Patent Grant
7735611
The present invention relates to the control of an elevator group.
When a passenger wishes to travel by elevator, he/she calls an elevator by pressing a landing call button installed at the floor. The control system of the elevator receives the call and tries to determine which one of the elevators in the elevator group is best able to serve the person having issued the call. This activity is call allocation. The problem to be solved by allocation is to select for each call an elevator that will minimize a preselected cost function.
The elevator group control system is typically configured to control the elevators in accordance with preselected control algorithms. The control algorithm selected depends on the traffic type prevailing in the building at the time. Therefore, the elevator group control system often comprises a traffic type detector. The traffic types identified by a basic traffic type detector are e.g. “normal traffic”, “incoming peak traffic”, “outgoing peak traffic” and “two-way peak traffic”. Fast and reliable detection of an incoming peak traffic condition is particularly important. In office buildings, incoming peak traffic conditions may arise in the morning during a few minutes as people arrive at their jobs within a short time. An example of typical incoming traffic in an office building is presented in
During incoming peak traffic, the primary function to be fulfilled by the group control system is to return elevators to the entrance floors in a suitable proportion. If in normal-traffic operating mode one elevator is returned for each call issued, then in incoming peak-traffic conditions elevators are returned directly to the entrance floors without a separate call until the system establishes that the peak traffic condition has ceased to exist. The operation of the system can not be influenced by allocation decisions made on the basis of landing calls, because on the entrance floors typically only one landing call, usually an up call, is valid. If direct return of elevators were not activated during incoming peak traffic, there would arise a situation where only two elevators for each entrance floor would be operating; one loaded with passengers and delivering them to their destination floors and another empty and returning to the entrance floor on the basis of a call issued from there. If incoming peak traffic is not identified quickly, long queues will build up in the lobby or in general on the entrance floor of the building and passenger waiting times will become longer. Long waiting times may cause dissatisfaction with the operation of the elevators.
On the other hand, the incoming peak mode should not be activated unnecessarily because direct return of elevators to the entrance floors is a strong measure and its uncalled-for activation will significantly interfere with the rest of elevator service in the building. In that case, calls issued from floors other than the entrance floors are obviously served more slowly than during normal traffic. The algorithm controlling the return of the elevators must be so designed that, during a long-lasting incoming peak traffic situation, calls issued from other floors will be served, although with a delay. Identification of an incoming peak traffic condition involves two partially contrary objectives. The identification must work as fast as possible, but it must not produce incorrect identification results.
In traditional identification of incoming peak traffic, the number of calls is monitored as passengers are entering an elevator in a lobby area (in this case, this comprises each entrance floor of the building). Among the calls, expressly the number of calls with a destination outside the lobby area are considered. When the number of calls exceeds a preset threshold value, the elevator in question is interpreted as a peak elevator and the situation as a potential incoming peak traffic condition.
A threshold value of a corresponding type is also set for the car load. When the elevator leaves the lobby area and its load exceeds the threshold value, the elevator is interpreted as a peak elevator and the situation as a potential incoming peak traffic condition. When two or more peak elevators are detected within a given time window, an incoming peak traffic mode is activated, which in turn starts direct return of elevators to the entrance floors. Two peak elevators at a given predetermined time are required to ensure that peak hour identification will not occur unnecessarily on the basis of occasional peak elevators outside actual peak traffic hours. On the other hand, this retards the identification of a real peak traffic situation at the early stage of a real peak traffic condition.
During actual peak traffic hours, it would be advantageous if the incoming peak traffic mode could be activated already on the basis of a single peak elevator identified. For this purpose, it is possible to set in the control system two separate time windows, typically for morning and lunch-time peak hours, during which the identification of a single peak elevator is sufficient for the activation of the incoming peak traffic mode. A problem with this solution is that it involves the necessity to know the building and its users' times of elevator utilization well enough to allow the aforesaid time windows to be set at the most probable times of beginning of peak traffic conditions. In addition, there should preferably be a possibility to set the time windows separately for each day of the week because the usage profile of the elevators of the building is typically different during the weekend as compared with weekdays. Weekdays again are mutually very similar to each other. However, in practice it is not possible to set the time windows separately for each day of the week because the control logic of the elevator system typically only allows two fixed time windows to be set.
Traffic Forecaster-based identification of peak traffic conditions (TF) calculates the numbers of passengers arriving to and leaving each floor of the building and maintains statistics of these numbers. The calculation is done during the time when the elevator is standing at the floor while passengers are leaving and entering the car. The calculation is based on the use of a car load weighing device and a light cell provided in the elevator door.
TF-based peak hour identification collects statistics of two different types: Long Term Statistics (LTS) and Short Term Statistics (STS). The unit measure used in LTS statistics is e.g. “number of passengers in 15 minutes” and in STS statistics “number of passengers in 5 minutes”. LTS statistics are generated for each floor i. For each floor there are four traffic components k: passengers arriving to the floor from below, passengers arriving to the floor from above, passengers leaving the floor in the downward direction and passengers leaving the floor in the upward direction. In LTS statistics, the day is divided into 96 time slices t of 15 minutes each: the first slice covers the time from 00:00 to 00:15, the next from 00:15 to 00:30 and the last slice from 23:45 to 00:00. Thus, LTS statistics is a three-dimensional matrix Li,k,t. During the day, the passengers are collected into daily statistics Li,k,t*. At midnight, the collected diurnal statistics are subjected to statistical approval tests to ensure that the day collected is not e.g. a midweek holiday. If the diurnal statistics pass the approval tests, then the LTS statistics will be updated e.g. as follows:
Li,k,t=(1−α)·Li,k,t+α·Li,k,t*, (1)
where α is an update factor (0<α<1). The selected α-value is generally small (0.1 . . . 0.2). With typical α values, the method preserves most of the old data and adds some new data. Depending on the school, this up-dating method is called exponential equalization or linear IIR (IIR, Infinite Impulse Response) low-pass filtering. Equation (1) yields a floating average of traffic component k for floor i of the building during time slice t. It describes a past situation, in other words, it gives the average number of passengers having moved before on floor i during the time slice t in question.
The floors comprised in the lobby area of the building being known, it is possible to produce from LTS statistics a traffic profile as shown in
In an attempt to solve the above-mentioned problem, short-term STS statistics have been introduced. STS statistics differ from LTS statistics in that they form a two-dimensional matrix Si,k, where i represents the floor and k the traffic component. The time index t is missing because the number of passengers is calculated and included in STS statistics in a floating manner for the time of five minutes preceding the current instant. In other words, passengers having used the elevator over five minutes ago are removed from the statistics. To identify the traffic type currently prevailing in the building, STS statistics are subjected to the same aforementioned fuzzy logic deduction procedure as LTS statistics.
After this, the information contained in the LTS and STS statistics are combined via a fairly complicated chain of inferences. In this connection, the traffic types given by the statistics are compared to each other, the traffic intensities measured by STS are compared to the transport capacity of the system and the LTS statistics are utilized to obtain confirmation of the traffic type given by the STS statistics.
There are two problems of principle associated with this method. First, LTS and STS statistics are not mutually comparable because the length of the period under consideration is not the same: typically 15 minutes in LTS and 5 minutes in STS. In addition, the time slices in LTS statistics are stationary and have a permanent length of 15 minutes, whereas in STS statistics the time window floats steplessly over the entire diurnal cycle. Second, particularly in view of incoming peak traffic, the five-minute time window of STS statistics is still too long to be used for the activation of an incoming peak mode.
A third problem is associated with practical implementation. The complicated deduction procedure for combining the traffic types produced by STS and LTS require many threshold values to be separately adjusted. Also, trimming and testing the set of rules themselves is a laborious task.
The object of the present invention is to overcome the above-mentioned drawbacks or at least to significantly alleviate them. A specific object of the invention is to achieve faster and more reliable identification of an incoming peak traffic condition than before. As for the features of the invention, reference is made to the claims.
The present invention discloses a method, a computer program product and a system for the identification of an incoming peak traffic condition in an elevator system.
The present invention combines information obtained from statistics with real-time information obtained from traditional peak hour identification. LTS (Long Term Statistics) statistical data collected over a long time span chart the passenger flows observed at different times of the day in the elevators of the building under consideration. Typically, queues build up on the lobby floors in the morning and around the end of the lunch break. From the statistics it is possible to distinguish the most probable times when congestion begins to develop on the lobby floors. In traditional elevator control, a call given by pressing a call button is served by one elevator, which remains stationary after the trip, waiting for the next call. This method works clumsily in a peak traffic situation. The service is slow and the customers are dissatisfied. There is a need to develop an algorithm that would allow faster detection of an incoming peak traffic condition and permit direct return of the elevators to the lobby floors to be activated without a separate press on a call button.
By using the present invention, it is possible to achieve faster identification of an incoming peak traffic condition. In an embodiment of the invention, statistics are utilized to determine the potential peak times when the lobby floors are typically congested. At the same time, the elevators in the elevator system are observed in real time via traditional monitoring of car calls and car load, and when a given threshold value is exceeded, the elevator is assigned a peak elevator status. Threshold value refers e.g. to the total weight or number of the elevator passengers. In addition, in this embodiment, one peak elevator is already sufficient to activate the incoming peak traffic mode, i.e. direct return of the elevators.
In another embodiment of the invention, the number of passengers gathered on the lobby floor is forecast by utilizing statistics and a theoretical so-called time interval between the times when elevators leave the lobby floor. If the number of customers given by the forecast exceeds the car load threshold value for traditional peak hour identification, the situation is considered as a potential peak time, in which situation even one peak elevator detected is sufficient to activate direct return of the elevators.
As an extension of the basic idea of the invention, it is also possible to include in the forecast the time windows preceding the time window for the moment under consideration and/or the time window following it. In this case, the method in a way takes a “lookahead” into the future and accelerates the identification of an incoming peak traffic condition when it is known on the basis of statistics that a peak time is just beginning.
The present invention has several advantages as compared with prior art. Fast identification of an incoming peak traffic condition is achieved, as a consequence of which, the incoming peak traffic mode being activated at the beginning of the peak time, the queues in the lobbies are shorter as compared with traditional peak hour identification. In this way, better service is provided and passengers are kept more satisfied. During statistically recorded peak hours, the system identifies a peak traffic condition already on the basis of a single peak elevator. In the most favorable case, the incoming peak traffic mode can be activated via inferences made from a large number of car calls even while the first peak elevator is only just being loaded at a lobby floor.
A second significant advantage of the present invention is that reliable identification of an incoming peak traffic condition is achieved. The system also identifies an “unexpected” peak traffic condition within a reasonable time on the basis of two peak elevators outside statistically unrecorded peak hours. After the initial start-up (during about a few weeks), the elevator system is not able to utilize LTS statistics because the system has not yet been in operation long enough to allow collection of statistics. In this case, optimal peak hour identification is achieved without help from statistics on the traditional principle whereby peak hour identification is only activated after two peak elevators have been detected.
A third advantage of the invention is that the function can be automated. The statistics collected are day-specific and the statistically recorded traffic profiles especially for weekends differ clearly from the corresponding profiles for weekdays. If the potential peak hours have been set manually, they are valid on every day of the week during the same times of the day and they cannot be modified to make them day-specific. This is naturally a definite disadvantage. In addition, typically a maximum of only two manually set potential peak times can be set for the diurnal cycle. Statistics again may in principle contain an unlimited number of potential peak times. Moreover, an automated identification function involves a great advantage of adaptability related to usability. If significant changes occur in the traffic situation in the building, these changes will appear before long in the LTS statistics and thus peak hour identification is always adapted to the prevailing passenger behavior. Furthermore, the delivery of an elevator system to a client is simplified by the fact that, using the new peak hour identification method, two parameters to be configured at delivery time or on site can be discarded.
In an embodiment of
In another embodiment of
nP=tI·(Li,up>,t+Li,dn>,t), (2)
where i is an index of the lobby floor, up> and dn> are indices referring to the traffic components 10 directed away from the floor and t is an index for the current 15-minute time slice. If the forecast number of passengers nP exceeds the predetermined car load threshold value for traditional peak hour identification, the situation will be interpreted as being a potential peak time. In this case, one peak elevator is sufficient for identification of incoming peak traffic. Otherwise, two peak elevators are required.
The above-described embodiments differ from each other among other things in that, in the latter embodiment, the fuzzy-logic deduction from LTS statistics can be omitted. In both of the above-mentioned embodiments, the traffic type 16 given by STS 15 is used if the traditional traffic detector 14 gives a traffic type other than incoming peak traffic. This selection is made in block 17.
In the identification of a potential peak traffic condition, it is possible to include in the processing, in addition to the 15-minute time window, even the preceding time window (with index ‘T−1’) and the next time window (with index ‘T+1’). In this case, the number of passengers gathering in the elevator queue can be forecast as follows:
nP1=tI·(Li,up>,t−1+Li,dn>,t−1)·β
nP2=tI·(Li,up>,t+Li,dn>,t)
nP3=tI·(Li,up>,t+1+Li,dn>,t+1)·χ, (3)
where β and χ are configuration coefficients (0≦β≦1 and 0≦χ≦1). If one of the calculated queue lengths nP1, nP2 or nP3 exceeds the car load threshold value, then the situation can be interpreted as being a potential peak time, from which again a transition to the incoming peak traffic mode is inferred as described above. The consideration is based on anticipating future events by having a lookahead into the next time window. If the next time window represents a peak time according to statistics but the current moment is still within normal traffic time, then it can be assumed to be very probable that a peak elevator detected at the current moment indicates the onset of an incoming peak traffic condition. A corresponding inference can be made from the time window preceding the current moment. If according to the statistics the preceding time window represents an incoming peak traffic condition, then it is very probable that a peak elevator detected at the current moment still means an actual incoming peak traffic situation. The configuration coefficients β and χ can be used to adjust the sensitivity of the ‘lookahead’.
In an elevator group there often occur situations where all the elevators in the group are not serving normal passenger traffic. Elevators may be undergoing maintenance, they may be serving special calls or being used for some other special purposes. In these situations, the transport capacity of the rest of the elevator group is reduced and lower-than-normal absolute traffic intensities lead to peak traffic situations. When elevators are missing from the service of normal traffic, the time interval tχincreases. Thus, according to (2) and (3), nP, increases, from which it again follows that the car load threshold value is reached sooner. The reduced transport capacity of the elevator group is thus automatically taken into account, because the peak hour identification system transits into a potential peak traffic mode at traffic intensities lower than normal.
In an embodiment of
In an embodiment of
In an embodiment of
In an embodiment of
In an embodiment of
In an embodiment of
In an embodiment of
The means described above are implemented using e.g. a control logic 26. The means can also be implemented as a combination of software and hardware.
A peak hour identification principle operating in the above-described manner can be compared to automatic parking of elevators. Typically, the parking floors are determined manually at the time of delivery of the elevator or they are configured on site. In automatic parking, the building is divided on the basis of LTS statistics into parking zones based on the traffic components directed away from the floors. Within each zone, the floor with the most intense traffic away from the floor is selected as the main parking floor. The zones again are defined in such manner that the intensity of the total traffic away from the floors of different zones is equal in each zone. Thus, the floors with quiet traffic form higher zones as compared to the floors with intense traffic. The actual dispatching of the elevators to the parking floors is done in the same way as in the case of manually defined floors.
In a manner corresponding to the above-described automatic parking, wherein statistics are used to determine where the elevators should preferably be parked and the actual parking is carried out by a traditional method, in peak hour identification the statistics are read in block 13 to see when a potential incoming peak traffic situation is to be expected and the actual incoming peak traffic condition is identified by a traditional method in block 14. Thus, the statistics have a role that is the most natural to them. They serve as an aid in actual decision-making, which again works in accordance with information on occurrences actually taking place in the system at the present moment.
The invention is not limited to the embodiment examples described above; instead, many variations are possible within the scope of the inventive concept defined in the claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 20030972 | Jun 2003 | FI | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/FI2004/000232 | 4/15/2004 | WO | 00 | 12/18/2006 |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2005/000726 | 1/6/2005 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 3967702 | Iwaska et al. | Jul 1976 | A |
| 3973649 | Iwasaka et al. | Aug 1976 | A |
| 4044860 | Kaneko et al. | Aug 1977 | A |
| 5229559 | Siikonen et al. | Jul 1993 | A |
| 5260526 | Sirag, Jr. | Nov 1993 | A |
| 5260527 | Sirag, Jr. | Nov 1993 | A |
| 5274202 | Kezer et al. | Dec 1993 | A |
| 5276295 | Kameli | Jan 1994 | A |
| 5511635 | Kameli | Apr 1996 | A |
| 5544059 | Hikita et al. | Aug 1996 | A |
| 5679932 | Kim | Oct 1997 | A |
| 6257373 | Hikita et al. | Jul 2001 | B1 |
| 6401874 | Siikonen | Jun 2002 | B2 |
| 20090133968 | Smith et al. | May 2009 | A1 |
| Number | Date | Country |
|---|---|---|
| 0 427 992 | May 1991 | EP |
| 0 739 848 | Oct 1996 | EP |
| Number | Date | Country | |
|---|---|---|---|
| 20070084674 A1 | Apr 2007 | US |
