JPH07102945B2 - Group management control method for elevators - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial field of application) The present invention relates to elevator group management control, and in particular, the application of knowledge engineering makes it possible to achieve favorable target values in various group management controls. The present invention relates to an elevator group management control method that can be made to follow.

(Prior Art) In recent years, when a plurality of elevators are installed side by side, in order to improve the operation efficiency of the elevators and the service to the elevator users, a response machine to the hall call on each floor is a small computer such as a microcomputer. Allocation is performed reasonably and promptly using a computer.

That is, this is a control method called group management control,
When a hall call occurs, select the most suitable elevator to handle the hall call, assign an elevator that responds to the hall call early, and prevent other elevators from responding to the hall call, or Efficient while switching to a preset operation mode according to the traffic changes unique to the building, such as when the traffic demand increases, such as when going to work or leaving the office in the morning, or during lunch, and when traffic decreases, such as at night Each elevator is controlled to ensure proper operation.

By the way, due to the cost reduction of small computers accompanying the remarkable development of small computers in recent years, small computers have come to be used not only for group management devices but also for elevator control devices for controlling a single elevator.
A serial transmission method is becoming the mainstream as a method of transmitting information to these computers. And
The group management control device and the elevator single control device can freely exchange group management data with each transmission device simply by wiring (or optical cable connection) by a certain procedure by software.

In addition, even in a large-scale building in which a large number of elevators are group-controlled, a building management computer or OA (office automation) computer manages the entire building and exchanges information between floors. There is. Many of these information are related to building traffic. The information of the group management control computer also includes hall data (for example, passenger detection in the hall) necessary for building management. However, such building management data and group management control data are hardly exchanged. For this reason, there are cases in which sensors and notification devices for obtaining information are attached to the respective cases.

However, in general, elevator group management control is centered on control of allocation for hall calls and special operation control for high traffic demand.

The allocation control for hall calls was conventionally performed by evaluation using various calculation data such as arrival time. For this reason, the allocation may have failed due to the prediction failure.

In order to eliminate such a situation, there have been studies such as learning the daily traffic flow and demanding the reliability of the prediction data.

(Problems to be solved by the invention) However, in an elevator system, it is impossible to completely predict stochastic hall calls and derived car calls, no matter how much learning is performed as described above. Yes, the data is unreliable. However, even if the reliability of the data is low, a value of some order is obtained by the prediction calculation, and it is unavoidable to incorporate it into the evaluation calculation.

This means that there are many possible failures in the evaluation decision. This also indicates that there is a discrepancy between the conditions set using the predicted values and the conditions that the true designer thinks, and the designer's condition settings are more stringent and complete. When it comes to aiming, it tends to be a conservative slow motion rather than a light motion that the designer intended.

Also, when incorporating the designer's idea into the evaluation operation of the allocation, it is not possible to wipe out the tendency that the function value is changed, the weight is changed, etc., and the function is indirectly incorporated and becomes sluggish.

In addition, this indicates that the tendency is less likely to appear in the verification of the effect.

Even in high-demand special operations of elevator group management control, due to the ambiguity of forecast data, it tends to be a conservative control rather than a fine-grained control as described above.

Therefore, the purpose of the present invention is to directly use the knowledge of the designer, that is, the expert of group management control for control, to realize fine control, and at the same time, make the ambiguity of the condition that is the basis of the control. This makes it possible to reduce the number of allocation control failures, and also enables fine-grained control that can consider the effect on the system of each floor when predicting the ambiguity and correctness of the control instruction. Another object of the present invention is to provide a group management control method for elevators.

[Configuration of the Invention] (Means for Solving Problems) In order to achieve the above object, the present invention is as follows. That is, in group management control of elevators that collectively control multiple elevators, when deciding various control instructions such as assigning a call to a generated hall call, use a rule of thumb to achieve the intended control purpose. Based on the control instruction contents based on various conditions, and determine the importance of each floor and the certainty factor for the various conditions, prepare a membership function that matches the certainty factor, and for each generated hall call Obtain the index value when performing the target control including the importance of each floor, to the control instruction corresponding to the condition determined by this index value using the membership function corresponding to the condition and its certainty The effectiveness is calculated by inference when the degree of belonging is obtained and it is assumed that the control instruction is given, and the control instruction is determined according to the result.

(Operation) That is, when determining various control instructions for the hall call generated in the group management control, the control instruction content under each condition based on the empirical rule set to achieve the intended control purpose If the index value when provisionally assigned is obtained including the importance of the floor, the above-mentioned condition determined by the above-mentioned index value is determined using the membership function that matches the importance of the floor where the hall call occurs and the certainty to various conditions. The degree of belonging to the control instruction is calculated, and the effectiveness when assuming that the control instruction is given is obtained by inference that takes this degree of belonging into account, and the final control instruction of each unit is determined according to the result. .

In this way, the final control instruction for the generated hall call is determined using the control instruction content and the certainty factor based on the empirical rule, and the membership function that matches the certainty factor, but the empirical rule and certainty factor are specialized. By considering the experience and knowledge of the house, that is, the expert of group management and control, the tendency of the change in the operation mode of the elevator based on the experience of the expert is added to the ambiguity, and the demand floor You will be able to infer including the importance. Therefore, it is possible to judge a situation that cannot be inferred from only actual information.

As a result, since the knowledge of the designer, that is, the expert of group management control can be directly used for control, fine control can be realized, and at the same time, selection can be made in consideration of the ambiguity of the conditions that are the basis of the control. Therefore, it is possible to reduce the failure of allocation control, and in predicting the ambiguity and correctness of the control instruction, the elevator group management control method that enables fine control that can consider the influence on the system of each floor Will be able to provide.

(Example) Hereinafter, one example of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a group management control system according to an embodiment of the present invention.

In the figure, 1 is a group management control device that controls the entire group management control, 2 is an elevator single control device that controls the operation of a single elevator, and 3 is a transmission controller that exchanges data with the outside. Is a monitor for monitoring the operation status of the elevator. The group management control device 1 is connected to the elevator control device 2, the transmission controller 3, and the elevator monitoring monitor 4 by a system bus using a serial transmission line using a dedicated LSI for transmission. Also,
I / O (input / output) coupling between the hall gate, lamp sensor, display, etc. and the hall input / output device 5 is performed by a serial transmission line. This depends on the transmission-dedicated LSI and general-purpose transmission software. Further, the in-car controller 6 provided in the car and the elevator control device 2 corresponding to the in-car controller 6 are also connected by a serial transmission line. Further, the data of the building management computer 11, the data of the OA computer 12, the data of the time recorder 13, and the information provided from the notification data input terminal device 14 via the I / O controller 15, for example, the use of the conference room, the end, etc. The notification data and the entrance counter data (the number of persons entering and leaving) are combined by the interface of the transmission controller 3 and transmitted to the serial system bus on the elevator system side. As a result, the group management control device 1 can be provided with the various types of data described above, as well as data relating to group management to the computers 11 and 12.

This system is an example close to the maximum scale. Therefore, the present invention can be applied even to a system in which a part is deleted.

Next, the software configuration of the group supervisory control device 1 will be described. As shown in FIG. 2, the software configuration for group management control comprises various functional tasks, their sub-routines, and a task management program for managing these tasks. Then, a small computer (generally a microcomputer is used in the group management control device 1,
When the microcomputer is started, the microcomputer first starts and then executes the task management program 20 to determine which task, that is, the software module separated according to the function. A task is a software module for each function and is activated by a condition.

Here, each task will be briefly described. Reference numeral 32 is an initialization task. This task initializes RAM (random access memory) and CPU (processor) registers, and LSI (large scale integrated circuit) initialization tasks. Is activated when is switched.

21 is an external input task, which is an input task for setting external inputs such as CCT and KCTHCT on a RAM (random access memory). This task has a high priority, and when it is started, it will restart every 100 ms.

Also, the trigger check and the like are performed in this task.

Here, the HCT is a hall condition table to which the data registered by the elevator controller for hall calls is input. CCT is a car condition table and KCT is a car condition table. Now, assuming that there are four elevators A to D in the group and the service floors are 1 to 8 floors, the above HCT, CCT, KCT Have bit configurations as shown in FIGS. 3, 4, and 5, respectively.

That is, in the HCT showing the hall state shown in FIG. 3, 8-bit information from the 8th floor down (8D) to the 7th floor up (7U) for the Hall sub index (HS) of 0 to 13 is obtained. It is stored. The state of the hall on each floor will be specifically described.

For example, when the ascending call hall button is pressed in the elevator hall on the 5th floor, the 7th bit of ll (5U) of HS becomes 1, and the service elevator corresponding to this hall call is determined to be Unit A by the method described later. Then, the 0th bit and the 6th bit in ll of HS become 1. When Unit A arrives on the 5th floor, all bits 0, 6, and 7 in ll of HS are reset to 0. That is, the 0th to 3rd bits indicate the elevator machine set, the 6th bit indicates whether or not the elevator corresponding to the hall call is allocated, and the 7th bit indicates the presence or absence of the hall call.

In the CCT representing the car state in FIG. 4, 16-bit information for each of elevators A to D is stored for indexes 0 to 3. That is, 0-3
The load condition of the car is indicated in binary notation on the number bit.
The meaning of these 0th to 3rd bits is "0001", "0010", "001"
1 "," 0100 "," 010 "," 0110 "," 0111 "," 1000 "," 1001 ",
0-10 for "1010", "1011", and "1110" respectively
%, 11-20%, 21-30%, ~ 31-40%, 41-50%, 51-60
%, 61-70%, 71-80%, 81-90%, 91-100%, 101-110
%, 111% or more.

The 5th bit shows the running state of the car, "1" is running,
“0” indicates deceleration. Bit 7 indicates the open / closed state of the door, "1" indicates opening and "0" indicates closing. Bits 8 to 13 represent the car position in binary notation. Bits 14 and 15 indicate the moving direction of the car. “10” indicates rising, “01” indicates falling, and “00” indicates no direction, that is, stopped.

Fig. 5 shows the KCT that represents the car call state.
In the same manner as in the HCT shown in FIG. 3, 0 to 3 bits indicate whether or not there is a car call to elevators A to D.

As a result, the elevator and hall call states are input.

22 is an assignment task, which is a task for assigning a car. This task checks the newly generated hall call every 100msec, and if it occurs, evaluates the damage due to the estimated unresponsive time, fullness, etc., and decides the best machine for evaluation.

26 is an assignment review task, and this assignment review task 26
Is a low-level task that is activated approximately once per second, and is used to change the allocation of long-waiting, full-occupied, or predicted hall calls.

28 is each single elevator communication task for communicating with each single elevator. In addition to cyclic data transmission, this task includes control output (allocation, allocation cancellation, etc.) and Data requests such as the number of passengers, the number of passengers getting off, and the call for a new car are made. These are performed using the buffer, and the data shown in FIG. 6 is carried as shown in FIG.

29 is an annual timer and various timers, which are 10msec, 100
It is a routine of various interval timers such as msec and 1 second, and an annual timer combined with them. Also, these data are corrected by an external timer. The yearly timer has information such as month, date, day of the week, holidays, six days, and other events. Information is provided by the second I / O task for the floppy disk and the first I / O task for the CRT. Will be updated.

Reference numeral 30 is a first I / O (input output) task for CRT (character display terminal) transmission, and this task is used for transmitting information to an external terminal, another computer or the like. This task is time-sliced and activated at a low interrupt level so as not to harm other group management masks. 31 is the second I / O task for floppy (flexible) disk control,
This task is activated when learning data and the like are stored in an external floppy disk. First I / O task for CRT transmission
It is activated at a low interrupt level similar to 30.

27 is a learning data processing task.This learning data processing task sets the state at that point in the data table according to external input or data from a single elevator, and when it changes to the next state, etc. A task for exchanging data, which is activated when the data changes or the state changes. Also, this task is a low interrupt level task and is activated so as not to harm the high interrupt level group management task. However, it changes when a special flag or priority is changed. The learning data is classified into several equivalent traffic modes according to the elements such as month, day of the week, six days off, holidays, and time zone (time band) as shown in Fig. 8, and the following data is provided for each mode. .

Examples of these are shown in FIGS. 9 and 10.

Where HCT $ RAT is the average number of hall calls in 15 minutes, KCT $ RAT is the average number of car calls, IN $ RAT is the average number of passengers, OUT $ RAT is the average number of descendants, and KCT $ SET is the average number of car calls for each floor. Indicates the incidence.

HCT $ RAT, ~ OUT $ RAT is the index floor index HS
(Hole Sub Index).

KCT $ RAT is indicated by a matrix of A and B to store individual car call occurrence rates from the A floor to the B floor, for example.

Moreover, when the demand is high, those changes are learned at fine intervals. This is AV $ MENSP for each HS and t
It is indicated by (HS, t).

However, t is time.

These learning data do not have to be directly used in the present invention, but they contribute to the improvement of accuracy in performing various prediction calculations, so that better results can be obtained. Needless to say, even if there is no learning data, control can be performed with a certain degree of accuracy.

As other tasks shown in FIG. 2, the data collection task 34, which is activated every 1 second and communicates with an external computer for data input and output and data collection by the building computer etc. There is a traffic demand forecasting task 33 (starting every 100 msec) that uses data to anticipate demand, predict traffic demand, and determine a driving model. There is a driving task 35.

In the system of the present invention, the group management control device 1 is further provided with a hall call assignment control function and a traffic model adaptive control function using inference described below.

Here, inference means that the "certainty" of data conditions is taken into consideration when performing such control, and the degree of contribution given to the instruction content given by the conditions including these and the judgment based on the instruction content is determined by the expert of elevator control. It refers to a method of determining the final instruction by making a comprehensive judgment based on the knowledge possessed.

The traffic model adaptive control is, for example, data of the demand situation (including learning data) that changes due to work hours, lunch, weekdays, holidays, nights, etc. Based on the data such as the special demand of the floor where the conference room is installed (this transient data is obtained by, for example, making a reservation to the building management computer 11), the specific floor of the building for each time zone of the day This refers to performing demand forecasting such as concentration of users and distribution from specific floors to various floors, and performing control that is in conformity with a pre-fetched demand forecasting model to cover the forecasted demand.

Next, a description will be given of a control routine, which is the core of the present invention, by inference from direct expression of expert knowledge.

Group management control is a major control element that involves allocation control for hall calls and special operations when demand is high. In addition, there are those in which the demand is concentrated and those in which the demand is diverged, or a combination thereof, and appropriate allocation control can be considered for each. Therefore, in the present invention, as shown in FIG. 11, the F1 routine that determines the operation mode correspondingly, and the allocation control corresponding to each operation mode OP1 to OPn F2-1 to F2- In controlling m, a control routine based on inference of a direct expression of expert knowledge is also used.

In the following, the control method is the allocation control F2 during balancing.
The explanation will focus on -1.

In the present invention, the inference routine is the main part and is commonly used for F1, F2-1 to F2-m in FIG. 11, so this part will be explained first.

In the inference routine, an error e with respect to the target value if a control instruction is given to a certain machine
(Deviation) ratio, that is, the ratio of the possibility of long wait occurrence and the error increment Δe generated due to the control instruction, that is, how much the possibility will be increased Is configured to determine its output Δu. Further, regarding e and Δe, there are some predicted values depending on the subjectivity of the expert and the certainty factor based on the probability data.

In the allocation control routine at the time of the balance pattern, as the target value, (1) decrease long waiting calls (0 value for waiting 60 seconds or more) (2) increase good calls (total number of calls that responded within 30 seconds) ( 3) Maintain good service on high demand floors.

(4) Eliminate full capacity. (When the load is 80% or more) is set.

Here, the above (1) will be described in detail below. still,
The same applies to (2) to (4). First, with respect to the target value of (1) (reduce long waiting calls), whether or not to reduce the value according to the error amount e and the error increment Δe is shown in FIG. It is determined from the condition-instruction table of the long waiting part, which is an artificial criterion obtained from experience. In Fig. 13, PB, PM and ZZ are
It is the level of e, Δe, PB is positive and large, PM is positive and medium, and ZZ is almost zero. Further, the output Δu indicates the determination content, PO is assigned, PS may be assigned,
ZO usually means that NS does not need to be allocated, NE does not.

Further, FIG. 12 shows a flowchart of the inference part.

That is, if a hall call occurs on a certain floor, the group management control device 1 responds to this hall call by step ST12.
In -1, a control instruction is temporarily set. For example, suppose that a hall call is assigned to Unit A. Next, in the routine of step ST12-2, how many long-waiting hall calls will occur in Unit A due to the allocation,
The amount of increase is predicted for each hall call based on the degree of error e and the degree of error data Δe with respect to each condition.

Here, a description will be given of a portion for obtaining the degree of belonging of data to each condition of each error amount e and error increment Δe.
Here, the degree of belonging is calculated from the predicted value (TP2) (that is, the certainty of waiting for a long time) of a long wait for a hall call on each floor and the importance (HS $ W) of those floors affecting the system.

The importance of this system is that it is expected that high demand will occur in the floor, etc., and the long waiting time of that floor will cause frequent car calls, which may impair the linearity of the system or make it impossible to predict. On the other hand, in consideration of the influence, the value is replaced with a numerical value. This is the HCT learned
It is effective to predict from data such as $ RAT, KCT $ RAT. When the learning data is not used, the designer's subjective amount or experience value is used.

By the above TP2 (possibility of long wait) and HS $ W (importance of floor), the density of states x is However, i = 1 to n: Means to perform all hall calls. HSi: Hall sub-index with a hole designation.

In addition, the degree of belonging y is y = MS by the degree of belonging function MS (x).
It is calculated by (x). An example will be described with reference to FIG. (MS is, for example, as shown in Fig. 15 (c), various types are prepared according to the certainty of the data, and the most suitable MS is used depending on the conditions.) Basket a in Fig. 15 (a) The status is shown. Also in b
TP2 (possibility of long wait), c to floor importance HS $ W, d
The floor utilization is shown in. In this example, the fifth floor is a floor with a reception desk, so the importance is high.

In addition, there is a hall call on the 4th floor up (HS = OEh; h is a hexadecimal number) and the 5th floor down (HS = 07h). Here, when the degree of belonging, which is an index of long-term waiting, is calculated, y ₁ = MS (x ₁ ) = MS (0 + 30 × 1.2) = MS (36) = 0.07.

Next, Fig. 15 (b) shows a case where a newly generated hall call for the 6th floor up direction is provisionally assigned. The degree of membership y _{2 in} this case is y ₂ = MS (x ₂ ) = MS (0 + 60 × 1.2) = MS (72) = 0.91. This is shown by a membership function as shown in FIG. 15 (c).

Next, inference is performed based on these, and the hall call assignment control based on this inference will be specifically described.

The example here is to reduce the number of long waiting calls. This is based on the goal of "zeroing long waiting calls for 60 seconds or more" and assuming that a newly generated hall call is tentatively assigned. It is necessary to investigate the magnitude of the possibility of long wait, and as a result, it is difficult to assign to a machine that has a high possibility of long wait. Therefore, the strength to be assigned to each car is obtained as a value, and the car having the strongest strength (the largest value) among the cars is finally allocated for the hall call.

This allocation control, when a new hall call occurs,
Elevator group currently operating under group management, for example A,
Assuming that there are three units, B and C, the procedure is as follows. That is, which of the three cars is to be assigned a new hall call is determined by inference operation.

The inference calculation of the allocation control is started by first performing temporary allocation to the newly generated hall call from, for example, the No. A machine. That is, y ₁ , y ₂ , ...
Start by asking for.

Then, the error e and the error increment Δe with respect to the target value are obtained for all hall calls including the already allocated portion of the machine that is changed by performing this temporary allocation, y ₁ , y ₂ ...
It is determined from the degree of membership function shown in FIG. 21 by using the combined value F which is the added value of

However, the membership function shown in the figure is an example and is artificially determined, and the optimal one is artificially selected for use.

That is, the combined value F before and after the temporary allocation to the machine A is performed.
Then, the increment ΔF _a of the values F _a1 and F _a2 and the combined value F _a2 is obtained.

For example, the value F _a1 before temporary assignment, increment [Delta] F _a new value of the temporary assignment from the value F _a2 after temporary allocation can be determined by ΔF _a = F _a2 -F _a1. Then, from that F _a2 and ΔF _a , the first
The condition-instruction corresponding to the error e, which is based on the artificially set empirical rule as shown in FIG.
The error increment Δe is calculated.

If the membership function used at that time is as shown in FIG. 21 and if the above F _a2 , ΔF _a is like the value shown in the figure, the error increment Δe is calculated from this F _a2 , ΔF _a. Has a degree of membership of 0.5 for condition ZZ and a degree of membership for condition PM.
0.5, and the error e has a degree of membership of 0.1 for condition ZZ,
Then, the degree of belonging is 0.9 for the condition PM.

That is, here, when considering whether or not a certain object is an element of the set A, it is not strictly divided, but is a function indicating the degree of membership to consider the degree of being an element of the set A. To use. This is as shown in FIG. 21, in which the horizontal axis shows the above-mentioned combined value F and the increment ΔF of the combined value, and the vertical axis shows the error amount e and the error increment Δe. Also, in the figure, as a set, set ZZ, set PM and set
The respective membership functions (one example) of PB are shown. set
ZZ indicates a set that is “generally zero”, set PM indicates a set that is “positive and medium”, and set PB indicates a set that is “positive and large”.

Each membership function is a set of respective values ZZ, PM, PB at all combined values F or increments ΔF of combined values.
Gives the degree included in. This degree indicates the degree of belonging to the above set, and is also called the degree of belonging, and is `` 0.
It is indicated by a number between "0" and "1.0".

For example, a degree of membership of 1.0 indicates that the target is a complete element of a set called A, and a degree of membership of 0.0 indicates that the target is not a member of a set called A. ing.

Here, let us consider the degree of belonging when the combined value F is f. As can be seen from FIG. 21, the degree of membership of the set PB with respect to the combined value f is 0.7, and the degree of membership of the set PM is 0.3.

That is, the combined value f belongs to the set PB that is "positive and large" with the degree of membership of 0.7, and belongs to the set PM that is "positive and medium" with the degree of membership of 0.3.

The membership function also changes depending on the overall confidence.
That is, when the data is accurate, the sections are clearly separated as shown in FIG. 22 (a), and conversely, when the data are incorrect, the sections are gently smoothed as shown in FIG.

When the applicable condition and the degree of belonging are obtained in this way, the degree of the control command is then determined from the error e and the error increment Δe described above. That is, since the newly assigned hall call is temporarily assigned as the control command Δu to the elevator No. A to be controlled, the strength of the hall call provisionally assigned to the no. Whether or not to assign is determined based on the condition-command table.

The condition-command table is as shown in FIG. 13 (a), which shows the contents of the control command .DELTA.u corresponding to the error e and the error increment .DELTA.e determined based on the experience and knowledge of the expert.

For example, when the error e is “approximately zero” (set ZZ) and the error increment Δe is “positive and large” (set PB), the control instruction Δu is PO, PS, ZO, NS, NE. There are 5 types, PO is "allocated", PS is "allocated",
ZO means “normal”, NS means “no need to allocate”, NE means “not allocate”.

The number of sets as error e is set PB, set PM, set
There are three types of ZO, and the same is the case of the error increment Δe. There are nine types of control commands Δu depending on the combination of the error e and the error increment Δe. Therefore, the nine types of rules are considered here.

These nine types of rules are determined by experience, and details are shown in FIG. 13 (b). In the figure, for example, when "rule 1" is "error e is" positive and large "(PB)" and "error increment Δe is" positive and large "(PB)", the control command Δu is " Not assigned. ”(NE).

Already the combined value F _a2 and increment ΔF _a for the temporary allocation of Unit A
The degree of membership of each set is obtained, and the combined value F _a2 has a degree of membership of the set PM of 0.9 and a degree of membership of the set ZZ of 0.1. The degree of membership of the increment ΔF _a with respect to the set PM is 0.5, and the degree of membership with respect to the set ZZ is 0.5.

Therefore, the combined value F _a2 belongs to the set PM and the set ZZ, and the combined value increment ΔF _a also belongs to the set PM and the set ZZ. Therefore, in order to determine the control command Δu from the error e and the error increment Δe, the “rule 5”, the “rule 6” in FIG.
"Rule 8" and "Rule 9" are used.

As shown in FIG. 14 when these rules are shown in the figure, with respect to “Rule 5”, for error e, the set PM is satisfied with a degree of 0.9, and for the error increment Δe, the set PM is with a degree of 0.5. It is filled. The degree to which "Rule 5" is satisfied is
The smaller value of the two sets that is satisfied is adopted. Therefore, the degree of strength of the control command Δu is limited to 0.5 in this case.

In the same manner, a set indicating the control command Δu is obtained for “Rule 6,” “Rule 8,” and “Rule 9” including the degree of strength.

Then, the logical sum of the sets of the control commands Δu respectively obtained by these four rules is calculated, and the weighted averages are calculated according to the degree of belonging to the set to obtain the final control command strength U. Here, as shown in FIG. 14, the value of the control-specified strength U is -0.69.

This control command is referred to as "hard to allocate", meaning that the larger the value of the strength U, the easier the allocation, and the smaller the value, the harder the allocation.

As described above, the strength of assigning a newly generated hall call to Unit A is required.

Hereinafter, the strength U to be assigned to other units is obtained by the same method.

This completes the explanation of the inference part.

In this way, conditions-based on the expert's knowledge-obtain instructions for each condition using the rule of thumb, and obtain the obtained instructions to the extent of reflecting the ambiguity of the conditions. Based on the logical sum of the instructions, it is possible to determine a comprehensive instruction based on experience and ambiguity. This routine can be used for F1, F2-1 to F2-m in FIG. 11 by replacing the reference table in step ST12-3.

Next, the allocation control will be described, including the part for obtaining the predicted value and the certainty factor of the data for performing the above inference.

This routine will be described by taking an example of a predicted arrival calculation in hall call allocation control. The algorithm of this routine can be used as it is for predictive load calculation.

FIG. 16 shows a flowchart of the hall call assignment control routine. The inference part described above corresponds to step ST16-4.

A group management control device 1 detects a hall call and detects it
Enters the hall call allocation control routine. Then, step ST16-1 is executed first, and it is determined here which machine is not in a state such as out-of-group or in a rest state, that is, which machine is incapable of hall call allocation control. And
For the impossible units, set them so that they are not subject to calculation. Next, in step ST16-2, various prediction calculations are performed. Here, a predicted arrival time, a predicted load calculation, etc. are performed and used as inference data.

A flowchart of the predicted arrival time routine is shown in FIG.

In FIG. 20, when step ST20-1 is executed, the hole sub index (HS) where the machine is located is set. Next, the process proceeds to step ST20-2, where a derived call such as a derived car call for the hall call or the like is generated (if there is learning data, it is generated by referring to it).

Then, the process proceeds to step ST20-3, where the predicted arrival time is set. The estimated arrival time (RESPT (HS)) is calculated by the following formula.

However, RANT (Stai / Endi) is the i-th start floor
The running time from Stai to the next stop floor Endi is shown, and LOST (Endi) shows the loss time at the scheduled stop floor Endi. In addition, KEIKAT (HS) indicates the elapsed time since the assignment was set for the HS that generated the hall call. l indicates the repetition up to that floor.

Here, in order to obtain the possibility that the car will arrive within the predetermined time on the floor with the hall call "Yes", step ST20-5
At, set the maximum and minimum arrival times and their probability distribution modes. The minimum arrival time is the minimum arrival time for an existing call, and the maximum arrival time is the maximum arrival time when a call occurs on each demand floor. The probability distribution mode is a part for setting a mode indicating the probability distribution of arrival time (see FIG. 18). Assuming that there are two kinds of arrival time probability distributions as shown in FIG. 18, the selection is performed as follows. In other words, as a result of responding to the hall call assigned to this car, there is a possibility that the car call that indicates the destination floor specified by the passenger who boarded the car and the derived call obtained by prediction based on the learned data may match. Determined by the size of.

For example, as shown in Fig. 17a, as a result of responding to the hall call of 5FUP (5th floor ascending call) assigned to this car, the car call on the destination floor designated by the boarding passenger is 12F.
If it is (12th floor), the possibility that this will coincide with the derived call is predicted because this 12th floor is the top floor, so there is little demand, and in this case, the possibility of the 12F derived call is small. The probability distribution of arrival time is in the mode of (a) in FIG.

That is, the probability distribution of the predicted arrival time is averaged over the length T _{L2 on} the left and right of the data Tx of the predicted arrival calculation time RESPT, and is between the minimum arrival time Tmin and the maximum arrival time Tmax.
However, since Tmin is closer to the data Tx of the predicted arrival operation time RESPT, it has the form of (a) in FIG. Further, since it is a probability, the area S ₁ · S ₁ ′ is S ₁ = S ₁ ′ in order to make the distribution state symmetrical. Moreover, it is normalized so that S ₁ + S ₁ ′ = 1.

Also, when the car is in the situation shown in Fig. 17b, considering the derivative call of 1F (1st floor) to the 8FDN (8th floor descending call) hall call assigned to this car, The probability of matching the actual car call to the derived call is because the derived call is likely to be due to the base floor.
T _L2 = 0, and therefore the form shown in FIG. 18 (b) is obtained.

As described above, the probability distribution of the predicted arrival time is obtained.
It should be noted that these shapes are examples and may be changed depending on the state.

The data thus obtained is set in the RAM of the group management control device 1 in the format shown in FIG. 19 (a) and used as information on the possibility of long wait.

Next, the process proceeds to step ST20-6 in FIG. 20, and this is performed for each floor, and the process ends when one round has been completed. The same applies to the predictive load routine. Thereby, information on the possibility of long waiting for each floor is obtained.

As a result, the various prediction calculations in step ST16-2 of FIG. 16 are completed, the preprocessing of step ST16-3 is performed, and the data is arranged in the format of (a) of FIG. 19 for convenient inference. .

In the inference example described above, the estimated arrival time was used within the order of 30 seconds, 31 seconds to 59 seconds, and 60 seconds or more, so this possibility is shown in FIG. 19 (b). , TP1, TP
Process as 2.

Based on the strength of the allocation thus obtained, in order to know the one with the largest value and to formally allocate the hall call to that car, the allocation is registered in the elevator controller of the car.

Of course, when an evaluation value or the like in consideration of various other conditions is also used, the hall call is assigned to a machine having a high overall evaluation including those values.

As described above, according to the present invention, various group management controls are performed by inference based on the direct expression of the empirical rule of knowledge of experts. Moreover, the strength of the control is changed according to the degree of ambiguity in establishing the empirical rule. For this reason, fine-tuned group management control is possible, and since the ambiguity of data is used at that level, allocation failures are reduced and ideal control is realized. Also,
Since the inference is performed considering the influence of the system on each floor, the direct macro control instruction appears as a fine-grained control effect for each floor.

It should be noted that the present invention is not limited to the embodiments shown in the drawings and can be carried out by appropriately modifying within the scope of the invention without changing the gist thereof. As described above, the heuristic reasoning routine can be used for the special operation selection routine for high demand, and fine control can be performed. In this case, traffic demand is classified into concentrated demand, divergent demand, and combined demand, and the size and duration of the demand are predicted by macro data such as learning data or micro data currently being collected. Then, the degree of certainty (the degree of belonging) is obtained, and the same inference is performed to make a selection decision.

In addition, as an application of the inference part, if a direct expression of these empirical rules is used and it is a self-growth type system that modifies and adds itself according to good or bad results, the performance becomes higher. In addition, it is more effective to prepare various kinds of membership functions for obtaining the degree of membership in the inference part described above so that the functions are suitable for the purpose.

[Effects of the Invention] As described in detail above, according to the present invention, the knowledge of the designer, that is, the expert of group management control is directly used for the control, and the fine control is realized, and at the same time, the basic control is performed. Since it is possible to make selections that take into consideration the ambiguity of the existing conditions, it is possible to reduce allocation control failures, and also consider the effect on the system of each floor when predicting the ambiguity and correctness of the control instruction. It is possible to provide a group management control method for elevators that enables fine control.

[Brief description of drawings]

FIG. 1 is a block diagram of a system used in the present invention, FIG. 2 is a diagram showing a configuration example of its program module, and FIGS. 3 to 6 and 8 to 10 are data used by this system. FIG. 7 is a diagram showing an example of the table, FIG. 7 is a diagram for explaining the data communication buffer, and FIGS. 11 to 22 are diagrams for explaining the operation of the main part of the present invention. 1 ... Group management control device, 2 ... Elevator control device, 4
...... Elevator monitoring monitor, 6 …… In-car controller.

Claims (1)

[Claims] 1. In an elevator group management control for controlling a plurality of elevators in a unified manner, in order to achieve an intended control purpose when determining various control instructions such as assigning a generated hall call to the hall call. Set the control instruction contents under various conditions based on the empirical rule, determine the importance of each floor and the certainty factor for each condition, prepare the membership function that matches the certainty factor, and generate halls for each unit. The index value when the target control is performed for the call is obtained by including the importance of each floor, and the condition is determined by the index value using the corresponding membership function corresponding to each condition and its certainty factor. A feature is that the degree of belonging to the corresponding control instruction is obtained, the effectiveness when assuming that the control is performed is obtained by inference, and the control instruction of each unit is determined according to the result. Beta group management control method.