PatentDe  


Dokumentenidentifikation EP1125881 22.12.2005
EP-Veröffentlichungsnummer 0001125881
Titel STEUERUNGSGERAET FUER AUFZUGSGRUPPE
Anmelder Mitsubishi Denki K.K., Tokio/Tokyo, JP
Erfinder HIKITA, Shiro, Tokyo 100-8310, JP;
TAJIMA, Shinobu, Tokyo 100-8310, JP
Vertreter derzeit kein Vertreter bestellt
DE-Aktenzeichen 69928432
Vertragsstaaten DE, NL
Sprache des Dokument EN
EP-Anmeldetag 03.08.1999
EP-Aktenzeichen 999332539
WO-Anmeldetag 03.08.1999
PCT-Aktenzeichen PCT/JP99/04186
WO-Veröffentlichungsnummer 0001010763
WO-Veröffentlichungsdatum 15.02.2001
EP-Offenlegungsdatum 22.08.2001
EP date of grant 16.11.2005
Veröffentlichungstag im Patentblatt 22.12.2005
IPC-Hauptklasse B66B 1/18
IPC-Nebenklasse B66B 1/20   

Beschreibung[en]
Technical Field

The present invention relates to an elevator group managing system for managing and controlling efficiently a plurality of elevators in a group.

Background Art

In general, in the system in which a plurality of elevators go into commission, the group management control is carried out. There are carried out therein the various types of controls such as the assignment control for selecting the optimal assigned elevator in response to a call which has occurred in a hole, the forwarding operation which is carried out in a peak time for the specific floor differently from the occurrence of the call, or the division of the service zone.

In recent years, for example, as disclosed in Japanese Patent No. 2664766 or Japanese Patent Application Laid-open No. Hei 7-61723, there has been proposed a method of predicting for the control result of the group management, i.e., the group management performance such as the waiting time and the like to set the control parameters.

In accordance with the above-mentioned two prior arts, there is stated a system in which the neural net for receiving as its input the traffic demand parameters and the evaluation arithmetic operation parameters when carrying out the call assignment to output the group management performance is employed, and the output result of the neural net is evaluated to set the optimal evaluation arithmetic operation parameter.

However, in the above-mentioned two articles relating to the prior art, the parameter which is set on the basis of the group management performance prediction result is limited to the single evaluation arithmetic operation parameter when carrying out the assignment. Thus, carrying out the arithmetic operation employing such a single evaluation arithmetic operation parameter when carrying out the call assignment leads to the limitation to the enhancement of the transport performance. That is, the various rule sets such as the forwarding operation and the zone division needs to be utilized depending on the traffic situation and hence the really excellent group management performance can not be obtained.

In addition, while the neural net has the advantage that its accuracy of the arithmetic operation can be enhanced by the learning, at the same time, it has also the disadvantage that it takes a lot of time for the accuracy of the arithmetic operation to reach the practical level.

In the system which is disclosed in the above-mentioned two articles relating to the prior art, it is impossible to obtain the expected group management performance unless the learning of the neural net is previously carried out in the factory. In addition, in the case where the traffic demand is abruptly changed due to the change or the like of tenants in an associated building, it is possible to cope speedily with such a change.

In the light of the foregoing, the present invention has been made in order to solve the above-mentioned problems associated with the prior art, and it is therefore an object of the present invention to provide an elevator group managing system which can select the optimal rule set in accordance with the performance prediction result to provide the excellent service at all times.

US 5,233,138 describes an elevator control apparatus that includes a fuzzy rule base having fuzzy rules stored therein which govern the selection of an elevator cage to be assigned to respond to a call. A reasoning unit is provided for selecting the appropriate fuzzy rule to be applied to a cage. The reasoning unit selects the appropriate fuzzy rule according to evaluation factors such as the miss forecast rate and the estimation rate of the cages.

US 5,412,163 describes an elevator control apparatus that determines the time required for a call to reach a hall and controls an operation of the car using the obtained estimated travel time. The elevator control apparatus includes an input data conversion unit for converting traffic data, including car position, car direction data, and data regarding car calls and hall calls into data that can be used as input data to a neural network. An estimated travel time operation unit including an input layer is provided for taking in the input data. An output layer is provided for outputting the estimated travel time. An intermediate layer is provided between the input and output layers in which a weighting factor is set. The estimated travel time operation unit comprises a neural network and an output data conversion unit for converting the estimated travel time output from the output layer into data that can be used for a predetermined control operation.

Disclosure of the Invention

According to an elevator group managing system of one aspect of the present invention, an elevator group managing system for managing a plurality of elevators in a group, includes: traffic situation detecting means for detecting the current traffic situation of a plurality of elevators; a rule base for storing therein a plurality of control rule sets; performance predicting means for predicting the group management performance which is obtained when applying an arbitrary rule set stored in the rule base to the current traffic situation; rule set selecting means for selecting the optimal rule set in accordance with the prediction result obtained from the performance predicting means; and operation control means for carrying out the operation control for each of the elevator cars on the basis of the rule set which has been selected by the rule set selecting means.

In addition, an elevator group managing system further includs a weight database for storing therein weight parameters of a neural net corresponding to an arbitrary rule set stored in the rule base, and the system is characterruzed in that the performance predicting means, for the specific rule set stored in the rule base, fetches the weight parameters of the neural net corresponding to the specific rule set from the weight database to carry out the prediction of the group management performance by the neural net using the weight parameters thus fetched.

In addition, an elevator group managing system further includs performance learning means for comparing the prediction result provided by the performance predicting means with the actual group management performance after having applied the specific rule set to carry out the learning of the neural net to correct the weight parameters stored in the weight database in accordance with the learning result, and the system is characterized in that the performance predicting means carries out the prediction of the group management performance by the neural net using the corrected weight parameters.

In addition, an elevator group managing system is characterized in that the performance predicting means, on the basis of the mathematical model, predicts the group management performance which is predicted when applying an arbitrary rule set stored in the rule base to the current traffic situation.

Furthermore, according to an elevator group managing system of another aspect of the present invention, an elevator group managing system for managing a plurality of elevators in a group, includs: traffic situation detecting means for detecting the current traffic situation of a plurality of elevators; a rule base for storing therein a plurality of control rule sets; first performance predicting means for on the basis of a neural net, predicting the group management performance which is obtained when applying an arbitrary rule set stored in the rule base to the current traffic situation; a weight database for storing therein weight parameters of the neural net corresponding to the arbitrary rule set stored in the rule base; and performance learning means for comparing the prediction result provided by the first performance predicting means with the actual group management performance after having applied the specific rule set to carry out the learning of the neural net to correct the weight parameters stored in the weight database in accordance with the learning result, wherein the first performance predicting means carries out the prediction of the group management performance by the neural net using the corrected weight performance, and wherein the system further includs: second performance predicting means for on the basis of the mathematical model, predicting the group management performance which is predicted when applying an arbitrary rule set stored in the rule base to the current traffic situation; performance prediction accuracy evaluating means for comparing the prediction results provided by the first and second performance predicting means with the actual group management performance to determine which of the first or second performance predicting means is employed in accordance with the comparison result; rule set selecting means for selecting the optimal rule set in accordance with the prediction result, from either the first or second performance predicting means, which has been determined by the performance prediction accuracy evaluating means; and operation control means for carrying out the operation control for each of the elevator cars on the basis of the rule set which has been selected by the rule set selecting means.

Brief Description of the Drawings

  • Fig. 1 is a block diagram showing a configuration of an elevator group managing system according to the present invention;
  • Fig. 2 is a functional association diagram of constituent elements provided in the elevator group managing system shown in Fig. 1;
  • Fig. 3 is a flow chart useful in explaining the schematic operation of the control procedure in the group managing system in an embodiment of the present invention; and
  • Fig. 4 is a flow chart useful in explaining the schematic operation of the learning procedure in the group managing system in an embodiment of the present invention.

Best Mode for carrying out the Invention Embodiment 1

An embodiment of the present invention will hereinafter be described with reference to the accompanying drawings.

Fig. 1 is a block diagram showing a configuration of an elevator group managing system according to the present invention, and Fig. 2 is a functional association diagram of constituent elements provided in the elevator group managing system shown in Fig. 1.

In these figures, reference numeral 1 designates a group managing system for managing a plurality of elevators in a group, and reference numeral 2 designates an associated elevator control apparatus for controlling an associated one of the elevators.

The above-mentioned group managing system 1 includes: communication means 1A for communicating with associated elevator control apparatuses 2; a control rule base 1B for storing therein a plurality of control rule sets, required for the group management control, such as a rule for allocation of elevators by zone based on the forwarding operation and the zone division/assignment evaluation system; traffic situation detecting means 1C for detecting the current traffic situation such as the number of passengers getting on and off the associated one of the elevators; first performance predicting means 1D for predicting the group management performance such as the waiting time distribution which is obtained when applying the specific rule set stored in the above-mentioned rule base 1B using the neural net under the traffic situation which is detected by the above-mentioned traffic situation detecting means 1C; a weight database 1E for storing therein the weight parameters of the neural net corresponding to an arbitrary rule set stored in the above-mentioned control rule base 1B; and second performance predicting means 1F for on the basis of the mathematical model, predicting the group management performance which is obtained when applying an arbitrary rule set containing the probability model under the traffic situation which has been detected by the above-mentioned traffic situation detecting means 1C.

The above-mentioned group managing system 1 further includes: performance learning means 1G for carrying out the learning for the neural net of the above-mentioned first performance predicting means 1D to enhance the accuracy of predicting the group management performance; performance prediction accuracy evaluating means 1H for comparing the prediction results provided by the above-mentioned first performance predicting means 1D and the above-mentioned second performance predicting means 1F with the actually measured group management performance to evaluate the prediction accuracy of the first performance predicting means 1D; rule set selecting means 1J for selecting the optimal rule set in accordance with the prediction results provided by the above-mentioned first performance predicting means 1D and the above-mentioned second performance predicting means 1F; rule set carrying out means 1K for carrying out the rule set which has been selected by the above-mentioned rule set selecting means 1J; operation controlling means 1L for carrying out the overall operation control for each of the elevator cars on the basis of the rule which has been carried out by the above-mentioned rule set carrying out means 1K; and learning database 1M for storing therein the learning data.

The group managing system 1 is configured by including the above-mentioned constituent elements and also each of the constituent elements is constructed in the form of the software on the computer.

Next, the operation of the present embodiment will hereinbelow be described with reference to the associated figures.

Fig. 3 is a flow chart useful in explaining the schematic operation in the control procedure of the group managing system 1 of the present embodiment, and Fig. 4 is likewise a flow chart useful in explaining the schematic operation in the learning procedure of the group managing system 1.

First of all, the description will hereinbelow be given with respect to the schematic operation in the control procedure with reference to Fig. 3.

In Step S101, the demeanor of each of the elevator cars is monitored through the communication means 1A, and also the traffic situation, e.g., the number of passengers getting on and off the associated one of the elevators in each of the floors is detected by the traffic situation detecting means 1C. For the data describing this traffic situation, for example, the accumulated value per time (e.g., for five minutes) of the number of passengers getting on and off the associated one of the elevators in each of the floors. Alternatively, the OD (Origin and Destination: the movement of passengers from one floor to another floor) estimate may also be employed which is obtained on the basis of the well known method as disclosed in Japanese Patent Application Laid-open No.Hei 10-194619 for example.

Next, in Step S102, an arbitrary rule set is fetched from the control rule base 1B to be set. In subsequent Step S103, it is judged whether the neural net prediction is valid or invalid to the rule set thus set (in this connection, in Fig. 3, reference symbol NN represents the neural net). As a result of the judgement, if invalid (NO in Step S103), then the processing proceeds to Step S104, while if valid (YES in Step S103), then the processing proceeds to Step S105.

In this connection, in the above-mentioned Step S103, the procedure of judging whether the neural net is valid or invalid is carried out, as one example, on the basis of a result of judging whether or not the prediction accuracy is ensured now after the neural net has completed the learning. More specifically, it is judged on the basis of the value of a neural net prediction flag which is set in Step S207 in the learning procedure shown in Fig. 4 which will be described later.

When it is judged in the above-mentioned Step S103 that the neural net prediction is invalid, in Step S104, the prediction of the group management performance based on the mathematical model is carried out by the second performance predicting means 1F. While in this procedure, the queue theory or the like may be employed, that prediction may also be calculated on the basis of the iteration method as hereinbelow shown instead. RTT = f(RTT)

Now, RTT represents a Round Trip Time of the elevator car. Then, for example, it is described in Japanese Patent Examined Publication No.Hei 1-24711 that the relation between the mean waiting time and the number of floors in which the associated one of the elevators is stopped is obtained due to the elevator car round trip time RTT. That is, f(RTT) is the function of calculating the group management performance such as the elevator car service intervals at which the associated one of the elevator cars reaches an arbitrary floor, the stop probability, the probability of the passengers getting on and off the associated one of the elevators and the waiting time from the restriction of the elevator car demeanor due to the application of the elevator car round trip time RTT which has been set, the traffic situation data and the rule set. Then, these factors can be calculated on the basis of the theory of probability. As for the prior art showing one example of the calculation method relating thereto, there is given an article of "Theory and Practice of Elevator Group Managing System": 517th short course teaching materials of the Japan Society of Mechanical Engineers (Theory and Practice of Control in Traffic Machine, March 9, 1981, Tokyo).

On the other hand, when it is judged in the above-mentioned Step S103 that the neural net prediction is valid, first of all, in Step S105, the weight parameters of the neural net corresponding to the rule set which has been set are fetched from the weight database 1E to be set. Then, in Step S106, there is carried out the prediction of the group management performance by the neural net using the weight parameters which have been set by the first performance predicting means 1D.

The neural net which is used in the first performance predicting means 1D sets the group management performance such as the traffic situation data as its input and the waiting time distribution as its output to carry out the learning in Step S203 in the learning procedure shown in Fig. 4 which will be described later, whereby the prediction becomes possible with accuracy of some degree.

The procedures ranging from Step S102 to Step S106 are carried out for a plurality of rule sets which are previously prepared within the control rule base 1B, respectively.

Next, in Step S107, the performance prediction result for each of the rule sets is evaluated by the rule set selecting means 1J to select the best rule set of them. Then, in Step S108, the rule set which has been selected in Step S107 is carried out by the rule set carrying out means 1K to transmit the various kinds of instructions, the constraint condition and the operation method to the operation controlling means 1L so that the operation control based on the instructions and the like which have been transmitted by the operation controlling means 1L is carried out.

Above, the description of the schematic operation of the control procedure in the present embodiment has been completed.

Subsequently, the description will hereinbelow be given with respect to the schematic operation of the learning procedure with reference to Fig. 4.

First of all, in Step S201, the result of the group management performance which has been obtained through the control procedure shown in Fig. 3 by the performance learning means 1G, the traffic situation at that time and the applied rule set are stored at regular intervals. Then, after the applied rule set, the traffic situation to which that rule set has been applied, and the group management performance after the application of that rule set are put in order in the form of the data set, a part of the data set is stored as the data for the test in the subsequent learning procedure in the learning database 1M and also the remaining data set is stored as the learning data therein.

Next, in Step S202, each of the learning data which has been stored in Step S201 is read out from the learning database 1M by the performance learning means 1G to be inputted. Then, in Step S203, the weight parameters corresponding to the used rule set is set in the neural net using each of the learning data by the performance learning means 1G to carry out the learning of the neural net with the traffic situation data as the input and the measured group management performance as the output. In this connection, for the learning of this neural net, the well known Back Propagation Method may be employed. In addition, in this Step S203, the weight parameters which have been corrected by the learning are stored in the weight database 1E. The procedures in the above-mentioned Step S202 and S203 are carried out with respect to each of the learning data.

After the learning of the neural net and the correction of the weight parameters by the learning have been completed with respect to each of the learning data on the basis of the procedure as described above, subsequently, in order to check the ability of the rule sets, each of the data for the test is temporarily inputted to obtain the predictor thereof.

That is, in Step S204, by using the data for the test which has been stored in the learning database 1M in the above-mentioned Step S201, the prediction of the group management performance made by the neural net in which the learning has been carried out for the corresponding rule set and traffic situation is carried out by the first performance predicting means 1D.

In addition, in Step S205, the prediction of the group management performance based on the mathematical model is carried out by the second performance predicting means 1F.

The procedures in Step S204 and Step S205 are carried out for each of the data for the test.

Next, in Step S206, each of the prediction results which have been predicted in Step S204 and Step S205 and the performance which has been measured are compared with each other by the performance prediction accuracy evaluating means 1H. For this comparison, for example, the following error may be made the index. That is, the performance predicting means having the smaller error ERR obtained on the basis of the following expression is regarded as the performance predicting means having the more excellent prediction accuracy. ERR =Σ | Xk - Yk |2 /N (k = 1, 2, ..., N) where ERR represents the error, N represents the number of data for the test, Xk represents the performance measured value vector, and Yk represents the performance predicted value vector.

Then, in Step S207, when as a result of the comparison in the above-mentioned Step S206, the first performance predicting means 1D has the more excellent prediction accuracy, a neural net prediction flag is set to the valid state by the performance prediction accuracy evaluating means 1H. Otherwise, the neural net prediction flag is set to the invalid state. This neural net prediction flag is used in the judgement in Step S103 of the control procedure shown in Fig. 3. In this connection, the procedures of the above-mentioned Steps S202 to S207 are carried out every rule set.

As set forth hereinabove, according to the present invention, in an elevator group managing system for managing a plurality of elevators in a group, a rule base for storing therein a plurality of control rule sets such as a rule for allocation of elevators by zone is prepared, group management performance such as the waiting time distribution which is obtained when applying an arbitrary rule set stored in the rule base to the current traffic situation is predicted, and the optimal rule set is selected in accordance with the performance prediction result. Therefore, there is offered the effect that the optimal rule set can be applied at all times to carry out the group management control and hence it is possible to provide the excellent service.

The elevator group managing system further includes a weight database for storing therein weight parameters of a neural net corresponding to an arbitrary rule set stored in the rule base, wherein for the specific rule set stored in the rule base, the weight parameters of the neural net corresponding to the specific rule set are fetched from the weight database, and the prediction of the group management performance by the neural net using the weight parameters thus fetched is carried out. Therefore, there is offered the effect that the learning of the neural net can be carried out every part corresponding to the associated one of the rule sets and hence it is possible to enhance the prediction accuracy.

The elevator group managing system further includes performance learning means for comparing the prediction result of the group management performance with the actual group management performance after having applied the specific rule set to carry out the learning of the neural net to correct the weight parameters stored in the weight database in accordance with the learning result, wherein the prediction of the group management performance by the neural net using the corrected weight parameters. As a result, there is offered the effect that it is possible to enhance the prediction accuracy in correspondence to the actual operating situation of a plurality of elevators.

In addition, the round trip time of each of the elevator cars which is predicted when applying an arbitrary rule set stored in the rule base to the current traffic situation is mathematically calculated and the group management performance such as the waiting time is predicted on the basis of the mathematical model from the round trip time and the traffic situation. As a result, there is offered the effect that the group management performance can be predicted without carrying out the prediction by the neural net and also it is possible to enhance the prediction accuracy thereof.

Furthermore, an elevator group managing system for managing a plurality of elevators in a group includes: traffic situation detecting means for detecting the current traffic situation of a plurality of elevators; a rule base for storing therein a plurality of control rule sets; first performance predicting means for on the basis of a neural net, predicting the group management performance which is obtained when applying an arbitrary rule set stored in the rule base to the current traffic situation; a weight database for storing therein weight parameters of the neural net corresponding to the arbitrary rule set stored in the rule base; and performance learning means for comparing the prediction result provided by the first performance predicting means with the actual group management performance after having applied the specific rule set to carry out the learning of the neural net to correct the weight parameters stored in the weight database in accordance with the learning result, wherein the first performance predicting means carries out the prediction of the group management performance by the neural net using the corrected weight parameters, the system further including: second performance predicting means for on the basis of the mathematical model, predicting the group management performance which is predicted when applying an arbitrary rule set stored in the rule base to the current traffic situation; performance prediction accuracy evaluating means for comparing the prediction results provided by the first and second performance predicting means with the actual group management performance to determine which of the first or second performance predicting means is employed in accordance with the comparison result; rule set selecting means for selecting the optimal rule set in accordance with the prediction result, from either the first or second performance predicting means, which has been determined by the performance prediction accuracy evaluating means; and operation controlling means for carrying out the operation control for each of the elevator cars on the basis of the rule set which has been selected by the rule set selecting means. As a result, there is offered the effect that it is possible to enhance the accuracy of the performance prediction in accordance with the actual operating situation of a plurality of elevators, even when the traffic situation is abruptly changed due to the change in the initial state or the change of tenants within an associated building in which a plurality of elevators are installed, it is possible to carry out the performance prediction with high accuracy, and also on the basis of that prediction, the group management control can be carried out using the optimal rule set at all times.

Industrial Applicability

According to the present invention, a rule base for storing therein a plurality of control rule sets is prepared, group management performance such as the waiting time distribution which is obtained when applying an arbitrary rule set stored in the rule base to the current traffic situation is predicted, and the optimal rule set is selected in accordance with the performance prediction result, whereby the optimal rule set can be applied at all times to carry out the group management control and hence it is possible to provide the excellent service.


Anspruch[de]
  1. Aufzuggruppen-Management-System (1) zum Steuern einer Vielzahl von Aufzügen in einer Gruppe, mit

    Verkehrsituationserfassungsmitteln (1C) zum Erfassen der aktuellen Verkehrssituation bei einer Vielzahl von Aufzügen;

    einer Regelbasis (1B) zum Speichern einer Vielzahl von Steuerregelsätzen;

    Leistungsvorhersagemitteln (1D, 1F) zur Vorhersage der Gruppensteuerungsleistung, welche ermittelt wird, wenn ein beliebiger Regelsatz, der in der Regelbasis (1B) gespeichert ist, auf die aktuelle Verkehrsituation angewendet wird;

    Regelsatzauswahlmitteln (1J) zur Auswahl des optimalen Regelsatzes in Übereinstimmung mit den Vorhersageergebnissen, die von den Leistungsvorhersagemitteln (1D, 1F) ermittelt wurden; und

    Betriebssteuerungsmitteln (1L) zur Ausführung der Betriebskontrolle für jeden Aufzugwagen auf der Basis des Regelsatzes, der durch die Regelsatzauswahlmittel (1J) ausgewählt wurde.
  2. Aufzuggruppen-Management-System nach Anspruch 1, weiterhin mit

    einer Gewichtsdatenbank (1E) zum Speichern von Gewichtsparametern eines neutralen Nettowertes, der einem beliebigen, in der Regelbasis gespeichert Regelsatz entspricht,

    dadurch gekennzeichnet, dass

    die Leistungsvorhersagemittel (1D, 1F) für einen bestimmten in der Regelbasis (1B) gespeicherten Regelsatz die Gewichtsparameter des neutralen Nettowertes heranziehen, die dem spezifischen Regelsatz der Gewichtsdatenbank (1E) entsprechen, um die Vorhersage der Gruppensteuerungsleistung durch den neutralen Nettowert auszuführen, wobei dazu die so herangezogenen Gewichtsparameter genutzt werden.
  3. Aufzuggruppen-Management-System nach Anspruch 2, weiterhin mit Leistungslernmitteln (1G) zum Vergleichen der Vorhersageergebnisse, die durch die Leistungsvorhersagemittel (1D, 1F) zur Verfügung gestellt werden, mit der Gruppensteuerungsleistung, nachdem der spezifische Regelsatz angewendet worden ist, um das Lernen des neutralen Nettowertes auszuführen, um die in der Gewichtsdatenbank (1E) gespeicherten Gewichtsparameter in Übereinstimmung mit den Lernergebnissen zu korrigieren,

    dadurch gekennzeichnet, dass

    die Leistungsvorhersagemittel (1D, 1F) eine Vorhersage der Gruppensteuerungsleistung durch einen neutralen Nettowert ausführen, wobei die korrigierten Gewichtsparameter genutzt werden.
  4. Aufzuggruppen-Management-System nach Anspruch 1,

    dadurch gekennzeichnet, dass

    die Leistungsvorhersagemittel (1D, 1F) die Gruppensteuerungsleistung auf der Basis eines mathematischen Modells vorhersagen, wenn ein beliebiger, in der Regelbasis (1B) gespeicherter Regelsatz auf die aktuelle Verkehrssituation angewendet wird.
  5. Aufzuggruppen-Management-System nach Anspruch 1, bei dem die Leistungsvorhersagemittel (1D, 1F) weiterhin folgendes enthalten:
    • erste Leistungsvorhersagemittel (1D) zur Vorhersage der Gruppensteuerungsleistung auf der Basis eines neutralen Nettowertes, die erhalten wird, wenn ein beliebiger, in der Regelbasis (1B) gespeicherter Regelsatz auf die aktuelle Verkehrssituation angewendet wird; und
    • zweite Leistungsvorhersagemittel (1F) zur Vorhersage der Gruppensteuerungsleistung auf der Basis eines mathematischen Modells, die erhalten wird, wenn ein beliebiger, in der Regelbasis (1B) gespeicherter Regelsatz auf die aktuelle Verkehrssituation angewendet wird; und
    wobei das System weiterhin enthält:
    • eine Gewichtsdatenbank (1B) zur Speicherung von Gewichtsparametern eines neutralen Nettowertes, der dem beliebigen, in der Regelbasis (1B) gespeicherten Regelsatz entspricht; und
    • Leistungslernmittel (1G) zum Vergleichen der Vorhersageergebnisse, die von den Leistungsvorhersagemitteln (1D) mit der aktuellen Gruppensteuerungsleistung nach der Anwendung des spezifischen Regelsatzes bereitgestellt werden, um das Lernen über den neutralen Nettowert durchzuführen, um die in der Gewichtsdatenbank (1E) gespeicherten Gewichtsparameter in Übereinstimmung mit den Lernergebnissen zu korrigieren, wobei die ersten Leistungsvorhersagemittel (1D) eine Vorhersage der Gruppensteuerungsleistung durch den neutralen Nettowert machen, wobei die korrigierten Gewichtsparameter genutzt werden; und
    • Leistungsvorhersagegenauigkeitsabschätzungsmittel (1H) zum Vergleichen der Vorhersageergebnisse, die durch die ersten und zweiten Leistungsvorhersagemittel (1D, 1F) verfügbar gemacht werden, mit der aktuellen Gruppensteuerungsleistung, um zu bestimmen, welcher der ersten und zweiten Leistungsvorhersagemittel (1D, 1F) in Übereinstimmung mit den Vergleichsergebnissen eingesetzt ist, wobei die Regelsatzauswahlmittel (1J) einen optimalen Regelsatz in Übereinstimmung mit den Vorhersageergebnissen entweder von den ersten oder zweiten Leistungsvorhersagemitteln (1D, 1F) auswählen, die durch die Leistungsvorhersagegenauigkeitsabschätzungsmittel (1H) ermittelt wurden.
Anspruch[en]
  1. An elevator group managing system for managing a plurality of elevators in a group, said elevator group managing system (1) comprising:
    • traffic situation detecting means (1C) for detecting the current traffic situation of a plurality of elevators;
    • a rule base (1B) for storing therein a plurality of control rule sets;
    • performance predicting means (1D, 1F) for predicting the group management performance which is obtained when applying an arbitrary rule set stored in said rule base (1B) to the current traffic situation;
    • rule set selecting means (1J) for selecting the optimal rule set in accordance with the prediction result obtained from said performance predicting means (1D, 1F) ; and
    • operation controlling means (1L) for carrying out the operation control for each of the elevator cars on the basis of the rule set which has been selected by said rule set selecting means (1J).
  2. An elevator group managing system according to claim 1, further comprising a weight database (1E) for storing therein weight parameters of a neural net corresponding to an arbitrary rule set stored in said rule base, said system characterized in that said performance predicting means (1D, 1F), for the specific rule set stored in said rule base (1B), fetches the weight parameters of the neural net corresponding to the specific rule set from said weight database (1E) to carry out the prediction of the group management performance by the neural net using the weight parameters thus fetched.
  3. An elevator group managing system according to claim 2, further comprising performance learning means (1G) for comparing the prediction result provided by said performance predicting means (1D, 1F) with the actual group management performance after having applied the specific rule set to carry out the learning of the neural net to correct the weight parameters stored in said weight database (1E) in accordance with the learning result, said system characterized in that said performance predicting means (1D, 1F) carries out the prediction of the group management performance by the neural net using the corrected weight parameters.
  4. An elevator group managing system according to claim 1,characterized in that said performance predicting means (1D, 1F), on the basis of the mathematical model, predicts the group management performance which is predicted when applying an arbitrary rule set stored in said rule base (1B) to the current traffic situation.
  5. An elevator group managing system according to claim 1,

    wherein said performance predicting means (1D, 1F) further comprises

    first performance predicting means (1D) for on the basis of a neural net, predicting the group management performance which is obtained when applying an arbitrary rule set stored in said rule base (1B) to the current traffic situation; and

    second performance predicting means (1F) for on the basis of the mathematical model, predicting the group management performance which is predicted when applying an arbitrary rule set stored in said rule base (1B) to the current traffic situation; and

    wherein said system further comprises

    a weight database (1E) for storing therein weight parameters of the neural net corresponding to the arbitrary rule set stored in said rule base (1B); and

    performance learning means (1G) for comparing the prediction result provided by said first performance predicting means (1D) with the actual group management performance after having applied the specific rule set to carry out the learning of the neural net to correct the weight parameters stored in said weight database (1E) in accordance with the learning result, wherein said first performance predicting means (1D) carries out the prediction of the group management performance by the neural net using the corrected weight parameters; and

    performance prediction accuracy evaluating means (1H) for comparing the prediction results provided by said first and second performance predicting means (1D, 1F) with the actual group management performance to determine which of said first or second performance predicting means (1D, 1F) is employed in accordance with the comparison result, wherein said rule set selecting means (1J) selects the optimal rule set in accordance with the prediction result, from either said first or second performance predicting means (1D, 1F), which has been determined by said performance prediction accuracy evaluating means (1H).
Anspruch[fr]
  1. Système de gestion de groupe d'ascenseurs pour gérer une pluralité d'ascenseurs dans un groupe, ledit système de gestion de groupe d'ascenseurs (1) comprenant :
    • un moyen de détection de situation de circulation (1C) pour détecter la situation de circulation actuelle d'une pluralité d'ascenseurs ;
    • une base de règles (1B) pour y stocker une pluralité d'ensembles de règles de commande ;
    • un moyen de prédiction de performance (1D, 1F) pour prédire la performance de gestion du groupe qui est obtenue en appliquant un ensemble arbitraire de règles stocké dans ladite base de règles (1B) à la situation de circulation actuelle ;
    • un moyen de sélection d'ensemble de règles (1J) pour sélectionner l'ensemble optimal de règles en fonction du résultat de prédiction obtenu à partir dudit moyen de prédiction de performance (1D, 1F) ; et
    • un moyen de commande de fonctionnement (1L) pour effectuer la commande de fonctionnement de chacune des cabines d'ascenseur sur la base de l'ensemble de règles qui a été sélectionné par ledit moyen de sélection d'ensemble de règles (1J).
  2. Système de gestion de groupe d'ascenseurs selon la revendication 1, comprenant de plus une base de données de poids (1E) pour y stocker des paramètres de poids d'un réseau neuronal correspondant à un ensemble arbitraire de règles stocké dans ladite base de règles, ledit système étant caractérisé en ce que ledit moyen de prédiction de performance (1D, 1F), pour l'ensemble spécifique de règles stocké dans ladite base de règles (1B), extrait les paramètres de poids du réseau neuronal correspondant à l'ensemble spécifique de règles dans ladite base de données de poids (1E) pour effectuer la prédiction de performance de gestion du groupe par le réseau neuronal en utilisant les paramètres de poids ainsi extraits.
  3. Système de gestion de groupe d'ascenseurs selon la revendication 2, comprenant de plus un moyen d'apprentissage de performance (1G) pour comparer le résultat de prédiction fourni par ledit moyen de prédiction de performance (1D, 1F) avec la performance réelle de gestion du groupe, après avoir appliqué l'ensemble spécifique de règles pour réaliser l'apprentissage du réseau neuronal afin de corriger les paramètres de poids stockés dans ladite base de données de poids (1E) en fonction du résultat d'apprentissage, ledit système étant caractérisé en ce que ledit moyen de prédiction de performance (1D, 1F) réalise la prédiction de performance de gestion du groupe au moyen du réseau neuronal utilisant les paramètres de poids corrigés.
  4. Système de gestion de groupe d'ascenseurs selon la revendication 1, caractérisé en ce que ledit moyen de prédiction de performance (1D, 1F), sur la base du modèle mathématique, prédit la performance de gestion du groupe qui est prédite en appliquant un ensemble arbitraire de règles stocké dans ladite base de règles (1B) à la situation de circulation actuelle.
  5. Système de gestion de groupe d'ascenseurs selon la revendication 1,

       dans lequel ledit moyen de prédiction de performance (1D, 1F) comprend de plus

       un premier moyen de prédiction de performance (1D) pour, sur la base d'un réseau neuronal, prédire la performance de gestion du groupe qui est obtenue en appliquant un ensemble arbitraire de règles stocké dans ladite base de règles (1B) à la situation de circulation actuelle ; et

       un second moyen de prédiction de performance (1F) pour, sur la base du modèle mathématique, prédire la performance de gestion du groupe qui est prédite en appliquant un ensemble arbitraire de règles stocké dans ladite base de règles (1B) à la situation de circulation actuelle ; et

       dans lequel ledit système comprend en outre

       une base de données de poids (1E) pour y stocker des paramètres de poids du réseau neuronal correspondant à l'ensemble arbitraire de règles stocké dans ladite base de règles (1B) ; et

       un moyen d'apprentissage de performance (1G) pour comparer le résultat de prédiction fourni par ledit premier moyen de prédiction de performance (1D) avec la performance réelle de gestion du groupe après avoir appliqué l'ensemble spécifique de règles pour réaliser l'apprentissage du réseau neuronal afin de corriger les paramètres de poids stockés dans ladite base de données de poids (1E) en fonction du résultat d'apprentissage, dans lequel ledit premier moyen de prédiction de performance (1D) réalise la prédiction de performance de gestion du groupe au moyen du réseau neuronal en utilisant les paramètres de poids corrigés ; et

       un moyen d'évaluation de la précision de la prédiction de performance (1H) pour comparer les résultats de prédiction fournis par lesdits premier et second moyens de prédiction de performance (1D, 1F) avec la performance réelle de gestion du groupe afin de déterminer lequel dudit premier ou second moyen de prédiction de performance (1D, 1F) est employé en fonction du résultat de comparaison, dans lequel ledit moyen de sélection d'ensemble de règles (1J) sélectionne l'ensemble optimal de règles en fonction du résultat de prédiction, à partir de l'un dudit premier ou second moyen de prédiction de performance (1D, 1F), qui a été déterminé par ledit moyen d'évaluation de la précision de la prédiction de performance (1H).






IPC
A Täglicher Lebensbedarf
B Arbeitsverfahren; Transportieren
C Chemie; Hüttenwesen
D Textilien; Papier
E Bauwesen; Erdbohren; Bergbau
F Maschinenbau; Beleuchtung; Heizung; Waffen; Sprengen
G Physik
H Elektrotechnik

Anmelder
Datum

Patentrecherche

Patent Zeichnungen (PDF)

Copyright © 2008 Patent-De Alle Rechte vorbehalten. eMail: info@patent-de.com