Textile air conditioner remote monitoring and intelligent regulation and control system based on Internet of Things

Abstract

The invention discloses a textile air conditioner remote monitoring and intelligent regulation and control system based on the Internet of Things. The system comprises an Internet of Things collecting and preprocessing module, an intelligent control module and a control module. The Internet of Things collecting and preprocessing module constructs a multivariable time sequence sample; the time sequence prediction module is used for reasoning the improved PatchTST model; the state prior construction module is used for constructing state prior information; the control parameter generation module is used for performing sequence modeling processing by the Mamba network; the consistency verification module is used for carrying out consistency verification on the control parameter adjustment sequence; the security constraint and screening module is used for executing textile exclusive security constraint and screening and outputting a screened control parameter adjustment sequence; and the industrial control and closed-loop regulation module is used for executing closed-loop regulation on the textile air conditioning system. The method combines Internet of Things acquisition and a depth time sequence prediction model, and has the advantages of high prediction precision, high response speed, safe and stable control process, adaptability to complex working conditions and the like.

Description

Technical Field

[0001] This invention relates to the fields of Internet of Things (IoT) technology and industrial environmental control technology, and in particular to an IoT-based remote monitoring and intelligent control system for textile air conditioning. Background Technology

[0002] With the continuous improvement of automation in the textile industry, textile workshops are placing higher demands on the stability, energy efficiency, and intelligent control capabilities of air conditioning systems. Traditional textile air conditioning systems typically employ local sensors and industrial controllers (such as PLCs or PID controllers) to construct closed-loop control logic, adjusting environmental parameters such as temperature and humidity in real time. However, these systems generally suffer from limitations such as fixed control methods, unpredictability of future operating conditions, and a lack of intelligent operating strategies. They are ill-suited to the complex and ever-changing textile process environment, especially under conditions of multi-station, multi-material, and multi-equipment collaborative operations. Delayed response or improper control of the air conditioning system can easily lead to sudden drops in humidity and temperature fluctuations, thereby affecting the quality of textiles.

[0003] In recent years, the integration of IoT and AI technologies in industrial scenarios has become increasingly widespread. Some research has attempted to introduce predictive models into air conditioning systems, such as using LSTM and GRU models to predict environmental parameter trends. However, in actual industrial environments, these methods are often not customized for the specific process characteristics of the textile industry, and the models have weak generalization ability for specific scenarios. Meanwhile, existing solutions often rely on empirical rules or fixed templates for control strategy generation, lacking deep linkage and adaptive adjustment mechanisms with prediction results, making it difficult to achieve dynamic control optimization under complex constraints. Furthermore, existing systems generally lack risk assessment, consistency verification, and multi-level screening mechanisms for control strategies, which can easily lead to abnormal control strategy execution and affect system stability and security.

[0004] Against this backdrop, there is an urgent need to develop a dedicated air conditioning control system for the textile industry. This system should integrate IoT data acquisition capabilities, an environmental prediction model with strong generalization capabilities, a state-space-based control strategy generation network, risk perception and multi-dimensional constraint mechanisms, and have full-process closed-loop feedback capabilities to achieve truly intelligent control.

[0005] Therefore, how to provide a remote monitoring and intelligent control system for textile air conditioning based on the Internet of Things is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a remote monitoring and intelligent control system for textile air conditioning based on the Internet of Things (IoT). This invention fully integrates IoT sensing technology, time-series predictive model (PatchTST), and state-space control strategy model (Mamba) to construct a closed-loop predictive control system for textile workshop operating conditions.

[0007] This invention uses an improved PatchTST model to predict the operating environment of textile workshops, including temperature and humidity. It then outputs future trends based on confidence level information. Based on the prediction results, operating status, and prior constraints, a continuous control parameter adjustment sequence is generated using a Mamba network, which is ultimately executed by an industrial controller for closed-loop regulation. Simultaneously, the system incorporates a control consistency verification mechanism and a dual-mode screening strategy driven by confidence level and textile-specific safety constraints to ensure the stability and reliability of the control strategy.

[0008] This invention has the following advantages: accurate prediction, real-time response, interpretable strategy, high security, and strong adaptability. It is especially suitable for the textile industry, which has strict environmental requirements and high sensitivity to humidity control. It can effectively reduce the risk to textile quality caused by sudden drops in humidity and temperature fluctuations, and significantly improve the operating efficiency and intelligence level of the air conditioning system.

[0009] According to an embodiment of the present invention, a remote monitoring and intelligent control system for textile air conditioning based on the Internet of Things includes: IoT data acquisition and preprocessing module: This module is used to install IoT data acquisition devices in the textile workshop to collect real-time data on the operating status of the textile air conditioning system and upload the data to a remote monitoring platform; it also preprocesses the uploaded data to construct multivariate time series samples. The time series prediction module is used to perform inference on the improved PatchTST model based on multivariate time series samples. During the inference stage, the multivariate time series samples are divided into patches, and patch embedding vectors are generated through sub-time series coding units. The patch embedding vectors are subjected to affine modulation and input into the coding network for feature extraction. Without performing model parameter updates, the module outputs the temperature prediction sequence, humidity prediction sequence and corresponding confidence sequence in the future prediction time domain. State Prior Construction Module: Used to construct state prior information based on temperature prediction sequence, humidity prediction sequence, confidence sequence, target temperature and humidity setpoints, and the current operating status of textile air conditioning. Control parameter generation module: It is used to input the state prior information into the Mamba network for sequence modeling and generate gating coefficients based on the confidence sequence to constrain the amplitude and rate of the control parameter adjustment sequence; Consistency verification module: used to perform consistency verification on the control parameter adjustment sequence, generate candidate control sequences based on the control parameter adjustment sequence and perform consistency evaluation; Safety constraint and screening module: used to perform textile-specific safety constraints and screening on candidate control sequences that have passed the consistency check, and output the screened control parameter adjustment sequence; Industrial control and closed-loop regulation module: used to receive the filtered control parameter adjustment sequence, and the industrial controller performs closed-loop regulation on the textile air conditioning system according to the control parameter adjustment sequence.

[0010] Optionally, modules can be integrated using the following methods: Step 1: Install IoT data acquisition devices in the textile workshop to collect real-time data on the operating status of the textile air conditioning system and upload the data to the remote monitoring platform; preprocess the uploaded data to construct a multivariate time series sample. Step 2: Before deploying the improved PatchTST model, pre-training is completed based on the running state data, and prior constraints of textile process are introduced during the training phase. During the inference phase, multivariate time series samples are divided into patches and Patch embedding vectors are generated through sub-time series coding units. After affine modulation of the Patch embedding vectors, they are input into the coding network for feature extraction. Without updating the model parameters, the temperature prediction sequence, humidity prediction sequence and corresponding confidence sequence in the future prediction time domain are output. Step 3: Generate the required prior state information based on the temperature prediction sequence, humidity prediction sequence, confidence sequence, target temperature and humidity setpoints, and the current operating status of the textile air conditioning system. Step 4: Input the state prior information into the Mamba network to generate a continuous control parameter adjustment sequence, generate gating coefficients based on the confidence sequence, and constrain the amplitude and rate of the control parameter adjustment sequence according to the gating coefficients; Step 5: Perform consistency verification on the control parameter adjustment sequence, generate candidate control sequences based on the control parameter adjustment sequence and perform consistency evaluation. If the consistency verification does not meet the preset conditions, perform rejection or downgrade processing on the candidate control sequence. Step 6: Perform textile-specific safety constraints and screening on the candidate control sequences that have passed the consistency check, and output the screened control parameter adjustment sequences; Step 7: Send the selected control parameter adjustment sequence to the industrial controller, which will then perform closed-loop regulation of the textile air conditioning system based on the control parameter adjustment sequence.

[0011] Optionally, step one includes the following specific steps: Internet of Things (IoT) data collection devices are installed in the textile workshop and connected to the textile air conditioning system; Operational status data includes temperature, humidity, air volume, fan frequency, valve opening, fresh air ratio, power parameters, and operating and alarm status. The operational status data is uploaded to the remote monitoring platform via the communication network for storage and display; The operational status data uploaded to the remote monitoring platform is preprocessed, including data cleaning, outlier removal, time alignment, and resampling. Based on the preprocessed operational status data, a multivariate time series sample is constructed according to a preset time window.

[0012] Optionally, the step of completing pre-training based on runtime data and introducing prior constraints on textile processes during the training phase before deploying the improved PatchTST model includes the following specific steps: Before deploying the improved PatchTST model, the improved PatchTST model is pre-trained based on the runtime status data to obtain the pre-trained model parameters; During the training phase of the improved PatchTST model, a priori constraints on textile processes are introduced. These priori constraints are predetermined based on the process stability requirements of environmental temperature and humidity during textile production, and include constraints on the range of environmental parameter values ​​and the rate of change. During the training phase, the temperature and humidity predictions output by the improved PatchTST model are compared with the predetermined process allowable range, and predictions that exceed the process allowable range are included in the value range constraint term. The magnitude of change between the predicted temperature and humidity values ​​at adjacent time steps is compared with a predetermined process change threshold, and the magnitude of change exceeding the process change threshold is included in the rate of change constraint term. Optionally, the step of dividing multivariate time series samples into patches during the inference phase and generating patch embedding vectors through sub-temporal coding units, applying affine modulation to the patch embedding vectors and inputting them into the coding network for feature extraction, and outputting the temperature prediction sequence, humidity prediction sequence, and corresponding confidence sequence in the future prediction time domain without updating model parameters, includes the following specific steps: During the inference phase of the improved PatchTST model, multivariate time series samples are obtained; The multivariate time series sample is divided into N patches in the time dimension according to the preset patch length. Each patch corresponds to a subsequence of running state data continuously extracted from the multivariate time series sample. The subsequence of running status data is input into the sub-temporal coding unit for encoding processing to obtain the internal temporal features of the patch, and the corresponding patch embedding vector is generated based on the internal temporal features of the patch. Feature aggregation processing is performed on the running status data within a preset time window in the multivariate time series samples to obtain a fixed-length vector representation, and the scaling coefficient vector and translation coefficient vector are calculated based on the fixed-length vector representation; Affine modulation is performed on the Patch embedding vector, wherein the affine modulation includes multiplying the scaling coefficient vector element-wise with the Patch embedding vector and superimposing the translation coefficient vector. The affine-modulated Patch embedding vector is input into the encoding network for feature extraction, and the output is the temperature prediction sequence, humidity prediction sequence, and the confidence sequence corresponding to the temperature prediction sequence and humidity prediction sequence in the future prediction time domain. No model parameter update is performed during the inference stage.

[0013] Optionally, step three includes the following specific steps: Obtain the target temperature and humidity setpoints and the current operating status of the textile air conditioner; Calculate the temperature prediction deviation sequence based on the temperature prediction sequence and the target temperature and humidity setpoints, calculate the humidity prediction deviation sequence based on the humidity prediction sequence and the target temperature and humidity setpoints, and combine the temperature prediction deviation sequence and the humidity prediction deviation sequence to form the prediction deviation sequence. Obtain equipment operation constraint information, which includes parameter value range constraints and parameter change rate constraints; The strategy is built based on the predicted deviation sequence, confidence sequence, current control parameter status, and equipment operation constraint information to generate the necessary prior state information.

[0014] Optionally, step four includes the following specific steps: The Mamba network is used to perform sequence modeling on the state prior information and generate a continuous control parameter adjustment sequence. The control parameter adjustment sequence includes a fan frequency adjustment sequence, a valve opening adjustment sequence, and a fresh air to return air ratio adjustment sequence. Gating coefficients are generated based on the confidence sequence, and the gating coefficients include amplitude gating coefficients and rate gating coefficients; The adjustment ranges of the fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence are constrained based on the amplitude gating coefficient. The adjustment rates of the fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence are constrained according to the rate gating coefficient to obtain the gated control parameter adjustment sequence.

[0015] Optionally, step five includes the following specific steps: Candidate control sequences are generated based on the constrained control parameter adjustment sequence. The candidate control sequences include candidate fan frequency adjustment sequence, candidate valve opening adjustment sequence, and candidate fresh air to return air ratio adjustment sequence. The consistency between the candidate control sequence and the state prior information is evaluated. The consistency assessment includes back-substituting the predicted deviation sequence in the state prior information based on the candidate control sequence to obtain the back-substituting deviation sequence, and then verifying the consistency between the back-substituting deviation sequence and the predicted deviation sequence. When the consistency check fails to meet the preset conditions, the candidate control sequence is rejected or downgraded. When the consistency check meets the preset conditions, the candidate control sequence that passes the consistency check is output.

[0016] Optionally, step six includes the following specific steps: Based on parameter value range constraints, the time step control parameters in the candidate control sequences that have passed the consistency check are checked for range, and control parameters that exceed the parameter value range constraints are replaced to generate range-constrained candidate control sequences. Based on the parameter change rate constraint, the rate of change of control parameters in adjacent time steps in the range constraint candidate control sequence is verified, and the change of control parameters in adjacent time steps that exceed the parameter change rate constraint is restricted to generate rate constraint candidate control sequences. The candidate control sequences with rate constraints are subjected to constraint processing to prevent humidity drop and humidity overshoot, resulting in candidate control sequences with textile-specific safety constraints. Obtain the confidence sequence and compare it with the preset confidence threshold. When the confidence sequence meets the confidence threshold condition corresponding to the fast correction screening mode, perform fast correction screening mode processing on the candidate control sequence after textile-specific safety constraints. When the confidence sequence meets the confidence threshold condition corresponding to the conservative control screening mode, the conservative control screening mode is applied to the candidate control sequence after the textile-specific safety constraints, and the screened control parameter adjustment sequence is output.

[0017] Optionally, step seven includes the following specific steps: The selected control parameter adjustment sequence is sent to the industrial controller; The industrial controller generates a sequence of control setpoints for closed-loop regulation based on the control parameter adjustment sequence. The control setpoint sequence includes a fan frequency setpoint sequence, a valve opening setpoint sequence, and a fresh air to return air ratio setpoint sequence. The industrial controller performs PID control on the fan according to the fan frequency setpoint sequence to adjust the fan frequency. The industrial controller performs PID control on the valve according to the valve opening setpoint sequence to adjust the valve opening, and performs PID control on the fresh air and return air ratio according to the fresh air and return air ratio setpoint sequence to adjust the ratio. During the closed-loop regulation process, the industrial controller samples the actual operating status of the fan, valves, and the ratio of fresh air to return air according to the preset control cycle and compares it with the corresponding set value. Based on the comparison result, it updates the control output for the next control cycle.

[0018] The beneficial effects of this invention are: Compared with the prior art, the present invention, based on the technical solution proposed in this invention, achieves significant performance improvement and system optimization in remote monitoring and intelligent control, and achieves the following beneficial effects: This invention introduces an improved PatchTST time-series prediction model, enabling multivariate and high-precision prediction of temperature and humidity trends in textile workshop air conditioning systems. By incorporating prior constraints from textile processes during the inference process, the model's predictive stability and process adaptability in practical applications are effectively enhanced. This predictive capability provides a solid data foundation for forward-looking decision-making in control strategies, overcoming the control bottleneck of traditional systems characterized by "sensing lag and response delay."

[0019] The collaborative design of the state prior construction module and the control parameter generation module enables intelligent mapping from prediction results to control strategies. By introducing a Mamba network to model the state prior information and dynamically generating amplitude and rate gating coefficients based on confidence sequences, the adjustment amplitude and rate of control parameters are effectively constrained. This not only ensures the smoothness and controllability of the control strategy but also enables the system to achieve a more robust dynamic response in the face of complex fluctuating environments.

[0020] This invention designs a multi-level control verification and safety screening mechanism. This includes prediction deviation back-substitution verification in consistency assessment, as well as safety constraints related to textile processes (such as humidity drop / overshoot limits, parameter change rate limits, etc.), and adaptively adjusts the aggressiveness or conservatism of the control strategy through confidence-driven screening modes. This mechanism significantly improves the system's fault tolerance to control failures and uncertainties, ensuring the continuity and stability of the air conditioning system in the textile production process.

[0021] This invention enables end-to-end closed-loop feedback and continuous optimization. By executing control parameters through an industrial controller and transmitting operational feedback data back for the next round of strategy iteration, it achieves data-driven intelligent adaptive control capabilities, which helps reduce energy consumption and improve environmental control accuracy and production efficiency. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0023] Figure 1This is a schematic diagram of the structure of a textile air conditioning remote monitoring and intelligent control system based on the Internet of Things proposed in this invention; Figure 2 This is an overall flowchart of a method for remote monitoring and intelligent control of textile air conditioning based on the Internet of Things proposed in this invention; Figure 3 This is a structural diagram of an improved PatchTST model for a remote monitoring and intelligent control system for textile air conditioning based on the Internet of Things proposed in this invention. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0025] refer to Figures 1-3 A remote monitoring and intelligent control system for textile air conditioning based on the Internet of Things, comprising: IoT data acquisition and preprocessing module: This module is used to install IoT data acquisition devices in the textile workshop to collect real-time data on the operating status of the textile air conditioning system and upload the data to a remote monitoring platform; it also preprocesses the uploaded data to construct multivariate time series samples. The time series prediction module is used to perform inference on the improved PatchTST model based on multivariate time series samples. During the inference stage, the multivariate time series samples are divided into patches, and patch embedding vectors are generated through sub-time series coding units. The patch embedding vectors are subjected to affine modulation and input into the coding network for feature extraction. Without performing model parameter updates, the module outputs the temperature prediction sequence, humidity prediction sequence and corresponding confidence sequence in the future prediction time domain. State Prior Construction Module: Used to construct state prior information based on temperature prediction sequence, humidity prediction sequence, confidence sequence, target temperature and humidity setpoints, and the current operating status of textile air conditioning. Control parameter generation module: It is used to input the state prior information into the Mamba network for sequence modeling and generate gating coefficients based on the confidence sequence to constrain the amplitude and rate of the control parameter adjustment sequence; Consistency verification module: used to perform consistency verification on the control parameter adjustment sequence, generate candidate control sequences based on the control parameter adjustment sequence and perform consistency evaluation; Safety constraint and screening module: used to perform textile-specific safety constraints and screening on candidate control sequences that have passed the consistency check, and output the screened control parameter adjustment sequence; Industrial control and closed-loop regulation module: used to receive the filtered control parameter adjustment sequence, and the industrial controller performs closed-loop regulation on the textile air conditioning system according to the control parameter adjustment sequence.

[0026] In this embodiment, the modules are interconnected using the following method: Step 1: Install IoT data acquisition devices in the textile workshop to collect real-time data on the operating status of the textile air conditioning system and upload the data to the remote monitoring platform; preprocess the uploaded data to construct a multivariate time series sample. Step 2: Before deploying the improved PatchTST model, pre-training is completed based on the running state data, and prior constraints of textile process are introduced during the training phase. During the inference phase, multivariate time series samples are divided into patches and Patch embedding vectors are generated through sub-time series coding units. After affine modulation of the Patch embedding vectors, they are input into the coding network for feature extraction. Without updating the model parameters, the temperature prediction sequence, humidity prediction sequence and corresponding confidence sequence in the future prediction time domain are output. Step 3: Generate the required prior state information based on the temperature prediction sequence, humidity prediction sequence, confidence sequence, target temperature and humidity setpoints, and the current operating status of the textile air conditioning system. Step 4: Input the state prior information into the Mamba network to generate a continuous control parameter adjustment sequence, generate gating coefficients based on the confidence sequence, and constrain the amplitude and rate of the control parameter adjustment sequence according to the gating coefficients; Step 5: Perform consistency verification on the control parameter adjustment sequence, generate candidate control sequences based on the control parameter adjustment sequence and perform consistency evaluation. If the consistency verification does not meet the preset conditions, perform rejection or downgrade processing on the candidate control sequence. Step 6: Perform textile-specific safety constraints and screening on the candidate control sequences that have passed the consistency check, and output the screened control parameter adjustment sequences; Step 7: Send the selected control parameter adjustment sequence to the industrial controller, which will then perform closed-loop regulation of the textile air conditioning system based on the control parameter adjustment sequence.

[0027] In this embodiment, step one includes the following specific steps: Internet of Things (IoT) data collection devices are installed in the textile workshop and connected to the textile air conditioning system; Operational status data includes temperature, humidity, air volume, fan frequency, valve opening, fresh air ratio, power parameters, and operating and alarm status. The operational status data is uploaded to the remote monitoring platform via the communication network for storage and display; The operational status data uploaded to the remote monitoring platform is preprocessed, including data cleaning, outlier removal, time alignment, and resampling. Based on the preprocessed operational status data, multivariate time series samples are constructed according to a preset time window, and prediction input data is generated.

[0028] In this embodiment, step two includes the following specific steps: Before deploying the improved PatchTST model, the improved PatchTST model is pre-trained based on the runtime status data to obtain the pre-trained model parameters; The improved PatchTST model is pre-trained using runtime data before deployment to form stable prior model parameters. During system operation, the model is only modulated based on real-time context information during the inference phase without updating parameters, so that the prediction results and confidence scores output by the model can serve as stable prior inputs for the generation of subsequent control strategies.

[0029] During the training phase of the improved PatchTST model, a priori constraints on textile processes are introduced. These priori constraints are predetermined based on the process stability requirements of environmental temperature and humidity during textile production, and include constraints on the range of environmental parameter values ​​and the rate of change. Among them, the environmental parameter value range constraint is used to limit the predicted temperature and humidity values ​​to be within the range allowed by the textile process; The rate of change constraint is used to limit the magnitude of change between the predicted temperature and humidity values ​​at adjacent time steps, so that the rate of change of the predicted results per unit time does not exceed the allowable change threshold of the textile process. The prior constraints of the textile process are as follows: they are predetermined based on the process stability requirements of environmental temperature and humidity during textile production, including constraints on the range of environmental parameter values ​​and the rate of change; in each training batch of the model training, the temperature prediction value sequence and humidity prediction value sequence output by the model are obtained. The predicted values ​​at each time step in the temperature prediction value sequence are compared with the pre-determined lower and upper limits of temperature. Temperature prediction values ​​that exceed the lower or upper limits of temperature are recorded as range violations and included in the corresponding range constraint terms in the training loss. The predicted values ​​at each time step in the humidity prediction value sequence are compared with the predetermined lower and upper limits of humidity. Humidity prediction values ​​that exceed the lower or upper limits of humidity are recorded as range violations and included in the corresponding range constraint terms in the training loss. In the same training batch, the difference between the predicted values ​​of adjacent time steps in the temperature prediction value sequence is compared with a pre-determined temperature change threshold. Adjacent time step pairs whose difference exceeds the temperature change threshold are recorded as rate violation terms and the corresponding rate of change constraint terms are included in the training loss. The difference between the predicted values ​​of adjacent time steps in the humidity prediction sequence is compared with a predetermined humidity change threshold. Adjacent time step pairs whose difference exceeds the humidity change threshold are recorded as rate violation terms and included in the corresponding rate of change constraint terms in the training loss.

[0030] During the inference phase of the improved PatchTST model, multivariate time series samples are obtained; The multivariate time series sample is divided into N patches in the time dimension according to the preset patch length. Each patch corresponds to a subsequence of running state data continuously extracted from the multivariate time series sample. The subsequence of running status data is input into the sub-temporal coding unit for encoding processing to obtain the internal temporal features of the patch, and the corresponding patch embedding vector is generated based on the internal temporal features of the patch. Feature aggregation processing is performed on the running status data within a preset time window in the multivariate time series samples to obtain a fixed-length vector representation, and the scaling coefficient vector and translation coefficient vector are calculated based on the fixed-length vector representation; The specific process for generating the scaling factor vector and translation factor vector is as follows: The fixed-length vector representation is input into the modulation parameter generation module; the modulation parameter generation module includes at least one linear mapping unit or nonlinear mapping unit for mapping the fixed-length vector representation; the fixed-length vector representation is mapped through a first mapping path to obtain a scaling coefficient vector corresponding to the dimension of the Patch embedding vector; the fixed-length vector representation is mapped through a second mapping path to obtain a translation coefficient vector corresponding to the dimension of the Patch embedding vector.

[0031] In the process of generating the scaling factor vector, a weight mapping is performed on the fixed-length vector representation and a numerical transformation is performed in combination with the bias term. The mapping result is then subjected to numerical range constraints to generate a scaling factor vector for modulating the amplitude of the feature. In the process of generating the translation coefficient vector, a weight mapping independent of the scaling coefficient vector generation process is performed on the fixed-length vector representation, and a numerical transformation is performed in combination with the bias term to generate the translation coefficient vector used to modulate the feature offset.

[0032] The scaling factor vector and translation factor vector are consistent with the patch embedding vector in terms of vector dimension, and are generated step by step in time based on the fixed-length vector representation during the inference phase, without performing model parameter updates during the generation process.

[0033] Affine modulation is performed on the Patch embedding vector, wherein the affine modulation includes multiplying the scaling coefficient vector element-wise with the Patch embedding vector and superimposing the translation coefficient vector. The affine-modulated Patch embedding vector is input into the encoding network for feature extraction, and outputs the temperature prediction sequence, humidity prediction sequence, and confidence sequence corresponding to the temperature prediction sequence and humidity prediction sequence in the future prediction time domain. No model parameter update is performed during the inference phase.

[0034] During the inference phase of the improved PatchTST model, the Patch embedding vector, after being processed by the sub-temporal coding unit and subjected to affine modulation, is input into the coding network. The coding network performs forward feature calculation on the Patch embedding vector based on the model parameters obtained from pre-training, and outputs a coded feature representation that characterizes the state of each time step in the future prediction time domain.

[0035] Based on the encoded feature representation, the encoded features are mapped by the prediction output unit to generate temperature and humidity prediction values ​​corresponding to each time step in the future prediction time domain according to the time dimension. The temperature prediction values ​​are arranged in time order to form a temperature prediction sequence, and the humidity prediction values ​​are arranged in time order to form a humidity prediction sequence.

[0036] During the generation of the temperature prediction sequence and humidity prediction sequence, based on the intermediate features or the output features of the prediction output unit corresponding to the coded feature representation of each time step, confidence values ​​corresponding one-to-one with the temperature prediction value and humidity prediction value are generated synchronously, and the confidence values ​​corresponding to each time step are arranged in chronological order to form a confidence sequence corresponding to the temperature prediction sequence and humidity prediction sequence.

[0037] In the above reasoning process, the improved PatchTST model only performs forward feature calculation, feature mapping and sequence generation operations, and does not perform backpropagation or update operations of model parameters. The temperature prediction sequence, humidity prediction sequence and confidence sequence are all obtained based on the pre-trained model parameters.

[0038] In this embodiment, step three includes the following specific steps: Obtain the target temperature and humidity setpoints and the current operating status of the textile air conditioner; Calculate the temperature prediction deviation sequence based on the temperature prediction sequence and the target temperature and humidity setpoints, calculate the humidity prediction deviation sequence based on the humidity prediction sequence and the target temperature and humidity setpoints, and combine the temperature prediction deviation sequence and the humidity prediction deviation sequence to form the prediction deviation sequence. Obtain equipment operation constraint information, which includes parameter value range constraints and parameter change rate constraints; The strategy is built based on the predicted deviation sequence, confidence sequence, current control parameter status, and equipment operation constraint information to generate the necessary prior state information.

[0039] In this embodiment, step four includes the following specific steps: The Mamba network is used to perform sequence modeling on the state prior information and generate a continuous control parameter adjustment sequence. Inside the Mamba network, the input mapping process is sequentially performed on the state prior information sequence to obtain the input feature sequence for state space recursion, and at each time step, the state update process is performed based on the hidden state of the previous time step and the input features of the current time step to obtain the hidden state of the current time step. Output mapping processing is performed based on the hidden state of each time step. A control parameter adjustment vector sequence corresponding to each time step is generated through a predefined multi-channel output mapping structure. Different output channels correspond to fan frequency adjustment parameters, valve opening adjustment parameters, and fresh air to return air ratio adjustment parameters, respectively, forming a fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence arranged in time order, which constitute a continuous control parameter adjustment sequence.

[0040] The control parameter adjustment sequence includes a fan frequency adjustment sequence, a valve opening adjustment sequence, and a fresh air to return air ratio adjustment sequence. Gating coefficients are generated based on the confidence sequence, and the gating coefficients include amplitude gating coefficients and rate gating coefficients; The process of calculating the gating coefficient based on the confidence sequence includes: for the confidence value corresponding to each time step in the confidence sequence, the confidence value is first input into the gating mapping module; the gating mapping module performs a numerical transformation on the confidence value according to a preset linear mapping rule, that is, multiplies the confidence value with a preset proportional coefficient and adds a preset bias to obtain the gating intermediate value for the corresponding time step; the gating intermediate value is subjected to numerical range restriction processing so that it falls within a preset gating value range, and the processed value is used as the gating coefficient corresponding to that time step.

[0041] Specifically, the confidence values ​​are mapped according to the first preset mapping rule to generate amplitude gating coefficients corresponding to each time step; the confidence values ​​are mapped according to the second preset mapping rule to generate rate gating coefficients corresponding to each time step; the first preset mapping rule and the second preset mapping rule use different scaling factors and biases.

[0042] The adjustment ranges of the fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence are constrained based on the amplitude gating coefficient. The adjustment rates of the fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence are constrained according to the rate gating coefficient to obtain the gated control parameter adjustment sequence.

[0043] In this embodiment, step five includes the following specific steps: Candidate control sequences are generated based on the constrained control parameter adjustment sequence. The candidate control sequences include candidate fan frequency adjustment sequence, candidate valve opening adjustment sequence, and candidate fresh air to return air ratio adjustment sequence. Based on the current control parameter status obtained in step three, the fan frequency adjustment, valve opening adjustment, and fresh air to return air ratio adjustment corresponding to each time step in the gated control parameter adjustment sequence are read. Using the current control parameter status as the initial setpoint, the control parameter adjustment is recursively updated in the time dimension according to the time sequence of the control parameter adjustment, to obtain the fan frequency setpoint sequence, valve opening setpoint sequence, and fresh air to return air ratio setpoint sequence corresponding to each time step. The setpoint sequence is then used as the candidate fan frequency adjustment sequence, candidate valve opening adjustment sequence, and candidate fresh air to return air ratio adjustment sequence to form the candidate control sequence.

[0044] The consistency between the candidate control sequence and the state prior information is evaluated. The consistency assessment includes back-substituting the predicted deviation sequence in the state prior information based on the candidate control sequence to obtain the back-substituting deviation sequence, and then verifying the consistency between the back-substituting deviation sequence and the predicted deviation sequence. When the consistency check does not meet the preset conditions, the candidate control sequence is rejected or downgraded. The rejection process includes stopping the output of the candidate control sequence, and the downgrade process includes reducing the adjustment range and adjustment rate of the fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence in the candidate control sequence. The preset conditions include at least one of the following: the difference in the deviation amplitude between the back-substitution deviation sequence and the predicted deviation sequence at the corresponding time step does not exceed a preset deviation range; the change direction of the back-substitution deviation sequence and the predicted deviation sequence in the time dimension is consistent; and the evolution direction of the back-substitution deviation sequence is consistent with the deviation evolution direction indicated by the predicted deviation sequence.

[0045] When the consistency check meets the preset conditions, the candidate control sequence that passes the consistency check is output.

[0046] In this embodiment, step six includes the following specific steps: Based on parameter value range constraints, the time step control parameters in the candidate control sequences that have passed the consistency check are checked for range, and control parameters that exceed the parameter value range constraints are replaced to generate range-constrained candidate control sequences. Based on the parameter change rate constraint, the rate of change of control parameters in adjacent time steps in the range constraint candidate control sequence is verified, and the change of control parameters in adjacent time steps that exceed the parameter change rate constraint is restricted to generate rate constraint candidate control sequences. The candidate control sequences with rate constraints are subjected to constraint processing to prevent humidity drop and humidity overshoot, resulting in candidate control sequences with textile-specific safety constraints. The constraint processing includes verifying the direction and magnitude of change of control parameters at each time step in the candidate valve opening adjustment sequence and the candidate fresh air to return air ratio adjustment sequence based on the candidate valve opening adjustment sequence and the candidate fresh air to return air ratio adjustment sequence, and restricting the change of control parameters that cause a sudden drop in humidity or humidity overshoot, so as to obtain a candidate control sequence with textile-specific safety constraints.

[0047] Obtain the confidence sequence and compare it with the preset confidence threshold. When the confidence sequence meets the confidence threshold condition corresponding to the fast correction screening mode, perform fast correction screening mode processing on the candidate control sequence after textile-specific safety constraints. The rapid correction and screening mode processing includes, without violating the parameter value range constraints, using a first preset change rate limitation rule and a first preset adjustment range limitation rule to screen candidate fan frequency adjustment sequences, candidate valve opening adjustment sequences, and candidate fresh air to return air ratio adjustment sequences. When the confidence sequence meets the confidence threshold condition corresponding to the conservative control screening mode, the conservative control screening mode is applied to the candidate control sequence after the textile-specific safety constraints, and the screened control parameter adjustment sequence is output.

[0048] The conservative control screening mode processing includes using a second preset rate of change limit rule and a second preset adjustment range limit rule to screen candidate fan frequency adjustment sequences, candidate valve opening adjustment sequences, and candidate fresh air to return air ratio adjustment sequences. The second preset rate of change limit rule and the second preset adjustment range limit rule are more stringent than the first preset rule, and the screened control parameter adjustment sequences are output.

[0049] In this embodiment, step seven includes the following specific steps: The selected control parameter adjustment sequence is sent to the industrial controller; The industrial controller generates a sequence of control setpoints for closed-loop regulation based on the control parameter adjustment sequence. The control setpoint sequence includes a fan frequency setpoint sequence, a valve opening setpoint sequence, and a fresh air to return air ratio setpoint sequence. The industrial controller performs PID control on the fan according to the fan frequency setpoint sequence to adjust the fan frequency. The industrial controller performs PID control on the valve according to the valve opening setpoint sequence to adjust the valve opening, and performs PID control on the fresh air and return air ratio according to the fresh air and return air ratio setpoint sequence to adjust the ratio. During the closed-loop regulation process, the industrial controller samples the actual operating status of the fan, valves, and the ratio of fresh air to return air according to the preset control cycle and compares it with the corresponding set value. Based on the comparison result, it updates the control output for the next control cycle.

[0050] Example 1: To verify the feasibility of this invention in practice, it was applied to a smart textile manufacturing plant. To address the higher demands placed on stable temperature and humidity control in the textile workshop during the hot and humid summer months, the project team deployed an IoT-based remote monitoring and intelligent control system for textile air conditioning. The factory's air conditioning system is a multi-zone, multi-unit collaborative control structure, encompassing multiple actuators such as a central fan, fresh air system, return air system, and electric valves. Its operation is complex, and control strategies rely heavily on manual experience, resulting in issues such as adjustment lag, high energy consumption, and poor product process stability.

[0051] In the first phase of the project, IoT data acquisition devices with edge computing capabilities were deployed at key air conditioning nodes to achieve minute-level real-time acquisition and remote transmission of multi-dimensional data such as temperature, humidity, air supply volume, valve opening, fan frequency, and fresh air to return air ratio. The system automatically preprocesses the collected data, including outlier removal, missing value imputation, time alignment, and sliding window resampling, to construct continuous multivariate time series samples as prediction input.

[0052] In the second phase, the improved PatchTST model was pre-trained based on one week's worth of historical data (a total of 10,080 data points). Prior knowledge of textile processes was incorporated for constraint optimization; for example, the temperature range was constrained to 24°C–27°C, and the humidity range to 60%–70%, with the rate of change controlled to no more than 1.2°C and 5% every 5 minutes, respectively. After training, the model was deployed for online inference. Through patch partitioning and sub-time series encoding, each multivariate time series was efficiently encoded, and affine modulation was used to enhance the predictive adaptability. In actual operation, the model generates a temperature and humidity prediction sequence and corresponding confidence score sequence for the next 60 minutes every 10 minutes.

[0053] In the third stage, the system automatically constructs prior state information based on the current operating status, prediction deviation, confidence level, target setpoint, and constraints. This information is then input into a Mamba network for dynamic sequence modeling, generating a sequence of control parameter adjustments that includes multiple dimensions such as fan frequency, valve opening, and fresh air / return air ratio. Gating coefficients are generated based on the prediction confidence level, and the system applies real-time constraints to the adjustment amplitude and rate to ensure stable and controllable adjustments.

[0054] Under intelligent control, the system performs consistency back-substitution verification on the output control parameters. If the predicted value and the back-substitution error exceed a set threshold, a degradation strategy will be triggered, automatically reducing the control signal. Through continuous screening and safety constraints, the risk of sudden humidity drops caused by a sudden increase in fresh air or excessive valve adjustment is avoided.

[0055] After a week of comparative testing, compared with the traditional fixed-parameter PID control method, the system showed significantly improved stability in temperature and humidity control under typical daytime high-load conditions: the standard deviation of temperature deviation decreased from 1.3°C to 0.4°C, and the standard deviation of humidity deviation decreased from 7.2% to 2.1%. In terms of energy consumption, the fan frequency variation became more stable, resulting in a reduction of approximately 8.5% in power consumption, and an average reduction of 11.3% in the proportion of fresh air intake, effectively suppressing unnecessary cooling load consumption. Simultaneously, the system possesses online learning and closed-loop data upload capabilities, allowing the control strategy to continuously self-optimize based on historical feedback, demonstrating good scalability and sustainable operation capabilities.

[0056] Therefore, the system scheme proposed in this invention achieves remote, intelligent, and stable optimization of textile air conditioning operation under complex actual working conditions, significantly improving the limitations of traditional control methods that rely on manual experience, suffer from control lag, and have weak energy-saving capabilities, thus possessing significant application value. Specific experimental data are shown in Table 1:

[0057] Table 1. Comparison of Air Conditioning Operation Effects under Different Control Methods

[0058] As can be seen from Table 1, after introducing the intelligent control system described in this invention, the textile air conditioning system shows significant optimization in terms of temperature and humidity control accuracy, energy consumption level, response time, and system stability, which fully verifies the effectiveness of this invention in practical applications.

[0059] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A remote monitoring and intelligent control system for textile air conditioning based on the Internet of Things, characterized in that, include: IoT data acquisition and preprocessing module: This module is used to install IoT data acquisition devices in the textile workshop to collect real-time data on the operating status of the textile air conditioning system and upload the data to a remote monitoring platform; it also preprocesses the uploaded data to construct multivariate time series samples. The time series prediction module is used to perform inference on the improved PatchTST model based on multivariate time series samples. During the inference stage, the multivariate time series samples are divided into patches, and patch embedding vectors are generated through sub-time series coding units. The patch embedding vectors are subjected to affine modulation and input into the coding network for feature extraction. Without performing model parameter updates, the module outputs the temperature prediction sequence, humidity prediction sequence and corresponding confidence sequence in the future prediction time domain. State Prior Construction Module: Used to construct state prior information based on temperature prediction sequence, humidity prediction sequence, confidence sequence, target temperature and humidity setpoints, and the current operating status of textile air conditioning. Control parameter generation module: It is used to input the state prior information into the Mamba network for sequence modeling and generate gating coefficients based on the confidence sequence to constrain the amplitude and rate of the control parameter adjustment sequence; Consistency verification module: used to perform consistency verification on the control parameter adjustment sequence, generate candidate control sequences based on the control parameter adjustment sequence and perform consistency evaluation; Safety constraint and screening module: used to perform textile-specific safety constraints and screening on candidate control sequences that have passed the consistency check, and output the screened control parameter adjustment sequence; Industrial control and closed-loop regulation module: used to receive the filtered control parameter adjustment sequence, and the industrial controller performs closed-loop regulation on the textile air conditioning system according to the control parameter adjustment sequence.

2. The textile air conditioning remote monitoring and intelligent control system based on the Internet of Things as described in claim 1, characterized in that, The modules are connected in the following way: Step 1: Install IoT data acquisition devices in the textile workshop to collect real-time data on the operating status of the textile air conditioning system and upload the data to the remote monitoring platform; preprocess the uploaded data to construct a multivariate time series sample. Step 2: Before deploying the improved PatchTST model, pre-training is completed based on the running state data, and prior constraints of textile process are introduced during the training phase. During the inference phase, multivariate time series samples are divided into patches and Patch embedding vectors are generated through sub-time series coding units. After affine modulation of the Patch embedding vectors, they are input into the coding network for feature extraction. Without updating the model parameters, the temperature prediction sequence, humidity prediction sequence and corresponding confidence sequence in the future prediction time domain are output. Step 3: Generate the required prior state information based on the temperature prediction sequence, humidity prediction sequence, confidence sequence, target temperature and humidity setpoints, and the current operating status of the textile air conditioning system. Step 4: Input the prior state information into the Mamba network to generate a continuous control parameter adjustment sequence, generate gating coefficients based on the confidence sequence, and constrain the amplitude and rate of the control parameter adjustment sequence according to the gating coefficients; Step 5: Perform consistency verification on the control parameter adjustment sequence, generate candidate control sequences based on the control parameter adjustment sequence and perform consistency evaluation. If the consistency verification does not meet the preset conditions, perform rejection or downgrade processing on the candidate control sequence. Step 6: Perform textile-specific safety constraints and screening on the candidate control sequences that have passed the consistency check, and output the screened control parameter adjustment sequences; Step 7: Send the selected control parameter adjustment sequence to the industrial controller, which will then perform closed-loop regulation of the textile air conditioning system based on the control parameter adjustment sequence.

3. The textile air conditioning remote monitoring and intelligent control system based on the Internet of Things according to claim 2, characterized in that, Step one includes the following specific steps: Internet of Things (IoT) data collection devices are installed in the textile workshop and connected to the textile air conditioning system; Operational status data includes temperature, humidity, air volume, fan frequency, valve opening, fresh air ratio, power parameters, and operating and alarm status. The operational status data is uploaded to the remote monitoring platform via the communication network for storage and display; The operational status data uploaded to the remote monitoring platform is preprocessed, including data cleaning, outlier removal, time alignment, and resampling. Based on the preprocessed operational status data, a multivariate time series sample is constructed according to a preset time window.

4. The remote monitoring and intelligent control system for textile air conditioner based on Internet of Things according to claim 2, characterized in that, The process of pre-training the improved PatchTST model based on runtime data and introducing prior constraints on textile processes during the training phase before deployment includes the following specific steps: Before deploying the improved PatchTST model, the improved PatchTST model is pre-trained based on the runtime status data to obtain the pre-trained model parameters; During the training phase of the improved PatchTST model, a priori constraints on textile processes are introduced. These priori constraints are predetermined based on the process stability requirements of environmental temperature and humidity during textile production, and include constraints on the range of environmental parameter values ​​and the rate of change. During the training phase, the temperature and humidity predictions output by the improved PatchTST model are compared with the predetermined process allowable range, and predictions that exceed the process allowable range are included in the value range constraint term. The magnitude of change between the predicted temperature and humidity values ​​at adjacent time steps is compared with a predetermined process change threshold, and the magnitude of change exceeding the process change threshold is included in the rate of change constraint term. 5.The remote monitoring and intelligent control system based on the Internet of Things for the textile air conditioner according to claim 2, characterized in that, The process of dividing multivariate time series samples into patches during the inference phase and generating patch embedding vectors through sub-temporal coding units, applying affine modulation to the patch embedding vectors, inputting them into the coding network for feature extraction, and outputting the temperature prediction sequence, humidity prediction sequence, and corresponding confidence sequence in the future prediction time domain without updating model parameters includes the following specific steps: During the inference phase of the improved PatchTST model, multivariate time series samples are obtained; The multivariate time series sample is divided into N patches in the time dimension according to the preset patch length. Each patch corresponds to a subsequence of running state data continuously extracted from the multivariate time series sample. The subsequence of running status data is input into the sub-temporal coding unit for encoding processing to obtain the internal temporal features of the patch, and the corresponding patch embedding vector is generated based on the internal temporal features of the patch. Feature aggregation processing is performed on the running status data within a preset time window in the multivariate time series samples to obtain a fixed-length vector representation, and the scaling coefficient vector and translation coefficient vector are calculated based on the fixed-length vector representation; Affine modulation is performed on the Patch embedding vector, wherein the affine modulation includes multiplying the scaling coefficient vector element-wise with the Patch embedding vector and superimposing the translation coefficient vector. The affine-modulated Patch embedding vector is input into the encoding network for feature extraction, and the output is the temperature prediction sequence, humidity prediction sequence, and confidence sequence corresponding to the temperature prediction sequence and humidity prediction sequence in the future prediction time domain. No model parameter update is performed during the inference stage.

6. The remote monitoring and intelligent control system for textile air conditioner based on Internet of Things according to claim 2, characterized in that, Step three includes the following specific steps: Obtain the target temperature and humidity setpoints and the current operating status of the textile air conditioner; Calculate the temperature prediction deviation sequence based on the temperature prediction sequence and the target temperature and humidity setpoints, calculate the humidity prediction deviation sequence based on the humidity prediction sequence and the target temperature and humidity setpoints, and combine the temperature prediction deviation sequence and the humidity prediction deviation sequence to form the prediction deviation sequence. Obtain equipment operation constraint information, which includes parameter value range constraints and parameter change rate constraints; The strategy is built based on the predicted deviation sequence, confidence sequence, current control parameter status, and equipment operation constraint information to generate the necessary prior state information.

7. The remote monitoring and intelligent control system for textile air conditioner based on Internet of Things according to claim 2, characterized in that, Step four includes the following specific steps: The Mamba network is used to perform sequence modeling on the state prior information and generate a continuous control parameter adjustment sequence. The control parameter adjustment sequence includes a fan frequency adjustment sequence, a valve opening adjustment sequence, and a fresh air to return air ratio adjustment sequence. Gating coefficients are generated based on the confidence sequence, and the gating coefficients include amplitude gating coefficients and rate gating coefficients; The adjustment ranges of the fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence are constrained based on the amplitude gating coefficient. The adjustment rates of the fan frequency adjustment sequence, valve opening adjustment sequence, and fresh air to return air ratio adjustment sequence are constrained according to the rate gating coefficient to obtain the gated control parameter adjustment sequence. 8.The remote monitoring and intelligent control system based on the Internet of Things for the textile air conditioner according to claim 2, characterized in that, Step five includes the following specific steps: Candidate control sequences are generated based on the constrained control parameter adjustment sequence. The candidate control sequences include candidate fan frequency adjustment sequence, candidate valve opening adjustment sequence, and candidate fresh air to return air ratio adjustment sequence. The consistency between the candidate control sequence and the state prior information is evaluated. The consistency assessment includes back-substituting the predicted deviation sequence in the state prior information based on the candidate control sequence to obtain the back-substituting deviation sequence, and performing consistency verification between the back-substituting deviation sequence and the predicted deviation sequence. When the consistency check fails to meet the preset conditions, the candidate control sequence is rejected or downgraded. When the consistency check meets the preset conditions, the candidate control sequence that passes the consistency check is output.

9. A remote monitoring and intelligent control system for textile air conditioning based on the Internet of Things as described in claim 2, characterized in that, Step six includes the following specific steps: Based on parameter value range constraints, the time step control parameters in the candidate control sequences that have passed the consistency check are checked for range, and control parameters that exceed the parameter value range constraints are replaced to generate range-constrained candidate control sequences. Based on the parameter change rate constraint, the rate of change of control parameters in adjacent time steps in the range constraint candidate control sequence is verified, and the change of control parameters in adjacent time steps that exceed the parameter change rate constraint is restricted to generate the rate constraint candidate control sequence. The candidate control sequences with rate constraints are subjected to constraint processing to prevent humidity drop and humidity overshoot, resulting in candidate control sequences with textile-specific safety constraints. Obtain the confidence sequence and compare it with the preset confidence threshold. When the confidence sequence meets the confidence threshold condition corresponding to the fast correction screening mode, perform fast correction screening mode processing on the candidate control sequence after textile-specific safety constraints. When the confidence sequence meets the confidence threshold condition corresponding to the conservative control screening mode, the conservative control screening mode is applied to the candidate control sequence after the textile-specific safety constraints, and the screened control parameter adjustment sequence is output. 10.The remote monitoring and intelligent control system based on the Internet of Things for the textile air conditioner according to claim 2, characterized in that, Step seven includes the following specific steps: The selected control parameter adjustment sequence is sent to the industrial controller; The industrial controller generates a sequence of control setpoints for closed-loop regulation based on the control parameter adjustment sequence. The control setpoint sequence includes a fan frequency setpoint sequence, a valve opening setpoint sequence, and a fresh air to return air ratio setpoint sequence. The industrial controller performs PID control on the fan according to the fan frequency setpoint sequence to adjust the fan frequency. The industrial controller performs PID control on the valve according to the valve opening setpoint sequence to adjust the valve opening, and performs PID control on the fresh air and return air ratio according to the fresh air and return air ratio setpoint sequence to adjust the ratio. During the closed-loop regulation process, the industrial controller samples the actual operating status of the fan, valves, and the ratio of fresh air to return air according to the preset control cycle and compares it with the corresponding set value. Based on the comparison result, it updates the control output for the next control cycle.

Images