Elevator group scheduling methods, electronic equipment, and computer-readable storage media

Abstract

This invention discloses a scheduling method, electronic device, and computer-readable storage medium for elevator groups. The method includes: detecting an elevator call request initiated to an elevator group; the elevator group including multiple elevators; determining a target request category corresponding to the elevator call request from multiple candidate request categories based on elevator call data associated with the elevator call request; determining a target model corresponding to the target request category from multiple candidate models; calculating the predicted execution time of each elevator responding to the elevator call request using the target model based on the operating status data of each elevator; and scheduling the elevator group according to the predicted execution time of each elevator. The technical solution provided by this invention aims to solve the technical problem of low overall operating efficiency of existing elevator systems.

Description

Technical Field

[0001] This invention relates to the field of elevator control technology, and more specifically to a method for scheduling elevator groups, electronic equipment, and computer-readable storage medium. Background Technology

[0002] Elevator systems are a core component of vertical transportation and play an increasingly important role in building design, especially in office buildings with high pedestrian traffic and peak elevator usage. Elevator traffic is characterized by fuzzy logic, nonlinearity, susceptibility to interference, incomplete information, and multi-objective nature, making elevator traffic planning extremely complex. While traditional elevator parallel and group control technologies can improve transport capacity to some extent, reduce repeated responses to the same passenger demand, and enable elevators to respond to calls more frequently, thus shortening waiting times, these methods may lead to difficulties for passengers on lower floors during peak hours, and frequent elevator stops reduce overall operational efficiency. Therefore, improving the overall operational efficiency of elevator systems has become a pressing issue. Summary of the Invention

[0003] The main objective of this invention is to provide a scheduling method for elevator groups, an electronic device, and a computer-readable storage medium, aiming to solve the technical problem of low overall operating efficiency of existing elevator systems.

[0004] To achieve the above objectives, one embodiment of the present invention proposes a scheduling method for elevator groups, comprising:

[0005] An elevator call request was detected to an elevator group; the elevator group includes multiple elevators.

[0006] Based on the elevator call data associated with the elevator call request, determine the target request category corresponding to the elevator call request from multiple candidate request categories;

[0007] Determine the target model corresponding to the target request category from among multiple candidate models;

[0008] Based on the operating status data of each elevator, the target model is used to calculate the predicted execution time of each elevator responding to the call request;

[0009] The elevator group is scheduled according to the predicted execution time of each elevator.

[0010] Optionally, the step of using the target model to predict the predicted execution time of each elevator's response to the call request based on the operating status data of each elevator includes:

[0011] Obtain the operating status data of each elevator;

[0012] For each elevator, a response feature vector is constructed based on the call data and the operating status data; the response feature vector is used to represent the theoretical execution time of the elevator responding to the call request.

[0013] For each elevator, the response feature vector of the elevator is input into the target model to obtain the predicted execution time.

[0014] Optionally, the plurality of candidate models correspond one-to-one with the plurality of candidate request categories. Each candidate model is trained using associated data of historical elevator call requests belonging to the corresponding candidate request category. The associated data of the historical elevator call requests includes one or more of the following data obtained from the historical elevator call requests: the historical elevator call data, the operating status data of the responding elevator, and the actual execution time of the responding elevator; the responding elevator is the elevator that actually responded to the historical elevator call request.

[0015] Optionally, the weights of each variable in the objective function of the multiple candidate models are set independently. The objective function includes at least one variable, and at least one of the variables is related to the actual execution time of the historical elevator call request.

[0016] Optionally, before determining the target request category corresponding to the elevator call request from multiple candidate request categories based on the elevator call data associated with the elevator call request, the method further includes:

[0017] Obtain multiple historical elevator call data of the elevator group, each of which corresponds to a historical elevator call request initiated to the elevator group.

[0018] Based on the similarity between any two historical elevator call data points, the historical elevator call requests are divided into a preset number of candidate request categories.

[0019] Optionally, the step of dividing the multiple historical elevator call requests into a preset number of candidate request categories based on the similarity between every two historical elevator call data points includes:

[0020] For each of the aforementioned historical elevator call requests, a historical elevator call feature vector is constructed based on the corresponding historical elevator call data.

[0021] Based on the distance between any two historical elevator call feature vectors, the multiple historical elevator call requests are clustered into a preset number of candidate request categories.

[0022] Optionally, determining the target request category corresponding to the elevator call request from multiple candidate request categories based on the elevator call data associated with the elevator call request includes:

[0023] Obtain baseline elevator call data for each of the candidate request categories; the baseline elevator call data for any candidate request category is used as a baseline sample for elevator call data associated with elevator call requests belonging to that candidate request category.

[0024] Calculate the similarity between each of the baseline elevator call data and the elevator call data;

[0025] The candidate request category corresponding to the benchmark call data with the highest similarity is determined to obtain the target request category.

[0026] Optionally, before obtaining the baseline elevator call data for each of the candidate request categories, the method further includes:

[0027] Obtain historical elevator call data corresponding to historical elevator call requests belonging to each of the aforementioned candidate request categories;

[0028] Calculate the statistical center sample of historical elevator call data corresponding to each of the candidate request categories to obtain the baseline elevator call data for each of the candidate request categories.

[0029] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the elevator group scheduling method provided in any embodiment of the present invention.

[0030] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the elevator group scheduling method as provided in any embodiment of the present invention.

[0031] An embodiment of the present invention also provides a computer program product, including a computer program that is executed by a processor to implement the elevator group scheduling method provided in any embodiment of the present invention.

[0032] An embodiment of the present invention also provides an elevator system, including multiple elevators, and an electronic device provided in the embodiment of the present invention, the electronic device being used to execute the elevator group scheduling method provided in the embodiment of the present invention.

[0033] In one or more technical solutions provided by the above embodiments of the present invention, after detecting an elevator call request initiated to an elevator group, the target request category corresponding to the elevator call request is determined from multiple candidate request categories based on the elevator call data associated with the elevator call request. The target model corresponding to the target request category can be determined from multiple candidate models. Then, based on the operating status data of each elevator, the predicted execution time of each elevator responding to the elevator call request is calculated using the target model. The elevator group is then scheduled according to the predicted execution time of each elevator. By classifying elevator call requests into different request categories and selecting the corresponding target model to predict the execution time of each elevator, more accurate prediction results can be obtained, and the user's elevator call request can be responded to more quickly, thereby improving the operating efficiency of the elevator group. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0035] Figure 1 A flowchart illustrating an elevator group scheduling method according to an embodiment of the present invention. Figure 1 ;

[0036] Figure 2 A flowchart illustrating an elevator group scheduling method according to an embodiment of the present invention. Figure 2 ;

[0037] Figure 3 This is a schematic diagram of the neural network model in an elevator group scheduling method provided in an embodiment of the present invention;

[0038] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0039] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0041] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the inventive concept is set forth in the claims.

[0042] The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items. Depending on the context, the word “if” as used herein can be interpreted as “when,” “in response to a determination,” or “when…”.

[0043] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, product, or apparatus that includes said element.

[0044] Furthermore, in this invention, descriptions involving "first," "second," etc., are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0045] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0046] One embodiment of the present invention provides a method for scheduling elevator groups, with reference to... Figure 1 The scheduling method for this elevator group includes the following steps:

[0047] Step 101: An elevator call request to the elevator group is detected; the elevator group includes multiple elevators.

[0048] Step 102: Based on the elevator call data associated with the elevator call request, determine the target request category corresponding to the elevator call request from multiple candidate request categories;

[0049] Step 103: Determine the target model corresponding to the target request category from multiple candidate models;

[0050] Step 104: Based on the operating status data of each elevator, use the target model to calculate the predicted execution time of each elevator responding to call requests;

[0051] Step 105: Schedule elevator groups according to the predicted execution time of each elevator.

[0052] An elevator group may include multiple elevators. The elevator described in this embodiment is an electrically driven device for vertically or inclinedly transporting passengers or goods. Elevator types include vertically ascending passenger elevators, freight elevators for transporting goods, vehicle transport equipment in multi-level parking garages, escalators, and moving walkways. Multiple elevators in the elevator group can be scheduled and operated by a control system. A group call control device can be installed on each service floor of the elevator group. Each floor may have one or more group call control devices; for example, one group call control device can be installed next to each elevator door. Each group call control device may have the same or different operation panels. For example, next to an elevator that does not provide upward service, the operation panel of the group call control device may only have a downward button. Users of the group call control device can call an elevator that can serve their current floor by operating the device from any floor. A service floor refers to the floor where the elevator can provide service. Optionally, the service floors of different elevators in the elevator group may be the same or different. The floor issuing the call request can call an elevator that can serve its current floor. In some embodiments, the operation panel of the elevator group control device may be equipped with buttons for moving upwards and / or downwards, providing users with the function of calling for upward and / or downward movement of the elevator. In other embodiments, the operation panel of the elevator group control device may provide buttons for setting a specific destination floor, thereby providing users with the function of calling for movement to a specific floor. When any user registers an elevator call request through the elevator group control device and sets the upward / downward / destination floor, the executor of this application embodiment detects the elevator call request initiated to the elevator group. In some embodiments, the executor of this application embodiment may be a control system capable of controlling the elevator group, obtaining the elevator call request through a communication connection with each elevator group control device.

[0053] Based on the elevator call data associated with the elevator call request, the target request category can be determined from multiple candidate request categories. The elevator call data associated with the request describes the demand information for the request, including but not limited to: the direction of the request, the calling layer (i.e., the current layer), the destination layer (i.e., the layer the user wants to go to), the time the request was initiated (specifically including but not limited to the exact time or time period of the day, day of the week, day of the month, etc.), and the identity of the passenger initiating the request (e.g., if the passenger initiates the request by swiping a card, the passenger's identity is determined based on the card's identifier). Classification is performed based on the elevator call data to determine the category to which the elevator call request belongs from multiple pre-provided candidate request categories. In some implementations, the elevator call data can be divided into multiple ranges based on at least one dimension; for example, the elevator call time can be divided into different time periods of the day. The elevator call request category can then be classified based on the specific range to which the elevator call data belongs in that dimension. It is understood that other dimensions of data can be combined to determine the specific category. In other implementations, historical elevator call data from multiple historical requests can be obtained to train an AI-based classification model, and then the trained classification model can be used to determine the category of the current elevator call request.

[0054] After determining the target request category to which the elevator call request belongs, a target model corresponding to the target request category is selected to predict the elevator's response to the call request. Different candidate request categories correspond to different candidate models. Each candidate model can be a trained artificial intelligence model. In some embodiments, each candidate model can be a different type of artificial intelligence model based on a different framework. In some embodiments, each candidate model can be based on the same framework but with different specific structures, such as different numbers of network layers in a neural network model. In still other embodiments, each candidate model can be based on the same framework and structure but have different weights / parameters after training. Optionally, the network structure of the candidate model can be a fuzzy neural network. A fuzzy neural network combines fuzzy theory with a neural network, using activation functions to map input values, realize fuzzy membership degree calculation and fuzzy inference, and output predicted values ​​with fuzziness and historical data patterns.

[0055] Specifically, elevator call request data samples belonging to each candidate request category can be used to train the corresponding candidate model. One elevator call request data sample can correspond to one historical elevator call request. The elevator call request data sample can include monitoring data of the responding elevator. The responding elevator is the elevator that actually responded to the historical elevator call request. The monitoring data includes, but is not limited to, the elevator call data, operation status data, and actual execution time of the historical elevator call request. Among them, the elevator call data is the data obtained when the user initiates the elevator call request, including but not limited to the time of initiating the elevator call request, the calling floor, the destination floor, and the user's identity (for example, the user calls the elevator by swiping a card, and the user's identity is determined based on the card identifier). The operation status data is the elevator's operation parameters monitored when the elevator call request is initiated. It can be understood that initiating an elevator call request can refer to the time when the elevator initiates the elevator call request. The actual execution time is related to the time of arrival at the departure floor and the time of arrival at the destination floor after the elevator responds to the elevator call request. It can be understood that when the elevator call request data sample corresponds to a historical elevator call request, the above-mentioned elevator refers to the monitoring data of the elevator in the elevator group that actually responded to the elevator call request. After obtaining the above monitoring data, an input vector can be constructed based on the elevator call data and operating status data, and an output vector can be constructed based on the actual execution time. The model can be trained with the actual execution time as the output training target, so that the trained model can be used to predict the execution time of the elevator responding to the elevator call request.

[0056] After selecting the target model corresponding to the target request category of the elevator call request, the predicted execution time for each elevator to respond to the current elevator call request is predicted using the target model based on the operating status data of each elevator. Operating status data refers to parameters collected within a preset time period before and after receiving the elevator call request, representing the elevator's operating status. This includes, but is not limited to, the elevator's current operating or idle status, current floor / operating position / next floor, whether the elevator door is open, whether the elevator is available (e.g., unavailable if under maintenance), call data received inside the elevator car, call data obtained from any available floor outside the elevator car, and call data from elevator calls awaiting allocation. Specifically, for each elevator, a feature vector is constructed based on its operating status data. This feature vector is input into the target model. Optionally, a feature vector input to the target model can also be constructed by combining call data. Call data includes, but is not limited to, the calling floor, destination floor, and call time in the call data.

[0057] After constructing feature vectors for each elevator, these feature vectors are input into the target model. The target model then outputs the predicted execution time for that elevator based on the input feature vector. It is understood that when inputting feature vectors into the target model, each elevator can be relatively independent. For example, multiple threads can be initiated, each independently calculating the feature vector of one elevator using the target model. Any thread can sequentially calculate the feature vectors of different elevators. In other implementations, the target model structure can be configured to simultaneously receive feature vectors from multiple elevators at the input layer. Correspondingly, the training data for the input and output can be constructed according to the corresponding structure when training the target model. The predicted execution time for any elevator refers to the time it takes for that elevator to respond to a call request from an elevator currently awaiting scheduling or allocation. For example, the passenger waiting time when the elevator reaches the calling floor. In applications where the elevator group control device is equipped with a call destination floor function, the predicted execution time can also include the passenger travel time from the calling floor to the destination floor, as well as the total passenger waiting time and passenger travel time. In a specific application example, the predicted execution time output by the target model is = w1 × passenger waiting time + w2 × passenger riding time, where w1 and w2 are the weights assigned to the two durations, and the weight values ​​can be predetermined or obtained by training the target model.

[0058] After calculating the predicted execution time for each elevator, the corresponding elevator can be assigned to the call request based on the prediction results, and the assigned elevator can be scheduled to stop at the calling floor. Optionally, if a destination floor can be set when initiating an elevator call request, a scheduling instruction can be issued to the assigned elevator after the assigned elevator is determined, controlling it to stop at both the calling floor and the destination floor. In other embodiments, if no destination floor is set when initiating an elevator call request, the scheduling instruction issued to the assigned elevator after the assigned elevator is determined may include controlling it to stop at the calling floor. After the calling passenger boards the elevator, the instruction receiving device in the elevator car receives the destination floor specified by the passenger, and then the elevator stops at the destination floor according to the passenger's instruction in the elevator.

[0059] In one or more technical solutions provided by the above embodiments of the present invention, after detecting an elevator call request initiated to an elevator group, the target request category corresponding to the elevator call request is determined from multiple candidate request categories based on the elevator call data associated with the elevator call request. The target model corresponding to the target request category can be determined from multiple candidate models. Then, based on the operating status data of each elevator, the predicted execution time of each elevator responding to the elevator call request is calculated using the target model. The elevator group is then scheduled according to the predicted execution time of each elevator. By classifying elevator call requests into different request categories and selecting the corresponding target model to predict the execution time of each elevator, more accurate prediction results can be obtained, and the user's elevator call request can be responded to more quickly, thereby improving the operating efficiency of the elevator group.

[0060] In some optional technical solutions of this embodiment, step 104, based on the operating status data of each elevator, uses the target model to predict the predicted execution time of each elevator responding to a call request. Specifically, it may include performing the following steps:

[0061] Obtain the operating status data of each elevator;

[0062] For each elevator, a response feature vector is constructed based on call data and operating status data; the response feature vector is used to represent the theoretical execution time of the elevator responding to a call request.

[0063] For each elevator, the elevator's response feature vector is input into the target model to obtain the predicted execution time.

[0064] The operational status data of each elevator can be automatically uploaded by each elevator. Specifically, it can communicate with the executor of this embodiment of the invention based on a pre-configured communication method, or it can be acquired through monitoring sensors pre-configured on each elevator. This embodiment of the invention does not impose any restrictions on this. After acquiring the operational status data, a response feature vector can be constructed based on the call data and operational status data to be input to the target model. For example, the theoretical execution time represented by the response feature vector includes, but is not limited to, the following for the elevator: the waiting time for the current call request, the riding time for the current call request, the increment of the waiting time for the allocated call request, the increment of the riding time for the allocated call request, the increment of the riding time for passengers in the elevator, and the increment of the elevator's single round-trip travel time. When calculating the above-mentioned theoretical execution time, it can be calculated in conjunction with the following data, at least one of the following elevator operational status data: elevator running direction, current position, whether the door is open, the floor to be stopped, and the calling floor and destination floor in the call data.

[0065] Since the target model outputs the predicted execution time of the elevator response to the call request, setting the feature vector input to the target model as a response feature vector related to the theoretical execution time can make the input vector and output vector closely related and matched in feature dimension. This improves the interpretability of the target model's decision, yields more accurate prediction results, and enhances the training efficiency of the target model.

[0066] In some optional technical solutions of this embodiment, multiple candidate models correspond one-to-one with multiple candidate request categories. Each candidate model is trained using associated data of historical elevator call requests belonging to the corresponding candidate request category. The associated data of historical elevator call requests includes one or more of the following data obtained based on the historical elevator call requests: historical elevator call data, operating status data of the responding elevator, and actual execution time of the responding elevator. The responding elevator is the elevator that actually responded to the historical elevator call request, and it is the elevator that was actually assigned to the historical elevator call request at that time, scheduled to the calling floor, and running to the destination floor. Specifically, before performing step 103 to determine the target model corresponding to the target request category among the multiple candidate models, the method may further include the following steps:

[0067] For historical elevator call requests of each candidate request category, obtain historical elevator call data, elevator operation status data, and actual execution time;

[0068] For each candidate request category, a candidate model corresponding to the candidate request category is trained based on the historical elevator call data, the operating status data of the responding elevator, and the actual execution time of each historical elevator call request.

[0069] The aforementioned technical solution, by customizing training models for different candidate request categories, enables each trained candidate model to more accurately capture the unique patterns and characteristics of various elevator call requests. This targeted training method allows the model to learn the data features of specific request categories. Through refined model training, the accuracy of the prediction results of each candidate model is optimized. By using more accurate prediction results to allocate elevators, the overall operating efficiency of the elevator system can be improved.

[0070] Furthermore, regarding the training steps corresponding to the candidate models for candidate request categories, in some optional technical solutions, the weights of each variable in the objective function of each candidate model can be set for each candidate model corresponding to the candidate request category. The objective function includes at least one variable, and at least one of these variables is related to the actual execution time of historical elevator call requests. In other words, the weights of the objective functions of multiple candidate models are set relatively independently, and the objective function weights include the weights of each variable in the objective function of the candidate models, and each variable in the objective function is related to the actual execution time of historical elevator call requests.

[0071] The objective function (also known as the loss function or cost function) is used to measure the difference between the model output and the actual target. In some implementations, the variables in the objective function may include passenger waiting time and passenger elevator travel time. The objective function can be w1 × passenger waiting time + w2 × passenger elevator travel time, where w1 and w2 are the weights of the two time periods, respectively. For each candidate request category, the weight values ​​w1 and w2 in the objective function can be set independently. A specific example is provided below:

[0072] Candidate request categories can include the following four types: morning peak, noon peak, evening peak, and off-peak. For requests in the morning peak category, the weight w1 is set to the highest value among multiple categories, exceeding the preset value w2. The weight w1 for noon and evening peaks is set to be lower than the morning peak weight w1, and close to or equal to w2. The weight w1 for off-peak categories is the lowest among multiple categories, and can be less than the weight w2. The weight w2 for off-peak categories is the highest among multiple categories. The principle is that the elevator travel time is usually shorter than the elevator waiting time. During the morning peak, people need to reach their destination floor more quickly. Therefore, setting a larger weight for elevator waiting time can give more consideration to the impact of elevator waiting time on the objective function result. During other peak periods, people's need to reach their destination floor quickly is reduced, which can reduce the impact of elevator waiting time on the objective function result. During off-peak periods, people may prioritize the elevator travel experience. Therefore, for off-peak periods, the weight of elevator travel time can be appropriately increased.

[0073] By assigning different objective function weights to different candidate request categories, the model can better align with the specific needs and priorities of each candidate request category in real-world application scenarios, thereby improving prediction accuracy and scheduling efficiency. For example, during morning rush hour, increasing the weight of waiting time allows the model to more effectively reduce waiting time and quickly respond to peak passenger flow demands; while during off-peak hours, increasing the weight of riding time can better optimize the passenger experience. This differentiated weighting allows the model to flexibly adapt to the characteristics of different elevator call request categories, achieving more refined and personalized elevator scheduling management.

[0074] In some optional technical solutions of this embodiment, before executing step 102 to determine the target request category corresponding to the elevator call request from multiple candidate request categories based on the elevator call data associated with the elevator call request, multiple candidate request categories can be determined by performing the following steps: obtaining multiple historical elevator call data of the elevator group, where each historical elevator call data corresponds one-to-one with multiple historical elevator call requests initiated to the elevator group; and dividing the multiple historical elevator call requests into a preset number of candidate request categories based on the similarity between every two historical elevator call data. By collecting the elevator call data of historical elevator call requests in the elevator group, the historical elevator call requests can be clustered into a specified number of categories based on the similarity between the elevator call data. By specifying the number of categories, the classification results can meet the required discriminative power. Clustering is used to divide an unlabeled dataset into classes composed of multiple similar samples. Optionally, a hierarchical clustering method can be used for clustering.

[0075] Historical elevator call data can be a vector containing at least one dimension of data, and the similarity between any two historical elevator call data can be measured by metrics such as the distance between the two vectors, cosine similarity, or correlation coefficient. Further, the step of classifying multiple historical elevator call requests into a predetermined number of candidate request categories based on the similarity between every two historical elevator call data can specifically include the following steps: for each historical elevator call request, constructing a historical elevator call feature vector based on the corresponding historical elevator call data; and clustering multiple historical elevator call requests into a predetermined number of candidate request categories based on the distance between any two historical elevator call feature vectors. It can be understood that the data in each dimension of the constructed historical elevator call feature vector can include the original data from the historical elevator call data, or processed data obtained by calculating and processing the historical elevator call data. Representing the similarity between any two historical elevator call requests by the vector distance between them provides an intuitive quantification of the similarity between any two historical elevator call requests, offering a clear mathematical basis for cluster analysis. Classification through clustering methods helps discover the correlations between historical elevator call data, thus naturally grouping the historical elevator call requests.

[0076] In some alternative technical solutions of this embodiment, step 102, which determines the target request category corresponding to the elevator call request from multiple candidate request categories based on the elevator call data associated with the elevator call request, may specifically include performing the following steps:

[0077] Obtain baseline elevator call data for each candidate request category; the baseline elevator call data for any candidate request category is used as a baseline sample for elevator call data associated with elevator call requests belonging to that candidate request category.

[0078] Calculate the similarity between each baseline elevator call data and the elevator call data;

[0079] The candidate request category corresponding to the baseline call data with the highest similarity is determined to obtain the target request category.

[0080] For each candidate request category, call data for each candidate request category can be provided in advance as a benchmark. In this way, by calculating the similarity between the benchmark call data and the call data of the current call request, it can be determined which candidate request category's benchmark is more similar to the current call request, and the candidate request category with the highest similarity is selected as the target request category corresponding to the current call request.

[0081] Optionally, each candidate request category can be a classification determined manually based on experience. Correspondingly, the baseline call data for each candidate request category can also be manually assigned based on experience. Alternatively, the baseline call data can be obtained by selecting the data with the most repeated calls or the data located at the median / mean from the historical call data belonging to each candidate request category. Optionally, before performing the above steps to obtain the baseline call data for each candidate request category, the historical call data corresponding to the historical call requests belonging to each candidate request category can be obtained, and the statistical center sample corresponding to the historical call data for each candidate request category can be calculated to obtain the baseline call data for each candidate request category. The statistical center sample can be the mean, median, mode, centroid, etc. When the historical call data includes more than two data points, one historical call data point can be regarded as a vector with more than two dimensions. The statistical center sample of multiple historical call data points can be obtained by calculating the geometric center of multiple vectors. Using the statistical center sample of historical elevator call data corresponding to each candidate request category as the benchmark elevator call data for each candidate request category can capture the typical characteristics of each candidate request category, which helps to improve the accuracy of elevator call request classification.

[0082] Figure 2This is a flowchart illustrating a specific application example of an elevator group scheduling method provided by the present invention. In this application scenario, the elevator system supports destination floor group control, allowing passengers to obtain their destination floor information via a call device outside the elevator at the calling floor before boarding. Current elevator group scheduling methods include: expert system-based control algorithms, fuzzy logic-based control algorithms, and neural network-based control algorithms. However, in expert system-based and fuzzy logic-based control algorithms, the rules or weights are determined by experts based on experience. Since passenger needs differ at different times—for example, during morning rush hour, passengers are less concerned about elevator congestion and more focused on getting on the elevator, while during evening rush hour, passenger tolerance for congestion decreases—using the same allocation algorithm for both periods may result in low satisfaction for some passengers. Therefore, using the same weight allocation algorithm for different building configurations and different time periods with fixed parameters lacks rationality.

[0083] refer to Figure 2 The technical solution of the elevator group scheduling method provided in this example can be summarized as follows: After preprocessing the historical elevator call data, it is divided into different subsets using hierarchical clustering. The historical elevator call data of each subset is then used to train a fuzzy neural network model, resulting in a fuzzy neural network model suitable for each subset. The model's independent variables consider various factors such as the waiting time and riding time of passengers in different states, including newly calling passengers, passengers already waiting, and passengers already using the elevator. The model is used to predict the waiting time and riding time of passengers in each state when using each elevator within the elevator group. When predicting the current elevator call request, the distance between the elevator call data of newly calling passengers and various request categories is used to determine the classification of the newly calling passenger's elevator call request. The corresponding neural network model is then selected, and the operating status data of each elevator is obtained. Feature vectors are generated for each elevator to be input into the neural network model. After calculating the prediction results for each elevator using the neural network model, the elevator with the smallest predicted value is selected and assigned to the currently newly added elevator call request.

[0084] Specifically, refer to Figure 2 The elevator group scheduling method provided in this example includes the following steps:

[0085] Step 201: Utilizing the data recording function of the group controllers set up on each service floor of the elevator system, collect elevator call data, operational status data, actual waiting time, and actual riding time (actual execution time) from the elevator system installation site. This data can be processed and saved via cloud and / or local devices. Call data may include the calling floor, destination floor, and call time. Elevator operational status data may include the assigned elevator's current floor, next stopping floor, direction of travel, door status, external call information, internal call information, and pending call information (i.e., call requests not yet assigned to an elevator). Actual waiting time and actual riding time are information monitored after the elevators in the elevator group actually respond to the call requests.

[0086] Step 202: Preprocess the historical elevator call data to construct a feature vector for classifying elevator call requests. The feature vector for classification may include variables in the following dimensions: direction of travel, elevator call type (entering, exiting, inter-floor), weekday, etc.

[0087] In one example, a dataset S = {X} is constructed. ij}, 1≤i≤N, 1≤j≤K+3, treat each historical elevator call request as a sample X. i The dataset contains N samples, each containing K+3 variables (or data dimensions). The first K variables in each sample are the original collected data, while the (K+1)th to (K+3)th variables are generated from the original data. For example, the elevator direction X is determined based on the call level and the destination level. i,K+1 Determine the day of the week X based on the time. i,K+2 Determine the elevator call type X based on the calling floor and the destination floor. i,K+3 It can be classified as outbound, inbound, or inter-level. If it is an upward trip from the entrance level, it is inbound data; if the destination level is a downward trip from the entrance level, it is outbound data; otherwise, it is inter-level data.

[0088] Step 203: Cluster the above feature vectors using hierarchical clustering.

[0089] Define two samples and The distance between them is:

[0090]

[0091] The inter-class distance between two classes S1 and S2 is defined as the maximum distance between sample points in the two classes, that is, the inter-class distance between class S1 and class S2 is:

[0092]

[0093] Initially, each sample point is assigned a class. The two classes with the smallest inter-class distance are merged sequentially until the final number of classes matches the set value. The final division is specified as 4 classes, resulting in 4 data subsets. Descriptive analysis is performed on these 4 data subsets, and the characteristics of each group can be preliminarily analyzed based on factors such as the proportion of different call types.

[0094] Step 204: Model training is performed using four subsets of data, with each subset employing a fuzzy neural network. Independent and dependent variables are constructed for each data sample in each class to obtain training data for the model. The independent variables are constructed based on historical elevator operation status data at the time of call initiation. Specifically, based on the allocated elevator operation status data, the theoretical waiting time Z1 and riding time Z2 for currently calling passengers, the waiting time increment Z3 and riding time increment Z4 for those who have already called, the riding time increment Z5 for those currently inside the elevator, and the time increment Z6 for a single round trip of the elevator are calculated, resulting in six independent variables. An example fuzzy neural network structure is shown below. Figure 3 As shown, there are 5 layers in total. The first layer is the input layer, which takes input variables Z1, Z2, Z3, Z4, Z5, and Z6. The second layer is the membership function layer, specifically, a Gaussian function can be selected as the membership function. ii Represents a membership function:

[0095]

[0096] Where u ij z is the output value of the membership function layer node. ij For input values, Here, is the parameter of the membership function, and m is the number of fuzzy levels. The third layer is the rule layer, a j This represents the j-th rule, i.e.:

[0097] a j =Πu ij

[0098] The fourth layer is the normalization layer, which normalizes the rule output value a. j Normalization is performed, that is:

[0099]

[0100] The fifth layer is the output layer, which performs a weighted summation of the normalized values ​​before outputting them.

[0101]

[0102] The dependent variable is constructed by weighting the actual waiting time and elevator travel time. The dependent variable Y = W1 × the passenger's actual waiting time + W2 × the passenger's actual elevator travel time. For datasets biased towards the morning rush hour, the proportion of actual waiting time in the dependent variable is larger; for datasets biased towards the evening rush hour, the proportion of actual elevator travel time increases; for other scenarios, the ratio of waiting time to travel time is consistent. During model training, the difference between the model output y and the constructed dependent variable Y can be used as feedback to adjust the model's internal parameters.

[0103] Step 205: Deploy the trained model into the elevator system.

[0104] Step 206: Received a new elevator call request.

[0105] Step 207: Calculate the similarity between the elevator call data of newly added elevator call requests and each data subset, and select the subset with the highest similarity to train the model.

[0106] Calculate the dataset S respectively k The class centers of |k=1..4 are denoted as respectively Regarding the newly added elevator call request X p Calculate X p With class center Distance d(X) p ,S k Select the data with the smallest distance. The corresponding model is FNN. k .

[0107] Step 208: For each elevator in the elevator group, calculate the theoretical duration values ​​of elevators such as Z1 to Z6 based on the current operating status data of the elevators, use them as independent variables, and substitute them into the model for prediction to predict the predicted value of each elevator.

[0108] Step 209: Allocate elevators according to the predicted values ​​and issue dispatch instructions to the corresponding elevators to respond to the call requests. The elevator with the smallest predicted value is assigned to the new call request; alternatively, taking the inverse of the predicted value yields the degree of matching between the elevator and the new call request, and the elevator with the highest degree of matching is assigned to that call request.

[0109] In other application scenarios, for elevator systems that do not support destination floor group control but only support ordinary group control, since the elevator call data lacks information on newly calling passengers and destination floor information of waiting passengers, the number of independent variables can be reduced. For example, the independent variables may include: the waiting time of newly calling passengers, the increase in waiting time of people waiting at the calling floor, the increase in riding time of passengers in the elevator car, and the increase in time of a single round trip of the elevator, which can then be substituted into the model for training and prediction.

[0110] Using specific models for prediction under different data patterns can more effectively meet passenger needs and improve passenger satisfaction. Fuzzy neural networks combine the advantages of fuzzy logic and neural networks, enabling them to handle uncertainty and nonlinear problems, and possess adaptive learning capabilities. Neural network models can learn patterns from historical data, introduce fuzzy rules, and mine these rules from the data, giving the model a degree of interpretability. Models trained using similar data can predict the optimal elevator for new passengers, considering the optimal efficiency for both new passengers and the overall passenger flow.

[0111] The present invention also provides an embodiment of an elevator group scheduling device, which includes an execution module that can be used to perform each step of the elevator group scheduling method provided in the embodiment of the present invention. For details, please refer to the relevant description in the embodiment of the elevator group scheduling method, which will not be repeated here.

[0112] Figure 4 This is a block diagram of an electronic device used to implement the elevator group scheduling method provided in the embodiments of this application. Figure 4 As shown, the electronic device includes a memory 401 and a processor 402. The memory 401 stores a computer program that can run on the processor 402. When the processor 402 executes the computer program, it implements the method described in the above embodiments. The number of memories 401 and processors 402 can be one or more. In a specific implementation, the electronic device may also include a communication interface 403 for communicating with external devices and performing data exchange and transmission.

[0113] In practical implementation, if the memory 401, processor 402, and communication interface 403 are implemented independently, they can be interconnected via a bus to communicate with each other. This bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0114] Optionally, in a specific implementation, if the memory 401, processor 402 and communication interface 403 are integrated on a single chip, the memory 401, processor 402 and communication interface 403 can communicate with each other through an internal interface.

[0115] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the elevator group scheduling method as provided in any embodiment of the present invention.

[0116] One embodiment of the present invention also provides a computer program product, including a computer program that is executed by a processor to implement the elevator group scheduling method provided in any embodiment of the present invention. Similar to a computer-readable storage medium, the computer program can also be executed by an electronic device or an electronic device communicating with an electronic device, as will not be elaborated further.

[0117] One embodiment of the present invention also provides an elevator system, including multiple elevators, and an electronic device provided in the embodiment of the present invention, which is used to execute the elevator group scheduling method provided in the embodiment of the present invention.

[0118] The above description is merely an optional embodiment of the present invention and does not limit the patent scope of the present invention. Any modifications, equivalent substitutions, improvements, etc., made under the concept of the present invention and using the contents of the present invention specification and drawings, or directly / indirectly applied to other related technical fields, should be included within the patent protection scope of the present invention.

Claims

1. A method for scheduling elevator groups, characterized in that, The method includes: An elevator call request was detected to an elevator group; the elevator group includes multiple elevators. Based on the elevator call data associated with the elevator call request, determine the target request category corresponding to the elevator call request from multiple candidate request categories; The target model corresponding to the target request category is determined from multiple candidate models; each candidate model is obtained by training a neural network model. The operation status data of each elevator is obtained. For each elevator, a response feature vector is constructed based on the call data and the operation status data. The response feature vector is used to represent the theoretical execution time of the elevator in response to the call request. The theoretical execution time is calculated based on the operation status data of the elevator. For each elevator, the response feature vector of the elevator is input into the target model to obtain the predicted execution time; The elevator group is scheduled according to the predicted execution time of each elevator.

2. The method according to claim 1, characterized in that, The multiple candidate models correspond one-to-one with the multiple candidate request categories. Each candidate model is trained using the associated data of historical elevator call requests belonging to the corresponding candidate request category. The associated data of historical elevator call requests includes one or more of the following data obtained from the historical elevator call requests: historical elevator call data, operating status data of the responding elevator, and the actual execution time of the responding elevator. The responding elevator is the elevator that actually responded to the historical elevator call request.

3. The method according to claim 2, characterized in that, The weights of each variable in the objective function of the multiple candidate models are set independently. The objective function includes at least one variable, and at least one of the variables is related to the actual execution time of the historical elevator call request.

4. The method according to claim 1, characterized in that, Before determining the target request category corresponding to the elevator call request from multiple candidate request categories based on the elevator call data associated with the elevator call request, the method further includes: Obtain multiple historical elevator call data of the elevator group, each of which corresponds to a historical elevator call request initiated to the elevator group. Based on the similarity between any two historical elevator call data points, the historical elevator call requests are divided into a preset number of candidate request categories.

5. The method according to claim 4, characterized in that, The step of classifying the multiple historical elevator call requests into a preset number of candidate request categories based on the similarity between every two historical elevator call data points includes: For each of the aforementioned historical elevator call requests, a historical elevator call feature vector is constructed based on the corresponding historical elevator call data. Based on the distance between any two historical elevator call feature vectors, the multiple historical elevator call requests are clustered into a preset number of candidate request categories.

6. The method according to claim 1, characterized in that, The step of determining the target request category corresponding to the elevator call request from multiple candidate request categories based on the elevator call data associated with the elevator call request includes: Obtain baseline elevator call data for each of the candidate request categories; use the baseline elevator call data for any candidate request category as a baseline sample for elevator call data associated with elevator call requests belonging to the candidate request category; calculate the similarity between each of the baseline elevator call data and the elevator call data; The candidate request category corresponding to the benchmark call data with the highest similarity is determined to obtain the target request category.

7. The method according to claim 6, characterized in that, Before acquiring baseline call data for each of the candidate request categories, the method further includes: Obtain historical elevator call data corresponding to historical elevator call requests belonging to each of the aforementioned candidate request categories; Calculate the statistical center sample of historical elevator call data corresponding to each of the candidate request categories to obtain the baseline elevator call data for each of the candidate request categories.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the program to implement the method as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by a processor to implement the method as described in any one of claims 1-7.

10. An elevator system, characterized in that, It includes multiple elevators and also includes the electronic device of claim 8, the electronic device being used to perform the method of any one of claims 1-7.

Images