JPH02261785A - Group-controlled elevator device - Google Patents

Abstract

PURPOSE:To improve evaluation accuracy, by deciding a parameter for optimal control index with the use of a net formed by an evaluation operation network means and, based on the above, executing assignment control for the optimal elevator No. for a call from a hall. CONSTITUTION:A processing device B1 for each elevator No. executes a near future prediction for the each one, predicts its scheduling, and estimates a predictive arrival time to each floor by means of a prediction operating part B1-1, and based on these data, operates an evaluation value of each evaluation index with the use of the evaluation operation network, by means of an evaluation operation unit B1-2. By using each control index parameter which is regarded as the evaluation value for each evaluation index and the weighted value for each evaluation index the total evaluation value is found out and an assigned elevator selection unit B2 assigns an elevator No. having the minimum evaluation value as a responding one for a call from a hall, by using an optimal control index parameter which is real-timely set by a control index parameter a setting unit B3, according to the demand per each time zone. Thus it is possible to perform the optimal assignment control.

Description

[Detailed description of the invention] [Purpose of the invention] (Industrial application field) The present invention relates to a group management control elevator system.

(Prior art) In recent years, when multiple elevators are installed in parallel, in order to improve the operating efficiency of the elevators and improve the service to elevator users, the elevators that respond to hall calls on each floor have been equipped with a microcomputer. Currently, small-sized computers such as the above are used to allocate information rationally and quickly.

That is, when a hall call occurs, the elevator of the optimum number to service the hall call is selected and assigned, and elevators of other numbers are not allowed to respond to the hall call.

Recently, in this type of group management control elevator equipment, it has become possible to understand the traffic flow between floors of each building by measuring car call registration data when responding to each hall call in real time, measuring boarding/alighting load data, etc. Based on the measurement data, building-specific demand is grasped and used to predict hall call allocation times.

In such a situation, the elevator hall call allocation control performs prediction based on the building-specific demand data from the condition of each elevator with respect to the occurrence of a hall call.
The schedule for each aircraft is expressed as a forecast arrival time, and each control index for group management performance is modeled by an objective function based on the predicted arrival time. This is done by converting it into a comprehensive evaluation value as a parameter and determining the optimum model number.

(Problem to be Solved by the Invention) However, it is necessary to select the optimum value by adjusting the parameters for each control index according to the traffic flow of each building and the demand for each time period. Even if demand is understood, it is difficult to directly determine optimal control index parameters from demand data.

In addition, in recent years, methods that incorporate fuzzy control, which directly incorporates experts' empirical rules into control, have appeared, but there is no guarantee that the optimal solution is always selected to determine control index parameters using an empirical approach. Furthermore, it is difficult to determine control index parameters based on the traffic flow and demand of each building by using an empirical approach, even if you are an expert.

This invention was made in view of these conventional problems, and it relates control index parameters and response time cumulative distributions through a net relationship, and optimizes each demand parameter using response time cumulative distributions. It is an object of the present invention to provide a group management control elevator system that can determine control index parameters that have been used, improve allocation evaluation accuracy, and improve performance.

[Structure of the Invention] (Means for Solving the Problems) A group management control elevator apparatus according to claim 1 of the present invention includes means for storing demand parameters, and demand parameters stored in the demand parameter storage means and each demand parameter stored in the demand parameter storage means. An evaluation calculation network means that forms a net relationship between the control index parameter in the evaluation calculation and the hall call response time cumulative distribution, and an optimization calculation for each value of each demand parameter using the net formed by the evaluation calculation network means. The apparatus comprises: a control index parameter setting means for determining a control index parameter; and a group management control means for controlling the assignment of an optimal car to the hall call using the optimal control index parameter determined by the control index parameter setting means. be.

The group management control elevator system according to claim 2 of the present invention is the group management control elevator system according to claim 1, further storing the control index parameter and actual hall call response time cumulative distribution data for the control index parameter. response time cumulative distribution data storage means, and after the elevator is in operation, the evaluation calculation network means calculates the demand parameters, control index parameters, and hall call response times that it has formed based on the data stored in the response time cumulative distribution data storage means. and network correction means for correcting the relationship between the cumulative distribution and the cumulative distribution.

(Operation) In the group management control elevator system of the present invention, the evaluation calculation network means changes the control index parameter within a set range for each value of each demand parameter, and calculates the cumulative distribution of hall call response times for each parameter value. Seek from relationships through the internet. The control index parameter setting means selects an optimal cumulative distribution of hall call response times from the distribution data corresponding to each parameter combination using average unanswered time, variance, maximum unanswered time, long waiting rate, etc. as evaluation indices, and The control index parameter corresponding to is selected as the optimal parameter.

Then, the group management control means uses the control index parameters determined by the control index parameter setting means to perform control of assigning the optimum car to the generated hall call.

(Example) Hereinafter, embodiments of the present invention will be explained in detail based on the drawings.

FIG. 1 is a block diagram showing the configuration of a group management control elevator system to which the present invention is applied.

In FIG. 1, reference numeral 1 denotes a group management control section, and this group management control section 1 includes single control sections 2-1. . . , 2-N are connected by a high-speed transmission line 6, which is a first transmission control means.

Group management control section 1 and single control section 2-1. ...2-N
It is composed of one or more small computers such as microcomputers, and operates under the control of software.

Further, 3 is a hall call button provided on each floor, and 4 is a hall call input/output control section for inputting and outputting hall calls.

Group management control section 1, single control section 2-1. ..., 2-N and each hall call input/output control unit 4 are connected via a low-speed transmission line 7, which is a second transmission control means.

The high-speed transmission line 6 includes the group management control section 1 and the single control section 2-1.
..., 2-N, that is, mainly between the control computers in the machine room, and is connected by a high-speed, highly intelligent network. Control information necessary for group management control is exchanged at high speed between the group management control section 1 and each of the individual control sections 2-1...2-N.

The low-speed transmission line 7 is a control system that mainly transmits information sent via the hoistway, such as the hall call buttons 3 of each hall and the monitoring panel 5 of the monitoring room, and is compared to the high-speed transmission line 6. Because of its low speed and long distance, it is composed of optical cables and the like. ...,
It is connected to 2-N and exchanges data.

When the group management control unit 1 is normal, the hall call button 3 is controlled by the group management control unit 1 via the low-speed transmission line 7, and when the hall call button 3 is pressed, the hall call gate is closed and the registration lamp is turned on. At the same time as setting, the single control unit 2-1. ... Determine the optimum machine based on the information data of 2-N, and issue a control command to that single machine.

Then, the unit control unit that receives the control command performs unit control of its own elevator using this control command as hall call information.

FIG. 2 shows the group management control section 1 and the single control section 2-1.・
..., 2-N software system configuration. The software configuration is the operating system (O5)
Single control function task by real-time O88,
Tasks 9 to 12 of the group management control main function task, group management control sub function task, and transmission control task are managed, and each task 9 to 12 is started or held by the scheduler in real-time 03g. There is.

Among these tasks 9 to 12, the unit control function task 9 includes the unit control unit 2-1. . . , is the core function in 2-N, and each single control unit 2-1. ...2-N is given a high priority as a task for operating it.

The group management control main function task 10 is a central function of the group management control unit 1, and is a main function of the group management control unit 1. ...,
The information data for each machine is collected from the group management control subfunction task 11 distributed in 2-N, and the optimum machine is determined by comparing and calculating, and control commands are given to the individual control unit of the relevant machine, and the hall It also controls the call button 3.

The group management control sub-function task 11 is a function of processing information for each machine of the group management control section 1, and processes the information under the control of the group management control main function task 10. That is, the configuration is such that a computer having a group management control main function task 10 manages the startup and termination of tasks via a high-speed transmission line 6. The configuration is such that distributed processing is performed in units and data is transferred to the group management control main function task 10 at the time of completion of the processing.

The transmission control task 12 controls the transmission and reception of data on the high-speed transmission line 6 and the activation and termination of the group management control subfunction task 11.

FIG. 3 is a block diagram showing the system configuration of the high-speed transmission line 6 of FIG. 1. Transmission control is by microprocessor 13
However, for example, a data link controller 14 and a media access controller 15 configured with hardware are used as parts for controlling the data link layer of the LAN network model layer proposed by ISO (International Organization for Standardization). It has a configuration that enables highly intelligent data transmission. A configuration is adopted in which the proportion of transmission control software managed by the microprocessor 13 is reduced for high-speed transmission control.

For example, the data link controller 14, which is a controller that realizes the above-mentioned highly intelligent transmission control, is an LSI 182586 manufactured by Intel Corporation, and the 182501, also manufactured by Intel Corporation, is used as a media access controller. By using these, a high-speed transmission function of 10Mt/sec can be performed relatively easily with a reduced support ratio of the microprocessor 13.

Note that 16 is a system bus, 17 is a control line, and 18 is a serial transmission system.

Next, the operation of the above group management control elevator system will be explained.

FIG. 4 is a functional block diagram showing the hall call assignment process according to an embodiment of the present invention, and shows the flow of each control process when the hall call button 3 generates a call.

In FIG. 4, each machine unit process B1 is performed by each unit control section 2-1. . . , carried out in the group management control subfunction task 11 in 2-N, and other processing B2. B3 is the group management control main function task 10 in the group management control section 1.
It will be held in

Now, when the hall call button 3 is registered and a hall call occurrence detection input is given, the group management control section 1 sends each unit control section 2-1. ..., 2-N is requested to start the machine unit processing B1. This machine unit processing B1 is a process for performing predictive evaluation calculation for each machine, and is a process for performing predictive evaluation calculation for each machine.
1, the near future of each car is predicted, the scheduling is predicted, and the predicted arrival time to each floor is determined.

Next, an evaluation calculation is performed in the evaluation calculation unit B1-2 based on the predicted arrival time determined by this prediction calculation.

This evaluation calculation is generally performed for a plurality of evaluation indicators m. For example, for car C, gt (c), g2
(c)+...+ go cc).

The evaluation value g+(c) for each evaluation index j is generally a function of the predicted arrival time of the calculation unit. Therefore, as an example, if we consider the index "prevention of long life" as the evaluation index j, the predicted unresponse time of the floor to which the i-th car is already assigned and the floor where the hall call occurs is "01tin j2".
+...+ The maximum value of tI can be considered as an example. (Here, the predicted δ-l non-response time is the sum of the predicted arrival time and the elapsed time of the hall call.) Evaluation value for each evaluation index gr (c) T (j-
When the calculation of 1n 2゜..., m) is completed,
Each unit control section 2-1. ..., 2-N sends the evaluation value g+ for each evaluation index, which is the calculation result, to the group management control unit 1 via the high-speed transmission line 6 (c); (c:1.2...
, N machine) are returned.

In the group management main function task 10 of the group management control unit 1, all the aircraft numbers 2-
1. ..., 2-N selects the allocated machine based on the evaluation value g+ (C) for each evaluation index returned.

The overall evaluation value that governs the comprehensive judgment of the allocation machine selection in the allocation machine selection section B2 is the evaluation value g+ (c
) and the weighting of each evaluation index is expressed using 1 for each control index parameter that is the value, but the assignment evaluation formula is
(c); (c: 1, 2°..., N machine), then E (c) -Σ k+ *g+ (C)
... (1) Here, each control index parameter is a parameter that depends on the traffic flow of each building and the traffic demand of the time period, and in the group management main function task 10 of the group management control unit 1, The control index parameter setting unit B monitors the traffic flow of the building online and sets the optimal control index parameters according to the demand for each time zone.
3, it is set in real time.

Then, the allocation machine selection unit B2 selects the allocation evaluation formula E.
(c); The machine having the minimum evaluation value determined in (c: 1, 2, . . . , N machine) is assigned as the hall call answering machine.

FIG. 5 is an operation flowchart of the control index parameter setting section B3, and FIG. 6 is a control index parameter setting section B3 using a net relation model between the control index parameters, demand parameters, and hall call response time cumulative distribution. 3 shows a flowchart of the evaluation operation.

In Fig. 5, in order to find the optimal control index parameters for a predetermined time period for each building, we first calculate a minute change in each control index parameter within the possible range for a predetermined time period. Δ is changed by 1, and a finite number of combinations p max times are determined for each 1 (
Step Sl).

Each control index parameter P (k, ,, k, □ kl, ......,
), the hall call response time cumulative array T is input to the neural network BIO along with the demand parameters of the current time period through the forward signal transmission process.

is calculated (steps 32 and 33).

Here, T is array data indicating a statistical cumulative distribution of the time spent responding to hall calls in a predetermined time period. In the example, the response time is increased to 2g), and the response time elements that satisfy the range determined in 2 second increments are counted up for each response time.

The hall call response time cumulative array T, (n, □ n2., ... + n1p) for the control index parameter combination P found by this neural network BIO controls the combination P for the demand in the current time zone. This data represents group management control performance when selected as an index parameter.

Next, based on this hall call response time cumulative array data T, a performance evaluation value PE is modeled as a mathematical expression model as an objective function indicating group management control performance evaluation, and the performance evaluation is quantified. In other words, the performance evaluation value for combination P is P
Let Ep be the average waiting time E, , standard deviation SD, , long waiting rate jjp, and maximum waiting time t, , then it can be modeled as PE, -α, *E, +α2*SDe +αN"tle+α4 * i maze (Step S4).

Performance evaluation value PE for all combinations P.

When (p-0 to p□a) is found, select the combination P that minimizes this PE. Binding, this combination P. The control index parameter P O (k+pol k2++o + “””+
kmeo) is set as the optimum parameter for the time period, and the setting is made to the assigned machine selection section B2 (step S6).

However, in reality, the hall call response time cumulative distribution takes various forms for each building, and cannot necessarily be accurately derived from the preset equation (2).

Therefore, it is necessary to use the self-learning algorithm of neural net BIO to correct each coefficient α1 to α4 in equation (2) based on simulation or actual data. Explain the process.

Fig. 7 is a block diagram of the group management neural network showing details of the neural network BIO in Fig. 6, Fig. 8 is a block diagram of the neuron, and Fig. 9 is a diagram of the input/output characteristics of the neuron. The operation of neural net BIO will be explained in detail.

As shown in FIG. 7, the neural network BIO has an input layer 21, an intermediate layer 22 . It is composed of an output layer 23, and each layer has a large number of neurons nu forming a unit as shown in FIG.
The input layer 21 has a network configuration with t.
- intermediate layer 22 - coupled in the direction of output layer 23;

As shown in Figure 6, the group management neural network takes control index parameters and demand parameters as input, and the cumulative distribution of hall call response times as output. In Figure 7, each of these input and output parameters is , control index parameters: 11, 12+..., i.

Demand parameter: I,, +1. ...i,.

Hall call response time cumulative distribution 2 〇１． 02 +... 0.

It is expressed in .

Each neuron nu+ receives inputs 01 + . . . , 05 . . . from other units nLIl as shown in FIG. ...l
When it receives o and l, it converts the input according to certain rules and outputs the result. A variable weight w is attached to each connection portion with another unit nu4. This is called a synapse,
It is used to express the strength of the bond, and changing this value changes the structure of the net. Then, learning of the net can be performed by changing this value.

The output 0+ when the unit nLI+ receives input from a plurality of units null is given by the following equation.

o + = f I (net I)
-(3) Here, net I -Σ WIIOI And this interval f is generally expressed by a 51gl1old function having input/output characteristics shown in FIG. Therefore, each unit nut has nonlinear input/output characteristics.

Next, we will explain pack propagation, which is a self-learning function of a neural network that converges to a constant value while changing the connection strength of each neuron.

The learning process from the output layer 23 to the human power layer 21 in FIG. 7 is pack propagation. The purpose is to converge the strength of the connection of each weight W,'' of the net between the index parameter and the hall call response time cumulative distribution to a constant value.

This pack propagation learning algorithm generally consists of a certain input cover turn O1+'2+ . . . 08 . ..., the actual output when 01 is given. , and the desired output 1+, the weight "W" is changed so as to reduce the root mean square difference E, , E, = (t+ O+) 2- (4).

The amount of change 6w H in the weight connection "wII" of the pack propagation process in which a desired teacher signal corresponding to the input signal in each learning process is given and the weight connection is changed is given by the following equation (5). It will be done.

ΔWz(n+1)−ηδ,01+aΔWz(n)−(4) Here, oI is unit nu, to unit nu.

is an input value to δ, and δ is a learning signal that determines the learning rule, and is calculated by the following formula.

δ+-(t+-Ol) ft (net+)(
(When l is an output unit) δ+-f+ (net+)ΣδmWll (When l is an intermediate unit) However, fl- is the differential value of f.

From the above, the learning signal δi becomes a recursive function, and the learning signal is calculated sequentially from the output layer 23 to the input layer 21 using the output of the output layer 23 and the teacher data as initial values. In addition,
In the above equation (5), η is a learning constant, α is a stabilization constant,
n represents the number of times of learning.

The qualitative meaning of equation (5) is that when a certain input is given, the amount of change Δw11(n+1) in the weight “WI,” is determined by the learning of each unit nul based on the state of each neuron due to the input and the teaching data. The signal δ is determined recursively, and the value obtained by multiplying the amount of change Δw+1(n) up to the previous time by the stabilization constant α,
This means that it is determined by the sum of the current change amount ηδ10.

Therefore, the amount of change ΔW of the weight 'w,1 decreases due to the input/output relationship of the neural network, and the output value 0.
The weight value converges to such that there is no difference between the output value and the desired output value of 1+.

The operation of the learning algorithm by pack propagation of the group management neural network described above will be explained in more detail based on the flowchart of FIG.

For the group management neural network BIO shown in Figure 6,
For each demand parameter value for each set predetermined component building, the combination P of control index parameters is set as the control index parameter in the same manner as shown in the flowchart of FIG.
is changed by a minute amount of change Δ within the range that can be taken, and a combination P is set (steps Sll to 513).

A computer simulation is executed for a predetermined period of time in the constituent buildings for the demand parameter q1 control index parameter combination P (step 514), and the hall call response time cumulative array Tp obtained as the execution result is used as teacher data to direct the learning process. The net weight change amount ΔW is calculated, and the weight values are converged (step 515).

The weight change algorithm calculates the weight change amount ΔW for the 6 weights W of each neuron nLI+ in the neural network BIO shown in FIG. 7 using the above equation (5), and changes the structure of the net (step 516). .

Next, demand parameter q1 control index parameter combination P
(step S17,518).

In this way, for each control index parameter setting in the evaluation calculation in allocation control, by forming a net relationship between the control index parameter and the hall call response time cumulative distribution via the demand parameter, each It is possible to set the optimal control index parameters based on the performance evaluation of the cumulative distribution of hall call response times for the demand for each time zone, so it is possible to perform optimal allocation control for each demand parameter value for each building. Management performance can be improved.

In the above embodiment, the configuration of the group management neural network was configured such that the demand parameter and the control index parameter are input as shown in FIG. 6, and the cumulative distribution of hall call response times is output. It is also possible to adopt a configuration in which the input of the net is only the control index parameter, the output is the cumulative distribution of hall call response times, and the neural network is set for each value of several types of demand parameters, such as long or short average non-response times.

In addition, in the above embodiment, the formation of the neural network is
As shown in Figure 0, we have shown a method of forming a net before the building is in operation by using the results obtained by computer simulation as training data, but even after the building is in operation, the hall call response time, It is also possible to store control index parameters and demand parameters online, modify the weights of the net based on the stored data at regular intervals, and constantly perform self-organization of the neural network. In that case, the same process can be performed by replacing the computer simulation part in step S14 of the flowchart of FIG. 10 with monitoring of the operating state.

[Effects of the Invention] As described above, according to the present invention, there is a net difference between each control index parameter corresponding to a large number of evaluation items in assignment control and the cumulative distribution of hall call response times that directly express group management performance. By constructing a relationship and evaluating the performance of the hall call response time cumulative distribution, it is possible to select a control index parameter corresponding to the distribution value, so that the optimal control index parameter can be set for each value of each demand parameter. I can do it. Moreover, by changing the performance evaluation of the hall call response time cumulative distribution value according to the usage of each building, it becomes possible to set the optimal control index parameters with flexibility to suit the function of each building, and to achieve optimal allocation. Control can be realized.

In addition, the relationships created by the network can be automatically learned in a form that matches the characteristics of each building even after the building is put into operation, making it possible to automatically modify the self-organized net, realizing allocation control that is more tailored to the actual situation. can.

[Brief explanation of drawings]

Fig. 1 is a block diagram of an embodiment of the present invention, Fig. 2 is a software system configuration diagram of the group management control unit and single control unit in the above embodiment, and Fig. 3 is the system configuration of the high-speed transmission line in the above embodiment. 4 is a block diagram of the hall call allocation process of the above embodiment, FIG. 5 is a flowchart showing the operation of the control index parameter setting section of the above embodiment, and FIG. 6 is a block diagram of the group management neural network of the above embodiment. , FIG. 7 is an explanatory diagram of the operation of the group management neural network of the above embodiment, FIG. 8 is a neuron configuration diagram of the above embodiment, and FIG.
The figure shows the input/output characteristics of the above neuron s1gIo1
The graph showing the d function and FIG. 10 are flowcharts showing the operation of the learning algorithm of the group management neural network of the above embodiment. 1...Group management control unit 2-1. ..., 2-N...Single control unit 3...Hall call button 4...Hall call input/output control unit 5...Monitoring panel 6...High speed transmission line 7...
・Low speed transmission line 21. ... Input layer 22 ... Middle layer 23 ... Output layer B1 ... Machine unit processing section B1-1 ... Prediction calculation section B1-2 ... Evaluation calculation section B2 ... Allocation machine selection section B3... Control index parameter setting section BIO... Neural network

Claims (2)

[Claims] (1) In a group management control elevator system that operates a plurality of elevators for a plurality of floors, performs a predetermined predictive evaluation calculation on a common hall call that occurs, and assigns the hall call to the optimum elevator, means for storing demand parameters; demand parameters stored in the demand parameter storage means and control index parameters in each evaluation calculation;
an evaluation calculation network means that forms a net relationship with the hall call response time cumulative distribution; and a control index that determines an optimal control index parameter for each value of each demand parameter using the net formed by the evaluation calculation network means. A group management control elevator apparatus comprising: a parameter setting means; and a group management control means for controlling the assignment of an optimum car to the hall call using the optimum control index parameter determined by the control index parameter setting means. (2) response time cumulative distribution data storage means for storing the control index parameter and actual hall call response time cumulative distribution data for the control index parameter; and data stored in the response time cumulative distribution data storage means after elevator operation. 2. The group according to claim 1, further comprising network correction means for correcting the relationship between the demand parameters and control index parameters formed by the evaluation calculation network means and the hall call response time cumulative distribution based on the evaluation calculation network means. Management control elevator equipment.