiment of the present invention, and is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added.

14 is a graph showing a membership function representing a fuzzy amount used for calculation of the elevator group management device in FIG.

15 is a flowchart showing the operation of the elevator group management device in FIG.

FIG. 16 is a diagram showing a seventh embodiment of the present invention and is a configuration diagram of a neural network of an elevator group management device.

17 is a flowchart showing the operation of the elevator group management device in FIG.

FIG. 18 is a diagram showing an eighth embodiment of the present invention and is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added.

FIG. 19 is a diagram showing an example of a display screen of the monitoring means in the elevator group management device of FIG. 18.

FIG. 20 is a diagram showing the ninth embodiment of the present invention and is a block diagram showing a configuration of an elevator group management device to which an elevator configuration is added.

[Explanation of symbols]

1 car 3 control means, 4 operation control means, 6 call allocation means, 7 traffic data collection means, 8 traffic flow discrimination means, 9 teacher data extraction means, 10 neural network, 11 learning means, 12 driving pattern selection means, 1
3 memory, 14 timers, 15 group management device, 16 task management means, 17 peak detection means, 18 service status measurement means, 19 neural network, 20 filtering means, 21 external communication means, 22 monitoring means, 23 reservation information input means .

Claims (9)

[Claims] 1. A control means for controlling the operation of each of a plurality of cars, an operation control means for managing and controlling the operation of all the cars as a group, and a call generated at a landing provided corresponding to the cars. Call allocation means for deciding an assigned car based on a predetermined evaluation method, traffic data collection means for collecting traffic data generated in the building in which the car is installed, and traffic required by the traffic data collection means Traffic flow discriminating means for discriminating a flow by applying a neural network, and teacher data for extracting teacher data necessary for learning the neural network based on the traffic data collected from the traffic flow discriminating means. An extracting means, a learning means for learning the neural network based on the teacher data, a timer connected to the operation control means for measuring time, At least the memory for storing the traffic data, the teacher data, and the information configuring the neural network, and the driving pattern selecting means for selecting the driving pattern corresponding to the traffic flow based on the output from the traffic flow determining means. Elevator group management device. 2. A group management device for elevators, which manages and operates a plurality of cars as a group, and a traffic data collecting means for collecting traffic data generated in a building in which the cars are installed, and from this traffic data collecting means. A traffic flow discriminating means for discriminating a required traffic flow by applying a neural net, and extracting teacher data necessary for learning the neural network based on the traffic data collected from the traffic flow discriminating means Teacher data extraction means for
The learning means for learning the neural network based on the teacher data and the traffic flow discriminating means are first-priority computing means, the learning means is third-priority computing means, and the traffic flow discriminating means and the learning means. An elevator group management device, comprising: task management means for managing the operation of each of the computing means in accordance with the order in which the other computing means is the second-priority computing means. 3. A group management device for elevators, which manages and operates a plurality of cars as a group, and a traffic data collecting means for collecting traffic data generated in a building in which the cars are installed, and from this traffic data collecting means. A traffic flow discriminating means for discriminating a required traffic flow by applying a neural net, and extracting teacher data necessary for learning the neural network based on the traffic data collected from the traffic flow discriminating means Teacher data extraction means for
Learning means for learning the neural network based on the teacher data, driving pattern selecting means for selecting a driving pattern corresponding to the traffic flow based on the output from the traffic flow determining means, and traffic for each driving pattern A group management device for an elevator, comprising: a memory that stores a plurality of neural nets for discriminating a flow. 4. In a group management device for elevators, which manages and operates a plurality of cars as a group, traffic data collecting means for collecting traffic data generated in a building where the cars are installed, and from this traffic data collecting means A traffic flow discriminating means for discriminating a required traffic flow by applying a neural network, and a peak time detecting means for detecting a time zone at which the traffic volume peaks from the traffic data collected by the traffic data collecting means. A teacher data extracting means for extracting, as teacher data, continuous traffic data a predetermined time before the peak time detected by the peak detecting means, and a learning means for learning the neural network based on the teacher data. Elevator group management device characterized by 5. A traffic data collection means for collecting, at a first time interval, traffic data generated in a building in which the above-mentioned cars are installed, in a group management device for elevators that manages and operates a plurality of cars as a group. A traffic flow discriminating means for discriminating a traffic flow obtained by the traffic data collecting means by applying a neural network at a second time interval shorter than the first time interval, and collected by the traffic data collecting means. A training data extracting unit that extracts training data necessary for learning the neural network based on the obtained traffic data; and a learning unit that learns the neural network based on the training data. Elevator group management device. 6. A group management device for elevators which manages and operates a plurality of cars as a group, and a traffic data collecting means for collecting traffic data generated in a building in which the cars are installed, and from this traffic data collecting means. Traffic flow discriminating means for discriminating a required traffic flow by applying a neural network, and extracting teacher data necessary for learning the neural network based on the traffic data collected by the traffic data collecting means. Teacher data extraction means, learning means for learning the neural network based on the teacher data, and driving pattern selection means for selecting a driving pattern corresponding to the traffic flow based on the output from the traffic flow determination means. , Based on the service status measuring means for measuring the room service status of each car and the room And an operation control means for controlling each of the cars in the operation pattern selected by the operation pattern selection means by applying a fuzzy theory. 7. A group management device for elevators, which manages and operates a plurality of cars as a group, and a traffic data collecting means for collecting traffic data generated in a building in which the cars are installed, and a traffic data collecting means for collecting the traffic data. A traffic flow discriminating means for discriminating a required traffic flow by applying a neural network, and a filtering means for outputting as unidentifiable or unidentifiable if the output result from the neural network does not clearly appear. , A traffic on the day when the neural network is discriminated for each collection day on the basis of the traffic data collected by the traffic data collecting means for a predetermined period, and the number of times that the unidentifiable or unidentifiable is output is the smallest Teacher data extraction means for extracting data as teacher data, and the neural network based on the teacher data. An elevator group management device comprising: learning means for learning a net; and driving pattern selection means for selecting a driving pattern corresponding to the traffic flow based on an output from the traffic flow determination means. 8. In a group management device for an elevator that manages and operates a plurality of cars as a group, traffic data collecting means for collecting traffic data generated in a building in which the cars are installed, and from this traffic data collecting means Traffic flow discriminating means for discriminating a required traffic flow by applying a neural network, and extracting teacher data necessary for learning the neural network based on the traffic data collected by the traffic data collecting means. Teacher data extraction means, learning means for learning the neural network based on the teacher data, and driving pattern selection means for selecting a driving pattern corresponding to the traffic flow based on the output from the traffic flow determination means. , An external communication unit that enables connection between the group management device and an external device, and the neural network discrimination status and learning A group management device for an elevator, comprising: a monitoring means for displaying at least one of the situations on the external device. 9. A group management device for elevators that manages and operates a plurality of cars as a group, and a traffic data collecting unit that collects traffic data generated in a building where the cars are installed, and from this traffic data collecting unit. Traffic flow discriminating means for discriminating a required traffic flow by applying a neural network, and extracting teacher data necessary for learning the neural network based on the traffic data collected by the traffic data collecting means. Teacher data extraction means, learning means for learning the neural network based on the teacher data, reservation information input means for inputting reservation information from the outside, and reservation information input from the reservation information input means The memory and the reservation pattern are input, and the reservation pattern is given priority over the output from the traffic flow discrimination means. A group management device for an elevator, comprising: an operation pattern selection means for selecting a vehicle.