An elevator control system based on artificial intelligence

Abstract

The present invention discloses an elevator control system based on artificial intelligence, which belongs to the field of elevator control technology. The system covers a multi-layer architecture including data acquisition, transmission, processing, intelligent decision-making and execution control. The data acquisition layer collects multi-source data through a group of sensors inside and outside the car and the operating status; the data processing layer provides support for intelligent decision-making through pre-processing, analysis and mining, and model training optimization. The elevator dispatching module of the intelligent decision-making layer dynamically partitions peak hours and performs energy-saving dispatching during flat hours based on information such as passenger flow; the safety protection module monitors and handles hidden dangers in real time; and the personalized service module provides customized services based on passenger behavior and needs. The drive, door machine and car environment control systems of the execution control layer respectively realize smooth lifting and lowering of the car, safe operation of the door and intelligent adjustment of the environment. Experiments show that the system can significantly improve the operating efficiency, safety and passenger satisfaction of the elevator, reduce energy consumption, and has important application value.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator control, and in particular to an elevator control system based on artificial intelligence. Background Art

[0002] In modern buildings, elevators are key equipment for vertical transportation, and the performance of their control systems directly affects people's daily travel experience and the overall operating efficiency of the building. However, existing elevator control systems have obvious deficiencies in many aspects.

[0003] Traditional elevator control systems are mostly based on simple preset rules and fixed logic algorithms for operation scheduling. During peak passenger flow periods, call requests from each floor are frequent. Due to the lack of accurate analysis of the overall passenger flow distribution and passenger flow trends, elevators are often unable to make reasonable allocation decisions, resulting in some elevators being overly concentrated in high-flow areas, while passengers in other areas have to wait too long, resulting in low overall transportation efficiency, which can easily cause passenger congestion between floors, greatly affecting people's travel efficiency and mood.

[0004] In terms of safety protection, the existing elevator safety system mainly relies on limited sensors to monitor single operating parameters (such as speed, load, etc.). When abnormal situations occur, the protection measures are relatively simple and passive. For example, for the fault detection of the door machine system, it can only respond when there are serious problems such as obvious jamming of the door or failure to close normally. It is difficult to prevent potential door machine failure risks in advance, such as the unsmooth operation caused by slight wear of the door machine transmission parts, which poses a potential threat to the personal safety of passengers.

[0005] From the perspective of energy utilization, the existing elevator control system lacks the ability to manage energy in a refined manner. During the operation of the elevator, whether it is fully loaded or unloaded, the operating power of the motor usually remains relatively fixed and cannot be dynamically adjusted according to the actual load and operating conditions. This causes a lot of energy waste, especially in high-rise buildings, where many elevators are running frequently, and the accumulated energy consumption is considerable, which does not conform to the current development concept of energy conservation and environmental protection.

[0006] In addition, the existing elevator system also has great limitations in terms of passenger services. Since it is impossible to conduct in-depth analysis of passengers' behaviors and needs, elevators cannot provide personalized services. For example, for passengers carrying heavy objects, the elevator door will not automatically extend the opening time to facilitate their entry and exit; for passengers who frequently go to specific floors, it cannot be prioritized or intelligently docked according to their habits, making the passenger's elevator experience relatively monotonous and inconvenient, and unable to meet modern people's pursuit of high-quality travel services.

[0007] To sum up, the existing elevator control system has many shortcomings in terms of operating efficiency, safety, energy utilization and passenger service, and it is difficult to adapt to the needs of modern buildings for efficient, safe, comfortable and intelligent elevator systems. Therefore, the development of a new intelligent elevator control system has important practical significance and urgent market demand. Summary of the invention

[0008] The main purpose of the present invention is to provide an elevator control system based on artificial intelligence, which can effectively solve the problems mentioned in the background technology.

[0009] To achieve the above object, the technical solution adopted by the present invention is:

[0010] An elevator control system based on artificial intelligence includes a data acquisition layer, a data transmission layer, a data processing layer, an intelligent decision-making layer and an execution control layer, wherein the data acquisition layer includes an internal car sensor group, an external car sensor group and an elevator operation status sensor group, wherein the internal car sensor group includes a high-definition camera, a pressure sensor, and a sound sensor, the external car sensor group includes a floor call button sensor, an infrared human body sensor, and an environmental sensor, and the elevator operation status sensor group includes a position sensor, a speed sensor, and a door status sensor.

[0011] The data transmission layer includes a wired communication network and a wireless communication network.

[0012] The data processing layer includes:

[0013] Data preprocessing unit: performs preliminary processing on the raw data from the data acquisition layer by cleaning, filtering, denoising, and format conversion, removes invalid values, outliers, and interference signals, and unifies the data format; improves data quality and availability, and provides an accurate and reliable data foundation for subsequent analysis and artificial intelligence model training.

[0014] Data Analysis and Mining Unit: Use big data analysis technology and data mining algorithms to analyze and mine pre-processed data to extract valuable information and knowledge patterns; for example, correlate and analyze historical and real-time passenger flow data to mine passenger flow change patterns and correlations; analyze elevator operation status data trends to discover equipment performance change trends and potential fault hazards; cluster analyze passenger behavior data to identify behavioral characteristics and demand patterns, provide decision support and data basis for the intelligent decision-making layer, and optimize elevator scheduling, operating parameters and service models.

[0015] Artificial intelligence model training and optimization unit: Build artificial intelligence models based on deep learning algorithms and reinforcement learning algorithms, and use a large amount of historical and real-time data to train and optimize the models, adjust parameters and structures to enable them to accurately handle complex elevator operation situations, improve prediction accuracy and decision-making intelligence, and at the same time evaluate the performance of the monitoring model, improve and update it according to actual conditions, and adapt to changes in the operating environment and business needs.

[0016] The intelligent decision-making layer includes:

[0017] Elevator dispatch decision module: Based on the passenger flow forecast information, passenger distribution and current elevator operation status provided by the data analysis and mining unit, the intelligent dispatch algorithm is used to make elevator operation dispatch decisions; during peak hours, a dynamic zoning dispatch strategy is adopted to divide the service area according to the passenger flow demand of each floor, allocate the running direction and the stop floor, improve transportation efficiency and reduce passenger waiting time; during off-peak hours, an energy-saving dispatch mode is adopted to dynamically adjust the elevator sleep and wake-up strategies according to the real-time call situation to reduce energy consumption. The decision results are converted into control instructions and sent to the execution control layer to achieve intelligent and precise control.

[0018] Safety protection and emergency decision-making module: Real-time monitoring of elevator operation status data and abnormal detection information of data analysis and mining units. Once safety hazards or emergencies such as overspeed, stall, door failure, overload, etc. are found in the elevator, the corresponding safety protection measures and emergency decision-making plans will be immediately initiated; such as triggering the safety brake device, opening the car door for ventilation, starting the emergency lighting and alarm system, and guiding passengers to respond through voice and light, and sending fault and emergency handling situations to the remote monitoring center and property management personnel in time to ensure the safety of passengers and safe operation of elevators.

[0019] Personalized service decision module: Provide personalized elevator service decisions based on the analysis results of passenger behavior and needs; for example, when a passenger is carrying heavy objects, the elevator door is controlled to extend the opening time and voice prompts are given; the elevator automatically slows down and notifies assistance to passengers with limited mobility. The personalized needs (reservation of elevator rides, priority services, etc.) sent by passengers through the interactive terminal in the car are processed and reasonably arranged according to the overall operation situation to improve the passenger experience and satisfaction.

[0020] The execution control layer includes:

[0021] Elevator drive control system: Receives elevator operation control instructions (speed, position, direction, etc.) issued by the intelligent decision-making layer, controls the operation of the elevator drive motor through variable frequency speed regulation technology, and realizes smooth lifting and lowering movement of the car; has high-performance speed regulation and fast dynamic response capabilities, and adjusts the motor output power and speed in real time according to the working conditions to ensure the comfort and safety of elevator acceleration, uniform speed, and deceleration. At the same time, it uses optimization algorithms and energy recovery technology to reduce energy consumption and operating costs, and improve energy utilization efficiency.

[0022] Elevator door machine control system: It is responsible for the opening and closing of the elevator door according to the control instructions of the intelligent decision-making layer, and accurately controls the opening and closing speed, force and position of the door to ensure safe and reliable operation and avoid accidents of people or objects being caught. It has automatic detection and fault diagnosis functions, and monitors the door operation status in real time through the position, speed and force sensors on the door machine. In case of abnormality, it takes protective measures and feeds back fault information to the intelligent decision-making layer, and at the same time sends prompts to passengers in the car to ensure the safety of taking the elevator.

[0023] Car environment control system: According to the environmental sensor data of the data collection layer and the environmental control decision instructions of the intelligent decision layer, the ventilation, lighting, temperature and humidity environmental parameters in the car are intelligently adjusted. For example, ventilation and cooling and heating are automatically controlled according to the number of passengers and temperature, and the lighting brightness and color are adjusted according to the light. The car environment is automatically optimized under different operation stages and working conditions of the elevator to improve the comfort and satisfaction of passengers.

[0024] Preferably, the high-definition camera captures all-round visual information of passengers' movements, expressions, and items carried in the car, and the image data is transmitted to the data processing layer in real time via a high-speed data transmission line; the pressure sensor measures the weight distribution in the car, determines the number of passengers, overload conditions, and position changes, and the data is transmitted to the car controller in a wired manner and then forwarded to the data processing layer; the sound sensor collects ambient sound in the car for monitoring the environmental conditions and equipment operating status in the car, and transmits the sound data to the data processing layer for analysis via an audio transmission line.

[0025] The floor call button sensor records the floor, time and direction called by the passenger, and the data is fed back to the data processing layer in real time via the elevator control system communication line to provide basic information for elevator scheduling decisions; the infrared human body sensor detects the number of people in the elevator waiting area, the entry and exit of the car and the stay time, and transmits the data to the data processing layer in a wired or wireless manner to assist in judging passenger flow and passenger flow; the environmental sensor monitors the environmental parameters of the elevator waiting area on each floor, providing a reference for environmental adjustment and lighting control in the car, and the data is transmitted to the data processing layer for comprehensive analysis with the environmental data in the car.

[0026] The position sensor measures the real-time position of the elevator car in the shaft, and the data is transmitted to the core processor and data processing layer of the elevator control system through a dedicated data line, which is crucial to the elevator speed control, floor stop accuracy and operation safety; the speed sensor monitors the elevator running speed in real time, which is used for speed closed-loop control and operation status analysis. The data is transmitted to the elevator control system and data processing layer, and combined with other parameters to judge the operation performance and potential problems to ensure the smooth operation of the elevator; the door status sensor monitors the elevator door switch status, speed, position and door lock locking status to ensure the safety of passengers entering and exiting, and provide a basis for operation efficiency analysis. The data is transmitted to the elevator control system and data processing layer via wireless.

[0027] Preferably, the specific implementation method of the analysis and mining of the data analysis and mining unit includes the following steps:

[0028] ①Data integration and association

[0029] Integrate the various types of pre-processed data according to reasonable logic and dimensions to construct a comprehensive data set, so that the passenger flow, elevator operating status, car environment and passenger behavior data can be correlated and corresponded to each other.

[0030] The improved Apriori algorithm is used to mine association rules, and information entropy is used to measure the uncertainty and activity of data, so as to assist in dynamically adjusting the support threshold and more accurately mine valuable association information. The core lies in the information entropy formula: ,in Represents a random variable The larger the entropy value, the greater the uncertainty of the random variable and the greater the amount of information contained; the smaller the entropy value, the smaller the uncertainty of the random variable and the smaller the amount of information contained; Represents a random variable The value is The probability of Indicates an event The amount of information, the lower the probability of the event, the higher the amount of information; according to the entropy value of the data distribution, different data activity areas are divided, and then different support thresholds are adapted to different areas to explore more practical and valuable association rules between factors such as passenger flow and elevator operation, passenger behavior and cabin environment, providing a reliable basis for subsequent decision-making.

[0031] ②Time series analysis (for passenger flow and operation status data)

[0032] The Seasonal ARIMA model is used, focusing on the characteristics of the autocorrelation function (ACF) and the partial autocorrelation function (PACF), and the Akaike Information Criterion (AIC) is used to automatically determine the difference order, and a suitable ARIMA model structure is constructed to analyze time series data. The model parameters are optimized by minimizing the loss function, and the trained model is used to predict future passenger flow and elevator operation status, thereby providing a key reference for the early layout of the elevator scheduling strategy to cope with changes in passenger flow in different periods of time.

[0033] ③Cluster analysis (for passenger behavior data)

[0034] Using the fast search clustering algorithm based on density peak (DPC), by calculating the local density of each data point ( ,in is the distance between data points, Dynamically determined based on data characteristics) and the shortest distance to a higher density point , based on these two key indicators, the cluster center is determined to achieve effective clustering of passenger behavior data, distinguish different behavior categories such as normal elevator riding, carrying special items, and mobility impairment, and provide accurate classification support for personalized service decisions.

[0035] ④Abnormal detection (for elevator operation data)

[0036] Combine the autoencoder (AE) based on reconstruction error and the long short-term memory network (LSTM) to carry out anomaly detection. Use AE to perform unsupervised learning on normal operation data and calculate the reconstruction error formula: , is the original data, To reconstruct the data, is the number of data points, so as to determine the reconstruction error range of normal data; at the same time, the LSTM network is used to perform dynamic feature extraction and anomaly scoring on the time series data, and the reconstruction error of AE and the anomaly score of LSTM are combined to determine whether the elevator operation data is abnormal. Once an anomaly is found, it will be fed back to the safety decision-making module in time to ensure the safe and stable operation of the elevator system.

[0037] Preferably, the specific steps of the artificial intelligence model training and optimization unit include:

[0038] (1) Model construction

[0039] ① Deep learning model:

[0040] Passenger behavior recognition: Construct an architecture that integrates convolutional neural networks (CNN) and long short-term memory networks (LSTM). CNN is responsible for extracting the spatial features of images in the car, while LSTM focuses on processing the time series information of passenger behavior, thereby achieving more accurate recognition and understanding of passenger behavior and enhancing the model's ability to grasp complex behavior patterns.

[0041] Passenger flow prediction: Build a recurrent neural network (RNN) model based on the attention mechanism, use the attention mechanism to focus on different time periods and various key factors that affect passenger flow (such as surrounding activities, time periods, etc.), and dynamically allocate the weights of each factor through a unique calculation method, so as to accurately predict passenger flow and improve the accuracy and adaptability of the prediction.

[0042] ②Reinforcement Learning Model:

[0043] A deep Q network (DQN) is used to build a model for elevator dispatching decisions. The real-time operation state of the elevator is defined as the environment state, and the available operations are set as the action space. The state-action value function is estimated through the Q network, and the optimal action is selected based on the estimated Q value, so as to realize intelligent elevator dispatching strategy selection, and the stability and effectiveness of training are guaranteed by the target Q network mechanism.

[0044] (2) Model training

[0045] ①Data preprocessing and division:

[0046] The collected historical and real-time data are standardized to unify the data feature dimensions and distribution range. The data set is then divided into 70% training set, 15% validation set, and 15% test set to ensure that the model can effectively verify the generalization ability on different data subsets.

[0047] ② Training algorithm selection and optimization:

[0048] Deep learning model: Use improved algorithms of stochastic gradient descent (SGD) (such as the Adam algorithm) to update parameters. The core of this algorithm is to be able to adaptively adjust the learning rate, dynamically optimize the update step size according to the gradient changes of different parameters, improve training efficiency and model convergence effect, and use the early stopping method to prevent overfitting, that is, terminate training in time according to changes in the performance indicators of the validation set to retain the optimal parameter state.

[0049] Reinforcement learning model: A multi-step learning mechanism is used to accelerate the training process. By considering the reward feedback of multiple steps in the future to update the Q value more efficiently, the model is prompted to learn high-quality scheduling strategies more quickly and enhance the model's decision-making ability in complex elevator operating environments.

[0050] (3) Model evaluation and optimization

[0051] ①Performance index selection:

[0052] Passenger behavior recognition model: We select accuracy, recall rate, F1 value, etc. as key evaluation indicators to measure the accuracy of the model's passenger behavior classification, its ability to capture positive samples, and its overall performance from different angles, so as to comprehensively evaluate the effectiveness of the model.

[0053] Passenger flow prediction model: Use indicators such as root mean square error (RMSE) and mean absolute error (MAE) to accurately quantify the degree of deviation between the predicted passenger flow and the actual passenger flow, reflecting the accuracy of the model prediction.

[0054] Elevator dispatch decision model: focuses on indicators such as average passenger waiting time (AWT), average transport time (ATT), and elevator energy consumption (EC), which are directly related to the core efficiency and energy utilization of elevator operation, so as to judge the pros and cons of dispatch decisions.

[0055] ②Model improvement and update:

[0056] According to the performance of the model on the test set, if it does not meet expectations, the model hyperparameters are adjusted and optimized in a targeted manner. For example, the deep learning model can adjust the network structure (number of layers, number of neurons, etc.), activation function, etc.; the reinforcement learning model can optimize the reward function design, exploration rate decay strategy, etc. At the same time, the newly collected operation data is regularly integrated to retrain the model so that it can adapt to the elevator operation environment and business demand changes in real time, and always ensure the intelligent optimization and precise control capabilities of the model in the elevator control system.

[0057] Preferably, the specific implementation method of the elevator scheduling decision module is:

[0058] (1) Receive passenger flow forecast information from the data analysis and mining unit and construct a two-dimensional matrix , where the elements Indicates in time slice Inner floor The predicted passenger flow; at the same time, the current position vector of each elevator is obtained , running direction vector (element value -1 means downward, 0 means stationary, 1 means upward), the number of passengers in the car vector and the current velocity vector , and the number of passengers waiting on each floor .

[0059] Perform normalization on the passenger flow data: ,in and They are The minimum and maximum passenger flow values ​​in ; normalize the elevator position data: ,in is the total number of floors; normalize the speed data: ,in This is the maximum design speed of the elevator, ensuring that data is used in a unified dimension for subsequent calculations.

[0060] (2) Dynamic zone scheduling during peak hours

[0061] ①Regional division:

[0062] The improved K-Means clustering algorithm is used to partition the passenger flow on each floor, and the objective function is: ,in is the number of partitions, For the The set of floors contained in a partition, is the mean passenger flow of the partition, For floor The weight of the floor is determined by the importance of the floor (such as the office-concentrated floor has a higher weight) or the historical average passenger flow ratio. The objective function is optimized through iteration. Minimum, get the results of each partition, to ensure that high-flow area and low-flow area can be effectively distinguished.

[0063] ② Elevator allocation and dispatching:

[0064] Calculate the number of elevators that should be allocated to each zone : ,in is the average elevator inter-floor transportation time (based on historical data statistics), is the total number of elevators.

[0065] When there is a new call, calculate each elevator To the call floor Response time : ,in The additional stop time caused by factors such as the current direction of the elevator and the destination of the passengers in the car is taken into account. The calculation is based on the elevator operation rules and current status information. For example, if the elevator is currently running upward and there are passengers in the car going to a destination higher than the called floor, then Increase the corresponding stop time estimate; conversely, if the elevator running direction is consistent with the call direction and there is no internal passenger conflict, then Take the smaller value. Select

[0066] The minimum elevator response time is achieved, and the order of the stopping floors is planned according to the passenger flow trend and elevator load conditions in the partition to optimize the average waiting and transportation time. This is achieved by building a dynamic programming model based on real-time passenger flow and elevator status. The state transfer equations and decision rules of the model are set according to the elevator operation logic and efficiency optimization goals. By solving the model, the optimal stopping floor sequence is obtained to improve transportation efficiency.

[0067] (3) Energy-saving scheduling during off-peak hours

[0068] ① Sleep and wake-up strategy:

[0069] Setting idle time thresholds and passenger flow threshold When the elevator idle time exceeds And the current floor predicted passenger flow is lower than When a new call is received, the elevator is set to sleep mode and non-critical equipment is turned off to save energy. The wake-up decision value ,in The start time of the elevator from sleep to operation (determined by the hardware characteristics of the elevator, and the empirical value is obtained through testing). Minimal elevator wake-up to respond to calls ensures demand is met while energy consumption is reduced.

[0070] ②Operation speed optimization:

[0071] According to the number of passengers in the car and current floor passenger flow Dynamically adjust the elevator speed. The speed adjustment formula is: ,in is the speed adjustment factor (optimized and determined based on the elevator energy consumption curve and performance test data), For full load, In order to average passenger flow, the speed can be reduced in this way to save energy when the load and traffic are low, while ensuring basic service quality.

[0072] (4) Generation and sending of decision instructions

[0073] The dispatch decision is converted into control instructions, including the elevator running direction, target floor and speed instructions. For example, if the decision is that the elevator Go to floor And at speed Run to generate the instruction tuple: ,in Determined based on the relationship between the target floor and the current position (1 for upward, -1 for downward, and 0 for static). The command is sent to the execution control layer through the elevator control system communication interface to ensure that the elevator runs accurately according to the scheduling strategy, realizing efficient and intelligent elevator control, which is significantly different from the traditional scheduling method, improving the overall performance and energy efficiency of elevator operation, meeting the complex and changeable actual operation needs, enhancing the adaptability and stability of the system, providing passengers with a high-quality service experience, and reducing operating costs and energy consumption.

[0074] Preferably, the specific implementation method of the security protection and emergency decision-making module is:

[0075] (1) Data monitoring and acquisition: Connect with the elevator operation status sensor and data analysis and mining unit to collect data such as elevator speed, position, door status, car weight, and abnormal detection information such as fault risk indicators and passenger flow abnormality indicators in real time, and store and update them.

[0076] (2) Determination of safety hazards and emergency situations: Set safety thresholds such as speed, door opening and closing time, and load according to elevator design standards, and combine real-time data and abnormal indicators to determine whether there are overspeed, stall, door failure, overload, and potential safety hazards, such as abnormal motor temperature and sudden changes in passenger flow.

[0077] (3) Safety protection measures and emergency decision-making and execution:

[0078] When overspeeding, a braking command is sent to the drive control system, the brake is activated, the car alarm is sounded and fault information is transmitted to the remote monitoring center.

[0079] When a door fails, the door machine stops running and measures are taken according to the fault situation, such as eliminating the jam or maintaining the door status and alarming for maintenance, while transmitting the fault details.

[0080] When overloaded, the elevator is prohibited from running, the car will give prompts and voice broadcasts, and the operation will be restored after the weight returns to normal.

[0081] When there are potential serious hidden dangers, such as motor overheating, the speed will be reduced, ventilation and heat dissipation will be carried out, and parameters will be monitored; in case of fire evacuation, the elevator will be recalled to the refuge floor, the door will be opened, and the fire protection system will be linked to ensure safe operation and emergency response.

[0082] Preferably, the specific implementation method of the personalized service decision module is:

[0083] (1) Data collection and analysis: Collect passenger behavior and demand data, such as behavior recognition results, historical elevator riding habits, and real-time demand information, organize and analyze them, and store them in a standardized format.

[0084] (2) Service decision making:

[0085] Behavior-triggered service: Automatically generates corresponding service instructions after identifying specific behaviors, such as extending the door opening time, reducing the operating speed and notifying staff for assistance.

[0086] Habitual preference service: Based on historical habits, priority dispatch, automatic docking and personalized greeting services are provided at specific times or when approaching specific floors.

[0087] Real-time demand-responsive service: Receive and process passengers' real-time demands, such as making reservations for elevators, adjusting the elevator environment, etc., incorporate the demands into the scheduling plan and provide timely feedback.

[0088] (3) Service feedback and optimization: Collect passenger feedback, optimize service strategies based on feedback information, adjust parameter settings and evaluate service effectiveness, and continuously improve service quality.

[0089] Preferably, the control method of the elevator drive control system is:

[0090] (1) Instruction processing:

[0091] Build a reliable communication link with the intelligent decision-making layer to quickly and accurately receive elevator operation control instructions, covering core parameters such as speed, position, direction, and special working condition instructions. At the same time, strictly verify the instructions. Once data anomalies or errors are found, immediately feedback to the intelligent decision-making layer and request to resend the correct instructions to ensure the accuracy and completeness of the instructions, providing a basic guarantee for subsequent precise control.

[0092] (2) Motor drive and speed control:

[0093] Relying on advanced variable frequency speed regulation technology to drive the elevator motor, with vector control algorithm as the core, the motor current is accurately decomposed into two key components, torque and excitation, to achieve independent fine control. In the acceleration stage, the torque current is rapidly increased according to the preset curve to ensure that the motor quickly outputs strong torque and promotes the smooth acceleration of the car; in the uniform speed stage, the excitation and torque currents are strictly regulated to maintain the motor speed stable at the target value to ensure the smooth and comfortable operation of the car. At the same time, with the help of high-speed digital signal processors or programmable logic controllers, the motor operation status is monitored and feedback adjusted at high frequencies. The actual speed and position information are obtained in real time through high-precision sensors installed on the motor shaft. After comparing with the target value, the optimized closed-loop control algorithm is used to accurately calculate and adjust the inverter output in time, so that the motor can quickly and accurately track the target operation status, ensuring the efficiency and stability of the elevator operation.

[0094] (3) Adjustment to suit the working conditions:

[0095] The elevator operation conditions are fully monitored in real time, including parameters such as car load, number of passengers, current position, shaft environment, etc. The motor output power and speed are automatically and flexibly adjusted based on these real-time data. For example, under heavy load conditions, the acceleration and deceleration slopes are moderately slowed down, while the motor power output is increased to ensure smooth and safe operation; in high temperature environments, the load rate is reasonably reduced considering the changes in the motor's heat dissipation performance to ensure the reliable operation of the motor, so that the elevator can provide passengers with a safe and comfortable riding experience under various complex conditions.

[0096] (4) Energy management and optimization:

[0097] Equipped with efficient energy recovery devices, such as regenerative braking units, the regenerative electric energy generated by the motor is effectively fed back to the power grid during the deceleration and braking phase of the elevator, realizing energy recycling and reducing energy consumption. At the same time, the intelligent optimization algorithm is used to conduct all-round energy management of the elevator operation process. Through accurate prediction of the operation trajectory and working condition changes, the motor speed regulation strategy and energy recovery timing are planned in advance. Under the premise of ensuring the normal operation of the elevator, the energy utilization efficiency is maximized, the long-term operation cost of the elevator is reduced, and a win-win situation of economic and environmental benefits is achieved.

[0098] Preferably, the control method of the elevator door machine control system is:

[0099] (1) Command reception and analysis:

[0100] Establish reliable communication with the intelligent decision-making layer, quickly receive various control instructions of the elevator door, such as door opening, door closing, speed and position setting, and parse and convert them into executable signals. At the same time, strictly verify the command data to ensure its accuracy, laying a solid foundation for the subsequent precise control of the door machine operation.

[0101] (2) Door machine drive and control:

[0102] The motor is driven according to the instructions, and the advanced vector frequency conversion control technology is used to accurately control the motor speed and torque, so as to achieve precise control of the door opening and closing speed. Combined with the door machine controller and position, speed and other sensors, according to the preset operation curve and algorithm, the motor output is adjusted in real time when the door is opened and closed to ensure that the door moves smoothly and can stop accurately at the predetermined position, ensuring seamless connection with the car and floor door, creating comfortable and safe entry and exit conditions for passengers.

[0103] (3) Safety protection and fault diagnosis:

[0104] Relying on the position, speed and force sensors on the door machine, the door's operating status is monitored in all directions and in real time. Once the force sensor detects an obstacle when closing the door, it quickly triggers the resistance rebound function; if the speed and position parameters are abnormal, it immediately brakes and alarms. At the same time, through in-depth analysis of sensor data and comparison with normal modes, the fault type and location, such as motor failure, transmission problem or sensor failure, are quickly diagnosed, and detailed fault information is promptly fed back to the intelligent decision-making layer, and prompts are issued to passengers in the car, and it will also be uploaded to the remote monitoring center, effectively ensuring the safety of elevator riding and improving the timeliness of fault repair.

[0105] (4) Status feedback and communication:

[0106] The operating status of the door machine is regularly fed back to the intelligent decision-making layer, including information such as position, switch status, speed, and fault conditions, so that the intelligent decision-making layer can grasp the dynamics of the door machine in real time and flexibly adjust the control strategy when necessary. When an abnormality occurs, communication with the remote monitoring system is strengthened to allow maintenance personnel to quickly know the details of the fault, efficiently solve the problem, and enhance the reliability and stability of the overall operation of the elevator.

[0107] Preferably, the control method of the car environment control system is:

[0108] Data acquisition and command reception:

[0109] It is stably connected to the environmental sensors of the data acquisition layer to continuously collect information such as temperature, humidity, light intensity, air quality and number of passengers in the cabin. At the same time, it receives environmental control instructions such as ventilation, temperature, humidity, and lighting issued by the intelligent decision-making layer to ensure accurate transmission and timely acquisition of data and instructions.

[0110] Intelligent adjustment function:

[0111] Ventilation adjustment: Intelligently control the operating power and ventilation mode of the ventilation equipment according to the number of passengers and the air quality in the cabin. When the number of passengers increases or the air quality decreases, the ventilation volume is automatically increased to ensure fresh air; and according to the temperature conditions, the ventilation and cooling and heating collaborative modes can be flexibly switched to improve energy efficiency and create a comfortable air environment.

[0112] Temperature and humidity control: Through feedback from temperature and humidity sensors, the cooling and heating systems and humidity control equipment are precisely controlled. When the temperature deviates from the comfortable range, precise control strategies are used to quickly adjust it to the appropriate range; when the humidity is abnormal, the humidifier or dehumidifier is started in time to ensure that the temperature and humidity are maintained at a comfortable level for the human body, providing passengers with a pleasant cabin climate environment.

[0113] Lighting adjustment: According to the data from the light sensor, the lighting brightness is automatically adjusted to adapt to the light requirements in the car. At the same time, the lighting color can be intelligently switched according to the requirements of different time periods and scenes to create a comfortable visual environment and avoid discomfort to passengers due to excessive or dark light.

[0114] Compared with the prior art, the present invention has the following beneficial effects:

[0115] The present invention builds an intelligent elevator control system with artificial intelligence as the core. By integrating multiple data such as elevator operation status, car environment, and passenger behavior, it uses deep learning and reinforcement learning algorithms to build multiple models to achieve from precise decision-making to all-round control. Its beneficial effects are remarkable. Intelligent scheduling optimizes strategies based on real-time data, and dynamic zoning during peak hours and energy-saving scheduling during flat hours improve transportation efficiency; real-time monitoring combined with artificial intelligence analysis ensures safety and reliability, timely handles hidden dangers and conducts emergency linkage; personalized services are customized according to passenger conditions to improve experience and satisfaction; drive and environmental control use artificial intelligence to save energy and reduce consumption, reduce costs and be environmentally friendly; the model self-learns and adapts to improve system flexibility and intelligence, and reduces manual maintenance costs. It has broad application prospects and important value as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0116] Figure 1 It is a schematic diagram of the composition architecture flow of the present invention. DETAILED DESCRIPTION

[0117] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0118] like Figure 1 The system composition and process shown are further described in conjunction with a specific implementation case. Assume that an elevator control system based on artificial intelligence provided by the present invention is installed in a 30-story commercial office building. In actual operation:

[0119] Data collection and transmission

[0120] Car interior sensor set

[0121] The high-definition camera is installed in the top corner of the car, which can capture all-round visual information of passengers' movements, expressions, and items carried in the car. For example, during the morning rush hour, the camera captured a passenger carrying a large suitcase into the car, and the image data was transmitted to the data processing layer in real time via a high-speed data transmission line.

[0122] Pressure sensors are distributed at different positions at the bottom of the car to measure the weight distribution in the car, thereby determining the number of passengers, overload conditions, and position changes. When 10 passengers enter the car, the pressure sensor detects that the weight is close to but not exceeding the rated load, and the data is transmitted to the car controller via wired mode and then forwarded to the data processing layer.

[0123] The sound sensor collects the ambient sound inside the car and is used to monitor the environmental conditions and equipment operating status inside the car. For example, during the operation of the elevator, the sound sensor detects abnormal mechanical friction sound and transmits the sound data to the data processing layer for analysis through the audio transmission line.

[0124] Car external sensor group

[0125] The floor call button sensor records the floor, time and direction called by the passenger. When the call button on the 15th floor is pressed, the data is fed back to the data processing layer in real time via the elevator control system communication line, providing basic information for elevator dispatching decisions.

[0126] The infrared human body sensor detects the number of people in the elevator waiting area, the situation of entering and exiting the elevator car, and the length of time they stay. In the elevator waiting area on the 5th floor, the infrared human body sensor detects that there are 5 people waiting for the elevator, and transmits the data to the data processing layer by wire or wireless means to assist in determining the passenger flow and passenger mobility.

[0127] Environmental sensors monitor environmental parameters of the elevator waiting area on each floor, such as temperature and humidity, to provide a reference for environmental adjustment and lighting control in the car. The data is transmitted to the data processing layer for comprehensive analysis of the environmental data in the car.

[0128] Elevator operation status sensor group

[0129] The position sensor measures the real-time position of the elevator car in the shaft, and transmits the data to the core processor and data processing layer of the elevator control system through a dedicated data line. It is crucial to the elevator speed control, floor stop accuracy and operation safety. For example, when the elevator rises from the 10th floor to the 11th floor, the position sensor continuously updates the car position information.

[0130] The speed sensor monitors the elevator speed in real time for speed closed-loop control and operation status analysis. During the acceleration phase of the elevator, the speed sensor detects that the speed is gradually increasing, and the data is transmitted to the elevator control system and data processing layer, which is combined with other parameters to determine the operation performance and potential problems to ensure the smooth operation of the elevator.

[0131] The door status sensor monitors the elevator door switch status, speed, position and door lock status to ensure the safety of passengers entering and exiting, and provide a basis for operation efficiency analysis. The data is transmitted wirelessly to the elevator control system and data processing layer.

[0132] All the above data are transmitted to the data processing layer for analysis through the wired communication network of the data transmission layer and the wireless communication network.

[0133] Data processing and model training

[0134] Data preprocessing unit

[0135] The raw data from the data collection layer is cleaned, filtered, denoised, format converted, and processed. For example, for the weight data transmitted by the pressure sensor, the abnormal fluctuation value caused by the short shaking of the passenger is removed, the data format is unified into a standard numerical format, invalid values, abnormal values, and interference signals are removed, the data quality and availability are improved, and an accurate and reliable data foundation is provided for subsequent analysis and artificial intelligence model training.

[0136] Data Analysis and Mining Unit

[0137] Data integration and association: Integrate all kinds of pre-processed data according to logic and dimensions to construct a comprehensive data set. Through analysis, it is found that at 9-10 o'clock on weekdays, the passenger flow on floors 10-20 is large, and most passengers on these floors carry office supplies. After being associated with the elevator operation status and car environment data, it provides a basis for subsequent decision-making. Use the improved Apriori algorithm to mine association rules, and use information entropy to measure the uncertainty and activity of data. For example, it is calculated that the information entropy of the passenger flow on the 20th floor within a certain period of time is high, indicating that its uncertainty is large. A lower support threshold is set for it, and association rules such as the passenger flow on this floor and the use of surrounding conference rooms and the speed of the elevator are mined.

[0138] Time series analysis: Seasonal ARIMA model is used to analyze passenger flow and operation status data. Based on the characteristics of autocorrelation function and partial autocorrelation function, the difference order is automatically determined in combination with Akaike information criterion to construct the ARIMA model structure. By analyzing and training the passenger flow data of the past week, the passenger flow trend of each floor in different time periods in the next day is predicted, providing a key reference for the early layout of elevator scheduling strategy.

[0139] Cluster analysis: A fast search clustering algorithm based on density peaks is used to cluster passenger behavior data. The cluster center is determined by calculating the local density of each data point and the shortest distance to a higher density point, and different behavior categories such as normal elevator riding, carrying special items, and mobility impairment are distinguished. For example, the group of passengers who often carry breakfast into the elevator around 8 o'clock and act in a hurry can be identified, providing accurate classification support for personalized service decisions.

[0140] Anomaly detection: Anomaly detection is carried out by combining autoencoders based on reconstruction error and long short-term memory networks. AE is used to perform unsupervised learning on normal operation data to determine the reconstruction error range of normal data. At the same time, the LSTM network is used to perform dynamic feature extraction and anomaly scoring on time series data. For example, when the current data of the elevator door machine system fluctuates abnormally, the reconstruction error of AE and the anomaly score of LSTM are combined to determine the abnormality of the elevator operation data, and feedback is promptly provided to the safety decision-making module.

[0141] Artificial intelligence model training and optimization unit

[0142] Model building

[0143] Deep Learning Models:

[0144] The passenger behavior recognition model builds an architecture that integrates convolutional neural networks and long short-term memory networks. CNN extracts spatial features of images in the car, such as the appearance of passengers and the shape of their belongings, while LSTM processes the time series information of passenger behavior, such as the movement trajectory and dwell time of passengers in the car, to achieve more accurate recognition and understanding of passenger behavior.

[0145] The passenger flow prediction model builds a recurrent neural network model based on the attention mechanism, which uses the attention mechanism to focus on different time periods and various factors that affect passenger flow, such as activities held in surrounding office buildings and weather conditions. It dynamically allocates weights of various factors by calculation to accurately predict passenger flow.

[0146] Reinforcement learning model: A deep Q network is used to build a model for elevator scheduling decisions. The real-time operating state of the elevator is defined as the environmental state, and the available operations are set as the action space. The state-action value function is estimated through the Q network, and the optimal action is selected based on the estimated Q value.

[0147] Model Training

[0148] The collected historical and real-time data are standardized to unify the dimension and distribution range of data features, and then the data set is divided into 70% training set, 15% validation set, and 15% test set. For example, for passenger flow data, it is normalized to the range of 0-1, and for elevator location data, it is normalized according to the total number of floors 30.

[0149] The deep learning model uses the Adam algorithm to update parameters and adopts the early stopping method to prevent overfitting. The reinforcement learning model uses a multi-step learning mechanism to accelerate the training process.

[0150] Model evaluation and optimization

[0151] The passenger behavior recognition model selects accuracy, recall, and F1 value as key evaluation indicators. After testing, the accuracy rate reaches more than 90% when identifying the behavior of passengers carrying heavy objects. The passenger flow prediction model uses root mean square error and mean absolute error to quantify the degree of deviation between the predicted passenger flow and the actual passenger flow. After optimization, the root mean square error is controlled within a small range. The elevator scheduling decision model focuses on the average waiting time of passengers, the average transportation time, and the energy consumption indicators of elevators. By continuously adjusting the model parameters, the average waiting time of passengers has been shortened by 30%, the average transportation time has been shortened by 25%, and the energy consumption of elevators has been reduced by 20%.

[0152] Based on the performance of the model on the test set, the model hyperparameters are adjusted and optimized in a targeted manner, such as adjusting the number of network layers and neurons of the deep learning model, optimizing the reward function design of the reinforcement learning model, and regularly incorporating newly collected running data to retrain the model.

[0153] Intelligent decision making and execution control

[0154] Elevator dispatch decision module

[0155] Receive passenger flow forecast information from the data analysis and mining unit and construct a two-dimensional matrix , where the elements Indicates in time slice Inner floor The predicted passenger flow; at the same time, the current position vector of each elevator is obtained , running direction vector , the number of passengers in the car vector and the current velocity vector , and the number of passengers waiting on each floor After normalizing the relevant data, the scheduling decision process begins.

[0156] Dynamic zoning scheduling during peak hours: The improved K-Means clustering algorithm is used to partition the passenger flow of each floor. Assuming that the 30 floors are divided into 3 zones, high-flow zone (10-20 floors), medium-flow zone (5-9 floors and 21-25 floors), and low-flow zone (1-4 floors and 26-30 floors), the number of elevators to be allocated to each zone is calculated based on the average passenger flow of each zone and the floor weight. When there is a new call, the response time of each elevator to the called floor is calculated. For example, elevator 1 is currently running downward on the 8th floor at a speed of 1.2m / s. When it receives a call from the 12th floor, its response time is calculated according to the formula, and the additional stop time caused by factors such as the destination of the passengers in the car is taken into account. The elevator with the shortest response time is selected to respond, and the stop floor sequence is planned according to the passenger flow trend and elevator load conditions in the zone.

[0157] Energy-saving scheduling during off-peak hours: Set the idle time threshold to 5 minutes and the passenger flow threshold to 2 people. When the elevator idle time exceeds 5 minutes and the predicted passenger flow on the current floor is less than 2 people, the elevator is set to sleep. When there is a new call, calculate the wake-up decision value of each dormant elevator, and select the elevator with the smallest wake-up decision value to wake up and respond to the call. At the same time, dynamically adjust the elevator speed according to the number of passengers in the car and the passenger flow on the current floor. For example, at a certain moment, there are only 3 passengers in the car and the passenger flow on the current floor is small. The elevator speed is reduced to 80% of the rated speed according to the formula to save energy.

[0158] Decision instruction generation and sending: The dispatching decision is converted into control instructions, including the elevator running direction, target floor and speed instructions, etc., and the instructions are sent to the execution control layer through the elevator control system communication interface.

[0159] Security protection and emergency decision-making module

[0160] Data monitoring and acquisition: Connect with the elevator operation status sensor and data analysis and mining unit to collect data such as elevator speed, position, door status, car weight and abnormal detection information in real time. For example, it detects that the elevator speed suddenly exceeds 120% of the rated speed, and at the same time receives feedback from the data analysis and mining unit that the motor temperature is abnormally high.

[0161] Determination of safety hazards and emergency situations: Set safety thresholds such as speed, door opening and closing time, and load according to elevator design standards, and combine real-time data and abnormal indicators to determine safety hazards such as overspeed and potential motor failure.

[0162] Safety protection measures and emergency decision execution: In case of overspeed, a braking command is sent to the drive control system, the brake is activated, the alarm device in the car sounds, and the fault information is transmitted to the remote monitoring center. For potential motor failures, the elevator speed is reduced, the ventilation and cooling device is activated, and the motor parameters are closely monitored.

[0163] Personalized service decision module

[0164] Data collection and analysis: Collect passenger behavior and demand data, such as identifying the historical habit of a passenger who often works on the 18th floor to take the elevator at around 8:30 on weekdays, and another passenger's reservation for taking the elevator (from the 5th floor to the 20th floor at 9:00) sent through the interactive terminal in the elevator car, organize and analyze them, and store them in a standardized format.

[0165] Service Decision Making:

[0166] Behavior-triggered service: When a passenger carrying heavy objects is identified entering the elevator car, a service instruction is automatically generated to extend the elevator door opening time to 8 seconds and a voice prompt is given: "To facilitate your entry and exit, the elevator door will remain open for 8 seconds."

[0167] Habitual preference service: For the passenger who often takes the elevator to the 18th floor at around 8:30, when this time period approaches, if the elevator is idle or the running route is similar, the elevator will be dispatched to his floor first, and a personalized greeting message "Welcome, the elevator is ready for you" will be displayed in the car.

[0168] Real-time demand-responsive service: After receiving the passenger's request for an elevator reservation, the request is included in the scheduling plan, and the elevator's route and stop position are reasonably arranged before 9:00 to ensure that the elevator arrives on the 5th floor on time at the reservation time, and a reminder message is sent to the passenger that "the elevator you reserved is about to arrive, please be prepared."

[0169] Service feedback and optimization: After passengers have completed their ride, we collect feedback from them through the satisfaction survey button in the elevator. If a passenger complains that the elevator door takes too long to open, affecting efficiency, we optimize the service strategy based on the feedback information, adjust the parameter settings, shorten the elevator door opening time to 6 seconds when carrying heavy objects, and evaluate the service effectiveness again.

[0170] Elevator drive control system

[0171] Command processing: Establish a reliable communication link with the intelligent decision-making layer to quickly and accurately receive elevator operation control commands, such as receiving a command to increase the elevator speed to 1.5m / s to go to the 15th floor. At the same time, the command is rigorously verified to ensure the accuracy and completeness of the command.

[0172] Motor drive and speed control: Relying on advanced variable frequency speed regulation technology to drive the elevator motor, with vector control algorithm as the core, the motor current is accurately decomposed into two key components: torque and excitation. When the elevator starts, the torque current is quickly increased so that the motor can quickly output strong torque and promote the smooth acceleration of the car. In the uniform speed operation stage, the excitation and torque currents are strictly regulated to maintain the motor speed stable at the target value to ensure the smooth and comfortable operation of the car. With the help of high-speed digital signal processors, the motor operation status is monitored and feedback adjusted at high frequencies. The actual speed and position information are obtained in real time through high-precision sensors installed on the motor shaft. After comparing with the target value, the optimized closed-loop control algorithm is used to accurately calculate and adjust the inverter output in time.

[0173] Condition adaptation and adjustment: The elevator operating conditions are monitored in real time in all directions. When the car load reaches 80% of the rated load, the acceleration and deceleration slopes are automatically and appropriately slowed down, and the motor power output is increased to ensure smooth and safe operation. When the ambient temperature of the shaft is high, the motor heat dissipation performance changes are considered, and the load rate is reasonably reduced to ensure reliable operation of the motor.

[0174] Energy management and optimization: Equipped with an efficient energy recovery device, the regenerative electric energy generated by the motor is effectively fed back to the power grid during the deceleration and braking phase of the elevator to achieve energy recycling. Through intelligent optimization algorithms, the motor speed control strategy and energy recovery timing are planned in advance. For example, when the elevator is going down and there is no need to stop at the floor ahead, the motor operation mode is adjusted in advance to increase the energy recovery ratio and reduce the long-term operation cost of the elevator.

[0175] Elevator door machine control system

[0176] Command reception and analysis: Establish reliable communication with the intelligent decision-making layer, quickly receive various control commands of the elevator door, such as door opening commands, door closing commands and commands to set the door speed to 0.2m / s, and analyze and convert them into executable signals, while strictly verifying the command data.

[0177] Door machine drive and control: Drive the motor according to the command, use advanced vector frequency conversion control technology, accurately control the motor speed and torque, and achieve precise control of the door opening and closing speed. Combined with the door machine controller and position, speed and other sensors, according to the preset operation curve and algorithm, the motor output is adjusted in real time when the door is opened and closed to ensure that the door moves smoothly and can stop accurately at the predetermined position, ensuring seamless connection with the car and floor door.

[0178] Safety protection and fault diagnosis: Relying on sensors such as position, speed and force on the door machine, the door's operating status is monitored in all directions and in real time. When the force sensor detects an obstacle when closing the door, it quickly triggers the rebound function to prevent people or objects from being caught. If the closing speed of the door exceeds the set safety threshold, it will brake and alarm immediately. Through in-depth analysis of sensor data and comparison with normal modes, the fault type and location can be quickly diagnosed. For example, if the door machine motor is faulty, detailed fault information will be fed back to the intelligent decision-making layer in a timely manner, and a reminder will be sent to passengers in the car, "The elevator door is faulty, please wait patiently for maintenance", which will also be uploaded to the remote monitoring center.

[0179] Status feedback and communication: The operating status of the door machine is regularly fed back to the intelligent decision-making layer, including information such as position, switch status, speed, and fault conditions, so that the intelligent decision-making layer can grasp the dynamics of the door machine in real time. When the door machine is abnormal, it strengthens communication with the remote monitoring system so that maintenance personnel can quickly know the details of the fault and solve the problem efficiently.

[0180] Car environment control system

[0181] Data acquisition and command reception: Stable connection with the environmental sensors of the data acquisition layer, continuous collection of information such as temperature, humidity, light intensity, air quality and number of passengers in the cabin, and receiving ventilation, temperature, humidity, lighting and other environmental control commands issued by the intelligent decision-making layer. For example, it is detected that the temperature in the cabin is 28°C and the number of passengers is 8, and a command to adjust the temperature to 24°C is received.

[0182] Intelligent adjustment function:

[0183] Ventilation adjustment: Intelligently control the operating power and ventilation mode of the ventilation equipment according to the number of passengers and the air quality in the cabin. When the number of passengers increases or the air quality decreases, the ventilation volume is automatically increased to ensure fresh air. For example, when there are 10 passengers in the cabin, the speed of the ventilation equipment is increased by 30%. And according to the temperature conditions, the ventilation and cooling and heating collaborative modes can be flexibly switched to improve energy efficiency and create a comfortable air environment.

[0184] Temperature and humidity control: The cooling and heating systems and humidity control equipment are precisely controlled through feedback from temperature and humidity sensors. When the temperature is higher than 26°C, the cooling system is started to adjust the temperature to a comfortable range. When the humidity is abnormal, the humidifier or dehumidifier is started in time to ensure that the temperature and humidity are maintained at a comfortable level for the human body, providing passengers with a pleasant cabin climate environment.

[0185] Lighting adjustment: According to the data of the light sensor, the lighting brightness is automatically adjusted to meet the lighting requirements in the car. When the light is strong during the day, the lighting brightness is reduced; at night or on darker floors, the lighting brightness is increased. At the same time, the lighting color can be intelligently switched according to the needs of different time periods and scenes. For example, the lighting color can be switched to warm tones during holidays to create a comfortable visual environment and avoid discomfort to passengers due to strong or dark light.

[0186] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An elevator control system based on artificial intelligence, comprising a data acquisition layer, a data transmission layer, a data processing layer, an intelligent decision-making layer and an execution control layer, characterized in that: The data acquisition layer includes a car internal sensor group, a car external sensor group and an elevator operation status sensor group, wherein the car internal sensor group includes a high-definition camera, a pressure sensor, and a sound sensor, the car external sensor group includes a floor call button sensor, an infrared human body sensor, and an environmental sensor, and the elevator operation status sensor group includes a position sensor, a speed sensor, and a door status sensor; The data transmission layer includes a wired communication network and a wireless communication network; The data processing layer includes: Data preprocessing unit: performs preliminary processing such as cleaning, filtering, denoising, and format conversion on the raw data from the data acquisition layer, removes invalid values, abnormal values, and interference signals, and unifies the data format; Data Analysis and Mining Unit: Use big data analysis technology and data mining algorithms to analyze and mine pre-processed data and extract information and knowledge patterns; Artificial Intelligence Model Training and Optimization Unit: Builds artificial intelligence models based on deep learning algorithms and reinforcement learning algorithms, and uses historical and real-time data to train and optimize models, adjust parameters and structures, and evaluates and monitors model performance, improving and updating based on actual conditions; The intelligent decision-making layer includes: Elevator dispatch decision module: Based on the passenger flow forecast information, passenger distribution and current elevator operation status provided by the data analysis and mining unit, the intelligent dispatch algorithm is used to make elevator operation dispatch decisions, and the decision results are converted into control instructions and sent to the execution control layer; Safety protection and emergency decision-making module: real-time monitoring of elevator operation status data and abnormal detection information of data analysis and mining units. Once a safety hazard or emergency situation is found in the elevator, the corresponding safety protection measures and emergency decision-making plans will be immediately initiated; Personalized service decision module: Provides personalized elevator service decisions based on passenger behavior and demand analysis results; The execution control layer includes: Elevator drive control system: receives elevator operation control instructions from the intelligent decision-making layer, controls the operation of the elevator drive motor through variable frequency speed regulation technology, and realizes smooth lifting and lowering of the car; Elevator door machine control system: responsible for the opening and closing of elevator doors according to the control instructions of the intelligent decision-making layer, controlling the opening and closing speed, strength and position of the doors, taking protective measures and feeding back fault information to the intelligent decision-making layer when an abnormality occurs, and issuing prompts to passengers in the car at the same time; Car environment control system: intelligently adjusts the ventilation, lighting, temperature and humidity parameters in the car according to the environmental sensor data of the data acquisition layer and the environmental control decision instructions of the intelligent decision layer.

2. The elevator control system based on artificial intelligence according to claim 1, characterized in that: The high-definition camera captures the dynamics, expressions, and visual information of the passengers in the car in all directions; the pressure sensor measures the weight distribution in the car to determine the number of passengers, overload conditions, and position changes; the sound sensor collects ambient sound in the car; The floor call button sensor records the floor, time and direction called by the passenger; the infrared human body sensor detects the number of people in the elevator waiting area, the situation of entering and exiting the elevator car, and the length of stay; the environmental sensor monitors the environmental parameters of the elevator waiting area on each floor; The position sensor measures the real-time position of the elevator car in the shaft; the speed sensor monitors the elevator running speed in real time; the door status sensor monitors the elevator door switch status, speed, position and door lock status; The above data is transmitted to the data processing layer for analysis through the wired communication network of the data transmission layer and the wireless communication network.

3. The elevator control system based on artificial intelligence according to claim 1, characterized in that: The specific implementation method of the analysis and mining of the data analysis and mining unit includes the following steps: (1) Data integration and association Integrate various types of pre-processed data according to logic and dimensions to build a comprehensive data set so that passenger flow, elevator operation status, car environment and passenger behavior data can be correlated with each other; The improved Apriori algorithm is used to mine association rules, and information entropy is used to measure the uncertainty and activity of data. The information entropy formula is: ,in Represents a random variable The information entropy of Represents a random variable The value is The probability of Indicates an event Based on the amount of information, the data is divided into different data activity areas according to the entropy value of data distribution, and different support thresholds are adapted to different areas to mine more practical and valuable association rules between factors such as passenger flow and elevator operation, passenger behavior and car environment; (2) Time series analysis of passenger flow and operation status data The Seasonal ARIMA model is used to analyze time series data by focusing on the characteristics of the autocorrelation function and the partial autocorrelation function and automatically determining the difference order in combination with the Akaike information criterion. (3) Cluster analysis of passenger behavior data Using a fast search clustering algorithm based on density peaks, by calculating the local density of each data point : ,in is the distance between data points, Dynamically determine the shortest distance to a higher density point based on data characteristics , based on these two indicators, the cluster center is determined, the passenger behavior data is effectively clustered, different behavior categories are distinguished, and classification support is provided for personalized service decisions; (4) Abnormal detection of elevator operation data Combine the autoencoder based on reconstruction error and the long short-term memory network to carry out anomaly detection, use AE to perform unsupervised learning on normal operation data, and calculate the reconstruction error. The formula is: , is the original data, To reconstruct the data, is the number of data points, so as to determine the reconstruction error range of normal data; at the same time, the LSTM network is used to perform dynamic feature extraction and anomaly scoring on the time series data, and the reconstruction error of AE and the anomaly score of LSTM are combined to determine whether the elevator operation data is abnormal. Once an abnormality is found, it will be fed back to the safety decision module in time.

4. The elevator control system based on artificial intelligence according to claim 3, characterized in that: The specific steps of the artificial intelligence model training and optimization unit include: (1) Model construction ① Deep learning model: Passenger behavior recognition: Build an architecture that integrates convolutional neural networks and long short-term memory networks. The convolutional neural network is responsible for extracting the spatial features of the images in the car, while the long short-term memory network focuses on processing the time series information of passenger behavior. Passenger flow prediction: Build a recurrent neural network model based on the attention mechanism, use the attention mechanism to focus on different time periods and various factors that affect passenger flow, and dynamically allocate the weights of each factor through calculation; ②Reinforcement Learning Model: A deep Q network is used to build a model for elevator dispatching decision-making. The real-time operation state of the elevator is defined as the environmental state, and the available operations are set as the action space. The state-action value function is estimated through the Q network, and the optimal action is selected according to the estimated Q value to realize intelligent elevator dispatching strategy selection. (2) Model training ① Data preprocessing and division: Standardize the collected historical and real-time data, unify the data feature dimensions and distribution range, and then divide the data set into 70% training set, 15% validation set, and 15% test set; ② Training algorithm selection and optimization: Deep learning model: Use the improved algorithm of stochastic gradient descent to update parameters; Reinforcement learning model: uses a multi-step learning mechanism to accelerate the training process and update the Q value more efficiently by considering reward feedback in multiple steps in the future; (3) Model evaluation and optimization ①Performance index selection: Passenger behavior recognition model: Select accuracy, recall rate, and F1 value as key evaluation indicators to measure the accuracy of the model's passenger behavior classification, its ability to capture positive samples, and its overall performance from different perspectives; Passenger flow prediction model: Root mean square error and mean absolute error are used to quantify the degree of deviation between the predicted passenger flow and the actual passenger flow; Elevator dispatch decision model: Focuses on average passenger waiting time, average transportation time, and elevator energy consumption indicators, directly links the core efficiency and energy utilization of elevator operation, and judges the pros and cons of dispatch decisions; ②Model improvement and update: Based on the performance of the model on the test set, if it does not meet expectations, the model hyperparameters are adjusted and optimized in a targeted manner.

5. The elevator control system based on artificial intelligence according to claim 1, characterized in that: The specific implementation method of the elevator dispatch decision module is: (1) Receive passenger flow forecast information from the data analysis and mining unit and construct a two-dimensional matrix , where the elements Indicates in time slice Inner floor The predicted passenger flow; at the same time, the current position vector of each elevator is obtained , running direction vector , the number of passengers in the car vector and the current velocity vector , and the number of passengers waiting on each floor ; Perform normalization on the passenger flow data: ,in and They are The minimum and maximum passenger flow values ​​in ; normalize the elevator position data: ,in is the total number of floors; Normalize the velocity data: ,in is the maximum design speed of the elevator; (2) Dynamic zone scheduling during peak hours ①Regional division: The improved K-Means clustering algorithm is used to partition the passenger flow on each floor, and the objective function is: ,in is the number of partitions, For the The set of floors contained in a partition, is the mean passenger flow of the partition, For floor The weight of the objective function is optimized through iteration. Minimum, get the results of each partition; ② Elevator allocation and dispatching: Calculate the number of elevators that should be allocated to each zone : ,in is the average elevator inter-floor transportation time, is the total number of elevators; when there is a new call, calculate each elevator To the call floor Response time : ,in Considering the additional stop time caused by the current direction of the elevator and the destination of the passengers in the car, the calculation is based on the elevator operation rules and current status information. Minimum elevator response and planning of the order of floors to be stopped according to the passenger flow trend and elevator load in the partition; (3) Energy-saving scheduling during off-peak hours ① Sleep and wake-up strategy: Setting idle time thresholds and passenger flow threshold , when the elevator idle time exceeds And the current floor predicted passenger flow is lower than When there is a new call, calculate the sleep state of each elevator The wake-up decision value ,in For the start time of the elevator from sleep to operation, select Minimal elevator wake-up response calls; ②Operation speed optimization: According to the number of passengers in the car and current floor passenger flow Dynamically adjust the elevator speed, the speed adjustment formula is: ,in is the speed adjustment factor, For full load, is the average passenger flow; (4) Generation and sending of decision instructions The dispatching decision is converted into control instructions, including the elevator running direction, target floor and speed instructions, and the instructions are sent to the execution control layer through the elevator control system communication interface.

6. The elevator control system based on artificial intelligence according to claim 5, characterized in that: The specific implementation method of the security protection and emergency decision-making module is: (1) Data monitoring and acquisition: Connect with the elevator operation status sensor and data analysis and mining unit to collect elevator speed, position, door status, car weight data and abnormal detection information in real time, and store and update; (2) Determination of safety hazards and emergency situations: Set speed, door opening and closing time, and load safety thresholds according to elevator design standards, and combine real-time data and abnormal indicators to determine whether there are overspeed, stall, door failure, overload, and potential safety hazards; (3) Safety protection measures and emergency decision-making and execution: When overspeeding, a braking command is sent to the drive control system, the brake is activated, the car alarm is sounded and the fault information is transmitted to the remote monitoring center; When the door fails, the door machine stops running; When overloaded, the elevator is prohibited from running, the car prompts and voice broadcasts, and the operation is restored after the weight returns to normal; When there is a potential serious hidden danger, recall the elevator to the refuge floor for processing.

7. The elevator control system based on artificial intelligence according to claim 6, characterized in that: The specific implementation method of the personalized service decision module is: (1) Data collection and analysis: Collect passenger behavior and demand data, organize and analyze them, and store them in a standardized format; (2) Service decision making: Behavior-triggered services: automatically generate corresponding service instructions after identifying specific behaviors; Habitual preference service: Provides priority dispatch, automatic parking and personalized greeting services at set times or when approaching specific floors based on historical habits; Real-time demand response service: Receive and process passengers' real-time demands, incorporate them into the dispatch plan and provide timely feedback; (3) Service feedback and optimization: Collect passenger feedback, optimize service strategies based on the feedback information, adjust parameter settings, and evaluate service effectiveness.

8. The elevator control system based on artificial intelligence according to claim 1, characterized in that: The control method of the elevator drive control system is: (1) Instruction processing: Establish a reliable communication link with the intelligent decision-making layer, receive elevator operation control instructions, including speed, position, direction parameters and special working condition instructions, and verify the instructions at the same time. Once data abnormality or error is found, immediately feedback to the intelligent decision-making layer and request to resend the correct instructions; (2) Motor drive and speed control: Relying on variable frequency speed regulation technology to drive the elevator motor, with vector control algorithm as the core, the motor current is accurately decomposed into two components: torque and excitation, to achieve independent regulation; (3) Adjustment to suit the working conditions: Real-time monitoring of elevator operating conditions, including car load, number of passengers, current location, and shaft environmental parameters, and automatic adjustment of motor output power and speed based on these real-time data; (4) Energy management and optimization: Equipped with an energy recovery device, during the elevator deceleration and braking phase, the regenerative electrical energy generated by the motor is effectively fed back to the power grid to achieve energy recycling.

9. An artificial intelligence-based elevator control system according to claim 8, characterized in that: The control method of the elevator door machine control system is: (1) Command reception and analysis: Establish reliable communication with the intelligent decision-making layer, receive various control instructions of the elevator door, analyze and convert them into executable signals, and strictly verify the instruction data; (2) Door machine drive and control: Drive the motor according to the instructions, and use frequency conversion control technology to control the speed and torque of the motor; (3) Safety protection and fault diagnosis: The door machine relies on the position, speed and force sensors to monitor the door's operating status. At the same time, by analyzing the sensor data and comparing it with the normal mode, the fault type and location can be quickly diagnosed, and detailed fault information can be fed back to the intelligent decision-making layer in a timely manner. Passengers will be prompted in the car and the information will be uploaded to the remote monitoring center. (4) Status feedback and communication: The operating status of the gantry crane is regularly fed back to the intelligent decision-making layer, so that the intelligent decision-making layer can grasp the dynamics of the gantry crane in real time.

10. An elevator control system based on artificial intelligence according to claim 9, characterized in that: The control method of the car environment control system is: Data acquisition and command reception: Stably connect with the environmental sensors of the data collection layer to continuously collect information on temperature, humidity, light intensity, air quality, and number of passengers in the car, while receiving environmental control instructions from the intelligent decision-making layer; Intelligent adjustment function: Ventilation adjustment: intelligently control the operating power and ventilation mode of ventilation equipment according to the number of passengers and the air quality in the cabin; Temperature and humidity control: Control the cooling and heating system and humidity control equipment through feedback from temperature and humidity sensors; Lighting adjustment: Automatically adjust lighting brightness based on light sensor data.

Images