A building elevator operation optimization method based on a dynamic allocation algorithm
By optimizing the elevator system through intelligent recognition and dynamic allocation algorithms, the problems of improper elevator allocation during peak hours and energy consumption during low-traffic periods have been solved, achieving a dual improvement in elevator operating efficiency and energy saving.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU TIWEISHI NETWORK TECH CO LTD
- Filing Date
- 2025-10-14
- Publication Date
- 2026-06-30
AI Technical Summary
Existing elevator systems cannot intelligently allocate passengers based on actual passenger distribution during peak hours, resulting in frequent stops at each floor, which affects operational efficiency and increases energy consumption. Furthermore, it is difficult to dynamically adjust the number of elevators to reduce energy consumption during low-traffic periods.
By acquiring passenger behavior data through monitoring equipment and sensors, analyzing destination floor information using intelligent recognition algorithms, optimizing elevator allocation and stopping order by combining K-means clustering and genetic optimization algorithms, and predicting traffic changes by combining real-time monitoring data, the number of elevators and resource allocation are dynamically adjusted to form an optimization closed loop.
It improves elevator operating efficiency, reduces passenger waiting time, and achieves energy conservation and consumption reduction, providing an innovative solution for intelligent building management.
Smart Images
Figure CN121516671B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of elevator operation optimization technology, and in particular to a building elevator operation optimization method based on a dynamic allocation algorithm. Background Technology
[0002] Elevator operation optimization, as a crucial area of intelligent building management, directly impacts people's daily travel efficiency and energy use efficiency, especially in office buildings during peak hours. Its importance is self-evident. Rational allocation of elevator resources not only improves user experience but also effectively reduces operating costs, becoming a key aspect of modern urbanization. However, current elevator operation solutions generally have limitations, mostly employing fixed scheduling modes or simple first-come, first-served principles, which are insufficient to handle the complex demands of peak hours. During rush hour, multiple elevators often fail to intelligently allocate resources based on actual passenger distribution, leading to frequent stops at each floor, severely impacting operational efficiency and increasing unnecessary energy consumption.
[0003] The primary challenge lies in how to acquire and predict passengers' destination floors in real time. Due to a lack of accurate identification and prediction methods, elevators cannot plan optimal stopping routes in advance, leading to inefficient resource allocation. This problem further gives rise to another core challenge: how to intelligently group passengers based on their destination floors and assign them to different elevators during dynamically changing peak traffic periods, ensuring that some elevators can reach destinations directly or with fewer stops, thereby improving overall operational efficiency. Furthermore, how to dynamically adjust the number of operating elevators to reduce energy consumption when passenger traffic is low is also a related issue that urgently needs to be addressed. These challenges are progressive and collectively constitute the technological barriers in the field of elevator optimization. Therefore, how to achieve efficient elevator grouping and optimized resource allocation through intelligent identification of passengers' destination floors combined with dynamic allocation algorithms is a key problem that this research urgently needs to solve. Summary of the Invention
[0004] The purpose of this invention is to provide a building elevator operation optimization method based on a dynamic allocation algorithm. This method can apply an energy consumption optimization control algorithm to reallocate power resources and continuously update passenger behavior patterns to form an optimization closed loop. This effectively improves elevator operating efficiency, reduces passenger waiting and riding time, and simultaneously achieves energy conservation and consumption reduction. The invention provides the following technical solution:
[0005] In a first aspect, the present invention provides a building elevator operation optimization method based on a dynamic allocation algorithm, which includes acquiring passenger behavior-related data from monitoring equipment and sensor equipment, analyzing passenger destination floor information, and generating an initial destination floor distribution dataset;
[0006] Based on the initial destination floor distribution dataset, passengers are classified to determine the concentrated destination floor range and quantity statistics for each type of passenger.
[0007] Based on the statistical results of the range and number of floors of the central destination, passengers are assigned to different elevator units to generate a preliminary elevator allocation plan;
[0008] The order of elevator stops at the floors in the preliminary elevator allocation scheme is adjusted to generate an optimized elevator stop sequence scheme.
[0009] Based on the optimized elevator stopping sequence scheme and real-time operating status data, analyze the passenger flow trend and generate a traffic change prediction dataset.
[0010] Adjust the elevator operation mode based on the traffic flow change prediction dataset, and generate the adjusted elevator operation mode;
[0011] Based on the adjusted elevator operation mode, resources are allocated to generate an energy-saving and optimized elevator operation configuration scheme.
[0012] Based on the energy-saving and optimized elevator operation configuration scheme and real-time passenger behavior data, the passenger behavior pattern is updated, an updated target floor distribution dataset is generated, and the data is fed back to the initial data processing stage to form a continuous optimization closed loop.
[0013] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the generation of the initial target floor distribution dataset includes:
[0014] Passenger image and behavioral data are collected through video surveillance equipment and sensor devices;
[0015] Intelligent recognition algorithms are applied to the image data and behavioral data to extract the passenger's movement trajectory and behavioral characteristics;
[0016] Based on movement trajectory and behavioral characteristics, infer the passenger's destination floor information;
[0017] The destination floor information of the passengers is summarized and organized to generate an initial destination floor distribution dataset.
[0018] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the passenger classification includes:
[0019] Obtain the initial target floor distribution dataset as input data;
[0020] Clustering algorithms are used to group passenger destination floor information in the initial destination floor distribution dataset to determine the concentrated destination floor range for each group of passengers.
[0021] For each group of passengers, the number of passengers within each destination floor range is counted, the number of passengers is generated, the number of passengers is sorted, the main destination floor range is identified, and a classification result report is generated.
[0022] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the generation of the preliminary elevator allocation scheme includes:
[0023] Obtain the statistical results of the range and number of floors for the central destination, and use a dynamic allocation algorithm to allocate passengers to the corresponding elevator units according to the range of floors for the central destination;
[0024] The number of passengers allocated to each elevator unit is calculated and compared with the maximum load threshold. If overloading occurs, a secondary adjustment is automatically triggered through a feedback mechanism to rebalance the allocation results.
[0025] The adjusted allocation results will form a preliminary elevator allocation plan, and a simulation test will be conducted.
[0026] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the method for generating the optimized elevator stopping sequence includes:
[0027] For each elevator unit, extract the floor information that needs to be stopped in the preliminary elevator allocation plan;
[0028] A genetic optimization algorithm is applied to adjust the order of the floor information to minimize passenger waiting time and elevator travel distance as the objective function.
[0029] Determine the optimal stopping sequence, integrate the optimal stopping sequence into the preliminary elevator allocation plan, and generate an optimized elevator stopping sequence plan.
[0030] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the generated traffic flow change prediction dataset includes:
[0031] Elevator operation status data and floor call signal data are collected from the real-time monitoring system, and the optimized stopping sequence scheme is correlated and analyzed with the real-time elevator operation status data and floor call signal data.
[0032] The correlation analysis results are processed using a traffic prediction model to predict the trend of passenger traffic changes throughout the entire time period. The formula for calculating the predicted passenger traffic value is as follows:
[0033]
[0034] Where: P t P represents the predicted passenger flow value at time t. t-1The value represents the passenger flow at the previous moment, γ represents the attenuation coefficient of historical data, H represents the total number of floors, and R represents the passenger flow at the previous moment. h F represents the call signal frequency of the h-th layer. h This represents the actual traffic data for layer h;
[0035] Generate a traffic change prediction dataset based on passenger flow trends.
[0036] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the generation of the adjusted elevator operation mode includes:
[0037] Analyze the traffic change trend in the traffic change prediction dataset. If it is determined that the current period is a low traffic period, the number of elevator units in operation will be reduced through a dynamic adjustment mechanism.
[0038] Based on the reduced number of elevator units, operational tasks will be reallocated in accordance with minimum service guarantee requirements;
[0039] Based on the reassignment results after the task allocation, the adjusted elevator operation mode is generated.
[0040] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the generation of the energy-saving optimized elevator operation configuration scheme includes:
[0041] Analyze the operating status and load requirements of each elevator unit under the adjusted elevator operation mode;
[0042] An energy consumption optimization control algorithm is applied to redistribute power resources and calculate the minimum energy consumption requirement for each elevator.
[0043] Based on the minimum energy consumption requirements and peak load limits, the resource allocation strategy is adjusted, and an energy-saving optimized elevator operation configuration scheme is generated based on the adjusted resource allocation strategy.
[0044] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the generation and updating of the target floor distribution dataset includes:
[0045] Acquire energy-saving and optimized elevator operation configuration schemes, elevator operation record data and energy consumption data from the system log database, and passenger behavior data obtained through real-time monitoring;
[0046] Integrate and analyze elevator operation records, energy consumption data, and passenger behavior data;
[0047] The passenger distribution update model was used to process the integrated analysis results and reanalyze passenger behavior patterns.
[0048] Based on the analysis of passenger behavior patterns, an updated target floor distribution dataset is generated;
[0049] The updated data is fed back to the initial data processing stage.
[0050] As a preferred embodiment of the building elevator operation optimization method based on dynamic allocation algorithm described in this invention, the continuous optimization closed loop includes:
[0051] The updated target floor distribution dataset is used as the input for the next round;
[0052] The updated target floor distribution dataset is linked to the initial data processing stage to update the basic data in the system database;
[0053] Based on the updated baseline data, the passenger classification, allocation, and stop order optimization processes were re-executed.
[0054] The traffic forecast and operating mode were updated by adjusting the allocation scheme and docking sequence;
[0055] Adjust resource allocation strategies based on updated traffic forecasts and operating patterns;
[0056] The adjusted resource allocation strategy generates a new operating configuration scheme, which is then applied to the elevator control system.
[0057] The beneficial effects of this invention are as follows: This invention collects passenger behavior data and applies intelligent recognition algorithms to analyze destination floor information. Passengers are classified using K-means clustering, and a dynamic allocation algorithm assigns them to different elevator units. Subsequently, a genetic optimization algorithm is applied to adjust the elevator stopping order, and real-time monitoring data is used to predict passenger flow trends. During low-flow periods, this invention dynamically adjusts the number of operating elevators to save energy. Furthermore, this invention applies an energy consumption optimization control algorithm to reallocate power resources and continuously updates passenger behavior patterns, forming an optimization closed loop. This effectively improves elevator operating efficiency, reduces passenger waiting time, and achieves energy conservation and consumption reduction, providing an innovative solution for intelligent building management. Attached Figure Description
[0058] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0059] Figure 1 This is a flowchart of a building elevator operation optimization method based on a dynamic allocation algorithm.
[0060] Figure 2 A flowchart for generating the initial target floor distribution dataset.
[0061] Figure 3 A flowchart for generating a preliminary elevator allocation plan.
[0062] Figure 4 A flowchart for generating a traffic change prediction dataset.
[0063] Figure 5 To generate a flowchart of the adjusted elevator operation mode.
[0064] Figure 6 A flowchart for generating an energy-efficient and optimized elevator operation configuration scheme. Detailed Implementation
[0065] To make the above-mentioned objects, features, and advantages of the present invention more readily understood, specific embodiments of the present invention will be described in detail 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 them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0066] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0067] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.
[0068] Example 1
[0069] Reference Figure 1 - Figure 6 This is the first embodiment of the present invention, which provides a building elevator operation optimization method based on a dynamic allocation algorithm, including:
[0070] S101: Collect passenger behavior-related data from video surveillance equipment and sensor equipment, analyze passenger destination floor information using intelligent recognition algorithms, and generate an initial destination floor distribution dataset for subsequent passenger classification and allocation processing.
[0071] S102: For the initial destination floor distribution dataset, the K-means clustering algorithm is applied to classify passengers, determine the concentrated destination floor range of each type of passenger and the passenger count statistics of each category, which are used for subsequent elevator unit allocation calculations.
[0072] S103: Based on the range of destination floors and the statistical results of passenger numbers, a dynamic allocation algorithm is used to assign passengers to different elevator units, generating a preliminary elevator allocation plan while adhering to the maximum load limit of the elevators, for subsequent optimization of the stopping order.
[0073] S104: For the initial elevator allocation plan, a genetic optimization algorithm is applied to adjust the floor order of each elevator unit. The goal is to minimize passenger waiting time and elevator travel distance, and generate an optimized elevator stopping sequence plan for subsequent traffic flow analysis.
[0074] S105: Based on the optimized elevator stopping sequence scheme, combined with elevator operation status data and floor call signal data obtained from the real-time monitoring system, the passenger flow change trend throughout the day is analyzed using a traffic prediction model to generate a traffic change prediction dataset for subsequent operation mode adjustments.
[0075] S106: For the traffic flow change prediction dataset, if it is determined that the current period is a low traffic flow period, the number of elevator units in operation will be reduced through a dynamic adjustment mechanism to meet the minimum service guarantee requirements, and an adjusted elevator operation mode will be generated for subsequent energy resource allocation.
[0076] S107: Based on the adjusted elevator operation mode, the power resources are redistributed using an energy consumption optimization control algorithm. Taking into account the minimum energy consumption requirements and peak load limits of elevator operation, an energy-saving optimized elevator operation configuration scheme is generated for subsequent passenger behavior pattern updates.
[0077] S108: For energy-saving and optimized elevator operation configuration schemes, combine elevator operation record data and energy consumption data obtained from the system log database, as well as real-time monitored passenger behavior data, use the passenger distribution update model to re-analyze passenger behavior patterns, generate updated target floor distribution datasets, and feed this dataset back to S101 to form a continuous optimization closed loop.
[0078] The process of generating the initial destination floor distribution dataset includes: collecting passenger behavior data in the elevator lobby area through video surveillance equipment and sensor equipment, and using intelligent recognition algorithms to analyze passenger action characteristics and voice interaction information to obtain preliminary destination floor information.
[0079] Based on the initial destination floor information, combined with historical behavior data records, data cleaning and feature extraction are performed to determine the correlation pattern between passenger behavior and destination floor.
[0080] By standardizing the association patterns, a machine learning model is used to predict the destination floors of passengers, resulting in an initial destination floor distribution dataset.
[0081] The passenger classification process specifically includes: collecting passenger behavior data through video surveillance and sensor devices, analyzing passenger destination floor information using intelligent recognition algorithms, and generating an initial destination floor distribution dataset.
[0082] Based on the initial destination floor distribution dataset, the K-means clustering algorithm is applied to classify passengers, determine the concentrated destination floor range for each category of passengers, and obtain the passenger count statistics for each category.
[0083] Data cleaning techniques were used to preprocess the categorized passenger data, removing outliers and duplicate data to obtain a standardized passenger classification dataset.
[0084] The process of generating a preliminary elevator allocation scheme includes: preprocessing the passenger destination floor distribution dataset, extracting the departure floor and destination floor information for each passenger, and obtaining an initial passenger distribution matrix.
[0085] Based on the initial passenger distribution matrix, the K-means clustering algorithm is applied to classify passengers, determine the concentrated destination floor range of each type of passenger, and obtain the passenger count statistics for each category.
[0086] Based on the range of destination floors and the statistical results of passenger numbers, a dynamic allocation algorithm is used to assign passengers to different elevator units, generating a preliminary elevator allocation plan while adhering to the maximum load limit of the elevators.
[0087] By verifying the data of the initial elevator allocation plan, it is determined whether there is a situation where the load of a certain elevator unit exceeds the limit. If it does, the allocation plan is readjusted to obtain an updated allocation plan that meets the load limit.
[0088] The process of generating an optimized elevator stopping sequence plan includes: based on the preliminary elevator allocation plan, extracting the initial stopping floor list and corresponding passenger distribution data for each elevator unit to obtain the preliminary running path information for each elevator.
[0089] The genetic optimization algorithm iteratively calculates the floor order for each elevator unit, adjusts the floor access order, and determines the optimized combination of stopping paths.
[0090] Based on the optimized stop path combination, calculate the total passenger waiting time when each elevator stops at different floors, and obtain the waiting time evaluation results.
[0091] By simulating and calculating the total distance of each elevator's operating path, and combining the path combination data, a quantitative indicator of the operating distance is determined.
[0092] If neither the waiting time assessment result nor the running distance quantification index reaches the preset threshold, the genetic optimization algorithm will be re-triggered to perform a new round of docking order adjustment, resulting in an updated path combination.
[0093] If the waiting time assessment results and the running distance quantification indicators reach the preset threshold, the current path combination will be saved as an optimized elevator stopping sequence scheme, and the final stopping sequence data will be determined.
[0094] The process of generating a traffic flow change prediction dataset includes: acquiring elevator operation status data and floor call signal data from a real-time monitoring system based on an optimized elevator stopping sequence scheme, organizing them into a structured dataset, and obtaining a preliminary set of operation and call information.
[0095] By cleaning and extracting features from the initial operation and call information sets, key indicators related to passenger flow are screened out, and the set of input parameters for flow prediction is determined.
[0096] A traffic flow prediction model is used to analyze the input parameter set throughout the entire time period. By combining historical traffic flow data and real-time signal changes, the passenger flow trend characteristics of each time period are determined. The formula for calculating the predicted passenger flow value is as follows:
[0097]
[0098] Where: P t P represents the predicted passenger flow value at time t. t-1 The value represents the passenger flow at the previous moment, γ represents the attenuation coefficient of historical data, H represents the total number of floors, and R represents the passenger flow at the previous moment. h F represents the call signal frequency of the h-th layer. h This represents the actual traffic data for layer h;
[0099] If the traffic prediction model outputs trend characteristics indicating that peak-hour calls are concentrated on a specific floor, then the data for that floor is weighted to obtain an adjusted traffic distribution dataset. Based on the adjusted traffic distribution dataset, combined with elevator operating status data, the load situation of each elevator unit in different time periods is analyzed to determine the load balancing reference value.
[0100] By comparing load balancing reference values with real-time monitoring data, the time periods and floor locations where congestion may occur can be identified, thus obtaining congestion risk prediction results.
[0101] If the congestion risk prediction results show that the risk is higher than the threshold for a certain period of time, the dynamic adjustment mechanism is triggered to generate a temporary parking order adjustment plan and determine the optimized operation configuration.
[0102] Based on the optimized operating configuration, the traffic change prediction dataset is updated, and combined with the latest data from the real-time monitoring system, the final prediction result of traffic changes throughout the entire time period is obtained.
[0103] Based on the final prediction results, guidance data for subsequent operation mode adjustments is generated and stored in the system database to determine the basic information for the next cycle of optimization.
[0104] The process of generating the adjusted elevator operation mode includes: collecting elevator operation status data and floor call signal data through a real-time monitoring system, and using a traffic prediction model to analyze the trend of passenger traffic changes throughout the day to obtain a traffic change prediction dataset.
[0105] Based on the traffic flow change prediction dataset, analyze the passenger flow level for the current time period to determine whether it is a low-flow period.
[0106] If it is determined that the current period is a low-traffic period, a dynamic adjustment mechanism is activated to obtain data on the number of elevator units currently in operation and the minimum service guarantee requirements, and to determine the range of elevator units that can be reduced.
[0107] By using a dynamic adjustment mechanism and combining data on minimum service guarantee requirements, the number of elevator units in operation is initially adjusted to obtain a temporarily adjusted elevator operation mode.
[0108] Based on the temporarily adjusted elevator operation mode, the genetic optimization algorithm is invoked to recalculate the stopping order of the remaining operating elevators and determine the optimized stopping order scheme.
[0109] By combining the optimized floor stop sequence scheme with the floor call signal data updated by the real-time monitoring system, we can analyze whether the adjusted elevator operation mode meets the full floor coverage requirement and obtain the coverage verification results.
[0110] If the coverage verification results show that there are uncovered floors, the operation tasks of specific elevator units will be increased through a dynamic adjustment mechanism, the stopping floors will be reallocated, and the corrected elevator operation mode will be determined.
[0111] The process of generating an energy-efficient and optimized elevator operation configuration scheme includes: obtaining power resource allocation data and elevator operation parameters in the system based on the adjusted elevator operation mode, comprehensively analyzing the minimum energy consumption requirements and peak load limits, and determining the initial power allocation constraints.
[0112] By applying the initial power allocation constraints and the energy consumption optimization control algorithm, the power resources of each elevator unit are dynamically calculated to obtain an energy-saving power allocation scheme.
[0113] Based on the energy-saving power distribution scheme and combined with the load distribution data in the elevator operation mode, the energy consumption fluctuations in different time periods are analyzed to determine whether the peak load limit is met.
[0114] If energy consumption fluctuations exceed peak load limits, the power resource allocation is recalculated by adjusting the elevator operating frequency and stopping strategy to obtain a revised power allocation scheme.
[0115] If the energy consumption fluctuation does not exceed the peak load limit, the modified power distribution scheme will be adopted to generate an energy-saving and optimized elevator operation configuration scheme.
[0116] The process of generating the updated target floor distribution dataset includes: optimizing elevator operation configuration schemes through energy conservation, obtaining real-time monitoring data on elevator operation status and energy consumption, and determining the actual operating parameters of the configuration scheme.
[0117] Based on actual operating parameters and elevator operation record data extracted from the system log database, the matching degree between energy consumption and operating efficiency is analyzed to obtain the verification results of the optimized configuration.
[0118] Based on the verification results of the optimized configuration, the passenger distribution update model is applied to reanalyze the passenger behavior data and generate an updated destination floor distribution dataset.
[0119] Based on the updated target floor distribution dataset, feedback is sent to the initial adjustment stage of the elevator operation mode, triggering a new round of energy consumption optimization calculations, forming a continuous closed-loop processing flow.
[0120] The continuous optimization of the closed loop includes: extracting elevator operation record data and energy consumption data from the system log database, and collecting passenger behavior data through the real-time monitoring system to obtain a preliminary operation and behavior dataset.
[0121] Based on the preliminary operational and behavioral datasets, a passenger distribution update model is used to reanalyze passenger behavior patterns and determine the updated passenger behavior characteristic parameters.
[0122] By updating passenger behavior characteristic parameters and combining them with energy-saving and optimized elevator operation configuration schemes, an updated target floor distribution dataset is generated to obtain distribution data for feedback.
[0123] The updated target floor distribution dataset is fed back to the initial configuration module to obtain the comparison results of historical operating data related to S101 and determine the range of data deviation.
[0124] If the data deviation exceeds the preset threshold, the target floor distribution dataset is corrected using a deviation analysis algorithm to obtain the corrected distribution dataset.
[0125] If the data deviation is within the preset threshold, the corrected distribution dataset or the original updated dataset will be used directly to determine the final floor distribution optimization parameters.
[0126] Based on the final floor distribution optimization parameters and combined with the elevator operation status data obtained from the real-time monitoring system, a dynamically adjusted elevator operation mode configuration is generated.
[0127] By dynamically adjusting the elevator operation mode configuration, an energy consumption optimization control algorithm is applied to redistribute power resources, resulting in an energy-saving and optimized resource allocation scheme.
[0128] Based on the energy-saving optimized resource allocation scheme, update the operation records and energy consumption data in the system log database to generate a closed-loop optimized elevator operation dataset.
[0129] In summary, this invention collects passenger behavior data and applies intelligent recognition algorithms to analyze destination floor information. Passengers are then classified using K-means clustering, and a dynamic allocation algorithm assigns them to different elevator units. Subsequently, a genetic optimization algorithm is applied to adjust the elevator stopping order, and real-time monitoring data is used to predict passenger flow trends. During low-flow periods, this invention dynamically adjusts the number of operating elevators to save energy. Furthermore, this invention applies an energy consumption optimization control algorithm to reallocate power resources and continuously updates passenger behavior patterns, forming an optimization closed loop. This method effectively improves elevator operating efficiency, reduces passenger waiting time, and achieves energy conservation and consumption reduction, providing an innovative solution for intelligent building management.
[0130] Example 2
[0131] Reference Figure 1 - Figure 6 This is the second embodiment of the present invention. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.
[0132] Passenger waiting behavior data is collected by 1080P high-definition cameras and infrared sensors deployed in the elevator lobby. The YOLOv5 algorithm is used to detect passengers' button presses and facial orientation in real time. Combined with voice commands collected by the microphone array (such as "go to the 8th floor"), passenger behavior feature vectors are extracted.
[0133] Based on 100,000 passenger behavior records in the historical database, the PCA dimensionality reduction method was used to clean the noisy data, and a correlation matrix between passenger key press habits and destination floors was constructed. The random forest model was used to predict the initial destination floor, and the output dataset included timestamps, floor numbers, and confidence scores (threshold set at 0.85).
[0134] The dataset is stratified into morning peak (7:00-9:00) and midday (11:00-13:00) time periods. Outliers are detected using the 3σ principle. If the data deviates from the mean by more than twice the standard deviation in a certain time period, the outlier records are corrected using the isolated forest algorithm.
[0135] The cleaned data is input into K-means clustering (k=5, Euclidean distance), the floor intervals corresponding to the center points of each cluster are calculated (e.g., cluster 1 is concentrated on floors 5-8), and the proportion of passengers in each category is counted.
[0136] The heatmap visualizes passenger density on each floor. Combined with elevator logs showing operating intervals (average 90 seconds / trip) and energy consumption data (0.5 kWh / trip), a greedy algorithm is used to dynamically allocate elevator stopping priorities. Finally, the NSGA-II multi-objective optimization model is used to adjust operating parameters and output the scheduling scheme with the lowest energy consumption.
[0137] Passenger behavior data is collected through video surveillance equipment and sensor equipment. The YOLOv5 algorithm is used to identify passengers' key presses. Combined with floor sensor data, an initial target floor distribution dataset is generated. For example, if 10 passengers are detected going to different floors such as the 5th, 8th, and 12th floors, the dataset is generated.
[0138] The K-means clustering algorithm was applied to perform cluster analysis on the dataset. K=3 was set, and the Euclidean distance was calculated to divide the passengers into 3 categories. The concentrated destination floor range of each category of passengers was determined, such as floors 5-7, 8-10, and 12-15. The number of passengers in each category was counted as 4, 3, and 3 people, respectively.
[0139] The Pandas library was used to clean the classification data, removing outliers that were outside the floor range (such as -1 floor or 30 floors) to obtain a normalized passenger classification dataset.
[0140] Use Matplotlib to draw box plots to analyze the distribution characteristics of destination floors for each type of passenger, and extract core floor intervals such as floors 5-6, 9-10, and 13-14.
[0141] Priority is calculated using a weighted formula (weight = number of passengers × floor span) based on the core floor range and the number of passengers. For example, the weight of floors 5-6 is 4×2=8, and the weight of floors 9-10 is 3×2=6.
[0142] If the number of passengers in a certain category exceeds the threshold (e.g., ≥5 people in a single category), an independent elevator unit is allocated, generating a preliminary grouping scheme containing 2 elevators.
[0143] Based on the maximum load limit of the elevator (e.g., 13 people / elevator), a greedy algorithm is used to allocate passengers to elevator units according to priority, generating a preliminary allocation scheme, such as elevator A carrying 4 people on floors 5-6 and 3 people on floors 9-10, and elevator B carrying 3 people on floors 13-14.
[0144] By adjusting the distribution ratio through simulated load balancing tests, it was finally determined that elevator A carries 4 people on floors 5-6 and 2 people on floors 13-14, while elevator B carries 3 people on floors 9-10 and 1 person on floors 13-14.
[0145] The final allocation results are written to a MySQL database, recording the elevator number, the number of passengers carried, and the target floor list, thus completing digital management.
[0146] By preprocessing the passenger destination floor distribution dataset, the departure floor and destination floor information of each passenger are extracted. For example, the departure floor (floors 1-10) and destination floor (floors 11-30) are extracted from 1000 passenger records to form an initial passenger distribution matrix.
[0147] Based on the initial passenger distribution matrix, the K-means clustering algorithm was applied to classify passengers. The number of clusters k=3 was set, the Euclidean distance was calculated and the cluster centers were iteratively optimized. Finally, the concentrated destination floor ranges of the three types of passengers were determined to be floors 11-15, 16-20 and 21-25, respectively, and the number of passengers in each type was counted as 350, 400 and 250 people.
[0148] Based on the statistical results of the destination floors and the number of passengers, a dynamic allocation algorithm is used to assign passengers to different elevator units. For example, passengers on floors 11-15 are assigned to elevator A, passengers on floors 16-20 are assigned to elevator B, and passengers on floors 21-25 are assigned to elevator C. At the same time, it is ensured that the load of each elevator does not exceed 1000kg, thus generating a preliminary elevator allocation plan.
[0149] By verifying the data of the initial elevator allocation plan, it was found that the load capacity of elevator B reached 1050kg, which exceeded the limit. Therefore, the allocation plan was readjusted, and 50 passengers were transferred to elevator A, resulting in an updated allocation plan that complies with the load limit.
[0150] According to the updated allocation scheme, obtain the passenger distribution data for each elevator unit. For example, elevator A needs to stop at floors 11, 12, 13, 14, and 15; elevator B needs to stop at floors 16, 17, 18, 19, and 20; and elevator C needs to stop at floors 21, 22, 23, 24, and 25. Determine the floor list for each elevator.
[0151] By initially sorting the list of floors to be stopped by each elevator unit, for example, the initial stopping order of elevator A is floors 11, 15, 12, 14, and 13, a genetic optimization algorithm is used. The population size is set to 50, the crossover probability is 0.8, the mutation probability is 0.1, and after 100 iterations, an optimized elevator stopping order scheme is generated, such as floors 11, 12, 13, 14, and 15.
[0152] Based on the optimized elevator stopping sequence scheme, the total running distance of each elevator and the total waiting time of passengers are calculated. For example, if the running distance of elevator A is 200 meters and the waiting time is 300 seconds, the operating efficiency evaluation index is obtained.
[0153] By analyzing the operational efficiency evaluation indicators, if the travel distance of elevator B exceeds 250 meters or the waiting time exceeds 350 seconds, the genetic optimization algorithm will be triggered again for a second adjustment, and the final stopping order will be determined to be floors 16, 17, 18, 19, and 20.
[0154] Based on the final stopping sequence plan, the operation instruction data for each elevator unit is generated. For example, the instruction for elevator A is "go up to the 11th floor → 12th floor → 13th floor → 14th floor → 15th floor". After determining that the instruction data conforms to the system operation logic, the final operation instruction is output for the elevator control system to execute.
[0155] Extract the initial floor list of each elevator from the preliminary elevator allocation plan. For example, the stopping order of elevator A is 1F→5F→8F, and the passenger distribution data is 1F (3 people), 5F (2 people), and 8F (4 people), forming the preliminary operation path information.
[0156] A genetic optimization algorithm is used to iteratively calculate the stopping order. The population size is set to 50, the crossover probability is 0.8, and the mutation probability is 0.1. After 100 generations of iteration, an optimized path combination is generated, such as 1F→8F→5F.
[0157] Passenger waiting time is calculated based on the optimized route. It is assumed that the waiting time for passengers on 1F is 10 seconds, 8F is 15 seconds, 5F is 12 seconds, and the total is 37 seconds.
[0158] The elevator travel distance was simulated and calculated. If the original total path distance was 120 meters, it was shortened to 95 meters after optimization.
[0159] If the total waiting time exceeds the threshold of 40 seconds or the running distance exceeds 100 meters, the genetic algorithm will be retried to adjust the path.
[0160] If the threshold condition is met, the current path is saved as the final solution. Real-time monitoring data is used to calculate the frequency of elevator stops on each floor, such as 40% stopping on the 8th floor, 30% on the 5th floor, and 30% on the 1st floor.
[0161] By comparing frequency data with historical traffic, a linear regression model is used to predict the traffic trend for the next hour, such as the 8F passenger flow is expected to increase by 20%.
[0162] The model parameters are dynamically adjusted based on the floor call signals, such as setting the response time weight to 0.7, to generate the final prediction result.
[0163] Elevator operation status data (such as current floor, direction of travel, and load sensor values) and floor call signal data (such as up / down requests) are obtained from the real-time monitoring system, stored as a structured dataset using a time-series database, and the sampling interval data of the most recent 5 minutes are extracted.
[0164] The dataset is cleaned to remove outliers (such as invalid data where the load suddenly drops to 0), and principal component analysis (PCA) is used to screen key features, retaining dimensions with a contribution rate of more than 85% (such as peak call frequency and elevator response delay time).
[0165] An LSTM neural network is used to build a traffic prediction model. The input is historical traffic data for 7 days (time granularity of 15 minutes) and real-time signals. The output is the predicted value for the next 30 minutes. A threshold trigger mechanism is set (such as when the call volume of a certain floor exceeds twice the standard deviation of the mean).
[0166] If the forecast shows that the uplink requests from floor 3 account for 60% during the period from 8:30 to 9:00, then apply a weighting factor of 1.5 to the data of that floor.
[0167] Based on the elevator load rate (e.g., elevator No. 1 is currently 70% loaded), calculate the load balance index for each time period (formula: (actual load - average load) / standard deviation). When the index exceeds 1.2, it is marked as a congestion risk.
[0168] After dynamic adjustment is triggered, the genetic algorithm generates a temporary solution (such as skipping low-demand floors) by iterating 50 times with the objective function of minimizing the waiting time.
[0169] When updating the prediction dataset, a sliding window mechanism is introduced (window size 10 minutes, step size 5 minutes), and the final output includes the prediction results containing the confidence interval (e.g., the traffic prediction value from 9:00 to 9:15 ±8%), which is stored in the MongoDB forecast collection.
[0170] The system collects elevator operation status data (including direction of travel, current floor, and load information) and floor call signal data (such as the number of up / down requests per floor within 5 minutes) with a sampling period of 1 second through a real-time monitoring system. The system trains the historical traffic data using an LSTM neural network model to predict the passenger traffic change trend for each time period in the next 30 minutes and outputs a traffic change prediction dataset containing timestamps and predicted traffic values.
[0171] The average flow rate for the current time period (e.g., 14:00-14:30) is calculated based on the dataset. If it is lower than the preset threshold (e.g., 20 people per hour), a dynamic adjustment mechanism is triggered. The minimum service guarantee parameters configured by the system (e.g., at least 2 elevators should be running) are read, and the adjustment range of 1 to 2 elevators that can be shut down is calculated in combination with the 4 elevator units currently in operation.
[0172] A greedy algorithm is used to prioritize stopping elevator units with idle time exceeding 10 minutes, generating a temporary operation mode that retains 3 elevators. An optimization module with a genetic algorithm population size of 50 is then invoked, using the constraint that the average passenger waiting time does not exceed 60 seconds, to iteratively calculate the stopping order of the 3 elevators, outputting a stopping scheme containing [E1:8F→5F→1F, E2:3F→6F, E3:2F→4F→7F].
[0173] After the real-time monitoring system detects the newly added 9th floor call signal, it triggers the coverage verification process. The E1 elevator path is replanned as 8F→5F→9F→1F using the Dijkstra algorithm, the operating mode is corrected and updated to the elevator control system's scheduling queue.
[0174] The energy consumption allocation model is based on the corrected operating status data and uses linear regression to calculate the expected energy consumption of each elevator (e.g., E1: 15kWh, E2: 12kWh, E3: 10kWh), generating a final scheme for proportionally allocating electricity.
[0175] Based on the adjusted elevator operation mode, the current power distribution data and elevator operation parameters are extracted from the system database. The minimum energy consumption requirement is set to 5 kWh per hour, and the peak load is limited to 20 kW. Power distribution constraints are then established.
[0176] An energy consumption optimization control algorithm based on dynamic programming is used to calculate the power allocation weight of each elevator. For example, 70% of the power resources are allocated during off-peak hours and 90% during peak hours, generating an energy-saving power allocation scheme.
[0177] By combining load distribution data and analyzing energy consumption fluctuation curves, if energy consumption reaches 22kW in a certain period, exceeding the peak limit, the elevator running interval is adjusted from 30 seconds to 45 seconds through a genetic algorithm, unnecessary stops are reduced, and the power distribution scheme is recalculated.
[0178] If energy consumption remains stable below 18kW, the current scheme will be used directly to generate the optimized configuration. Real-time elevator operation data, such as motor current and speed curves, will be collected and compared with historical energy consumption records in the system log to calculate the energy efficiency deviation. If the deviation exceeds 5%, optimization adjustments will be triggered.
[0179] Based on the optimization results, a hidden Markov model is used to analyze passenger behavior data and predict the probability of destination floor distribution. For example, during the morning rush hour, 70% of passengers go to floors 10 and above. The distribution dataset is then updated and input into the elevator mode adjustment module to form a closed-loop optimization process.
[0180] Elevator operation records and energy consumption data for the most recent week were extracted from the system log database. For example, the average number of elevator runs per day was found to be 150, with an energy consumption of 500 kWh. Simultaneously, passenger behavior data, such as the frequency of passengers getting on and off at each floor, was collected through a real-time monitoring system. It was found that the 10th floor had the highest number of passengers getting on and off during the morning rush hour, reaching 200 passengers per day. These two types of data were combined to obtain a preliminary operation and behavior dataset.
[0181] A passenger distribution update model based on time series analysis is used to reanalyze passenger behavior patterns and determine updated passenger behavior characteristic parameters, such as increasing the proportion of passengers on the 10th floor to 30% during the morning peak hours. Using these characteristic parameters, combined with a pre-set energy-saving optimized elevator operation configuration scheme, an updated target floor distribution dataset is generated, such as setting the 10th floor as the priority stopping floor.
[0182] The dataset was fed back to the initial configuration module and compared with historical operating data, revealing a passenger distribution deviation of 15%.
[0183] If the deviation exceeds the preset 10% threshold, it will be corrected through a deviation analysis algorithm, such as adjusting the priority of the 10th floor to reduce the deviation to 8%.
[0184] If the deviation is within the threshold, the corrected dataset is used directly to determine the final floor distribution optimization parameters.
[0185] Based on these parameters and real-time monitoring of elevator operation status data, such as the current elevator load rate of 70%, a dynamically adjusted elevator operation mode configuration is generated, such as increasing the number of stops on the 10th floor.
[0186] With this configuration, an energy consumption optimization control algorithm based on genetic algorithms is applied to redistribute power resources, such as reducing the frequency of elevator operation during off-peak hours, resulting in an energy-saving optimized resource allocation scheme that reduces energy consumption to 450 kWh.
[0187] According to the scheme, the operation records and energy consumption data in the system log database are updated to generate a closed-loop optimized elevator operation dataset, providing a data foundation for the next optimization.
[0188] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for optimizing the operation of building elevators based on a dynamic allocation algorithm, characterized in that: include: Data on passenger behavior is acquired from monitoring and sensor devices, and passenger destination floor information is analyzed to generate an initial destination floor distribution dataset. Based on the initial destination floor distribution dataset, passengers are classified to determine the concentrated destination floor range and quantity statistics for each type of passenger. The classification of passengers includes: Obtain the initial target floor distribution dataset as input data; Clustering algorithms are used to group passenger destination floor information in the initial destination floor distribution dataset to determine the concentrated destination floor range for each group of passengers. For each group of passengers, the number of passengers within each destination floor range is counted, the number statistics are generated, the number statistics are sorted, the main destination floor range is identified, and a classification result report is generated. Based on the statistical results of the range and number of floors of the central destination, passengers are assigned to different elevator units to generate a preliminary elevator allocation plan; The generation of the preliminary elevator allocation plan includes: Obtain the statistical results of the range and number of floors for the central destination, and use a dynamic allocation algorithm to allocate passengers to the corresponding elevator units according to the range of floors for the central destination; The number of passengers allocated to each elevator unit is calculated and compared with the maximum load threshold. If overloading occurs, a secondary adjustment is automatically triggered through a feedback mechanism to rebalance the allocation results. The adjusted allocation results will form a preliminary elevator allocation plan, and a simulated operation test will be conducted. The stopping order of the preliminary elevator allocation scheme is adjusted to generate an optimized elevator stopping order scheme; Based on the optimized elevator stopping sequence scheme and real-time operating status data, analyze the passenger flow trend and generate a traffic change prediction dataset. Adjust the elevator operation mode based on the traffic flow change prediction dataset, and generate the adjusted elevator operation mode; Based on the adjusted elevator operation mode, resources are allocated to generate an energy-saving and optimized elevator operation configuration scheme. Based on the energy-saving and optimized elevator operation configuration scheme and real-time passenger behavior data, the passenger behavior pattern is updated, an updated target floor distribution dataset is generated, and the data is fed back to the initial data processing stage to form a continuous optimization closed loop.
2. The building elevator operation optimization method based on dynamic allocation algorithm as described in claim 1, characterized in that: The initial target floor distribution dataset includes: Passenger image and behavioral data are collected through video surveillance equipment and sensor devices; Intelligent recognition algorithms are applied to the image data and behavioral data to extract the passenger's movement trajectory and behavioral characteristics; Based on movement trajectory and behavioral characteristics, infer the passenger's destination floor information; The destination floor information of the passengers is summarized and organized to generate an initial destination floor distribution dataset.
3. The building elevator operation optimization method based on dynamic allocation algorithm as described in claim 2, characterized in that: The optimized elevator stopping sequence scheme includes: For each elevator unit, extract the floor information from the preliminary elevator allocation plan; A genetic optimization algorithm is applied to adjust the order of the floor information to minimize passenger waiting time and elevator travel distance as the objective function. Determine the optimal stopping sequence, integrate the optimal stopping sequence into the preliminary elevator allocation plan, and generate an optimized elevator stopping sequence plan.
4. The building elevator operation optimization method based on dynamic allocation algorithm as described in claim 3, characterized in that: The generated traffic change prediction dataset includes: Elevator operation status data and floor call signal data are collected from the real-time monitoring system, and the optimized stopping sequence scheme is correlated and analyzed with the real-time elevator operation status data and floor call signal data. The correlation analysis results are processed using a traffic prediction model to predict the trend of passenger traffic changes throughout the entire time period. The formula for calculating the predicted passenger traffic value is as follows: in: This represents the predicted passenger flow value at time t. This represents the passenger flow value at the previous moment. This represents the attenuation coefficient of historical data, where H represents the total number of floors. This represents the call signal frequency of the h-th layer. This represents the actual traffic data for layer h; Generate a traffic change prediction dataset based on passenger flow trends.
5. The building elevator operation optimization method based on dynamic allocation algorithm as described in claim 1, characterized in that: The generated and adjusted elevator operation mode includes: Analyze the traffic change trend in the traffic change prediction dataset. If it is determined that the current period is a low traffic period, the number of elevator units in operation will be reduced through a dynamic adjustment mechanism. Based on the reduced number of elevator units, operational tasks will be reallocated in accordance with minimum service guarantee requirements; Based on the reassignment results after the task allocation, the adjusted elevator operation mode is generated.
6. The building elevator operation optimization method based on dynamic allocation algorithm as described in claim 1, characterized in that: The generated energy-saving and optimized elevator operation configuration scheme includes: Analyze the operating status and load requirements of each elevator unit under the adjusted elevator operation mode; An energy consumption optimization control algorithm is applied to redistribute power resources and calculate the minimum energy consumption requirement for each elevator. Based on the minimum energy consumption requirements and peak load limits, the resource allocation strategy is adjusted, and an energy-saving optimized elevator operation configuration scheme is generated based on the adjusted resource allocation strategy.
7. The building elevator operation optimization method based on dynamic allocation algorithm as described in claim 1, characterized in that: The generated and updated target floor distribution dataset includes: Acquire energy-saving and optimized elevator operation configuration schemes, elevator operation record data and energy consumption data from the system log database, and passenger behavior data obtained through real-time monitoring; Integrate and analyze elevator operation records, energy consumption data, and passenger behavior data; The passenger distribution update model was used to process the integrated analysis results and reanalyze passenger behavior patterns. Based on the analysis of passenger behavior patterns, an updated target floor distribution dataset is generated; The updated data is fed back to the initial data processing stage.
8. The building elevator operation optimization method based on dynamic allocation algorithm as described in claim 1, characterized in that: The continuous optimization closed loop includes: The updated target floor distribution dataset is used as the input for the next round; The updated target floor distribution dataset is linked to the initial data processing stage to update the basic data in the system database; Based on the updated baseline data, the passenger classification, allocation, and stop order optimization processes were re-executed. The traffic forecast and operating mode were updated by adjusting the allocation scheme and docking sequence; Adjust resource allocation strategies based on updated traffic forecasts and operating patterns; The adjusted resource allocation strategy generates a new operating configuration scheme, which is then applied to the elevator control system.