JPS594583A - Predicting system of traffic demand of passenger of elevator - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to an elevator passenger traffic demand prediction method useful for elevator group management control.

[Technical background of the invention and its problems]

In elevator group management control where multiple elevators are managed as a group, it is important to decide which elevator cage will respond to a hall call that occurs on the service floor.
This is an important issue in achieving efficient operation of the above-mentioned electric pianos and evenings.

For example, if you simply assign the elevator car that can respond to a hall call the quickest, this will inevitably create an obligation to transport the passengers who board the car, and there is a risk that you will not be able to respond to other hall calls. . Therefore, when attempting to respond to a hall call, it is very important to accurately predict new calls, that is, derivative calls, by passengers boarding from the floor where the hall call was made.

However, if it is possible to accurately predict this derivative call, for example, the car call occurring in each of the multiple elevator cars that can respond to a hall call, and the derivative call predicted for the hall call. By contrasting this with the elevator car and causing the most matching elevator car to respond to the hall call, it becomes possible to respond to the existing car call and simultaneously process the derived call. In other words, the processing efficiency regarding the operation control of the elevator cage is increased, and the number of floors at which the elevator car stops can be reduced to improve the operation efficiency.

In addition, as mentioned above, the operation efficiency is high, and since an appropriate elevator car is assigned to a hall call, the average time from the generation of a hall call to the response of the elevator car (average unanswered time) can be shortened. It becomes possible. As shown in this example, if the derived elevator calls can be accurately predicted, elevator group management control can be performed extremely effectively.

However, accurate prediction of the above B1 derivative calls is extremely difficult, and various studies have been conducted in the past.

For example, the hall call obtained from the previous elevator operation and the derivative call that occurred when responding to the hall call are recorded, and the current derivative call is predicted based on the call. However, this prediction method is very useful when the elevator traffic volume is constant and stable, but when the traffic demand changes greatly depending on the time of day, such as in an office building, etc. However, there is a problem that the prediction accuracy is significantly reduced and it cannot be used effectively. Furthermore, there is a problem in that the above-mentioned traffic demand cannot be predicted because the transportation length cannot be accurately grasped only by recording derived calls.

[Purpose of the invention]

The present invention has been made in consideration of the above circumstances, and its purpose is to accurately predict derived calls when responding to a hall call, and to effectively perform group management control of elevators. The object of the present invention is to provide a highly practical method for predicting passenger traffic demand for elevators, which enables efficient operation of elevators.

[Summary of the invention]

The present invention measures the traffic volume of passengers using elevator evacuation between a pair of service floors, predicts the average traffic volume per unit time, and calculates the hall call from the hall call for the elevator cage and the time when the hall call occurs. The arrival time of the elevator car at the floor where the hall call occurs is predicted, and derived calls at the floor where the hall call occurs are predicted from these prediction results.

〔Effect of the invention〕

Therefore, according to the present invention, derived calls are predicted from the average traffic volume per unit time at the predicted arrival time on the floor where the hall call occurs, so the prediction results are extremely accurate. Therefore, it becomes possible to accurately allocate elevator treads to hall calls, effectively perform group management control of elevators, and operate the elevator efficiently and smoothly. Moreover, since the above-mentioned derived call prediction is performed using the average traffic volume per unit time for each service floor pair as a pace, highly accurate prediction results that accurately capture traffic demand fluctuations can be obtained. Therefore, its practical advantages are extremely high, and it has great effects on group management control of elevators.

[Embodiments of the invention]

FIG. 1 shows an example of fluctuations in traffic demand for passengers using elevators, and shows temporal changes between pairs of service floors. There are nine turns in elevator traffic demand that show a temporal tendency, and for example, the traffic demand increases at the time of dispatch, lunch break, and leaving work. In general, it can be said that there is almost no traffic demand at night.

In this method, such fluctuations in traffic demand are measured by measuring the inter-floor traffic volume of passengers using elevator cages, and then calculating the average traffic direction per unit time between service floor pairs according to the above-mentioned measured traffic volume. The prediction is held while updating, and this is used to predict derived calls.

FIG. 2 is a schematic configuration diagram showing an example of an elevator control device that implements the embodiment method, and is realized using, for example, a part of a computer system for controlling elevator operation. However, it is of course possible to configure the prediction device itself independently. A computer system 1 that controls the operation of an elevator generally includes a control section 2 equipped with an accumulator, various instruction registers, etc., a storage section 4 connected to the control section 2 via a line 3, and a passenger counter. It is composed of a bufferless recorder 5 and the like. Sensors provided in various elevator cages, external storage devices, etc. are connected through this crowd buffer register 5. Reference numeral 6 in the figure is an inter-floor traffic prediction device, which predicts and calculates the average traffic volume per unit time between a pair of service floors of the elevator car, according to the information given through the input/output/interface sensor 5. This is given to the control section 2. The data sent and received between the prediction device 6 and the control unit 2 has a format in which, for example, the upper 4 bits of the 8-bit data indicate the departure floor i, and the lower 4 bits indicate the arrival floor. Floor j is shown to indicate the floor pair where the elevator is used. Further, the transportation amount for each floor is expressed as, for example, the value (number of people) mh obtained by converting the passenger weight into the number of people.

However, in this example, since lXj is represented by 4 bits, the maximum number of service floors is 16, but by improving the data structure, it is possible to obtain a higher number of service floors. It may be applicable to In addition to the data described above, control signals such as REQ and ACK are also transmitted and received between the computer system I and the prediction device 6, but since they are not directly relevant to the main discussion here, their explanation will be omitted.

Therefore, the subtraction machine system 1 has a table as shown in FIG.
Corresponding to the floor pair denoted by and arrival floor j, the average traffic volume per unit time rrH measured between that floor pair.
j is stored. In addition, this storage unit 4
control information (described later)'+Si+'+jQ +il
, tk, and d are provided as shown in FIG.

Note that the inter-floor traffic prediction device 6 detects, for example, the daily inter-floor traffic volume together with the information on the time, weights the information obtained on a daily basis, and adds it to the past data. The average traffic volume mh is sequentially updated based on the average traffic volume mh.Then, along with the traffic demand pattern for the day, the average inter-floor traffic volume mtj for the 10-minute period, for example, is determined. table has 1@floor vs. 1%
The average traffic volume mij at the time when j is specified is read out from the prediction device 6 and stored, and the control unit 2 operates based on the information obtained as the average traffic volume Ii mij at that time.

Next, we will explain the control sequence of the computer system 10 that executes this method, and then explain the principle of operation of the derived call prediction of this method.

First, when the entire set of service floors of the elevator car is F, from the departure floor 1 (16F) to the arrival floor j・(jEF
) is the average traffic volume per unit time moving to λ1. It is defined as At this time, the average traffic volume λ1. A matrix regarding departure floor i and arrival floor j whose elements are defined as demand matrix A.

In addition, in a building where traffic demand is periodic on a daily basis, by selecting an appropriate time interval, the traffic demand within that interval can be considered to be constant. Now, if the driving direction of the elevator car is d, one elevator hall H is defined by the pair H of the floor f (fεF) and the driving direction d.
= (f, d) dε (down, up) ・・・・・・
It can be expressed as (1). Therefore, at time t.

Assume that a hall call occurs in elevator hall H at , and this state continues until time tr (ts > to ). At this time, the number of people X waiting for the elevator to arrive in the hall H at time t1 is predicted as follows.

Generally, the number of people arriving at Hall H can be approximated by a Poisson process, and as a property of the Poisson process, time t. It can be assumed that the number of people who arrived at Hall H in 2018 was one person. Furthermore, the number of people arriving at the hall H after that can be regarded as an event independent of the first person. Therefore, predict the number of people k who will arrive at time (to'''-t+),
By adding "1" to this predicted value, it is possible to predict the number of people X waiting for the elevator in Hall H. That is, when the elevator car descends (d = down
), λ=Σλ ・・・・・・・・・・・・(2a)
As l j × I IJ, and when the elevator car is rising (d=up)
=Σλ・・・・・・Song (2b)l
As j>, lj, the hall call occurrence rate λ in hall H can be determined from the reproducibility of the night.

In addition, the maximum likelihood estimated predicted value number for the above number of people is, when "α" is defined as a symbol indicating the integer part of α, money = [λ1 (
tt to)J (3). Therefore, the predicted value of the number of people waiting
) will be required.

Next, consider the destination of one person waiting in Hall H for the elevator car to arrive. If the number of people who wish to go to the first floor is uj, then when the elevator car descends (d = down)
is defined as , and when the elevator rolls up (d=up
) is defined as. Here, (λIJ/λ) is,
It can be statistically collected by automatically observing the inter-floor traffic volume of passengers using elevator cages over a long period of time, and has a meaning as the distribution ratio of X to (16F). However, the above uj is Σuj”! ・・・・・・・・・・・・
- If we have the relationship (6) and X is very large, then we can assume that the number of people who specify their destination on the first floor is 14. In this case, it is assumed that each passenger acts independently, which is generally acceptable. So now, if No is the set of all non-negative integers and day is the set of all real numbers, then the above problem can be solved by converting u:Fyoj −+ u, f− day to v:Fyoj −+ Vii, N .

It means to approximate by. This approximation is Σ+u
, −vj+→i as the objective function ΣV , = z ・・・・・・・・・
It is only necessary to satisfy the constraint (7). So now,
From equations (6) and (7) above, when [uj] is defined as the smallest integer greater than or equal to 05, vj = 1ujJ or r'uj'1
It can be expressed as (8). Also, y = x−Σ “U・” ・・・・・・・・・・・・
・・・(9)J Wj= uj−1ujJ ・・・・・・・・・
....(1()), VJ E:F is 0 <Wj< 1 -'-...
(11), and the problem is to find ΣZi=F Σ(2・-W) → mln . return. Therefore, the elements of F are arranged in descending order of Wj and in descending order of the value of tsumashi (1-cho), and the set is set as G. Then, the value of above 2 is jεG...zzj==
1 j Monkey G...Zj=0...
...α] From K, jεG ...... v, = 1,000 Luj'' j year G.
・・・・・・・・・vl=l J J ・・・・・・
・It becomes possible to find the solution as α◆. This kind of processing is essentially a proportional distribution of the number of people waiting,
This corresponds to the maximum likelihood estimation method and makes it possible to effectively predict derived calls.

Now, the above prediction processing is executed in the computer system I as follows. This computer system 1 performs normal elevator operation control and executes a control processing sequence as shown in FIG. 6, for example. That is, when the control program is started, first, an existing initial routine is executed in step II, and various initial settings for elevator operation control are performed. Thereafter, in step 12, the tables shown in FIGS. 4 and 5 are initialized, and preparations for elevator operation control and prediction processing are completed. Next, after a first process routine for operation control is performed in step 13, it is determined in step I4 whether or not there is a request to update the table. According to this determination result, the table update routine shown in Stella frs is executed, or this step 1
5 to 4 pinos (.

Then, in the next Stella 7'16, a second operation control processing routine is executed. Thereafter, in step 17 it is determined whether the prediction request has been received, and if there is a prediction request, step 18
The prediction @l calculation process is executed, and if there is no prediction request, the above Stella f18 is inserted into the pino and step 1 is executed.
9 is executed.

Thereafter, the processes from step I3 to step fI9 are repeatedly executed.

In other words, the prediction processing routine according to this method is incorporated as a subroutine in the main program for elevator operation control.

Now, step 12 which is the initialization routine is, for example, the seventh step.
The configuration is as shown in the figure, and first, in step 21, data i
, , jK are each set to "1". Next, in step 22, data "0" is set in mIj. As a result, (1,'') in the table shown in Figure 4 becomes (1,1)
mIj-0 is set in the data area indicated by.

This process is performed by comparing the values of i and j with the maximum value using the determination Stella f23.24 and incrementing the values of 1 and j using Stella f25.26 as described in 1. This process is repeated until the value of j reaches ('m1LX + jmaz). Through these processes, the value m1j=0 is sequentially stored in each data area of the table shown in Figure 4 corresponding to the value of (*, j), and the entire area of the cable is cleared to zero. Become. Then, in step 27,
"0" is set in each data area of the table shown in FIG. Furthermore, the information t shown in Figure 5
k is a counter area 4 for updating data every tko cycle time, f.

d is an area for storing floor names and direction data for prediction calculation instructions, respectively. Further, the data update indicated by tk is determined such that, for example, the cycle time is 1 second/2, tko=599, and data is updated once every 10 minutes.

On the other hand, step 14 for determining a table update request is configured as shown in FIG. 8, for example.

That is, stored in the table 1. The cycle count data tk is read out and compared with the maximum value tko in the stellar fzs to determine whether it is time to update the cycle time counter f'. When a match of the above data is detected, this is determined to be the time to update the data, and step 29
After resetting the data tk to zero, the table update program is started. If the data does not match, it is determined that the data update timing has not been reached (7), the r-data tk is incremented at step 30, and the process moves to the next processing step.

Now, the Stella frs table update process 1, which is started upon detection of the above-mentioned table update request, is shown in FIG. 9, for example. In this processing routine, first, in step 31, "IJ" is set as the data for the departure floor i, and then in step 27°32, "1" is set as the data for the arrival floor j. These data are then processed in step 33
The upper 8 bits (hn) of the 16 bits of output data
, the lower 8 bits (hL), and output to the prediction device 6 of the computer system 1 at the timing of step 33 as a prediction information request signal.

Upon receiving this predicted trepidation@request signal, the device 6 outputs (
Information mh corresponding to i, j) is captured in step 34. Then, this information mh is stored as predicted data m1j in the data area indicated by (i, j) of the diple in the process of step 35.

Thereafter, the value of the above is determined in step 37,
It is incremented in step 38, and the processing up to steps 33 to 36 of the upper 812 is repeated. When data storage of columns j(1 to Jma) is completed by this repeated process, the value of information l is determined in step 40, the value is incremented, and the process is performed in step 32.
The processes from 38 to 38 are repeatedly executed. As a result, predicted values corresponding to the floor pair (l, j)K at one point in time are sequentially stored in all data areas of the table shown in FIG. 4. Then, this process is executed at 600 by the determination of the table update request mentioned above.
This is performed every 1 cycle time, that is, every 10 minutes.

After that, the prediction request is determined in step 17. When issuing this prediction request, first set the desired floor base as f in the table shown in Figure 5 above, and at the same time set the direction as data d, such as d=o for ascending and d=1 for descending. set. Also, the call generation time t of the target hole. , and set the predicted response time of the elevator car to 1 to 1.

Note that the above to, tl data is, for example, 0:00 AM.
The time may be set as the reference time "0" and the time may be given as a seconds value. At this time, it is sufficient to prepare 3 bytes for each data area. Furthermore, it is sufficient to prepare 1 byte each for the data fXd and 2 bytes for tk. Further, the predicted response time t1 may be obtained by adding the predicted arrival time to the current time.

When the prediction request is not made, it is simple and effective to set "0" as the data f.

After obtaining such information, in step 17, for example, the first
As shown in FIG. 0, the presence or absence of a prediction request is identified by determining whether or not data f is "0".

Now, Stella 7 is executed in response to such a prediction request.
For example, the prediction process in Z8 is shown in FIG. 11(a) (
This is carried out as shown in b). That is, first, in step 4, information on the floor f where the hall call occurs is used to predict the traffic volume between floors.
1, it is stored as information on the departure floor l. after that,
At step 42, the cap calculator storage address S1 is cleared to 0, and at step 43, it is determined from the data d whether the direction of movement of the elevator car is upward or downward.

If it is determined that the data d is 10" and is rising, the Stellar 7644 sets the value of (i+1) to j, and then in step 45 the value is determined by the value of (1, j). Set the value of ml 1 to S. This data set is done by rewriting (Si+mu) to sl.

This process is repeated until the value of j is determined in step 46 and j is incremented in step 47 until j reaches its maximum value of 8.

On the other hand, when descending according to the above determination, in step 48,
After setting the value of j to #1, step 49 stores the data (Si+rnlj) in Si in the same way as in step 45. This process is then performed by the determination in step 50 and the determination in step 5I. The execution is repeated by incrementing the value of j until the value of j reaches (i-1).
Depending on the direction of movement of x, Σ1111j or Σtown can be found respectively. The data of Sl obtained in this way corresponds to λ shown in the above-mentioned equation (2).

Thereafter, in step 52, based on the data Si obtained as described above, the equation (4) is calculated as X←1+LSI*(tlto)J, and the number of people waiting is calculated. Predicting X.

Thereafter, j=1 is set in Stella 7D53, and in the next step 54, the determination of ascending and descending is performed again. Further, in steps 55.56, the magnitude relationship of (1, j) is determined, and steps 57.5B and 5
9.60, subtraction processing of uj is performed.

That is, when the condition is (i≦j) when rising, the process of uj←0 is performed in step 57, and when the condition is (i>j) when rising, the processing is performed in step 58 such that U''←X*mlj / S 1 processing is performed. Further, when the condition is (J<i) during descent, the process U-0 is performed in step 59, and when the condition is (j>1), the process uj4-x*mlj/S1 is performed in the Stella f6o. be exposed. This process is sequentially repeated through the determination in step 6I and the increment of the value of j in step 62, so that the value uj is obtained corresponding to each value of j. As a result, the number uj of passengers boarding from the hall-called floor for each destination floor is predicted and calculated.

After that, the process moves to the processing flow shown in Fig. 11 6),
First, in step 63, the value of y in the table is
cleared. Then, in step 64, 'l' is set to the value of j.
' is set, and in steps 65 and 66, the calculation y←y+luj'' w1←uj+L ujJ is performed. This process is repeated by determining the value of j by Step 6 and Proa, and incrementing the value of j by BVC in Step 6, until the value of j reaches its maximum value jma
This will be done until the As a result, the value is finally obtained and calculated in the next step 69. This process is performed using the above-mentioned equation (9) and the 0th equation.
This corresponds to the calculation shown in equation 0.

Note that these arithmetic operations can be easily realized by, for example, using floating point dL arithmetic processing and configuring a subroutine that extracts only the integer value of the value obtained thereby.

Thereafter, the value of y is determined in Stella 'j'' 70, and if it is '0#, step 7I is executed, and if it is not '0#, Stella 7' 72 is executed.

At this time, after executing step 7'77, step 73
After executing the process Vl←i+LujJ at , the value of is decremented at Stella 7u24 and the process goes to upper NI2 step yoVc.

Also, after executing step 72, it is determined whether the value of W is "o" in Stella 7'75, and the value is "
The above-described process is repeated until the value becomes 0''.

Here, the process in step 71 is a process when the condition jθG shown in the above-mentioned formula 01 is satisfied, and Vj==1 +Lujl is subtracted. In this process, for example, as shown in FIG.
The magnitude relationship between and wj is determined.

Then, when (W<Wj), in step & 4.85°86, set the wjO value to W, and then change the above wj to “
0" and then reset the value of j to j. When (W≧wj), the steps f84, 85.
86 is passed, and the value of j is determined in step 87, and j is incremented in step 88, and the process is repeated until the value of j reaches its maximum value jrflX. After completing this process, in step 89, the value of jx obtained as described above is set to j.
Then, the process in step 71 ends. In this step 71, select the maximum value of the town and send the candy to '0'.
′ is performed.

Step 72 searches for the WjO value that is not 10'', sets that value to '01, and performs the process v1←LujJ, which is executed, for example, by the flow shown in FIG. 13. That is, In step 9Z, set the value of j to 11.
#, and the wjO value indicated by j is determined in step 92. If the WjO value is not '0'', the value of Wj is set to 10'' in step 93, and then the integer part of uj is set to Vi in step 94, and the process ends. Also, in step 92, Wj
If it is determined that the O value is 'O', then in the next step 95 the value of j is compared with its maximum value and in step 96
The above process is repeated after incrementing the value of j at . Through the above processing, all of Wj will be set to '0#.

Thus, according to the processing by the computer system I described above, in the matrix vj from j=1 to jmaX, a hall call in the direction indicated by d occurs on the f floor at time t.

The passenger traffic demand in the state that would respond to time t1 when it occurs is predicted. In other words, the number of passengers in hall f for each destination floor is predicted, and a predicted value of inter-floor traffic volume can be obtained here.

The effects of the embodiment of this system described above can be summarized as follows.

(1) First, without installing special sensors in each elevator hall on each floor, we automatically and over a long period of time collect data while learning the inter-floor transportation system, and then select the most likely passenger based on this data. The destination floor can be predicted.

(1) In addition, by learning past inter-floor traffic data for each time period, it is possible to use this to predict the destination floor of passengers quite accurately0 (Il+) Furthermore, there is no prediction response. Depending on the length of time,
It is possible to reasonably predict the arrival of passengers, their number and destination floor.

11v) Using these prediction results, efficient group management control of the elevators can be carried out, resulting in effects such as smoother operation of the elevators. Therefore, the practical advantages thereof are enormous, but the present invention is not limited to the above embodiments. For example, the type of elevator and the number of parallel elevators.

Regardless of the number of service floors or the presence or absence of an express zone,
It is widely applicable. In addition, the inter-floor traffic prediction device 6
All of the information obtained may be stored in a computer system, and the device configuration is not particularly limited. Alternatively, the table update request may be issued at a predetermined time or every time when traffic demand fluctuates. In short, the present invention can be implemented in various modifications without departing from its gist.

[Brief explanation of the drawing]

The figures show one embodiment of the present invention. Fig. 1 is a diagram showing fluctuations and turns in traffic demand, Fig. 2 is a configuration diagram of a control device to which the method of the embodiment is applied, and Fig. 3 is a data format. FIG. 4 and FIG. 5 are diagrams showing an example of a memory table structure, FIG. 6 is a diagram showing the overall control φ41 processing sequence, and FIGS. 7 to 13 are diagrams showing each processing step. FIG. 2 is a diagram showing a processing sequence in FIG. I... tl computer system, 6... Inter-floor traffic prediction device, 12, 14.15.77.18... Prediction processing step. Figure 4 Figure 5 Figure 7

Claims (3)

[Claims] (1) means for measuring the traffic volume of passengers using an elevator car between a pair of service floors and predicting the average traffic volume per unit time between the pair of floors; means for detecting a hall call and the time at which the hall call occurs; means for predicting the arrival time of the elevator car at the hall call generation floor for the hall call; and the predicted arrival time and the predicted average. A method for predicting passenger traffic demand for elevators, comprising means for predicting the destination floor of passengers boarding the elevator cage arriving at the floor where the hall call occurs based on the traffic volume. (2) The means for predicting the average traffic volume per unit time is to measure the passenger transport performance between pairs of two-bis floors, and use the newly measured passenger transport performance values to replace the past passenger transport performance values. 2. The elevator passenger traffic demand prediction method according to claim 1, wherein the predetermined weighting is processed and updated to predict the average traffic volume. (3) The means for predicting the destination floor of passengers is a patent that predicts the number of passengers boarding an elevator, the destination floor of the boarding passengers, and the number of people transported to the destination floor. Elevator passenger traffic demand prediction method 0 according to claim 1