Multi-air-pipe valve self-adaptive control method and system for temperature and humidity uniformity

Abstract

The invention is suitable for the technical field of industrial environment control, and provides a multi-air-duct valve self-adaptive control method and system oriented to temperature and humidity uniformity, and the method comprises the steps: carrying out the dynamic characteristic modeling and constraint analysis of a tail end air valve during the shutdown period of the system, obtaining an opening mapping model and response time of each air valve, and carrying out the dynamic characteristic modeling and constraint analysis of the tail end air valve; determining a global adjustment period and set value compensation; calculating an environment uniformity index, and judging environment stabilization time; generating a control scheme based on the hybrid strategy; establishing a control scheme scoring model, scoring the generated control scheme, and predicting a control effect; the control scheme with the highest score in the fourth step is received, the control scheme is written into the OPC according to the minimum time interval and the air valve mapping relation, and the air valve is controlled. Batch control can be achieved, efficiency is improved, the labor burden is relieved, intelligent decision making based on a model is achieved, dependence on artificial experience is eliminated, self-optimization design is achieved, and the adaptability and robustness of the system are enhanced.

Description

Technical Field

[0001] This invention relates to the field of industrial environmental control technology, specifically to an adaptive control method and system for multi-duct valves aimed at ensuring temperature and humidity uniformity. Background Technology

[0002] Environmental temperature and humidity control in industrial plants is crucial for ensuring production process stability and product quality. Current mainstream industrial air conditioning systems generally employ a zoned air supply design, which involves installing multiple branch ducts at the end of the main air supply duct and installing electrically operated regulating valves on each duct to distribute airflow to different areas. Existing technical solutions are as follows:

[0003] 1. Manual experience control mode: Relies on the operator's observation of local temperature and humidity parameters and manual adjustment of valve opening based on personal experience;

[0004] 2. Basic Automatic Control Mode: Employs a traditional PID control algorithm, adjusting the opening of corresponding valves based on a single temperature feedback signal. For example, when the temperature in a certain area is too high, the opening of the corresponding valve in that area is automatically increased to increase the amount of cold air delivered.

[0005] In addition, there are some attempts at intelligent control in the market, such as zone control strategies based on fixed rules, or predicting valve opening by building linear regression models using limited sensor data. However, such solutions are mostly limited to laboratory environments or small spaces, and mature solutions for complex industrial scenarios have not yet been developed.

[0006] While the above technologies have been applied in certain scenarios, they still have significant shortcomings in actual deployment in industrial plants:

[0007] 1. Failure of multivariable coupling control: Industrial plants have large spaces and significant airflow coupling effects in the air supply of each duct. Traditional single-loop PID control only focuses on local temperature feedback and cannot coordinate the linkage effects between multiple valves, resulting in regulation oscillation or temperature and humidity conflicts between areas. For example, adjusting a valve may cause temperature fluctuations in adjacent areas, forming a dead loop of repeated adjustments.

[0008] 2. Poor environmental dynamic adaptability: Industrial plants often experience sudden disturbances such as equipment start-up and shutdown, personnel movement, and opening and closing of doors and windows, which lead to rapid changes in heat load distribution. Existing control strategies rely on static models or empirical rules and lack the ability to model dynamic nonlinear relationships. The adjustment response is significantly lagging and it is difficult to maintain the stability of uniformity indicators.

[0009] 3. High reliance on manual operation: The manual control mode requires operators to continuously monitor parameters in multiple areas, which can lead to subjective judgment bias and operational delays. Especially when multiple valves are adjusted in a coordinated manner, human experience is not enough to quantify the indirect impact of each valve on the remote area, which can easily lead to local over-adjustment or waste of resources.

[0010] Therefore, in view of the above situation, there is an urgent need to provide an adaptive control method and system for multi-duct valves that is oriented towards temperature and humidity uniformity, so as to overcome the shortcomings in current practical applications. Summary of the Invention

[0011] The purpose of this invention is to provide an adaptive control method and system for multi-duct valves aimed at achieving temperature and humidity uniformity, thereby addressing the problems mentioned in the background art.

[0012] This invention is implemented as follows: an adaptive control method and system for multi-duct valves oriented towards temperature and humidity uniformity, the method comprising the following steps:

[0013] Step 1: During system shutdown, perform dynamic characteristic modeling and constraint analysis on the terminal air valves, obtain the opening mapping model and response time of each air valve, and determine the global adjustment cycle and setpoint compensation.

[0014] Step 2: Calculate the environmental uniformity index and determine the environmental stabilization time to determine the minimum interval between the issuance of two control plans;

[0015] Step 3: Generate a control scheme based on a hybrid strategy, which includes an analysis method based on a physical model and a solution method based on an input convex neural network;

[0016] Step 4: Establish a control scheme scoring model, score the generated control schemes, and predict the control effect;

[0017] Step 5: Receive the control scheme with the highest score in Step 4, write the control scheme into the OPC according to the minimum time interval and the air valve mapping relationship, and control the air valve.

[0018] As a further aspect of the present invention: the dynamic characteristic modeling and constraint analysis of the terminal air valve specifically includes:

[0019] During non-production periods, the module is manually activated to automatically obtain the correspondence between the set opening degree and the actual opening degree of each air valve, as well as the air valve response time delay.

[0020] A multinomial regression model is used to construct the opening degree mapping model for each damper, with the input being the set opening degree and the output being the actual opening degree.

[0021] The response time T for each damper was measured from 0% to 100%.

[0022] Take the maximum value of the response time of all dampers as the global adjustment period, which is the minimum time interval between two damper controls.

[0023] By inversely solving the mapping model, the desired opening degree is converted into the actual setting command, and the setting value compensation is performed.

[0024] As a further aspect of the present invention: the calculation of the environmental uniformity index to determine the environmental stabilization time specifically includes:

[0025] The theoretical steady-state time is estimated using the following formula: Where V is the volume of the controlled space, Q is the total air volume, and K... turb This is the turbulence enhancement factor;

[0026] Real-time monitoring of temperature and humidity sensor data, and calculation of temperature standard deviation σ T Humidity standard deviation σ RH And the comprehensive evaluation index U, U = αU T +βU RH , where α+β=1, α is the temperature weight, and β is the humidity weight;

[0027] Balance is determined when the uniformity index remains within the target range and the rate of change approaches zero.

[0028] As a further aspect of the present invention: the analysis method based on the physical model includes:

[0029] By integrating multi-dimensional sensor data with physical laws, a control algorithm is constructed. The correlation between the spatial distribution patterns of sensors and air outlets is obtained, and the uniformity index U and the adjustment index γ of each sensor are calculated. i Set the basic step size and generate the opening scheme;

[0030] The solution method based on the input convex neural network includes:

[0031] The system is modeled using an input convex neural network, and the system model is embedded in the model predictive control framework to solve for the optimal control value of the damper.

[0032] As a further aspect of the present invention: the input features of the control scheme scoring model include the current environmental state, the opening degree of the air valve, and additional parameters. The output target of the control scheme scoring model is the predicted uniformity index. The lower the uniformity index obtained by the scheme input into the evaluation model, the higher the scheme score.

[0033] As a further aspect of the present invention: the current environmental state includes real-time measurement values ​​from eight temperature and humidity sensors;

[0034] The valve opening degree is the current opening degree of the 22 valves;

[0035] The additional parameters include the air conditioning supply temperature and humidity, the external ambient temperature and humidity, and the equipment load.

[0036] As a further aspect of the present invention, the hybrid strategy-based generation control scheme also includes a setting phase adaptive application mechanism. When the accumulation of effective historical data reaches a preset threshold, the system automatically activates the iterative training process of the solution method based on the input convex neural network, ultimately forming a dual-mode parallel generation architecture.

[0037] An adaptive control system for multi-duct valves oriented towards temperature and humidity uniformity, used to implement the adaptive control method for multi-duct valves oriented towards temperature and humidity uniformity as described above, the system includes:

[0038] Dynamic characteristic modeling and constraint analysis module for air valves: used to perform dynamic characteristic modeling and constraint analysis on terminal air valves during system shutdown, obtain the opening mapping model and response time of the air valves, and determine the global adjustment cycle and setpoint compensation;

[0039] Environmental uniformity index and environmental stabilization time determination module: used to calculate environmental uniformity index, determine environmental stabilization time, and determine the minimum interval between two control schemes.

[0040] Control scheme generation module: used to generate control schemes based on a hybrid strategy, which includes an analysis method based on a physical model and a solution method based on an input convex neural network;

[0041] Control scheme scoring model module: used to build a control scheme scoring model, score the generated control schemes, and predict the control effect;

[0042] Control distribution module: Used to receive the highest-scoring control scheme, write the control scheme into the OPC according to the minimum time interval and the air valve mapping relationship, and control the air valve.

[0043] As a further aspect of the present invention: the dynamic characteristic modeling and constraint analysis module of the air valve includes an execution mechanism, modeling output, and constraint application unit.

[0044] As a further aspect of the present invention, the dynamic characteristic modeling and constraint analysis module of the air valve, the environmental uniformity index and environmental stability time determination module, the control scheme generation module, the control scheme scoring model module, and the control distribution module perform a self-optimization update every 90 days to adapt to equipment aging and environmental changes.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] 1. Enable batch control, improve efficiency, and reduce manual workload:

[0047] This method significantly reduces the workload of manual operation by enabling batch issuance of control commands in the control issuance module. After the system goes live, users no longer need to manually adjust the opening of each terminal duct valve. The system can automatically identify the controlled object and issue matching valve adjustment commands, thereby improving control response speed and overall operating efficiency. It is especially suitable for large-scale system scenarios with a large number of devices.

[0048] 2. Model-based intelligent decision-making, eliminating reliance on human experience:

[0049] To break free from reliance on human experience and rules, this invention designs two control scheme generation modes: PHY-Model and IRCNN-Model;

[0050] PHY-Model is built based on physical principles, emphasizing model interpretability and safety boundary control, and is suitable for scenarios with high transparency requirements;

[0051] IRCNN-Model introduces a deep neural network structure, focusing on automatically learning control patterns from a large amount of historical data, making it more adaptable;

[0052] The two modes can be switched or combined according to actual needs to achieve more flexible, efficient and scientific control strategy generation.

[0053] 3. Self-optimizing design enhances system adaptability and robustness:

[0054] To enhance the system's adaptability to environmental changes, equipment aging, and other factors, each control module incorporates a self-optimization mechanism.

[0055] In the dynamic characteristic modeling and constraint analysis module of the air valve, the system supports users to start the update function after calibrating the valve, which automatically corrects the opening mapping relationship and response time parameters of each valve.

[0056] The environmental uniformity index and environmental stability time determination module, control scheme generation module, and control scheme scoring model module are set to automatically perform a "self-optimization" update (i.e., adaptive parameter re-estimation and model reconstruction) every 90 days to cope with equipment performance fluctuations and changes in the operating environment, and to ensure that the control strategy is always in the optimal state. Detailed Implementation

[0057] The technical solution of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] This invention uses intelligent algorithms to automatically regulate the opening of multi-duct valves, solving the problems of low efficiency relying on manual experience and the inability of traditional industrial control solutions to adapt to hardware aging. It also improves the uniformity of temperature and humidity in industrial environments. Specifically, it employs a multi-variable collaborative model to eliminate airflow coupling interference, combined with a dynamic environmental perception-based real-time adjustment strategy to ensure uniformity, and supports rapid deployment and adaptive control of large-scale valve / sensor clusters.

[0059] The present invention will be further explained below with reference to specific embodiments.

[0060] The adaptive control method for multi-duct valves aimed at temperature and humidity uniformity provided in this invention includes:

[0061] Step 1: During system shutdown, perform dynamic characteristic modeling and constraint analysis on the terminal air valves, obtain the opening mapping model and response time of each air valve, and determine the global adjustment cycle and setpoint compensation.

[0062] Step 2: Calculate the environmental uniformity index and determine the environmental stabilization time to determine the minimum interval between the issuance of two control plans;

[0063] Step 3: Generate a control scheme based on a hybrid strategy, which includes an analysis method based on a physical model and a solution method based on an input convex neural network;

[0064] Step 4: Establish a control scheme scoring model, score the generated control schemes, and predict the control effect;

[0065] Step 5: Receive the control scheme with the highest score in Step 4, write the control scheme into the OPC according to the minimum time interval and the air valve mapping relationship, and control the air valve.

[0066] The system includes a module for modeling and constraining the dynamic characteristics of air valves, a module for determining environmental uniformity index and environmental stability time, a module for generating control schemes, a module for scoring control schemes, and a module for issuing control commands.

[0067] The following provides further explanation for each module:

[0068] Module 1: Dynamic Characteristic Modeling and Constraint Analysis of Air Valves

[0069] During system shutdown, a dynamic behavior model of the terminal valves is established to quantify their physical constraints and provide underlying data support for real-time control.

[0070] 1. Implementation Mechanism

[0071] After the factory completes the valve calibration during non-production periods, this module is manually activated. This module can automatically obtain the correspondence between the set opening degree and the actual opening degree of each valve, as well as the valve response time delay.

[0072] 2. Modeling Output

[0073] Valve Opening Mapping Model: For each valve, a mapping model is constructed using multinomial regression. The model input is a set opening degree (γ). set The output is the actual opening degree (γ). real (For example: setting 20 corresponds to 17.5 in reality).

[0074] Valve response time: Measure the response time of each valve change to obtain the response time T of each valve from 0% to 100%.

[0075] 3. Constraint Application

[0076] Global control cycle: Take the maximum value of the response time of all valves as the minimum time interval between two valve controls (e.g., maximum response 12s → minimum time interval ≥ 12s).

[0077] Setpoint compensation: By inversely solving the mapping model, the desired opening degree is converted into the actual set command (if a true 20% opening degree is required, the setpoint needs to be compensated to 23%).

[0078] Module 2: Environmental Uniformity Indicators and Environmental Stabilization Time Determination Module

[0079] This module uses a dual mechanism of physical laws and data intelligence to obtain the time it takes for the uniformity of ambient temperature and humidity to stabilize after changing the air valve. This time is used as the minimum interval between two control schemes.

[0080] 1. Theoretical steady-state time estimation

[0081]

[0082] V: Controlled space volume (m³) 3 Q: Total air volume: m³ 3 / s;K turb Turbulence enhancement coefficient (1.5-4.0), the values ​​of which are shown in Table 1:

[0083] Table 1. Values ​​of turbulence enhancement coefficients

[0084] Spatial features <![CDATA[K turb Values]]> Empty factory building (accessible) 1.8 Medium-density equipment (<30% coverage) 2.5 High-density equipment (≥30% coverage) 3.2 Complex Piping Systems +0.5 additional item

[0085] 2. Monitoring of actual stabilization time

[0086] It monitors temperature and humidity sensor data distributed in the environment in real time and calculates how long it takes for the environmental temperature and humidity uniformity to stabilize after a control is issued.

[0087] (1) Calculation of environmental uniformity index:

[0088] Temperature standard deviation:

[0089]

[0090] Among them, T i For the temperature of each sensor, Here, N represents the average temperature of the sensors, and N is the total number of sensors.

[0091] Humidity standard deviation:

[0092]

[0093] Among them, RH i For the relative humidity of each sensor, The average humidity of the sensor;

[0094] Comprehensive evaluation indicators:

[0095]

[0096] U=αU T +βU RH ;

[0097] Among them, U T U is a temperature evaluation index. RH U is the humidity evaluation index, α is the temperature weight, β is the humidity weight, and α+β=1.

[0098] Summer mode: Temperature weighting α increased to 0.7;

[0099] Winter mode: Humidity weight β increased to 0.6.

[0100] (2) Determination of steady time

[0101] Balance is determined when the uniformity index remains within the target range and the rate of change approaches zero.

[0102] ∈ is a small value close to 0. The time Tstart is recorded when a control scheme is first issued, and the time Tend is recorded when the scheme reaches equilibrium. The difference between these two times is taken as the settling time T. To ensure the reliability of the results, the settling time needs to be tested at least 5 times, and the maximum value of all results should be taken.

[0103] Module 3: Control Scheme Generation Module

[0104] This module generates control schemes based on a hybrid strategy, and its core feature is the adoption of a phased collaborative control strategy generation mechanism.

[0105] 1. Control Strategy Generation Method

[0106] This system provides two control scheme generation modes:

[0107] (1) Physical Model-Based Analysis Method (PHY-Model): By fusing multi-dimensional sensor data with physical laws, an interpretable control algorithm is constructed. The PHY-Model combines the spatial distribution patterns of sensors and air vents to obtain the correlation between temperature and humidity sensors and air vents; it uses sensor samples within the space to calculate the uniformity index U, as well as the adjustment index (γ) of each sensor. i The opening scheme is obtained by comprehensively calculating the valve status, spatial temperature and humidity distribution gradient, and air supply status, plus some basic settings. For example, we set the basic step size to 5%, which corresponds to the uniformity index U. 0% corresponds to the position where U=0, and Ui corresponds to the opening size, which is distributed on the straight line where (U,5%) and (0,0%) are located.

[0108] (2) Solution method based on input convex neural network (IRCNN-Model): In this method, after the system is modeled using input convex neural network, the system model is embedded into the model predictive control framework to solve for the optimal valve control value.

[0109] 2. Stage-Adaptive Application Mechanism

[0110] In the initial stage of the system (when historical operating data is insufficient), the PHY-Model mode is activated first, which can effectively improve the limitations of traditional manual control strategies in terms of efficiency and accuracy. When the accumulation of effective historical data reaches a preset threshold (typically set to ≥720 device operating hours in a typical implementation case), the system automatically activates the iterative training process of IRCNN-Model, ultimately forming a dual-mode parallel generation architecture.

[0111] Module 4: Control Scheme Scoring Model

[0112] The system inputs control schemes and predicts their effectiveness under current operating conditions to determine which scheme is superior. A deep learning model is used to build the evaluation model, which is periodically self-optimized as data accumulates. Its input and output features are as follows.

[0113] 1. Input Feature Design

[0114] The model's input needs to include two types of information: the current environmental state and the valve opening degree, in order to reflect the system dynamics.

[0115] Environmental conditions (sensor data): Real-time measurements from 8 temperature and humidity sensors (16 dimensions: 8 temperature + 8 humidity).

[0116] Valve opening: Current opening of 22 valves (0-100%, normalized to 0-1).

[0117] Additional parameters (optional): air conditioning supply temperature and humidity, external ambient temperature and humidity, equipment load, etc.

[0118] 2. Output target

[0119] Predicted uniformity indices: overall uniformity index and deviation indices for each valve.

[0120] 3. Principles for Project Evaluation

[0121] The lower the uniformity index obtained for each scheme in the input evaluation model, the higher the environmental uniformity and the higher the scheme score.

[0122] Module 5: Control Distribution Module

[0123] This module receives the control scheme with the highest score from module 4, processes it according to the minimum time interval obtained from module 2 and the valve mapping relationship obtained from module 1, and writes it into the OPC to control the duct valves.

[0124] In summary, the present invention has at least the following advantages:

[0125] Advantage 1: Enables batch control, improves efficiency, and reduces manual workload.

[0126] This method significantly reduces the workload of manual operation by enabling batch issuance of control commands in module 5. After the system goes live, users no longer need to manually adjust the opening of terminal duct valves one by one. The system can automatically identify the controlled objects and issue matching valve adjustment commands, thereby improving control response speed and overall operating efficiency. It is especially suitable for large-scale system scenarios with a large number of devices.

[0127] Advantage 2: Model-based intelligent decision-making, eliminating reliance on human experience.

[0128] To break free from reliance on human experience and rules, we designed two control scheme generation modes: PHY-Model and IRCNN-Model.

[0129] PHY-Model is built based on physical principles, emphasizing model interpretability and safety boundary control, and is suitable for scenarios with high transparency requirements.

[0130] The IRCNN-Model introduces a deep neural network structure, focusing on automatically learning control patterns from a large amount of historical data, making it more adaptable.

[0131] The two modes can be switched or combined according to actual needs to achieve more flexible, efficient and scientific control strategy generation.

[0132] Advantage 3: Self-optimizing design enhances system adaptability and robustness.

[0133] To enhance the system's adaptability to environmental changes, equipment aging, and other factors, each control module incorporates a self-optimization mechanism.

[0134] In Module 1, the system allows users to activate the update function after calibrating the valves, which automatically corrects the opening mapping relationship and response time parameters of each valve.

[0135] Modules 2, 3, and 4 are set to automatically perform a "self-optimization" update (i.e., adaptive parameter reestimation and model reconstruction) every 90 days to cope with fluctuations in equipment performance and changes in the operating environment, ensuring that the control strategy is always in the optimal state.

[0136] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive control method for multi-duct valves aimed at ensuring temperature and humidity uniformity, characterized in that, The method includes the following steps: Step 1: During system shutdown, perform dynamic characteristic modeling and constraint analysis on the terminal air valves, obtain the opening mapping model and response time of each air valve, and determine the global adjustment cycle and setpoint compensation. Step 2: Calculate the environmental uniformity index and determine the environmental stabilization time to determine the minimum interval between the issuance of two control plans; Step 3: Generate a control scheme based on a hybrid strategy, which includes an analysis method based on a physical model and a solution method based on an input convex neural network; Step 4: Establish a control scheme scoring model, score the generated control schemes, and predict the control effect; Step 5: Receive the control scheme with the highest score in Step 4, write the control scheme into the OPC according to the minimum time interval and the air valve mapping relationship, and control the air valve.

2. The adaptive control method and system for multi-duct valves oriented towards temperature and humidity uniformity according to claim 1, characterized in that, The dynamic characteristic modeling and constraint analysis of the terminal air valve specifically includes: During non-production periods, the module is manually activated to automatically obtain the correspondence between the set opening degree and the actual opening degree of each air valve, as well as the air valve response time delay. A multinomial regression model is used to construct the opening degree mapping model for each damper, with the set opening degree as the input and the actual opening degree as the output. The response time T for each damper was measured from 0% to 100%. Take the maximum value of the response time of all dampers as the global adjustment period, which is the minimum time interval between two damper controls. By inversely solving the mapping model, the desired opening degree is converted into the actual setting command, and the setting value compensation is performed.

3. The adaptive control method and system for multi-duct valves oriented towards temperature and humidity uniformity according to claim 1, characterized in that, The calculation of the environmental uniformity index, which determines the environmental stabilization time, specifically includes: The theoretical steady-state time is estimated using the following formula: Where V is the volume of the controlled space, Q is the total air volume, and K... turb This is the turbulence enhancement factor; Real-time monitoring of temperature and humidity sensor data, and calculation of temperature standard deviation σ T Humidity standard deviation ρ RH And the comprehensive evaluation index U, U = αU T +βU RH , where α+β=1, α is the temperature weight, and β is the humidity weight; Balance is determined when the uniformity index remains within the target range and the rate of change approaches zero.

4. The adaptive control method and system for multi-duct valves oriented towards temperature and humidity uniformity according to claim 1, characterized in that, The analysis method based on the physical model includes: By integrating multi-dimensional sensor data with physical laws, a control algorithm is constructed. By combining the spatial distribution patterns of sensors and air outlets to obtain correlations, the uniformity index U and the adjustment index γ of each sensor are calculated. i Set the basic step size and generate the opening scheme; The solution method based on the input convex neural network includes: The system is modeled using an input convex neural network, and the system model is embedded in the model predictive control framework to solve for the optimal control value of the damper.

5. The adaptive control method and system for multi-duct valves oriented towards temperature and humidity uniformity according to claim 1, characterized in that, The input features of the control scheme scoring model include the current environmental state, the opening degree of the damper, and additional parameters. The output target of the control scheme scoring model is the predicted uniformity index. The lower the uniformity index obtained by the scheme input into the evaluation model, the higher the scheme score.

6. The adaptive control method and system for multi-duct valves oriented towards temperature and humidity uniformity according to claim 5, characterized in that, The current environmental status includes real-time measurements from eight temperature and humidity sensors; The valve opening degree is the current opening degree of the 22 valves; The additional parameters include the air conditioning supply temperature and humidity, the external ambient temperature and humidity, and the equipment load.

7. The adaptive control method and system for multi-duct valves oriented towards temperature and humidity uniformity according to claim 4, characterized in that, The hybrid strategy-based generation control scheme also includes a setting phase adaptive application mechanism. When the accumulation of effective historical data reaches a preset threshold, the system automatically activates the iterative training process of the solution method based on the input convex neural network, ultimately forming a dual-mode parallel generation architecture.

8. A multi-duct valve adaptive control system for temperature and humidity uniformity, used to implement the multi-duct valve adaptive control method for temperature and humidity uniformity as described in any one of claims 1-7, characterized in that, The system includes: Dynamic characteristic modeling and constraint analysis module for air valves: used to perform dynamic characteristic modeling and constraint analysis on terminal air valves during system shutdown, obtain the opening mapping model and response time of the air valves, and determine the global adjustment cycle and setpoint compensation; Environmental uniformity index and environmental stabilization time determination module: used to calculate environmental uniformity index, determine environmental stabilization time, and determine the minimum interval between two control schemes. Control scheme generation module: used to generate control schemes based on a hybrid strategy, which includes an analysis method based on a physical model and a solution method based on an input convex neural network; Control scheme scoring model module: used to build a control scheme scoring model, score the generated control schemes, and predict the control effect; Control distribution module: Used to receive the highest-scoring control scheme, write the control scheme into the OPC according to the minimum time interval and the air valve mapping relationship, and control the air valve.

9. The multi-duct valve adaptive control system for temperature and humidity uniformity according to claim 8, characterized in that, The dynamic characteristic modeling and constraint analysis module of the air valve includes an execution mechanism, modeling output, and constraint application unit.

10. The multi-duct valve adaptive control system for temperature and humidity uniformity according to claim 8, characterized in that, The dynamic characteristic modeling and constraint analysis module of the air valve, the environmental uniformity index and environmental stability time determination module, the control scheme generation module, the control scheme scoring model module, and the control distribution module perform self-optimization updates every 90 days to adapt to equipment aging and environmental changes.

Images