Cloud-side collaborative intelligent elevator group control system

Abstract

The invention discloses a cloud-side collaborative intelligent elevator group control system which is characterized in that the system comprises a cloud platform and an edge side, and the cloud platform comprises an AI model, a cloud computing module, a data circulation module, a decision module, a dispatching command module, an abnormal backflow module and an efficiency evaluation module; the edge side comprises a data acquisition module, a data cleaning processing module, a scheduling execution module and a man-machine interaction module. According to the invention, a system structure design in which knowledge in the traditional elevator group control field is fused with cloud computing, edge computing, artificial intelligence, Internet of Things and other technologies is created, a technical path of cloud-side cooperative allocation of elevator group control system services is opened, and the intelligence, automation and adaptive capabilities of a traditional isolated elevator group control system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of elevator technology, in particular to a cloud-edge collaborative intelligent elevator group control system. BACKGROUND

[0002] The application of existing cloud-edge collaborative control technology in the field of elevators is mostly limited to single-lift-side fault prediction or life analysis, and there is no case of combination with elevator group control systems. Chinese patent document 1 (CN116553323A) mentions a single-lift-side cloud-edge collaborative method, which downloads the trained model on the cloud to the edge side, and optimizes the fine-tuning of the model based on the collected data on the edge side, achieving the purpose of fault probability prediction, but the model described in document 1 can only process frequency domain data and fails to disclose the prediction processing effect on time series data, and at the same time, the application is also limited to single-lift-side audio analysis, resulting in limited application range. Chinese patent document 2 (CN118255219A) mentions cloud-edge collaboration and digital twinning, but on the one hand, it is still limited to single-elevator digital twinning and lacks applicability to multiple group control elevators, and on the other hand, the cloud-edge collaboration described in document 2 only involves building a more accurate observation system on the cloud and does not involve control and deployment of elevators, which is essentially one-way monitoring of the edge side from the cloud rather than two-way monitoring and control. The combination of artificial intelligence and cloud-edge collaboration in the field of elevator group control systems usually relies on image recognition technology of additional cameras, such as Chinese patent document 3 (CN119551518A), or simply uses artificial intelligence technology on the edge side for simple modeling, such as Chinese patent document 4 (CN118619024A) and Chinese patent document 5 (CN117985557A).

[0003] The existing methods mentioned in the above documents have one common point, which is that they do not completely integrate the knowledge of elevator group control system field, cloud computing, edge computing, artificial intelligence, and Internet of Things technology at the system level. At the same time, the models established in the above documents often cannot achieve automatic identification and feedback of abnormal data and automatic optimization and iteration of the model due to the lack of 7x24h timed collection of Internet of Things data. At the same time, the above documents lack an evaluation system for evaluating the pros and cons of the model using actual data, which belongs to open-loop control in classical control theory and has low practicality. At the same time, the artificial intelligence mentioned in most documents requires the addition of cameras or microphones to recognize images or sounds, which also limits the application range to the classic video and audio field, and lacks analysis and utilization of pure time series data of elevator group control systems. SUMMARY

[0004] To solve the above technical problems, the application provides a cloud-edge collaborative intelligent elevator group control system, which comprises a cloud platform and an edge side.

[0005] The data acquisition module uses Internet of Things technology to collect and transmit all elevator group control system operation data to the cloud computing module; the cloud computing module schedules the AI model to perform prediction, classification and evaluation tasks of the elevator group control according to the input information; the data flow module transfers the results calculated by the cloud computing module to subsequent other modules according to the purpose; the decision module makes a decision command according to the multi-model fusion or expert knowledge heuristic search method in part of the task; the scheduling command module schedules the relevant execution module to execute the specified task according to the decision command; the abnormal backflow module is responsible for judging whether the data after executing the task exceeds the expected need to remind experts to analyze; the efficiency evaluation module automatically calculates the group control deployment efficiency according to the collected group control state data; the cloud-edge collaborative intelligent elevator group control system uses Internet of Things, cloud computing and artificial intelligence technology to build a digital twin system for the elevator group control system on the cloud.

[0006] Preferably, the AI model is trained by the data collected by a single elevator group control system, or by the data collected by multiple similar or different types of elevator group control systems; the historical passenger flow data, holiday information and other key data of the group control system can be used to predict the future passenger flow of the elevator group control system through the model. At the same time, the peak historical data labeled by experts and the passenger flow prediction results of the group control system are input into the model, and combined with expert knowledge rules, various different passenger flow peaks or non-peak are automatically classified, realizing personalized elevator group control system and more accurate identification of passenger flow peak of the group control system to make preparation in advance and improve the deployment efficiency of the group control system.

[0007] Preferably, the group control deployment efficiency at least includes the average waiting time and the long waiting rate of a single group control system itself; the efficiency evaluation module compares and calculates the ranking median of the elevator group control system in a large number of similar buildings according to the same type of buildings accessed by the cloud platform; the efficiency evaluation module is also used to compare the efficiency indicators of different elevator group control systems in the same building accessed by the cloud platform.

[0008] Preferably, the data acquisition module only calculates passenger flow information according to the original operation data of the elevator group control system, and the passenger flow information at least includes the number of people getting on and off the elevator at each uplink and downlink stop; the passenger flow information is processed by the data cleaning processing module to restore the complete traffic flow data.

[0009] Preferably, the training of the electric AI model adopts the method of convolutional neural network deep learning and the method of decision tree traditional machine learning.

[0010] Preferably, the data acquisition module automatically acquires data for 7x24 hours.

[0011] Preferably, the AI model has a dynamic optimization iteration function, which automatically optimizes and iterates the AI model using difficult examples; the difficult examples are returned by the abnormal return flow module to identify cases that are not accurately classified by the model.

[0012] Preferably, the scheduling execution module generates a scheduling command based on the results analyzed by the AI model and sends it to the elevator group control system for deployment.

[0013] Preferably, the steps of the intelligent elevator group control system deployment in cloud-edge collaboration are as follows:

[0014] Step S1, collect group control running data on the edge side, including real-time floor, direction, in-car instructions, landing call, landing call waiting time, etc.

[0015] Step S2, the edge side uses edge technology to complete preliminary data cleaning and processing, and uses Internet of Things technology to transmit data to the cloud platform;

[0016] Step S3, the cloud platform selects features or directly constructs a data set based on the group control running data uploaded by the edge side;

[0017] Step S4, the cloud platform starts training the AI model based on the constructed data set or feature engineering and continuously iterates the training until the AI model with satisfactory precision and recall rate is obtained;

[0018] Step S5, the cloud platform schedules the computing resources on the cloud, deploys the AI model, and processes the real-time collected group control state data into the input data required by the AI model;

[0019] Step S6, the AI model automatically performs real-time inference based on the input data to obtain the prediction of future passenger flow;

[0020] Step S7, the AI model classifies whether there will be a passenger flow peak and what kind of passenger flow peak in a certain period of time based on the prediction result of passenger flow and multi-model fusion technology;

[0021] Step S8, the cloud platform decision module deeply processes the peak classification result according to statistical rules and expert rules, and finally obtains the deployment decision command that the control group control system should immediately enter or exit the peak or what kind of peak;

[0022] Step S9, the cloud platform scheduling command module calls the edge side scheduling execution module according to the command of the decision module to control the elevator group control system to enter or exit the peak or what kind of peak task;

[0023] Step S10, the efficiency evaluation module of the cloud platform automatically calculates the efficiency index according to the continuously collected group control state data, and quantitatively evaluates the group control running efficiency in multiple dimensions;

[0024] Step S11, after scheduling the edge side, the cloud platform compares the quantified index calculated by the efficiency evaluation module with the actual group control system performance, judges the rationality of this scheduling according to the expert system, and if it is not reasonable, returns the scheduling as a difficult example through the abnormal return flow module for optimization iteration of the AI model.

[0025] Preferably, the method for training the AI model comprises the following steps:

[0026] Step A1, collecting group control running data on the edge side, at least including real-time floor, direction, in-car instruction, landing call, landing call waiting time, etc.

[0027] Step A2, the edge side uses edge technology to complete preliminary data cleaning and processing, and uses Internet of Things technology to automatically send to the cloud platform;

[0028] Step A3, the cloud platform selects features based on the group control running data uploaded by the edge side, and attempts to use weather, holidays, building types, and time information as parameter variables to assist the model training link;

[0029] Step A4, using the collected group control running data as the main parameter variable and the weather assistance data as the secondary parameter variable, the model is trained according to the traditional machine learning decision tree method;

[0030] Step A5, continuously repeat the model training process, automatically adjust the weight of each variable, output the effect of the model on the historical data validation set, until the accuracy and recall rate of the model no longer improve, and the preliminary training of the model is completed.

[0031] Preferably, the method for training the AI model comprises the following steps:

[0032] Step B1, collecting group control running data on the edge side, at least including real-time floor, direction, in-car instruction, landing call, landing call waiting time, etc.

[0033] Step B2, the edge side uses edge technology to complete preliminary data cleaning and processing, and uses Internet of Things technology to automatically send to the cloud platform;

[0034] Step B3, the cloud platform is based on the group control running data uploaded by the edge side, without screening and distinction, all elevator running data are taken as main parameters, and weather, time, holidays, and building types are taken as secondary parameters to construct a data set;

[0035] Step B4, model training is performed according to a deep learning convolutional neural network method, and a node with the greatest influence factor is found by the neural network through iteration;

[0036] Step B5, the process of model training is repeatedly performed, the types of input data are adjusted and cropped, the effect of the output model on the historical data verification set is output, and the preliminary training of the model is completed until the accuracy and recall rate and other indicators of the model no longer improve.

[0037] Preferably, the method for training the AI model comprises the following steps:

[0038] Step C1, group control running data are collected by the edge side, at least including real-time floors, directions, in-car instructions, landing calls, and landing call waiting times;

[0039] Step C2, data preliminary cleaning and processing are completed by the edge side using edge technology, and the data are automatically sent to the cloud platform using Internet of Things technology;

[0040] Step C3, the cloud platform is based on the group control running data uploaded by the edge side, and weather, time, holidays, and building types are used to assist in constructing a data set;

[0041] Step C4, an edge side computing power module is added to the group control elevator, a distributed training method is adopted, the training task of the AI model is decomposed into subtasks, distributed computing technology is used to complete the training of the model on the edge side, or a personalized cloud-edge collaborative AI model training system is built by combining cloud computing power and edge computing power, in which cloud computing power is used for coarse adjustment of parameters, and edge computing power is used for fine adjustment of parameters;

[0042] Step C5, the process of model training is repeatedly performed, the effect of the output model on the historical data verification set is output, and the preliminary training of the model is completed until the accuracy and recall rate and other indicators of the model no longer improve.

[0043] Preferably, the intelligent elevator group control system is also used to realize adaptive deployment of elevator group control software functions based on actual collected data combined with the AI model, and the specific steps are as follows:

[0044] Step T1, group control running data are collected by the edge side, including real-time floors, directions, in-car instructions, landing calls, and landing call waiting times;

[0045] Step T2, data preliminary cleaning and processing are completed by the edge side using edge computing technology, and the data are automatically sent to the cloud platform using Internet of Things technology;

[0046] Step T3, the cloud platform automatically executes the efficiency evaluation module to evaluate the efficiency indicators of the current group control system and output the health degree of the project;

[0047] Step T4, compare the health degree with the historical health degree of the current group control system longitudinally to evaluate and identify the degradation trend;

[0048] Step T5, compare the health degree of the current elevator group control system with the health degrees of the same type of projects in a transverse manner;

[0049] Step T6, compare the health degrees of the remaining elevator group control systems in the same building in a cross-space manner to identify the health degree distribution difference and mark the same-space degradation degree of the group control system;

[0050] Step T7, according to the results of steps T4, T5 and T6, determine whether the elevator group control system has the demand for efficiency optimization; if there is the optimization demand, the actual collected state data of the elevator group control system can be input to automatically adjust different group control system function configuration parameters, repeatedly simulate automatically multiple times, and finally obtain a set of optimal solution function configuration parameter list in combination with the algorithm simulation module on the cloud platform;

[0051] Step T8, the optimal function configuration parameter list is sent to the edge side to modify the elevator group control system function configuration, which takes effect immediately and completes the efficiency optimization task.

[0052] Compared with the prior art, in one aspect, the present application creates a system structure design that integrates traditional elevator group control field knowledge with cloud computing, edge computing, artificial intelligence and Internet of Things technologies, and breaks through the technical path of cloud-edge collaborative deployment of elevator group control system business, thereby improving the intelligentization, automation and self-adaptation capability of traditional isolated elevator group control system.

[0053] On the other hand, the present application also applies advanced technologies in the field of artificial intelligence such as machine learning, deep learning, multi-model fusion and heuristic search in traditional elevator group control deployment, which can improve the accuracy and generalization of single model or traditional algorithm.

[0054] On the other hand, the present application also continuously collects data through Internet of Things technology and builds a group control efficiency evaluation system using data analysis technology, which solves the problems that the previous elevator group control system algorithm change is difficult to quantitatively evaluate the pros and cons, especially difficult to practically test and observe the effect except for theoretical simulation means, and difficult to tailor the efficiency evaluation for the passenger flow of a specific building, and finally creates a direct efficiency evaluation system that can be quantified, compared, ranked and traced across space and time on the cloud.

[0055] Finally, the present application also realizes the automatic backflow of abnormal data and the dynamic automatic optimization iteration of AI model through the integration of cloud computing technology and Internet of Things technology, which endows the AI model with the ability of continuous self-evolution and dynamic improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0057] Figure 1 This is a schematic diagram of the cloud-edge collaborative intelligent elevator group control system architecture of Example 1;

[0058] Figure 2 This is a schematic diagram of the specific steps for cloud-edge collaborative intelligent elevator group control deployment in Example 1. DETAILED DESCRIPTION

[0059] The following describes the implementation manner of the present invention through specific specific embodiments. Those skilled in the art can fully understand other advantages and technical effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through different specific implementation manners, and the various details in this specification can also be applied based on different viewpoints, and various modifications or changes can be made without deviating from the overall design concept of the invention. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. The following exemplary embodiments of the present invention can be implemented in a variety of different forms and should not be interpreted as being limited to the specific embodiments described herein. It should be understood that these embodiments are provided to make the disclosure of the present invention thorough and complete, and to fully convey the technical solutions of these exemplary embodiments to those skilled in the art.

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment provides a cloud-edge collaborative intelligent elevator group control system, including a cloud platform and an edge side; the cloud platform includes the AI ​​model, cloud computing module, data flow module, decision module, scheduling command module, abnormal reflux module, and efficiency evaluation module, and the edge side includes a data acquisition module, a data cleaning and processing module, a scheduling execution module, and a human-computer interaction module.

[0062] The AI ​​model includes but is not limited to traditional machine learning frameworks such as decision trees or deep learning frameworks such as CNN neural networks. The cloud computing module schedules the AI ​​model based on input information to perform advanced tasks such as prediction, classification, and evaluation of elevator group control.

[0063] The data flow conversion module converts the results calculated by the cloud computing module to subsequent modules according to the purpose. The decision module is responsible for making the final decision command in part of the task according to the multi-model fusion, expert knowledge heuristic search and other ways. The scheduling command module is responsible for scheduling the relevant execution module to execute the specified task according to the decision command. The abnormal backflow module judges whether there is a need to backflow to remind the expert analysis according to the data after executing the task. The efficiency evaluation module automatically calculates the group control deployment efficiency according to the collected group control state data. The intelligent elevator group control system of cloud edge collaboration uses Internet of Things, cloud computing and artificial intelligence technology to build a digital twin system on the cloud for the elevator group control system.

[0064] The AI model can be a personalized model trained only by the data of the current group control elevator, or a general model trained by the data of many similar or even different types of group control elevators. The AI model also has a dynamic optimization iteration function, which can automatically optimize and iterate the model according to the difficult examples of abnormal backflow.

[0065] The historical passenger flow data, holiday information and other key data of the input group control system can be used to predict the future passenger flow of the elevator group control system through the model. At the same time, the peak historical data of the group control system marked by experts and the passenger flow prediction results of the group control system are input, and combined with expert knowledge rules, various different passenger flow peaks or non-peak are automatically classified, realizing the personalized elevator group control system and more accurately identifying the passenger flow peak of the group control system to make deployment preparation in advance and improve the deployment efficiency of the group control system.

[0066] Embodiment 2

[0067] As shown in Figure 2 The steps of the cloud edge collaborative intelligent elevator group control system of embodiment 1 during deployment are as follows:

[0068] Step S1, edge side collects group control operation data, including but not limited to real-time floor, direction, in-car instruction, landing call, landing call waiting time and other key data;

[0069] Step S2, the edge side uses edge technology to complete data preliminary cleaning and processing, and uses Internet of Things technology to automatically send to the cloud platform;

[0070] Step S3, the cloud platform selects features based on the group control operation data uploaded by the edge side (traditional machine learning method) or directly constructs a data set (deep learning method);

[0071] Step S4, the cloud platform starts training the AI model based on the constructed data set or feature engineering and continuously iterates the training until the AI model with satisfactory precision and recall rate is obtained;

[0072] Step S5, the cloud platform schedules the computing power resources on the cloud, deploys the AI model, and processes the real-time collected group control state data into the input data required by the AI model;

[0073] Step S6, the AI model automatically performs real-time inference according to the input data to obtain the prediction of future passenger flow;

[0074] Step S7, the AI model classifies the prediction results of the future passenger flow and the multi-model fusion technology to obtain the prediction of whether there will be a passenger flow peak in a certain period of time and what kind of passenger flow peak;

[0075] Step S8, the cloud platform decision command module deeply processes the peak classification results according to the statistical law and expert rules, and finally obtains the deployment decision command of the control group control system, such as entering or exiting the peak / what kind of peak command;

[0076] Step S9, the cloud platform scheduling command module calls the edge side scheduling execution module to control the elevator group control system to enter or exit the peak / what kind of peak task according to the command of the decision command module;

[0077] Step S10, the efficiency evaluation module of the cloud platform automatically calculates the efficiency index according to the continuously collected group control state data, and quantitatively evaluates the group control operation efficiency in multiple dimensions;

[0078] Step S11, after the cloud platform schedules the edge side, the quantified index calculated by the efficiency evaluation module is compared with the actual group control system performance, and the rationality of this scheduling is judged according to the expert system. If it is not reasonable, the abnormal backflow module is used to return this scheduling as a difficult example for optimizing and iterating the AI model;

[0079] The present application creates a system structure design that integrates traditional elevator group control field knowledge with cloud computing, edge computing, artificial intelligence, Internet of Things and other technologies, and breaks through the technical path of cloud-edge collaborative deployment of elevator group control system business.

[0080] The traditional elevator group control system on the edge side can improve the processing capacity for complex scenes and the rationality, predictability and accuracy of processing in traditional ordinary scenes under the empowerment of sufficient computing power and rich AI models on the cloud. At the same time, it is easy to update the algorithm on the cloud platform, greatly reducing the optimization and update cycle of traditional algorithms. At the same time, the empowerment of the efficiency evaluation system based on actual collected data also makes the efficiency index of the elevator group control system quantifiable and traceable, greatly improving the reliability and authenticity of the optimization evaluation of traditional algorithms.

[0081] The present application only uses elevator group control system native data for data analysis, processing and deployment, which is different from the past need to additionally install cameras or microphones, reducing the hardware cost and complexity of the entire system and expanding the usable range of the system.

[0082] Embodiment 3

[0083] This embodiment is a refinement of the AI model training scene in the cloud-edge collaborative intelligent elevator group control system. Traditional machine learning methods in the field of artificial intelligence are used, and the specific steps are as follows:

[0084] Step A1, collect group control operation data on the edge side, including but not limited to real-time floor, direction, in-car instructions, landing call, landing call waiting time, and other key data;

[0085] Step A2, use edge technology to complete preliminary data cleaning and processing on the edge side, and use Internet of Things technology to automatically send to the cloud platform;

[0086] Step A3, based on the group control operation data uploaded by the edge side, perform feature engineering selection, and input weather, holidays, building type, and time information as parameter variables to assist model training;

[0087] Step A4, use the collected group control operation data as the main parameter variable and weather assistance data as the secondary parameter variable to train the model according to the traditional machine learning decision tree method;

[0088] Step A5, continuously repeat the model training process and automatically adjust the weights of each variable, output the model's effect on the historical data validation set, and complete the preliminary training of the model until the accuracy and recall rate of the model no longer improve.

[0089] The advantage of traditional machine learning methods is that the computing power requirement is low, and model training on CPU architecture can be achieved at low cost, greatly reducing the overall cost. After reasonable feature engineering selection, the performance on some time series data processing tasks such as prediction and classification is usually not worse than or slightly worse than deep learning methods, but does not affect engineering application, and is suitable for rapid construction or verification.

[0090] Embodiment 4

[0091] This embodiment is a refinement of the AI model training scene in the cloud-edge collaborative intelligent elevator group control system. Traditional machine learning methods in the field of artificial intelligence are used, and the specific steps are as follows:

[0092] Step B1, collect group control operation data on the edge side, including but not limited to real-time floor, direction, in-car instructions, landing call, landing call waiting time, and other key data;

[0093] Step B2, use edge technology to complete preliminary data cleaning and processing on the edge side, and use Internet of Things technology to automatically send to the cloud platform;

[0094] Step B3, the cloud platform is based on the group control running data uploaded by the edge side, without screening and distinction, all elevator running data as the main parameter variable, weather, time, holiday, building type as the secondary parameter variable, to build a data set;

[0095] Step B4, model training according to the deep learning convolutional neural network method, and the neural network iteratively finds out the node with the greatest influence factor;

[0096] Step B5, continuously repeat the model training process, and adjust the types of input data, output the effect of the model on the historical data validation set, until the accuracy and recall rate of the model no longer improve, and the preliminary training of the model is completed.

[0097] The model training effect of deep learning is usually closely related to the number of parameters and the number of neural network layers, and generally requires high computing power and cannot be completed on CPU, requiring professional GPU or NPU, but in specific businesses, it has the characteristics of simple and convenient training and real-time inference better than traditional machine learning methods, which can be used as formal use or large-scale use, or combined with traditional machine learning models for multi-model fusion algorithms to balance the input-output ratio of resources.

[0098] Example 5

[0099] This embodiment is a refinement of the AI model training scene of the cloud-edge collaborative intelligent elevator group control system, focusing on the use of computing power, which can be centralized cloud computing power or distributed edge-side cluster computing power, and the specific steps are as follows:

[0100] Step C1, the edge side collects group control running data, including but not limited to real-time floor, direction, in-car instructions, landing call, landing call waiting time, and other key data;

[0101] Step C2, the edge side uses edge technology to complete preliminary cleaning and processing of data, and uses Internet of Things technology to automatically send to the cloud platform;

[0102] Step C3, the cloud platform is based on the group control running data uploaded by the edge side, and weather, time, holiday, building type assistance data to build a data set;

[0103] Step C4, install an edge-side computing power module on the group control elevator, adopt a distributed training method, decompose the AI model training task into sub-tasks, and use distributed computing technology to complete model training on the edge side; or combine cloud computing power and edge-side computing power to build a personalized cloud-edge collaborative AI model training system with cloud computing power parameter coarse tuning and edge-side computing power parameter fine tuning;

[0104] Step C5, constantly repeat the process of model training, output the effect of the model on the historical data validation set, until the accuracy and recall rate of the model no longer improve, complete the preliminary training of the model.

[0105] The computing power on the cloud generally adopts a centralized and subscription payment mode. Pure use of the computing power on the cloud for large-scale promotion of a project will bring a high cost. The distributed computing mode using the existing computing power on the edge side in the embodiment can save a large amount of demand for newly added resources, and can solve the problem of high cost of traditional cloud computing.

[0106] Embodiment 6

[0107] The embodiment is a refinement of the specific application scene of the intelligent elevator group control system in the cloud-edge cooperation. The system can also realize adaptive adjustment of the elevator group control software function based on the actual collected data combined with the AI model. The specific steps are as follows:

[0108] Step T1, the edge side collects group control operation data, including but not limited to real-time floor, direction, in-car instruction, landing call, landing call waiting time, and other key data;

[0109] Step T2, the edge side uses edge computing technology to complete preliminary cleaning and processing of data, and automatically sends to the cloud platform using Internet of Things technology;

[0110] Step T3, the cloud platform automatically executes the efficiency evaluation module to evaluate the efficiency indicators of the current group control system and output the health degree of the project;

[0111] Step T4, compare the health degree with the historical health degree of the current group control system to evaluate and identify the degradation trend;

[0112] Step T5, compare the health degree of the current elevator group control system with the health degrees of similar projects, such as hotel type group control, office building type group control, and other scenes, to obtain the ranking in similar projects;

[0113] Step T6, compare the health degree of the current elevator group control system with the health degrees of other elevator group control systems (if any) in the same building to identify the difference in health degree distribution and mark the degradation degree of the same space of the group control system;

[0114] Step T7, according to the results of steps T4, T5, and T6, determine whether there is a need for efficiency optimization of the elevator group control system. If there is an optimization requirement, the actual collected state data of the elevator group control system can be input into the algorithm simulation module on the cloud platform to automatically adjust different group control system function configuration parameters, repeatedly simulate automatically multiple times, and finally obtain a list of optimal solution function configuration parameters;

[0115] Step T8, the optimal function configuration parameter list is issued to the edge side, the function configuration of the elevator group control system is modified, takes effect immediately, and the efficiency optimization task is completed.

[0116] The advantages and novelty of the embodiment are that the automatic optimization demand identification function across space-time on the cloud solves the problems of uncertainty and high labor cost of traditional artificial analysis optimization demand, the real automatic simulation function based on actual collected data solves the problem that the traditional group control elevator efficiency optimization problem solution cannot predict the effect according to the fact through continuous trial and error, and the elevator group control function parameter automatic adjustment function combined with the cloud-edge collaborative system realizes one-key implementation modification, reduces the time and labor cost of traditional group control elevator efficiency optimization problem requiring manual modification configuration.

[0117] The application is described in detail through the specific embodiments and examples, but these do not constitute a limitation on the application. Those skilled in the art can also make many modifications and improvements without departing from the principles of the application, and these should also be considered as the protection scope of the application.

Claims

1. A cloud-edge collaborative intelligent elevator group control system, characterized by: It includes a cloud platform and an edge side. The cloud platform includes an AI model, a cloud computing module, a data flow module, a decision-making module, a scheduling command module, an abnormal reflux module and an efficiency evaluation module; the edge side includes a data acquisition module, a data cleaning and processing module, a scheduling execution module and a human-computer interaction module. The data acquisition module uses the Internet of Things technology to collect and transmit all the operating data of the elevator group control system to the cloud computing module; The cloud computing module dispatches the AI ​​model to perform prediction, classification, and evaluation tasks for elevator group control based on input information; The data transfer module transfers the results calculated by the cloud computing module to other subsequent modules according to their purpose; The decision module makes decision commands in some tasks according to multi-model fusion or expert knowledge heuristic search methods; The scheduling command module schedules the relevant execution modules to perform the specified tasks according to the decision command; The abnormal reflux module is responsible for judging whether there is any abnormal reflux that requires expert analysis based on the data after the task is executed; The efficiency evaluation module automatically calculates the group control deployment efficiency based on the collected group control status data; The cloud-edge collaborative intelligent elevator group control system uses the Internet of Things, cloud computing and artificial intelligence technologies to build a digital twin system for the elevator group control system on the cloud.

2. The cloud-edge collaborative intelligent elevator group control system according to claim 1 is characterized in that: The AI ​​model is trained using data collected by a single elevator group control system, or is trained using a combination of data collected by multiple similar or different types of elevator group control systems; By inputting key data such as the group control system's historical passenger flow data and holiday information, the model can predict future passenger flow for the elevator group control system. Furthermore, by inputting the group control system's own expert-labeled historical peak flow data and its passenger flow forecast results, combined with expert knowledge rules, the model automatically categorizes various peak and off-peak passenger flow periods, creating a personalized elevator group control system. This allows for more accurate identification of peak passenger flow periods, enabling early deployment preparation and improving the system's deployment efficiency.

3. The cloud-edge collaborative intelligent elevator group control system according to claim 1 is characterized in that: The group control deployment efficiency at least includes the average waiting time and long waiting rate of a single group control system itself; The efficiency evaluation module calculates the median ranking of the elevator group control system among a large number of similar buildings based on a comparison of similar buildings connected to the cloud platform; The efficiency evaluation module is also used to compare the efficiency indicators of different elevator group control systems in the same building that have been connected to the cloud platform.

4. The cloud-edge collaborative intelligent elevator group control system according to claim 1 is characterized in that: The data acquisition module calculates passenger flow information based only on the original operation data of the elevator group control system, and the passenger flow information at least includes the number of people entering and exiting the elevator each time it stops up or down; After the passenger flow information is processed by the data cleaning processing module, the complete traffic flow data is restored.

5. The cloud-edge collaborative intelligent elevator group control system according to claim 2 is characterized in that: The electric AI model is trained using the convolutional neural network deep learning method and the decision tree traditional machine learning method.

6. The cloud-edge collaborative intelligent elevator group control system according to claim 4 is characterized in that: The data acquisition module automatically performs 24 / 7 data acquisition.

7. The cloud-edge collaborative intelligent elevator group control system according to claim 2 is characterized in that: The AI ​​model has a dynamic optimization and iteration function, which automatically optimizes and iterates the AI ​​model using difficult examples; the difficult examples are identified by the abnormal reflow module and are reflowed when the model predicts that the classification is inaccurate.

8. The cloud-edge collaborative intelligent elevator group control system according to claim 1 is characterized in that: The scheduling execution module generates a scheduling command based on the results analyzed by the AI ​​model and sends it to the elevator group control system for execution.

9. The cloud-edge collaborative intelligent elevator group control system according to claim 8 is characterized in that: The steps for deploying the cloud-edge collaborative intelligent elevator group control system are as follows: Step S1: The edge side collects group control operation data, which includes real-time floor, direction, in-car instructions, landing call, and landing call waiting time; Step S2: The edge side uses edge technology to complete preliminary data cleaning and processing, and uses IoT technology to transmit data to the cloud platform; Step S3: The cloud platform performs feature engineering selection or directly constructs a data set based on the group control operation data uploaded by the edge side; In step S4, the cloud platform starts training the AI ​​model based on the constructed data set or feature engineering and continuously iterates the training until an AI model with satisfactory precision and recall is obtained; Step S5: The cloud platform dispatches computing resources on the cloud, deploys the AI ​​model, and processes the real-time collected group control status data into the input data required by the AI ​​model; Step S6: The AI ​​model automatically performs real-time reasoning based on the input data to obtain a prediction of future passenger flow; In step S7, the AI ​​model classifies and predicts whether a passenger flow peak will occur within a certain period of time in the future and what type of passenger flow peak it will be based on the passenger flow prediction results and multi-model fusion technology; Step S8: The cloud platform decision module processes the peak classification results according to statistical laws and expert rules, and finally obtains a deployment decision command for the control group control system to immediately enter or exit a peak or which peak; Step S9: The cloud platform dispatching command module calls the edge-side dispatching execution module according to the command of the decision module to control the elevator group control system to enter or exit the peak or peak task; Step S10: The efficiency evaluation module of the cloud platform automatically calculates the efficiency index based on the continuously collected group control status data, and performs a multi-dimensional quantitative evaluation of the group control operation efficiency; In step S11, after scheduling the edge side, the cloud platform compares the quantitative indicators calculated by the efficiency evaluation module with the performance of the actual group control system, and uses the expert system to judge the rationality of this scheduling. If it is unreasonable, the abnormal return flow module will use this scheduling as a difficult case return to optimize the iterative AI model.

10. The cloud-edge collaborative intelligent elevator group control system according to claim 1 is characterized in that: The method for training the AI ​​model includes the following steps: Step A1: The edge side collects group control operation data, including at least real-time floor, direction, in-car instructions, landing call, and landing call waiting time; Step A2: The edge side uses edge technology to complete preliminary data cleaning and processing, and automatically sends it to the cloud platform using IoT technology; Step A3: The cloud platform performs feature engineering based on the group control operation data uploaded by the edge side, and uses weather, holidays, building type, and time information as parameters to assist in the input of model training. Step A4: Using the collected group control operation data as the main parameter variable and the weather assistance data as the secondary parameter variable, model training is performed according to the traditional machine learning decision tree method; Step A5: Repeat the model training process continuously, automatically adjust the weights of each variable, and output the model's effect on the historical data validation set until the model's accuracy and recall rate no longer improve, completing the initial model training.

11. The cloud-edge collaborative intelligent elevator group control system according to claim 1 is characterized in that: The method for training the AI ​​model includes the following steps: Step B1: The edge side collects group control operation data, including at least real-time floor, direction, in-car instructions, landing call, and landing call waiting time; Step B2: The edge side uses edge technology to complete preliminary data cleaning and processing, and automatically sends it to the cloud platform using IoT technology; Step B3: The cloud platform constructs a dataset based on the group control operation data uploaded by the edge side without filtering or distinguishing, using all elevator operation data as the main parameter variable and weather, time, holidays, and building type as secondary parameters; Step B4: Perform model training using a deep learning convolutional neural network method, and let the neural network iterate to find the node with the largest impact factor; Step B5: Repeat the model training process, adjust and tailor the type of input data, and output the model's effect on the historical data validation set until the model's accuracy and recall rate no longer improve, completing the initial model training.

12. The cloud-edge collaborative intelligent elevator group control system according to claim 1, characterized in that: The method for training the AI ​​model includes the following steps: Step C1: The edge side collects group control operation data, including at least real-time floor, direction, in-car instructions, landing call, and landing call waiting time; Step C2: The edge side uses edge technology to complete preliminary data cleaning and processing, and automatically sends it to the cloud platform using IoT technology; Step C3: The cloud platform constructs a data set based on the group control operation data uploaded by the edge side, as well as weather, time, holidays, and building type assistance data; Step C4: Install an edge computing module in the group-controlled elevators and adopt a distributed training approach to break down the AI ​​model training task into subtasks. Use distributed computing technology to complete model training on the edge. Alternatively, combine cloud computing power and edge computing power to build a personalized cloud-edge collaborative AI model training system that uses cloud computing power for coarse parameter adjustment and edge computing power for tailored parameter adjustment. Step C5: Repeat the model training process and output the model's effect on the historical data validation set until the model's accuracy and recall rate no longer improve, completing the initial model training.

13. The cloud-edge collaborative intelligent elevator group control system according to claim 1, characterized in that: The intelligent elevator group control system is also used to implement adaptive deployment of elevator group control software functions based on actual collected data combined with AI models. The specific steps are as follows: Step T1: The edge side collects group control operation data, including real-time floor, direction, in-car instructions, landing call, and landing call waiting time; Step T2: The edge side uses edge computing technology to complete preliminary data cleaning and processing, and automatically sends it to the cloud platform using IoT technology; Step T3: The cloud platform automatically executes the efficiency evaluation module to evaluate the efficiency indicators of the current group control system and output the health of the project; Step T4: Compare the health status with the historical health status of the current group control system to evaluate and identify degradation trends; Step T5: horizontally compare the health of the current elevator group control system with projects of the same category; Step T6: Compare the health of other elevator group control systems in the same building across different spaces, identify health distribution differences, and mark the same-space degradation degree of the group control system; Step T7: Based on the results of steps T4, T5, and T6, determine whether there is a need for efficiency optimization of the elevator group control system; if there is an optimization need, the algorithm simulation module on the cloud platform can be combined with the actual collected state data of the elevator group control system to automatically adjust different group control system function configuration parameters, and the automatic simulation can be repeated multiple times to finally obtain a set of optimal solution function configuration parameter lists; Step T8: Send the optimal function configuration parameter list to the edge side, modify the elevator group control system function configuration, and take effect immediately, completing the efficiency optimization task.

Images