Elevator intelligent air curtain and semiconductor refrigeration control method

Abstract

The invention provides an elevator intelligent air curtain and a semiconductor refrigeration control method. The method comprises the steps that the historical temperature and airflow distribution data in an elevator are obtained; according to historical temperature and airflow distribution data, training a temperature and airflow distribution prediction model by adopting a long-short-term memory network; acquiring real-time temperature and airflow distribution data in the elevator, inputting the temperature and airflow distribution prediction model, and predicting future temperature and airflow distribution; according to the predicted temperature and airflow distribution, an optimization algorithm is adopted to calculate the opening angle of the air curtain and the refrigeration power parameter of the refrigeration cold piece, and the optimal control strategy of the air curtain and the refrigeration cold piece is obtained with the temperature control precision and the system energy consumption as optimization targets; and parameters in the optimal control strategy are input into a controller of the air curtain and the refrigerating fins, the air speed and the air direction of the air curtain and the refrigerating power of the refrigerating fins are adjusted, and optimal control over the temperature and airflow distribution in the elevator is achieved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an elevator intelligent air curtain and a semiconductor refrigeration control method. Background Art

[0002] In the intelligent air curtain and semiconductor refrigeration control of elevators, it is necessary to monitor the temperature distribution data in the elevator in real time and analyze and obtain the airflow distribution data in the elevator. The airflow distribution data includes the average wind speed distribution in different areas, the volume proportion of the airflow short-circuit area, the volume proportion of the airflow dead zone, and the distribution of the airflow short-circuit area. The airflow short-circuit area refers to the phenomenon that the airflow flows rapidly in certain areas, resulting in the failure of the airflow to fully cover or mix other areas. The airflow dead zone refers to the area where the airflow cannot flow effectively or stagnates. Its impact includes local temperature anomalies, resulting in the inability to effectively adjust the temperature of the local area, and humidity problems, which may cause the humidity to be too high or too low, affecting the air quality and comfort. According to the airflow distribution data and the temperature and humidity data in the elevator, the air curtain and refrigeration plate need to be dynamically adjusted to achieve the best temperature control effect. In this process, there are several technical contradictions and difficulties. First, the internal space of the elevator is small, and the temperature and airflow distribution are complex. How to arrange sensors in a limited space to obtain accurate temperature and airflow distribution data is a technical challenge. Secondly, the adjustment of the air curtain and refrigeration plate needs to be dynamically optimized according to the real-time temperature and airflow distribution data. How to design an efficient control algorithm to achieve fast response and precise control is also a difficulty. Finally, during the operation of the elevator, the opening and closing of the door and the entry and exit of people will affect the temperature and airflow distribution. How to maintain a stable temperature control effect in a dynamically changing environment is also a technical problem that needs to be solved. Summary of the invention

[0003] The present invention provides an elevator intelligent air curtain and semiconductor refrigeration control method, which mainly includes:

[0004] Obtain historical temperature and airflow distribution data inside the elevator, wherein the airflow distribution data includes average wind speed distribution in different areas, volume proportion of airflow short-circuit areas, and volume proportion of airflow dead areas;

[0005] According to the historical temperature and airflow distribution data, a temperature and airflow distribution prediction model is trained using a long short-term memory network;

[0006] Acquire real-time temperature and airflow distribution data inside the elevator, input the temperature and airflow distribution prediction model, and predict future temperature and airflow distribution;

[0007] According to the predicted temperature and airflow distribution, the optimization algorithm is used to calculate the opening angle of the air curtain and the cooling power parameters of the refrigeration fins. The temperature control accuracy and system energy consumption are taken as the optimization goals to obtain the optimal control strategy of the air curtain and refrigeration fins.

[0008] The parameters in the optimal control strategy are input into the controllers of the air curtain and refrigeration fins to adjust the wind speed and direction of the air curtain and the cooling power of the refrigeration fins to achieve optimal control of the temperature and airflow distribution inside the elevator.

[0009] Furthermore, before obtaining the historical temperature and airflow distribution data inside the elevator, the method further includes:

[0010] The moving average method is used to fill in the missing values ​​in the historical temperature and airflow distribution data, and the outliers in the historical temperature and airflow distribution data are eliminated by the 3σ principle to obtain a standardized historical temperature and airflow distribution data set.

[0011] Furthermore, the optimization algorithm is used to calculate the air curtain opening angle and the cooling power parameters of the refrigeration plate according to the predicted temperature and airflow distribution, including:

[0012] The NSGA-II algorithm is used to establish a multi-objective optimization model for air curtain and refrigeration fin adjustment, with temperature control accuracy and system energy consumption as optimization objectives. The multi-objective optimization model is solved by an adaptive particle swarm optimization algorithm to obtain the optimal parameter combination of air curtain opening angle, wind speed, swing frequency and refrigeration fin cooling power.

[0013] Furthermore, after inputting the parameters in the optimal control strategy into the controllers of the air curtain and the refrigeration fins, the method further includes:

[0014] A distributed sensor network is used to obtain the actual temperature and airflow distribution data inside the elevator, and the actual temperature and airflow distribution data are processed by a Kalman filter algorithm; it is determined whether the processed actual temperature and airflow distribution data deviate from the predicted value by more than a set threshold. If the deviation exceeds the set threshold, retraining of the temperature and airflow distribution prediction model is triggered.

[0015] Furthermore, while obtaining the real-time temperature and airflow distribution data inside the elevator, it also includes:

[0016] The opening and closing status of the elevator door is detected by the access control sensor; if the elevator door is detected to be open, the wind speed, wind direction of the air curtain and the cooling power of the refrigeration plate are dynamically adjusted according to the duration of the door opening and the temperature and airflow changes before and after the door opening.

[0017] Furthermore, before calculating the air curtain opening angle and the refrigeration power parameters of the refrigeration fins using an optimization algorithm according to the predicted temperature and airflow distribution, the method further includes:

[0018] The geometric parameters and material property data of the air curtain and the refrigeration fins are obtained; a vibration model of the air curtain and the refrigeration fins is established according to the geometric parameters and material property data; the natural frequency, vibration mode and frequency response function of the vibration model are calculated through modal analysis and harmonic response analysis; the material parameters and structural dimensions of the air curtain and the refrigeration fins are optimized by using a genetic algorithm, with minimizing the maximum vibration amplitude of the vibration model as the objective function, to obtain the optimal material selection and structural design scheme of the air curtain and the refrigeration fins.

[0019] Furthermore, after obtaining the real-time temperature and airflow distribution data inside the elevator, the method further includes:

[0020] An infrared human body detector is used to obtain the data on the number of people in the elevator. According to the data on the number of people and the real-time temperature and airflow distribution data, the correlation model between the density of people and the changes in the temperature field and the airflow field is calculated. A people density influence controller based on fuzzy reasoning is designed, and the wind curtain speed and the power of the refrigeration plate are dynamically adjusted according to the correlation model, and the membership function parameters of the fuzzy controller are optimized by the gradient descent method.

[0021] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0022] The present invention discloses an intelligent wind curtain and semiconductor refrigeration control method for elevators. Accurate monitoring of temperature and airflow distribution is achieved. A prediction model is established based on historical data analysis to provide a basis for pre-adjustment of wind curtains and refrigeration fins. An innovative combination of airflow detection instruments, temperature sensors and sound level meters is used to identify airflow short-circuit areas, and a thermometer, hygrometer and anemometer are used to determine airflow dead zones. The material and structural design of wind curtains and refrigeration fins are optimized according to the vibration model to effectively reduce the vibration level. Sensors are arranged near the elevator doors to monitor environmental changes when the doors are opened and closed in real time, and equipment parameters are adjusted dynamically. Finally, a strategy based on fuzzy control is established to dynamically adjust the wind curtain wind speed and refrigeration fin power according to factors such as the number of people, thereby realizing intelligent and precise control of the elevator internal environment and improving passenger comfort and system energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The present invention is a flow chart of an elevator intelligent air curtain and semiconductor refrigeration control method.

[0024] Figure 2 It is a schematic diagram of an elevator intelligent air curtain and semiconductor refrigeration control method of the present invention.

[0025] Figure 3 This is another schematic diagram of an elevator intelligent air curtain and semiconductor refrigeration control method of the present invention. DETAILED DESCRIPTION

[0026] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0027] like Figure 1-3 In this embodiment, an elevator intelligent air curtain and semiconductor refrigeration control method may specifically include:

[0028] S101. According to the elevator's internal space structure and temperature and airflow distribution characteristics, three-dimensional modeling is performed on the elevator's internal space to simulate the temperature and airflow distribution under different sensor layout schemes. By comparing the simulation results of different schemes, the sensor layout scheme is determined.

[0029] A three-dimensional grid model is constructed according to the size and structural characteristics of the elevator's internal space, and the three-dimensional grid model is used to refine the elevator space; the temperature field and velocity field simulation is performed in ANSYS Fluent software using the three-dimensional grid model to obtain a temperature distribution cloud map and an airflow velocity vector map in the elevator space; the key areas in the elevator where the temperature gradient is greater than a preset temperature gradient threshold or the airflow velocity is greater than a preset velocity threshold are identified through the temperature distribution cloud map and the airflow velocity vector map; temperature sensors and airflow sensors are arranged in the key areas, wherein the spacing between the temperature sensors and the airflow sensors is not less than the preset spacing threshold, and the total number does not exceed the preset number threshold; multiple simulation calculations are performed for multiple arrangement schemes of the temperature sensors and the airflow sensors, the position parameters of the temperature sensors and the airflow sensors are adjusted, the relative error between the measured value of each position of the temperature sensor and the airflow sensor and the average value of the grid points within the surrounding preset range is calculated, and the temperature sensor and airflow sensor arrangement scheme with the smallest relative error is selected as the final scheme.

[0030] For example, a three-dimensional grid model is constructed according to the size and structural characteristics of the elevator's internal space, the elevator space is divided into fine parts, and the appropriate grid unit size is selected according to the size of the elevator to form a digital expression of the elevator space. The temperature field and velocity field of the elevator's internal space are simulated using ANSYS Fluent software. The elevator wall temperature is set to a constant value, the inlet boundary condition is the velocity inlet, and the outlet boundary condition is the pressure outlet. The temperature distribution cloud map and airflow velocity vector map in the elevator space are calculated. Through the analysis of the simulation results, the key areas in the elevator where the temperature gradient is greater than 5°C / m or the airflow velocity is greater than 0.5m / s are identified, and temperature sensors and airflow sensors are arranged in these areas. The sensor spacing is not less than 30cm, and the total number does not exceed 8. For multiple sensor layout schemes, multiple simulation calculations are performed, the sensor position parameters are adjusted, and the relative error between the measured value of each sensor position and the average value of the grid points within the surrounding 1m range is calculated. The sensor layout scheme with the smallest relative error is selected as the final scheme. The root mean square error between the sensor measurement value and the overall temperature field and velocity field under different schemes is compared, and the sensor layout scheme with the highest measurement accuracy is selected.

[0031] The digital expression of the elevator space is modeled using SolidWorks software. Taking the standard elevator size of 2m×2m×2.5m as an example, a grid unit size of 10cm×10cm×10cm is selected for space division to generate a 20×20×25 three-dimensional grid model. ANSYS Fluent software is used to simulate the temperature field and velocity field. The elevator wall temperature is set to 25℃, the inlet boundary condition is the velocity inlet, the velocity is 0.5m / s, and the outlet boundary condition is the pressure outlet, and the pressure is standard atmospheric pressure. By solving the Navier-Stokes equation and the energy equation, the temperature distribution cloud map and airflow velocity vector map in the elevator space are obtained. The simulation results are analyzed to identify the key areas where the temperature gradient is greater than 5℃ / m or the airflow velocity is greater than 0.5m / s. Eight sensors are arranged in these areas, including 4 temperature sensors and 4 airflow sensors, and the sensor spacing is set to 50cm. For multiple sensor arrangement schemes, 10 simulation calculations are performed, and the sensor position parameters are adjusted each time, with a range of ±10cm. The relative error between the measured value of each sensor position and the average value of the grid points within 1m around it was calculated, and the least square method was used to fit the error curve. The sensor layout scheme with the smallest relative error was selected as the final scheme. The root mean square errors of the sensor measurements and the overall temperature field and velocity field under different schemes were compared, and the sensors were divided into a temperature sensor group and an airflow sensor group using the K-means clustering algorithm. The average root mean square error of each group of sensors was calculated, and the sensor layout scheme with the smallest sum of the average root mean square errors of the two groups was selected as the optimal scheme.

[0032] S102. By analyzing the historical temperature and airflow distribution data, the airflow distribution data includes the average wind speed distribution in different areas, the volume ratio of the airflow short-circuit area, the volume ratio of the airflow dead zone, and the airflow short-circuit area distribution, a prediction model for temperature and airflow distribution is established. Based on the temperature and airflow distribution data obtained in real time, the prediction model is used to predict the temperature and airflow distribution, so as to provide a basis for the pre-adjustment of the air curtain and the refrigeration plate.

[0033] According to the acquired historical temperature and airflow distribution data, the moving average method is used to fill in the missing values, and the outliers are eliminated by the 3σ principle to obtain a standardized historical data set; based on the standardized historical data set, the average wind speed distribution in different areas is determined, if the airflow speed is less than 0.1m / s, it is determined to be an airflow dead zone, and if the airflow speed is greater than 1.5m / s, it is determined to be an airflow short-circuit area; the temperature and airflow distribution prediction model is trained using a long short-term memory network, and the airflow distribution prediction model is obtained by cross-validation training; the real-time temperature and airflow distribution data inside the elevator are obtained from the temperature sensor and the airflow sensor, and the real-time temperature and airflow distribution data are pre-processed and then input into the airflow distribution prediction model; the airflow distribution prediction model is used to predict the future temperature and airflow distribution, and the air curtain opening angle and the cooling power of the refrigeration plate are calculated according to the prediction results, and the calculation process adopts a PID control algorithm.

[0034] Exemplarily, the historical temperature and airflow distribution data are preprocessed by data cleaning method, the missing values ​​are filled by moving average method, the outliers are judged and eliminated by 3σ principle, and the data are normalized to obtain a standardized historical data set. According to the preprocessed historical data, the average wind speed distribution in different areas is calculated, and the area with airflow speed less than 0.1m / s is defined as the airflow dead zone, and the area with airflow speed greater than 1.5m / s is defined as the airflow short-circuit area. The volume proportion and distribution of the airflow short-circuit area and the airflow dead zone are calculated. A prediction model for temperature and airflow distribution is constructed by long short-term memory network, and the data set is divided into training set and validation set. The model is trained and validated by cross-validation method. The temperature and airflow distribution data inside the elevator are obtained in real time from the temperature sensor and airflow sensor, and the real-time data is preprocessed and feature extracted, and the processed data is input into the prediction model. The prediction model is used to predict the temperature and airflow distribution in the future period, and the target values ​​of temperature and airflow speed are set. The opening angle of the air curtain and the cooling power of the refrigeration plate are calculated by PID control algorithm, and the adjustment parameters of the air curtain and the refrigeration plate are obtained according to the prediction results.

[0035] When preprocessing the historical temperature and airflow distribution data, the moving average method of the 5-minute time window is used to fill the missing values, and the 3σ principle is used to judge the outliers. The data beyond the range of ±3 standard deviations of the mean value is considered abnormal and eliminated, and then the data is normalized by min-max to obtain a standardized historical data set. When calculating the average wind speed distribution in different areas, the elevator space is divided into 27 sub-areas of 3×3×3, and the average wind speed of each sub-area is obtained by the arithmetic mean of the wind speeds of all sampling points in the area. The area with an airflow speed less than 0.1m / s is defined as the airflow dead zone, and the area with an airflow speed greater than 1.5m / s is defined as the airflow short-circuit area. The volume proportion and spatial distribution of these areas are calculated. The long short-term memory network (LSTM) is used to construct a prediction model for temperature and airflow distribution. The network structure contains an LSTM layer with 64 neurons and a fully connected output layer. The data set is divided into a training set and a validation set in a ratio of 8:2. The model is trained and validated using the 5-fold cross-validation method, and the batch size of each epoch is set to 32. The temperature and airflow distribution data inside the elevator are obtained from the temperature sensor and airflow sensor in real time, with a sampling frequency of once per second. The real-time data is preprocessed and feature extracted in the same way as the historical data, and the processed data is input into the prediction model. The prediction model is used to predict the temperature and airflow distribution in the next 10 minutes, with the temperature target value set at 25°C and the airflow velocity target value set at 0.3m / s. The air curtain opening angle and the cooling power of the refrigeration plate are calculated through the PID control algorithm. The PID parameters are determined by the Ziegler-Nichols tuning method, and the specific values ​​of the air curtain opening angle and the cooling power of the refrigeration plate are finally obtained.

[0036] Use an airflow detection instrument to monitor the flow direction of the airflow. If it flows quickly from the air inlet to the air outlet, a short circuit occurs. Based on the existence of the short circuit phenomenon, further use temperature sensors to detect the temperatures of different areas, obtain areas with abnormal temperatures, and preliminarily determine the airflow short-circuit area. Based on the characteristic that the airflow short-circuit area causes increased airflow noise, use a sound level meter to monitor the noise level and identify the airflow short-circuit area.

[0037] A hot wire anemometer is used to monitor the airflow direction of the air inlet and outlet. The airflow velocity and airflow direction data are recorded by setting multiple monitoring points. It is determined whether there is an airflow short-circuit phenomenon based on the airflow velocity and the airflow direction data. According to the result of the airflow short-circuit phenomenon, the temperature sensor array is started to detect the temperature of different areas, and the temperature sensor array dynamically adjusts the grid division according to the size of the space. The temperature abnormality area is identified by temperature data analysis, and the area where the temperature deviation exceeds the set threshold is marked as a potential airflow short-circuit area. The temperature distribution thermodynamic map is generated by the Kriging interpolation algorithm. A sound level meter is arranged for the potential airflow short-circuit area, the noise level is monitored in real time, the noise data is obtained and the spectrum analysis is performed. If the noise level exceeds the preset threshold and the spectrum analysis result meets the airflow noise characteristics, the potential airflow short-circuit area is determined to be an airflow short-circuit area.

[0038] For example, a hot wire anemometer is used to monitor the direction of air flow. Multiple monitoring points are set at the air inlet and outlet to record the air flow speed and direction data. If the air flow speed is detected to be more than 2m / s and the direction is from the air inlet to the air outlet, it is determined that there is a short circuit. Based on the short circuit determination result, the temperature sensor array is activated to detect the temperature of different areas. The grid division is dynamically adjusted according to the size of the space, and the grid is divided into 50 grids every 5m. 3 Set up a temperature sensor to collect temperature data and compare it with the preset temperature threshold. Identify temperature anomaly areas through temperature data analysis, mark areas where the temperature deviation exceeds the set threshold as potential airflow short-circuit areas, and use the Kriging interpolation algorithm to generate a continuous temperature distribution heat map for visual display of temperature distribution. Pass the temperature anomaly area information to the sound level meter layout system, arrange sound level meters in the identified potential airflow short-circuit areas, monitor the noise level in real time, collect noise data and perform spectrum analysis. Take the average noise level during normal operation plus 5dB as the initial threshold, and dynamically adjust the threshold according to real-time data. If the noise level exceeds the threshold and the spectrum analysis results meet the airflow noise characteristics, the area is confirmed to be an airflow short-circuit area, and an airflow short-circuit area distribution map is generated. At a 200m 3In the space, a TSI9565 hot wire anemometer was used to set 5 monitoring points at the air inlet and outlet, and data was collected once a second. When any monitoring point detected an airflow velocity exceeding 2m / s and the direction was from the air inlet to the air outlet in three consecutive samplings, a short circuit judgment was triggered. Subsequently, 40 SHT31 temperature sensors were arranged according to the size of the space to form a 4×4×2.5 three-dimensional grid. Each sensor collected temperature data every 5 seconds and compared it with the preset 25℃ standard temperature. The kriging function in Python's scipy library was used to interpolate the temperature data to generate a 100×100×62.5 high-precision temperature distribution thermodynamic map. Areas with temperature deviations exceeding 3℃ were marked as potential airflow short-circuit areas, and the coordinate information of these areas was transmitted to the sound level meter layout system. A Brüel&Kjaer2250 sound level meter was placed in each potential short-circuit area, and the sampling frequency was set to 48kHz. The collected sound signals were spectrally analyzed using Fast Fourier Transform (FFT), focusing on the spectral features in the range of 100Hz-1kHz. The initial threshold was set to 50dB by measuring the average noise level during normal operation, such as 45dB. The threshold was dynamically adjusted using the exponential moving average algorithm, with the weight factor set to 0.1. When the noise level exceeded the threshold and the spectral energy in the range of 100Hz-1kHz accounted for more than 60%, it was confirmed as an airflow short-circuit area. Finally, the matplotlib library was used to draw a three-dimensional scatter plot to show the spatial distribution of the airflow short-circuit area.

[0039] At the same time, a thermometer and hygrometer are used to monitor the humidity and temperature of different areas in the elevator, and an anemometer is used to measure the air flow velocity in each area of ​​the elevator to obtain the temperature, humidity and air flow velocity data of different areas. The temperature, humidity and air flow velocity of each different area are judged by the preset air flow dead zone judgment rules. If the humidity in some areas is higher than the preset humidity threshold and the temperature is lower than the preset temperature threshold, and the air flow velocity is lower than the preset value, it is judged as an air flow dead zone.

[0040] The grid division scheme is dynamically adjusted according to the elevator size information, and a temperature and humidity sensor and a wind speed sensor are set in each divided grid unit, and the sensor collects temperature data, humidity data and wind speed data in real time; the temperature data, humidity data and wind speed data collected by the sensor are received, and the temperature data, humidity data and wind speed data are preprocessed by a data acquisition module, and the preprocessed data are filtered by a Kalman filter to obtain a filtered measurement result; the filtered measurement result is judged according to a preset airflow dead zone judgment rule, wherein the judgment includes: obtaining a preset humidity threshold, temperature threshold and wind speed threshold, judging whether the filtered measurement result meets the airflow dead zone condition, and if the airflow dead zone condition is met, marking the corresponding grid unit as a potential airflow dead zone and recording its confidence; receiving potential airflow dead zone information and its confidence, and transmitting the potential airflow dead zone information and its confidence to a cluster analysis module; using a DBSCAN algorithm to perform spatial clustering analysis on the potential airflow dead zone, merging adjacent potential airflow dead zones to form an airflow dead zone, and generating a three-dimensional distribution map of the airflow dead zone.

[0041] For example, multiple temperature and humidity sensors and wind speed sensors are installed inside the elevator, and the grid division is dynamically adjusted according to the size of the elevator, with each 1m 3A group of sensors are set up to collect temperature, humidity and wind speed data in real time, and the sampling frequency is set to 1 time per second. The data acquisition module is used to preprocess the temperature, humidity and wind speed data collected by the sensor, including removing outliers and smoothing data. The data is filtered by the Kalman filter to obtain stable measurement results. According to the preset airflow dead zone judgment rule, the humidity threshold is set to 70%, the temperature threshold is set to 20℃, and the wind speed threshold is set to 0.1m / s. The data of each grid area is judged, and the fuzzy logic is introduced to deal with the situation close to the threshold, and the confidence of the judgment result is calculated. If the humidity in a certain area is higher than 70% and the temperature is lower than 20℃, and the wind speed is lower than 0.1m / s, the area is marked as a potential airflow dead zone, and its confidence is recorded. The potential airflow dead zone information and its confidence are transmitted to the clustering analysis module, and the DBSCAN algorithm is used to perform spatial clustering analysis on the marked potential airflow dead zones, merge adjacent potential airflow dead zones, form a larger range of airflow dead zones, and generate a three-dimensional distribution map of the airflow dead zone, marking the location, size and shape of the dead zone. The judgment threshold is dynamically adjusted according to the clustering results to improve the accuracy of the judgment. In a standard elevator of 2m×2m×3m, 12 groups of SHT31-D temperature and humidity sensors and TSI9565 wind speed sensors are installed to form a 2×2×3 three-dimensional grid. Each group of sensors collects data once a second and transmits it to the central processor through the RS-485 bus. The Kalman filter is used to process the raw data, and the process noise covariance Q=0.001, the measurement noise covariance R=0.1, and the initial estimation error covariance P=1 are set. The filtered data is used to judge the airflow dead zone, and the Mamdani fuzzy inference system is used to deal with the situation close to the threshold. Three fuzzy sets are set: humidity (low, medium, high), temperature (low, medium, high), and wind speed (low, medium, high). Fuzzy rules are defined, such as "IF humidity is high AND temperature is low AND wind speed is low THEN the possibility of dead zone is high". The centroid method is used for defuzzification to obtain the possibility of dead zone, ranging from 0-1. When the possibility is greater than 0.8, it is marked as a potential airflow dead zone. The three-dimensional coordinates and probability values ​​of the potential airflow dead zone are passed into the DBSCAN algorithm, and the neighborhood radius eps = 0.5m and the minimum number of points minPts = 3 are set. DBSCAN outputs clustering results, including the center coordinates and range of each cluster. The Matplotlib library is used to generate a three-dimensional scatter plot, where the X, Y, and Z axes represent length, width, and height, respectively. Different colors are used to represent clusters, and the size of the points represents the possibility of dead zones. According to the clustering results, if the average value of the dead zone possibility in a certain area exceeds 0.9, the humidity, temperature, and wind speed thresholds of the area are adjusted by 5% respectively for the next round of judgment.

[0042] S103. According to the prediction model of temperature and airflow distribution, the adjustment of the air curtain and the refrigeration fins is optimized, with the temperature control accuracy and system energy consumption as the optimization targets, and the optimal adjustment strategy of the air curtain and the refrigeration fins is obtained through optimization solution.

[0043] According to the two sub-goals of temperature control accuracy and system energy consumption, the NSGA-II algorithm is used to establish the optimization objective function of the air curtain and refrigeration plate adjustment. The opening angle of the air curtain, the wind speed, the swing frequency and the range of the refrigeration power of the refrigeration plate are obtained to determine the decision variables of the optimization problem. The optimization objective function is solved by the adaptive particle swarm optimization algorithm, which includes setting the initial population size and the maximum number of iterations, and introducing adaptive inertia weights and learning factors. The optimization results are simulated and verified by ANSYS Fluent software to obtain the temperature and airflow distribution effects of the optimization parameters in the elevator environment. If the optimization parameters are verified, the optimal adjustment parameters are input into the controller of the air curtain and refrigeration plate, and the actual temperature and airflow distribution data are obtained by using a distributed sensor network, and the actual temperature and airflow distribution data are processed by the Kalman filter algorithm. It is judged whether the processed actual temperature and airflow distribution data deviate from the predicted value by more than the set threshold. If the deviation exceeds the set threshold, the re-optimization process is triggered.

[0044] Exemplarily, based on the temperature and airflow distribution data output by the prediction model, the optimization objective function of the wind curtain and refrigeration plate adjustment is established, and the NSGA-II algorithm is used to process the two sub-goals of temperature control accuracy and system energy consumption to better handle the nonlinear relationship and trade-off between the goals. The parameter value range is determined according to the elevator specifications and environmental conditions, and the value range of the wind curtain adjustment parameters is set, including the wind curtain opening angle, ranging from 0-90 degrees, the wind speed, ranging from 0-10m / s, and the swing frequency, ranging from 0-2Hz. At the same time, the refrigeration power adjustment range of the refrigeration plate is set within 0-1000W, and these parameters are used as decision variables for the optimization problem. The objective function is solved by the adaptive particle swarm optimization algorithm, the initial population size is set to 50, the maximum number of iterations is 100, the adaptive inertia weight and learning factor are introduced, and the parameters are dynamically adjusted according to the optimization process. The optimization results are simulated and verified using ANSYSFluent software to simulate the temperature and airflow distribution effects of the optimization parameters in the elevator environment. The verified optimal adjustment parameters are input into the controllers of the air curtain and refrigeration plate to perform the corresponding adjustment actions. At the same time, the distributed sensor network and Kalman filter algorithm are used to monitor the actual temperature and airflow distribution in real time. If the deviation from the predicted value exceeds the set threshold, the re-optimization process is triggered. Based on the actual operation data, the prediction model is continuously updated and fine-tuned using an online learning algorithm to improve the prediction accuracy of the model. In a standard elevator of 2m×2m×3m, the output temperature distribution range of the prediction model is 18-28℃, and the airflow velocity distribution is 0-1.5m / s. The NSGA-II algorithm is used to construct a dual-objective optimization problem. The temperature control accuracy objective function is Σ(Ti-Tset)2 / n, where Ti is the temperature of each measuring point, Tset is the set temperature of 25℃, and n is the number of measuring points; the system energy consumption objective function is Σ(Pf+Pc), where Pf is the air curtain power and Pc is the fin power. The parameter range of the wind curtain is set as: opening angle 0-90°, wind speed 0-10m / s, swing frequency 0-2Hz; the cooling power range of the cooling plate is 0-1000W. In the adaptive particle swarm optimization algorithm, the inertia weight w decreases linearly from 0.9 to 0.4, and the initial values ​​of the learning factors c1 and c2 are both 2, which are dynamically adjusted according to the degree of improvement of each generation of the optimal solution. The elevator CFD model is established using ANSYS Fluent software, with a grid number of 1 million, a k-ε turbulence model, and boundary conditions set as: wall adiabatic, inlet as velocity inlet, and outlet as pressure outlet. The optimal parameter combination obtained by optimization is: wind curtain opening angle 75°, wind speed 8m / s, swing frequency 1.5Hz, and cooling plate power 800W. 20 SHT31-D temperature and humidity sensors and TSI9565 wind speed sensors are arranged in the elevator, and a 5-state Kalman filter is used to fuse multi-sensor data. The observation noise covariance R is set to a diagonal matrix diag(0.1,0.1,0.01).If the measured temperature deviates from the predicted value by more than ±1°C or the airflow velocity deviation exceeds ±0.3m / s, re-optimization is triggered. The least squares method of the sliding time window is used, with a window length of 24 hours. The prediction model parameters are updated every hour to improve the model's adaptability to environmental changes.

[0045] S104. Obtain the physical structure of the wind curtain and the refrigeration fins, analyze their vibration characteristics, which include natural frequency, damping characteristics, and vibration amplitude, establish a vibration model of the wind curtain and the refrigeration fins, and use the vibration characteristics of natural frequency, damping characteristics, and vibration amplitude as model inputs. According to the output results of the vibration model, analyze the impact of different materials and structural design schemes on vibration, and reduce the vibration level of the wind curtain and the refrigeration fins by optimizing material selection and structural design.

[0046] A three-dimensional scanner is used to obtain a three-dimensional model of an air curtain and a refrigeration fin, and the scanning result is processed by computer-aided design software to obtain geometric parameters of the air curtain and the refrigeration fin; according to the geometric parameters, the elastic modulus and loss factor of the air curtain and the refrigeration fin are measured by a dynamic mechanical analyzer to obtain material property data of the air curtain and the refrigeration fin; the geometric parameters and material property data are input into a finite element analysis software to establish a vibration model of the air curtain and the refrigeration fin, and the natural frequency and vibration mode of the vibration model are obtained by modal analysis calculation; a harmonic response analysis is performed on the vibration model, excitations of different frequencies and amplitudes are input, the vibration response of the vibration model at each frequency is calculated, and the frequency response function and vibration amplitude distribution of the vibration model are obtained; according to the frequency response function and vibration amplitude distribution, a genetic algorithm is used to optimize the material parameters and structural dimensions of the air curtain and the refrigeration fin, and if the population size of the genetic algorithm is a preset value, and the crossover probability and mutation probability of the genetic algorithm are preset probabilities, then the maximum vibration amplitude of the vibration model is minimized as the objective function to obtain the optimized material selection and structural design scheme of the air curtain and the refrigeration fin.

[0047] Exemplarily, a 3D scanner is used to obtain an accurate 3D model of the wind curtain and refrigeration fins, and the scanning results are processed by computer-aided design software to obtain detailed geometric parameters. The elastic modulus and loss factor of the material are measured by a dynamic mechanical analyzer to obtain complete material property data. The vibration model of the wind curtain and refrigeration fin is established by finite element analysis software, and the geometric parameters and material properties are used as inputs. The natural frequency and vibration mode of the system are calculated by modal analysis. The modal damping ratio is determined by the half-power bandwidth method to improve the input parameters of the vibration model. Harmonic response analysis is performed to estimate the excitation force under working conditions through actual measurements or empirical formulas, input excitations of different frequencies and amplitudes, calculate the vibration response of the system at each frequency, and obtain the frequency response function and vibration amplitude distribution. The vibration response results are evaluated, and the maximum displacement and acceleration are defined as evaluation indicators. Based on the output results of the vibration model, the genetic algorithm is used to optimize the material parameters and structural dimensions. The population size is set to 100, real number coding is used, the crossover probability is 0.8, the mutation probability is 0.1, and the maximum vibration amplitude of the system is minimized as the objective function. At the same time, weight and cost constraints are considered. After 500 generations of iteration, the optimized material selection and structural design scheme are obtained. The optimized design scheme is verified and analyzed, and the vibration performance before and after optimization is compared to confirm the degree of reduction in vibration level. Taking a typical 1.5m×0.5m×0.3m wind curtain as an example, a FAROFocusS350 laser scanner is used for three-dimensional scanning with a resolution of 1mm to obtain point cloud data containing 5 million points. The point cloud is processed by GeomagicDesignX software to generate an accurate CAD model. The elastic modulus of the main material of the wind curtain is determined by TAInstrumentsQ800 dynamic mechanical analyzer, which is assumed to be ABS plastic, set to 2.3GPa, and the loss factor is set to 0.02. Import the CAD model into ANSYSWorkbench, set the material properties, divide the mesh, the unit type is SOLID186, the mesh size is 2mm, and the total number of units is about 1 million. Perform modal analysis and calculate the first 10 natural frequencies, ranging from 20-500Hz, and vibration modes. Use the half-power bandwidth method to calculate the modal damping ratio, and the result is 0.5%-2%. In the harmonic response analysis, according to the fan speed, it is set to 3000rpm, the estimated excitation force frequency is 50Hz, the amplitude is 10N, and the sweep range is set to 0-1000Hz. The evaluation results show that the maximum displacement at 50Hz is 0.5mm and the maximum acceleration is 5m / s 2During the genetic algorithm optimization process, the wall thickness of the wind curtain, ranging from 2-5mm, and the size of the supporting structure, ranging from 10-30mm, were encoded, and the fitness function was the inverse of the maximum vibration amplitude. After 500 generations of iterations, the optimal solution was obtained: the wall thickness was 3.5mm and the supporting structure size was 22mm. Verification analysis showed that after optimization, the maximum displacement was reduced to 0.3mm and the maximum acceleration was reduced to 3m / s 2 , vibration levels are reduced by 40%.

[0048] S105. Based on the temperature and airflow distribution data, temperature sensors and airflow sensors are arranged near the elevator door to monitor the temperature changes and airflow changes when the elevator door is opened and closed in real time. When the elevator door is detected to be open, the working parameters of the air curtain and the refrigeration plate are dynamically adjusted according to the door opening time and the temperature and airflow change amplitude.

[0049] A distributed data acquisition module is used to obtain the temperature and airflow data collected by a high-precision temperature sensor and a hot wire anemometer, and the sampling frequency of the temperature and airflow data is set to 1 Hz; the switch state of the elevator door is detected by the access control sensor, and if the door opening signal is detected and the duration exceeds the preset 3-second threshold, the data recording program is triggered; according to the temperature and airflow data recorded by the data recording program, a 60-second sliding time window is used to calculate the temperature change rate and the airflow velocity change rate before and after the door is opened; based on the temperature change rate and the airflow velocity change rate, a temperature and airflow change prediction model is established using a long short-term memory network (LSTM), and the network structure of the LSTM model is a double-layer LSTM plus a fully connected layer; according to the prediction results of the LSTM model, the wind speed and wind direction of the air curtain and the cooling power of the refrigeration plate are adjusted.

[0050] For example, high-precision temperature sensors and hot-wire anemometers are installed near the elevator door, and distributed data acquisition modules are used to collect temperature and airflow data in real time. The sampling frequency is set to 1Hz, and the collected data is subjected to sliding average filtering, and the processed data is transmitted to the central processing unit. The access control sensor is used to detect the switch state of the elevator door. When the door opening signal is detected, the data recording program is triggered to record the door opening time and duration, and the temperature and airflow data of the corresponding time period are marked. The minimum effective door opening time threshold is set to 3 seconds to filter out short-term door opening and closing interference. A 60-second sliding time window is used to analyze the temperature and airflow data in real time, and the temperature change rate and airflow velocity change rate before and after the door is opened are calculated. The fuzzy logic controller is used for preliminary adjustment. The input variables are the temperature change rate and the airflow velocity change rate, and the output variables are the wind curtain wind speed adjustment amount and the refrigeration plate power adjustment amount. 9 fuzzy rules are designed. A temperature and airflow change prediction model is established based on the long short-term memory network (LSTM). The network structure is a double-layer LSTM plus a fully connected layer. The data of the past 24 hours is used for training. The current state is input to predict the temperature and airflow change trend within the next 5 seconds. The weighted average method is used to fuse the output results of fuzzy logic and LSTM, and the weights are dynamically adjusted according to the historical accuracy. According to the fusion results, the wind speed of the air curtain is adjusted in the range of 0-10m / s, the wind direction in the range of -45° to 45°, and the cooling power of the refrigeration plate in the range of 0-1000W. Based on the actual adjustment effect, the prediction error is calculated, and the LSTM model parameters are updated every hour using an online learning algorithm. In a standard 2m×2m×3m elevator, three SHT31-D temperature sensors and three TSI9565 hot wire anemometers are installed on both sides of the door frame with a vertical spacing of 0.5m. The NIcDAQ-9185 data acquisition module is used to collect data at a frequency of 1Hz, and the data is transmitted to the central processing unit after a 5-point sliding average filter. The access control sensor uses a Hall effect switch. After detecting the door opening signal, the timer is started to record the door opening duration. If the door opening time exceeds 3 seconds, the data analysis program is triggered. The temperature change rate (ΔT / Δt) and air flow velocity change rate (Δv / Δt) are calculated using a 60-second sliding time window. The input variables "temperature change rate" and "air flow speed change rate" of the fuzzy logic controller are divided into three fuzzy sets (small, medium, and large), and the output variables "wind speed adjustment" and "cooling power adjustment" are divided into five fuzzy sets (large decrease, small decrease, unchanged, small increase, and large increase). The LSTM network structure is a 64-node double-layer LSTM plus a 32-node fully connected layer. The training is based on the temperature and air flow data of the past 24 hours and 1 minute intervals. The batch_size is set to 32 and the epoch is 100. The fusion weight is initially set to 0.7 for fuzzy logic and 0.3 for LSTM, and is updated every hour based on the prediction accuracy.The wind curtain has a wind speed adjustment range of 0-10m / s, a wind direction of -45° to 45°, and a cooling plate power of 0-1000W. The online learning algorithm uses stochastic gradient descent, with a learning rate of 0.001, and the LSTM model parameters are updated every hour using the data from the last hour.

[0051] S106. Establish a control strategy for wind curtains and refrigeration fins based on fuzzy control. According to the temperature, air flow distribution, and number of people in the elevator, monitor the changes in temperature and air flow distribution under different numbers of people, and dynamically adjust the wind speed of the wind curtain and the working power of the refrigeration fins.

[0052] The environmental data in the elevator collected by the temperature sensor array, the airflow sensor array and the infrared human detector are obtained, and the environmental data include temperature distribution, airflow distribution and the number of people; the environmental data are filtered by using a Kalman filter to obtain filtered data; if the filtered data is processed by a minimum-maximum normalization method, standardized data is obtained; the change rate of the temperature field and the airflow field is calculated by the change of the number of people data in the standardized data, and the correlation model between the density of people and the change of environmental parameters is determined; the fuzzy controller based on Mamdani reasoning is designed by using the correlation model, and the fuzzy controller includes a fuzzy rule base and a membership function affected by the density of people; according to the output result of the fuzzy controller, the wind speed of the wind curtain and the power of the refrigeration plate are adjusted, and the adjustment amount of the wind speed of the wind curtain and the power of the refrigeration plate is determined by the output language variable of the fuzzy controller; the membership function parameters are dynamically adjusted by the gradient descent method, and the adjustment of the membership function parameters is based on the evaluation results of the temperature uniformity and the predicted average voting index.

[0053] Exemplarily, a 3×3×3 temperature sensor array and an airflow sensor array are installed in the elevator, and an infrared human body detector is configured on the top of the elevator to collect temperature distribution, airflow distribution and personnel number data in the elevator in real time. The sampling frequency is set to 0.1Hz, and the collected data is transmitted to the central control unit for processing. The collected raw data is filtered by a Kalman filter to eliminate the influence of noise, and then the data is standardized using the minimum-maximum normalization method. The processed temperature data, airflow data and personnel number data are used as input for subsequent analysis. According to the change in the number of people, the change rate of the temperature field and the airflow field is calculated, and a correlation model between the change of personnel density and the change of environmental parameters is established. A fuzzy controller based on Mamdani reasoning is designed, and the temperature deviation, airflow speed deviation and personnel density are used as input language variables, and the wind curtain wind speed adjustment amount and the refrigeration plate power adjustment amount are used as output language variables. A fuzzy rule base and membership function containing the influence of personnel density are established. According to the output result of the fuzzy controller, the wind speed of the wind curtain and the working power of the refrigeration plate are adjusted by the actuator, and the membership function parameters are dynamically adjusted using the gradient descent method. The control effect is evaluated based on temperature uniformity and predicted mean voting index (PMV), and the control strategy is optimized through feedback mechanism. In a standard elevator of 2m×2m×3m, 27 SHT31-D temperature and humidity sensors and 27 TSI9565 hot wire anemometers are installed to form a 3×3×3 sensor array with a spacing of 0.67m. A Panasonic Grid-EYE infrared array sensor is installed in the center of the top of the elevator for personnel detection. All sensors are sampled at a frequency of 0.1Hz, and the data is transmitted to the central control unit through the RS-485 bus. The fifth-order Kalman filter is used to process the raw data, with the process noise covariance Q=0.001 and the measurement noise covariance R=0.1. The filtered data is normalized by minimum-maximum normalization and mapped to the interval [0,1]. Based on the change of personnel density, the change rate of temperature field and airflow field is calculated, and a multivariate linear regression model is established to describe the relationship between personnel density and environmental parameter changes. The fuzzy controller is designed with 7 input membership functions, 3 temperatures, 3 airflows, 1 personnel density, and 5 output membership functions, 2 wind curtains, 3 cooling sheets, and a total of 35 fuzzy rules. The wind curtain wind speed adjustment range is 0-10m / s, and the cooling sheet power range is 0-1000W. The centroid method is used for defuzzification. The random gradient descent method is used, the learning rate is 0.01, and the membership function parameters are updated every 100 control cycles. The control effect is evaluated using the temperature uniformity index, with a standard deviation of <1℃ as the best, and the PMV index, -0.5 to +0.5 as the comfort range. According to the evaluation results, the fuzzy rule weights are updated every hour, and the weight adjustment range is ±0.1.

[0054] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for optimizing the control of the temperature and airflow distribution inside an elevator, characterized in that: include: Obtain historical temperature and airflow distribution data inside the elevator, wherein the airflow distribution data includes average wind speed distribution in different areas, volume proportion of airflow short-circuit areas, and volume proportion of airflow dead areas; According to the historical temperature and airflow distribution data, a temperature and airflow distribution prediction model is trained using a long short-term memory network; Acquire real-time temperature and airflow distribution data inside the elevator, input the temperature and airflow distribution prediction model, and predict future temperature and airflow distribution; According to the predicted temperature and airflow distribution, the optimization algorithm is used to calculate the opening angle of the air curtain and the cooling power parameters of the refrigeration fins. The temperature control accuracy and system energy consumption are taken as the optimization goals to obtain the optimal control strategy of the air curtain and refrigeration fins. The parameters in the optimal control strategy are input into the controllers of the air curtain and refrigeration fins to adjust the wind speed and direction of the air curtain and the cooling power of the refrigeration fins to achieve optimal control of the temperature and airflow distribution inside the elevator.

2. The optimization control method according to claim 1, characterized in that: Before obtaining the historical temperature and airflow distribution data inside the elevator, the method further includes: The moving average method is used to fill in the missing values ​​in the historical temperature and airflow distribution data, and the outliers in the historical temperature and airflow distribution data are eliminated by the 3σ principle to obtain a standardized historical temperature and airflow distribution data set.

3. The optimization control method according to claim 1, characterized in that: The method of calculating the air curtain opening angle and the cooling power parameters of the refrigeration plate by using an optimization algorithm according to the predicted temperature and airflow distribution includes: The NSGA-II algorithm is used to establish a multi-objective optimization model for air curtain and refrigeration fin adjustment, with temperature control accuracy and system energy consumption as optimization targets. The multi-objective optimization model is solved by an adaptive particle swarm optimization algorithm to obtain the optimal parameter combination of the air curtain opening angle, wind speed, swing frequency and refrigeration power of the refrigeration plate.

4. The optimization control method according to claim 1, characterized in that: After the parameters in the optimal control strategy are input into the controllers of the air curtain and the refrigeration fins, the method further includes: A distributed sensor network is used to obtain actual temperature and airflow distribution data inside the elevator, and the actual temperature and airflow distribution data are processed by a Kalman filter algorithm; It is determined whether the processed actual temperature and airflow distribution data deviate from the predicted value by more than a set threshold value. If the deviation exceeds the set threshold value, retraining of the temperature and airflow distribution prediction model is triggered.

5. The optimization control method according to claim 1, characterized in that: While obtaining real-time temperature and airflow distribution data inside the elevator, it also includes: Detect the opening and closing status of the elevator door through the access control sensor; If it is detected that the elevator door is open, the wind speed, wind direction of the air curtain and the cooling power of the refrigeration plate are dynamically adjusted according to the duration of the door opening and the temperature and airflow changes before and after the door opening.

6. The optimization control method according to claim 1, characterized in that: Before calculating the air curtain opening angle and the refrigeration power parameters of the refrigeration fins by using an optimization algorithm according to the predicted temperature and airflow distribution, the method further includes: Obtain geometric parameters and material property data of air curtains and refrigeration fins; Establishing a vibration model of the air curtain and the refrigeration fin according to the geometric parameters and material property data; Calculating the natural frequency, mode shape and frequency response function of the vibration model through modal analysis and harmonic response analysis; Genetic algorithms are used to optimize the material parameters and structural dimensions of the air curtain and refrigeration fins, with minimizing the maximum vibration amplitude of the vibration model as the objective function, to obtain the optimal material selection and structural design scheme for the air curtain and refrigeration fins.

7. The optimization control method according to claim 1, characterized in that: After obtaining the real-time temperature and airflow distribution data inside the elevator, the method further includes: Use infrared human body detector to obtain data on the number of people in the elevator; According to the personnel quantity data and real-time temperature and airflow distribution data, the correlation model between personnel density and temperature field and airflow field changes is calculated; A personnel density influence controller based on fuzzy reasoning is designed, the wind speed of the wind curtain and the power of the refrigeration plate are dynamically adjusted according to the association model, and the membership function parameters of the fuzzy controller are optimized by the gradient descent method.

Images