A clean room control system based on multi-objective and multi-level model predictive control

Abstract

The present invention relates to an intelligent control system and control method for room pressure in a clean workshop, which aims to achieve precise regulation of room pressure and flow in a clean workshop through an automated state-space model and a multi-level model predictive control algorithm, thereby ensuring the stability of the room pressure difference. The control system includes a data processing module, a modeling and prediction module, a control strategy selection module, a real-time feedback adjustment module, a prediction and optimization allocation module, a control signal output module, an actual execution module, and a physical signal feedback module; the control strategy selection module includes a room pressure control module and an air valve flow control module, which generate global and refined control signals based on constrained MPC and nonlinear MPC algorithms, respectively; the control signal output module transmits the optimized signal to the actual execution module to adjust the opening of the supply and return exhaust valves to achieve precise pressure and flow control. The system of the present invention has high precision and rapid response capabilities, meeting the high requirements of clean workshops.

Description

Technical Field

[0001] The present invention relates to an automated control technology for a clean workshop, in particular to an intelligent pressure control method for a clean workshop based on a multi-objective hierarchical model predictive control (MPC) algorithm. Background Art

[0002] Cleanrooms often employ a traditional hierarchical control approach, where pressure control is typically divided into different zones or rooms. The upper level controls the overall pressure differential within each room, and further controls local equipment within the room to achieve precise pressure regulation. While this traditional hierarchical control approach can meet cleanroom requirements to a certain extent, it has significant limitations when dealing with multivariable coupling and nonlinear systems. For example, in complex environments where multiple rooms interact, traditional hierarchical control suffers from slow response speed and low regulation accuracy, making it difficult to ensure real-time stability of room pressure.

[0003] Cleanroom pressure control systems often involve the coupling of multiple variables, such as room pressure, damper opening, and air volume. In particular, the nonlinear relationship between damper opening and air volume makes it difficult for traditional linear control methods to achieve precise control within the full opening range. Furthermore, cleanrooms may be affected by changes in the external environment or other room coupling effects, requiring the control system to have adaptive capabilities to cope with dynamic changes. This complex nonlinear dynamic response makes it difficult for traditional PID or linear MPC control to effectively adapt, resulting in a decrease in system control accuracy.

[0004] In recent years, model predictive control (MPC) has been introduced for cleanroom pressure control to improve control accuracy and response speed. However, existing MPC methods often employ a single-level control strategy in practical applications. Specifically, they employ constrained optimal control (MPC) for overall room pressure control, but often use simple linear methods for specific damper control, failing to fully consider the nonlinear relationship between dampers and flow. This single-level MPC control scheme exhibits significant control deficiencies when dealing with nonlinear actuators, making it difficult to achieve precise regulation of pressure and flow over a wide operating range. Summary of the Invention

[0005] To address these issues, cleanroom pressure control systems urgently require a multi-objective, hierarchical control architecture to optimize different control objectives. Specifically, constrained MPC control is employed at the upper level to ensure that the overall room pressure meets cleanliness requirements, while simultaneously performing nonlinear MPC control of the damper opening and flow at the lower level to accommodate the characteristics of nonlinear actuators. This innovative multi-objective, hierarchical MPC control architecture not only provides higher response accuracy under multivariable coupling conditions, but also ensures that the system maintains pressure stability in the cleanroom environment during dynamic changes.

[0006] The present invention provides an intelligent control system for clean room pressure. Through a multi-level model predictive control (MPC) algorithm, combined with the automated modeling and dynamic updating of the state space model, it achieves precise control of the pressure and flow in the clean room. This improves the control accuracy and response speed while maintaining a stable pressure differential in the clean room, thus meeting the strict requirements of the clean environment.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] An intelligent control system for room pressure in a clean room includes a data processing module, a modeling and prediction module, a control strategy selection module, a prediction and optimization allocation module, a control signal output module, an actual execution module, a physical signal feedback module, and a real-time feedback adjustment module, wherein:

[0009] The data processing module includes an initialization module and a data acquisition module; the initialization module is used for the initial setup of the system and the acquisition of baseline data; wherein the baseline data acquisition refers to the acquisition of the initial environmental parameters and external factors of the clean room when the system is started, the establishment of the initial state space model and the setting of the initial control parameters; the data acquisition module is used to collect the dynamic environmental data of the clean room in real time, including the supply air volume, return air volume and room pressure;

[0010] The modeling and prediction module is used to generate a state space model based on the data of the data processing module and dynamically update the state space model through an adaptive algorithm;

[0011] The control strategy selection module includes a room pressure control module and an air valve flow control module; the room pressure control module, based on a constrained model predictive control algorithm, generates a global control signal for regulating the overall pressure of the room. This control signal directly acts on the supply air valve execution module and the return and exhaust air valve execution module to achieve overall pressure difference control of the room; the air valve flow control module, based on a nonlinear model predictive control algorithm, generates a refined control signal for individually adjusting the opening and flow of the supply air valve, ensuring precise regulation of the air supply volume in the room, thereby optimizing the stability of the room pressure and flow at a micro level;

[0012] The prediction optimization allocation module is used to receive the control signal from the control strategy selection module and dynamically allocate the control signal to different actuators after optimization;

[0013] The real-time feedback adjustment module is used to monitor the real-time feedback data in the prediction optimization allocation module and dynamically adjust the control signal of the control strategy selection module, so that the system can quickly respond to changes in the external environment and realize the real-time flow of feedback data and model update;

[0014] The control signal output module is used to transmit the optimized and allocated control signal to the actual execution module;

[0015] The actual execution module includes an air supply valve execution module and a return and exhaust air valve execution module; the air supply valve execution module is used to receive the global control signal from the room pressure control module and the refined control signal from the air valve flow control module, adjust the opening of the air supply valve, and realize the air supply control of the clean workshop; the return and exhaust air valve execution module is used to receive the global control signal from the room pressure control module, adjust the opening of the return and exhaust air valve, and realize the control of the overall pressure difference of the room;

[0016] The physical signal feedback module is used to collect physical feedback data generated after execution, including pressure feedback and flow feedback, and send the feedback data to the data processing module to adjust the control strategy in real time.

[0017] The specific control steps of the room pressure control module and the air valve flow control module of the control strategy selection module are as follows:

[0018] The room pressure control module generates a global control signal based on the constrained MPC algorithm, including obtaining the deviation between the current room pressure and the target pressure value, calculating the pressure adjustment amount, generating a global control signal, and sending the signal to the supply air valve execution module and the return air valve execution module for adjusting the supply air volume and exhaust air volume;

[0019] The air valve flow control module generates a refined control signal based on the nonlinear MPC algorithm, including obtaining the current opening of the air supply valve and the actual flow data, calculating the deviation between the target air supply flow and the actual flow, generating a refined flow control signal, and sending the signal to the air supply valve execution module for adjusting the opening of the air supply valve to accurately control the air supply flow.

[0020] The state space model generated by the modeling and prediction module consists of the following two parts:

[0021] The room pressure control model is used to describe the dynamic characteristics of the overall room pressure and adjust the global control variables of the supply air valve and the return air valve. The state vector, control input, and output vector of the room pressure control model are defined as follows:

[0022] The state vector x1(t), which includes the room pressure value (p(t)) and the pressure change rate Expressed as:

[0023]

[0024] Control input u1(t), including the air volume of the air supply system (q s (t)) and the air volume of the return air system (q r(t)), which are used to adjust the positive and negative pressures of the room, respectively, and are expressed as:

[0025] u1(t)=[q s (t) q r (t)] T

[0026] The output vector y1(t) is the real-time pressure value p(t) of the room, which is used to compare with the target pressure p ref (t) comparison, expressed as:

[0027] y1(t)=p(t)

[0028] Equation of state:

[0029] x1(t+1)=A1x1(t)+B1u1(t)+w1(t)

[0030] Where: A1 is the system matrix, used to describe the dynamic characteristics of room pressure and its gradient changes; B1 is the input matrix, used to describe the regulatory effect of supply and return air on room pressure; w1(t) is the system noise, including instantaneous pressure fluctuations caused by opening doors and windows and ambient pressure gradient fluctuations caused by starting and stopping equipment fans;

[0031] Output equation:

[0032] y1(t)=C1x1(t)+v1(t)

[0033] Where: C1 is the output matrix used to extract the real-time pressure value of the room; v1(t) is the measurement noise, which is mainly the measurement error of the pressure sensor;

[0034] The damper flow control model is used to describe the dynamic characteristics of the air supply valve. By fine-tuning the opening of the air supply valve, precise control of the air supply flow can be achieved. The state vector, control input, and output vectors of the damper flow control model are defined as follows:

[0035] The state vector x2(t), including the air volume q of the air supply system s (t) and the opening of the air supply valve θ s (t), expressed as:

[0036] x2(t)=[q s (t)θ s (t)] T

[0037] Control input u2(t), represents the adjustment amount Δθ of the air supply valve opening s (t), expressed as:

[0038] u2(t)=Δθ s (t)

[0039] Output vector y2(t), representing the actual air supply volume q at the current moment s (t), used to match the target air supply volume q s , ref (t) comparison, expressed as:

[0040] y2(t)=q s (t)

[0041] Equation of state:

[0042] x2(t+1)=A2x2(t)+B2u2(t)+w2(t)

[0043] Where: A2 is the system matrix, used to describe the dynamic characteristics of the air supply valve; B2 is the input matrix, used to describe the regulating effect of the air supply valve opening on the air supply volume; w2(t) is the system noise, including nonlinear interference such as air supply duct vibration and ambient temperature and humidity fluctuations;

[0044] Output equation:

[0045] y2(t)=C2x2(t)+v2(t)

[0046] Where: C2 is the output matrix used to extract the air supply volume; v2(t) is the measurement noise, which is mainly the measurement error of the air volume sensor.

[0047] The steps for the room pressure control module to generate the global control signal based on the constrained MPC algorithm are:

[0048] Establish a prediction model with constrained MPC algorithm:

[0049] x1(t+k+1|t)=A1·x1(t+k|t)+B1·u1(t+k|t)+w1(t+k|t)

[0050] y1(t+j|t)=C1·x1(t+k|t)+v1(t+k|t)

[0051] where x1(t+k+1|t) represents the predicted state of the system at time t+k+1 given the system state and control input known at time t, where k = 0, 1, ..., N p -1, N p represents the number of time steps for future prediction; u1(t+k|t), y1(t+k|t), w1(t+k|t) and v1(t+k|t) represent the control input, control output, system noise and measurement error of the system at time t+k, respectively, when the system state and control input are known at the current time t;

[0052] The optimization objective function J is:

[0053]

[0054] Among them, J1 is the optimization objective function value, which is the target to be minimized. Its value reflects the control performance of the system in the prediction time domain and consists of two parts: the output error term and the weight term of the control input; y1(t+k|t) is the predicted system output room pressure value at t+k when the current time t is known; y 1,ref is the target pressure value; Q is the output error weight matrix, which is used to weigh the deviation between the target pressure and the actual pressure. It is a positive semidefinite matrix. The larger the weight value, the more important the error in this dimension is. Q is adjusted according to the pressure stability requirements in the system design; u1(t+k|t) is the control input vector, which means that when t is known, the control variables at time t+k are predicted, which are the supply air volume and return air volume; R is the control input weight matrix, which is used to limit the range of control input changes. It is a positive definite matrix. The larger the value, the stricter the limit on signal changes.

[0055] Constraints include pressure range constraints, input amplitude constraints, and input change rate constraints:

[0056] Pressure range constraints:

[0057] p min ≤y1(t+k|t)≤p max

[0058] Among them, p min and p max is the allowable room pressure range, indicating the minimum and maximum values ​​respectively;

[0059] Input amplitude constraint:

[0060] q s,min ≤q s (t+k|t)≤q s,max

[0061] q r,min ≤q r (t+k|t)≤q r,max

[0062] Among them, q s,min is the minimum value of air supply volume; q s,max is the maximum value of air supply volume; q s (t+k|t) is the predicted state of the air supply volume at time t+k based on time t; q r,min is the minimum value of return air volume; q r,max is the maximum value of return air volume; q r (t+k|t) is the predicted state of return air volume at time t+k based on time t;

[0063] Input change rate constraint:

[0064] Δq s,min ≤q s (t+k|t)-q s (t+k-1|t)≤Δq s,max

[0065] Δq r,min ≤q r (t+k|t)-q r (t+k-1|t)≤Δq r,max

[0066] Where Δq s,min The minimum difference between the air supply volume at two predicted moments; Δq s,max The maximum difference between the air supply volume at two predicted moments; Δq r,min The minimum difference in return air volume between the two predicted moments; Δq r,max It is the maximum value of the difference in return air volume between the two predicted moments.

[0067] The steps for the air valve flow control module to generate refined control signals based on the nonlinear MPC algorithm are:

[0068] Establish a prediction model for the nonlinear MPC algorithm:

[0069] x2(t+1|t)=f(x2(t),u2(t))

[0070] y²(t)=h(x²(t))

[0071] Among them, x2(t+1|t) represents the predicted state of the system at time t+1 under the condition of the system state and control input known at the current time t; f(x2(t),u2(t)) represents the nonlinear function of state transition, the specific form of which is obtained by actual physical modeling or actual data-driven method, x2(t) is the internal pressure value and pressure change rate of the system at the current time t, u2(t) is the control input vector of the system, including the supply air volume and return air volume; y2(t) is the output vector of the system, which represents the measured output value of the system at the current time t, including the room pressure value; h(x2(t)) represents the nonlinear relationship between output and state;

[0072] Optimize the objective function J f for:

[0073]

[0074] Among them, J2 is the optimization objective function value, and the air supply volume is the target to be minimized; y2(t+k|t) is the predicted air supply volume value at the current time t predicted at the future time t+k; y 2,refis the target reference value of the air supply volume;

[0075] Q is the output error weight matrix, which is used to weigh the deviation between the predicted value and the target value of the air supply volume; Δ is the change, Δu2(t+k|t) is the change value of the air supply volume at the prediction moment; R is the control input weight matrix, which is used to limit the adjustment range of the air supply volume.

[0076] Use Newton's method to directly optimize the objective function J2;

[0077] The Newton method prioritizes setting the convergence threshold and establishing the objective function gradient. Use the Hessian matrix to calculate, and finally update the formula of u according to the convergence condition;

[0078] Constraints include flow range constraints and control input range constraints:

[0079] Flow range constraints:

[0080] q s,min ≤q s (t+k|t)≤q s,max

[0081] Where: q s,min is the minimum value of air supply volume; q s,max is the maximum value of air supply volume; q s (t+k|t) is the predicted state of the air supply volume at t+k based on time t;

[0082] Control input range constraints:

[0083] θ s,min ≤θ s (t+k|t)≤θ s,max

[0084] Where: θ s,min is the minimum opening of the air supply valve; θ s,max is the maximum opening of the air supply valve;

[0085] θ s (t+k|t) is the predicted state of the air supply valve opening at t+k based on time t.

[0086] The prediction optimization allocation module integrates the global and refined control signals to generate the final control instructions, which act on the air supply valve and return and exhaust valve execution modules. The final air valve control signal u final (t) is:

[0087] u final (t)=u1(t)+u2(t)

[0088] Among them, u1(t) is the global control signal generated by the pressure control module, and u2(t) is the refined control signal output by the air valve flow control module.

[0089] The global control signal u1(t) generated by the pressure control module and the refined control signal u2(t) output by the air valve flow control module are respectively:

[0090]

[0091] Among them, q s (t) is the global air supply volume; q r (t) is the global return air volume;

[0092] u2(t)=Δθ s (t)

[0093] Where Δθ s (t) is the opening adjustment of the air supply valve.

[0094] The data processing module further includes an external factor input interface for collecting data on external interference factors in the clean room environment, including temperature, humidity and other room pressure differences.

[0095] The present invention also discloses a method for intelligently controlling the pressure of a clean room. The method is based on the above-mentioned intelligent control system for the pressure of a clean room. The method comprises the following steps:

[0096] Step 1: Initialize the system through the data processing module and collect the baseline data of the clean room;

[0097] Step 2: Generate the initial state space model using the modeling and prediction module, and update the model by acquiring real-time data through the data acquisition module;

[0098] Step 3: The room pressure control module generates a global control signal based on the constrained MPC algorithm. Specifically, it obtains the deviation between the current room pressure and the target pressure, calculates the adjustment amount and generates a global control signal, and sends the signal to the supply air valve execution module and the return air exhaust valve execution module.

[0099] Step 4: The air valve flow control module generates a refined control signal based on the nonlinear MPC algorithm. Specifically, it obtains the current opening and actual flow of the air supply valve, calculates the flow deviation and generates a refined control signal, which is sent to the air supply valve execution module.

[0100] Step 5: The prediction optimization allocation module optimizes the control signal and allocates it to the corresponding actuator;

[0101] Step 6: The control signal output module transmits the optimized control signal to the actual execution module;

[0102] Step 7: The air supply valve execution module and the return and exhaust air valve execution module adjust the valve opening according to the control signal to adjust the supply air volume and return air volume;

[0103] Step 8: The physical signal feedback module collects the feedback data of the room pressure and flow after execution and sends it to the real-time feedback adjustment module to adjust the control strategy according to the real-time feedback;

[0104] Step 9: Return to step 2 for loop control to achieve adaptive control of the pressure and flow in the clean room.

[0105] The control signal output module is used to transmit the optimization signal generated by the prediction optimization allocation module to the actual execution module, so as to realize the effective output of the signal and the accurate issuance of the control instructions.

[0106] The actual execution modules include the supply air valve execution module and the return and exhaust air valve execution module, which receive control signals from the control strategy selection module and the predictive optimization allocation module, respectively. The supply air valve execution module adjusts the opening of the supply air valve based on the control signal to control the supply air volume. The return and exhaust air valve execution module also adjusts the opening of the return and exhaust air valve based on the control signal to regulate the return air volume, ensuring that the overall pressure control target of the room is achieved.

[0107] The physical signal feedback module is used to collect pressure feedback data and flow feedback data in the room, and feed this data back to the data processing module and the real-time feedback adjustment module so as to continuously adjust and optimize during the execution of the control strategy.

[0108] The significant benefit of this invention lies in its ability to precisely regulate pressure and flow within cleanroom rooms through the combination of automated state-space modeling and a multi-level MPC control strategy. This system improves system control accuracy and response speed while ensuring stable pressure differentials across rooms, meeting the stringent requirements of a cleanroom environment. Compared to traditional control systems, this system is more adaptable to complex multivariable coupling environments and exhibits superior real-time and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] Figure 1 It is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0110] The specific embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0111] like Figure 1As shown, the cleanroom intelligent pressure control system of the present invention consists of a data processing module, a modeling and prediction module, a control strategy selection module, a real-time feedback adjustment module, a prediction and optimization allocation module, a control signal output module, an actual execution module, and a physical signal feedback module. The system's primary function is to precisely control pressure and flow within the cleanroom using a multi-level model predictive control (MPC) algorithm and an automated state-space model, ensuring stable room pressure differentials and flow rates.

[0112] The data processing module includes an initialization module and a data acquisition module. The initialization module collects baseline data from the cleanroom during system startup and obtains initial environmental parameters (such as the room's initial pressure, supply air volume, and return air volume) to provide basic data for subsequent control strategies. The data acquisition module collects real-time data during system operation, obtaining environmental data such as the cleanroom's supply air volume, return air volume, room pressure, temperature and humidity, and transmits this data to the modeling and prediction module and the control strategy selection module for model prediction and control strategy calculation.

[0113] The modeling and prediction module is used to automatically build the state space model and dynamically update it based on real-time data. The automated state space model includes three parts: data collection and initialization, automated modeling, and state variable update:

[0114] Data acquisition and initialization: When the system starts up, the data processing module collects room environmental data through the initialization module and the data acquisition module. This includes the room's initial pressure, target pressure, supply air volume, return air volume, and external disturbances that affect the room pressure (such as door and window opening and closing status, and the pressure coupling effect of other rooms). This data serves as the basis for subsequent state-space modeling and control.

[0115] Automated modeling involves the modeling and prediction module automatically generating a state-space model describing the dynamic behavior of room pressure based on collected data. The model uses room pressure and its rate of change as core state variables. By analyzing the impact of the supply and return air systems on pressure, it automatically establishes a mapping relationship between state variables and control inputs. The system analyzes the response relationship between supply and return air valve openings and room pressure based on historical data, extracting dynamic characteristics and constructing a state-space model. During operation, the modeling and prediction module dynamically updates model parameters based on real-time data, ensuring that the model accurately reflects the actual dynamic characteristics of the room.

[0116] The state variable is updated at each time step. The modeling and prediction module calculates the state update value of the system based on the latest room pressure, supply air volume and return air volume data, and provides the updated model to the control strategy selection module to provide support for the hierarchical MPC algorithm.

[0117] The control strategy selection module includes a room pressure control module and an air valve flow control module, which respectively achieve precise control of the overall room pressure and the supply air valve opening. Based on a constrained MPC algorithm, the room pressure control module first obtains the deviation between the current room pressure and the target pressure, calculates the pressure adjustment amount, generates a global control signal, and sends this signal to the supply air valve execution module and the return and exhaust air valve execution module to adjust the supply and exhaust air volumes to ensure that the overall room pressure remains stable at the set value. The air valve flow control module generates refined control signals based on a nonlinear MPC algorithm, obtains the opening and actual flow data of the supply air valve, calculates the deviation between the target supply air flow and the actual flow, and generates a flow control signal for adjusting the supply air valve opening to achieve precise control of the supply air volume. Through a hierarchical control strategy, the system can coordinate the regulation of pressure and flow at both the macro and micro levels.

[0118] The MPC algorithm performs prediction and control based on a state-space model. The system uses the MPC algorithm to predict room pressure trends within a specified prediction horizon (5 to 30 seconds in the future) and makes control decisions based on the target pressure setpoint. This system specifically includes a room pressure control module (global control with constrained MPC) and a damper flow control module (fine-grained control with nonlinear MPC). The following describes the implementation of each component:

[0119] The room pressure control module uses a constrained MPC algorithm to calculate global control signals for the supply and return air valves to stabilize the overall room pressure. Based on the target pressure and real-time pressure data, the system generates global control signals to adjust the supply and return air volumes. The system sets the target pressure value through the user and, combined with the current room pressure, determines the adjustment range. Constraints are then set based on the system's maximum allowable supply and return air volumes and the range of valve opening to avoid excessive fluctuations or over-limit behavior during the control process. Finally, an optimization calculation is performed to obtain the optimal supply and return air control signals. The optimization results generate global control signals that directly act on the supply and return air valve actuation modules.

[0120] The system collects historical pressure data for a room, including external interference factors (such as door and window openings, HVAC system operating status, etc.), as well as the pressure coupling effects of other rooms. It analyzes this data using a machine learning algorithm to automatically generate a state-space model that describes the room's pressure changes.

[0121] During model predictive control, the state-space model is not fixed. The system continuously updates the model based on real-time pressure data and historical data, using online learning algorithms to dynamically modify model parameters to ensure that the model accurately reflects the dynamic changes in the system. This automated model update mechanism ensures that the system maintains a high level of control accuracy even in dynamic environments and nonlinear changes.

[0122] The control strategy selection module includes a room pressure control module and a damper flow control module, which are responsible for different control objectives respectively.

[0123] The room pressure control module generates a global control signal based on the constrained MPC algorithm, which specifically includes obtaining the deviation between the current room pressure and the target pressure value, calculating the pressure adjustment amount, generating a global control signal, and sending the signal to the supply air valve execution module and the return air valve execution module to adjust the supply air volume and exhaust air volume to ensure that the overall pressure of the room is maintained at the set value.

[0124] The MPC algorithm predicts and controls the room pressure over a given forecast horizon (5 to 30 seconds in the future) and makes control decisions based on the preset pressure target.

[0125] The specific process includes: generating a prediction model based on the state-space model, using the MPC algorithm to predict the pressure changes of the system over a period of time in the future and generate a prediction curve; determining the optimal control input sequence u(t) over a period of time in the future by solving a constrained optimization problem, so that the system pressure gradually approaches and remains near the target pressure value. The optimization goal is to minimize the deviation between the target pressure and the actual pressure, while constraining the control input (such as the rate of change and size of the air supply volume) to remain within the system's acceptable range. At each time step, the MPC algorithm recalculates the control strategy based on the latest measurement data to ensure that the system can respond to changes in the external environment in real time.

[0126] The real-time feedback adjustment module is used to receive feedback data from the physical signal feedback module in real time, and dynamically adjust the control strategy based on the actual pressure and flow feedback data, including adjusting the intensity or response speed of the control signal, to quickly respond to environmental changes and execution errors, ensuring that the pressure difference and flow in the clean room are always maintained within the set target range.

[0127] The predictive optimization allocation module is used to receive the global and refined control signals from the control strategy selection module, distribute the control signals through the optimization algorithm, and transmit the optimized signals to the supply and return air actuators respectively to achieve comprehensive control of room pressure and flow.

[0128] The global control signal is mainly used to adjust the overall pressure of the room. It is obtained by combining the room target pressure setting value and the actual pressure state through constrained MPC solution:

[0129]

[0130] Among them, q s (t) is the global air supply volume; q r(t) is the global return air volume;

[0131] The refined control signal is used to further refine the air supply control. Based on the optimization results of nonlinear MPC, the opening adjustment signal acting on the air supply valve is generated:

[0132] u2(t)=Δθ s (t)

[0133] Where Δθ s (t) is the opening adjustment of the air supply valve.

[0134] The mechanism for collaboratively generating control signals involves fusing global and refined signals and feedback adjustment;

[0135] The fusion of global and refined signals provides the basic adjustment quantity q for the global control signal s (t), the refined control signal u2(t) is further corrected to ensure the control accuracy; the final air valve control signal is u final (t)=u1(t)+u2(t);

[0136] Feedback adjustment adjusts global and refined control signals based on real-time feedback to ensure that the overall room pressure and air flow meet target requirements simultaneously;

[0137] The global control signal provides the basic adjustment amount, and the refined control signal is further corrected to ensure control accuracy and ultimately ensure the air valve control signal.

[0138] The air valve flow control module uses a nonlinear MPC algorithm to achieve refined control of the air supply valve opening. This is primarily used to address the nonlinearity of the air supply valve and further improve the accuracy of flow regulation. The flow control module calculates the deviation between real-time air supply flow data and the target air supply flow and generates a refined control signal. The data acquisition module obtains the current air supply valve opening and actual flow rate, recording the flow deviation. Based on the nonlinear relationship between air supply flow and air supply valve opening, the opening is then fine-tuned to ensure that the air supply flow meets the target value.

[0139] Implementation of hierarchical coordinated control: The global control signal, generated by the room pressure control module, serves as the system's primary regulation signal, controlling large-scale variations in supply and return air volume. The refined control signal, generated by the damper flow control module, further optimizes the regulation accuracy of the supply air valves. The system's predictive optimization allocation module integrates the global and refined control signals to generate the final control instructions, which are then applied to the supply and return air valve execution modules.

[0140] The air valve flow control module generates refined control signals based on the nonlinear MPC algorithm, which specifically includes obtaining the current opening of the air supply valve and the actual flow data, calculating the deviation between the target air supply flow and the actual flow, and generating a refined flow control signal for adjusting the opening of the air supply valve to achieve precise control of the air supply flow.

[0141] The steps for the air valve flow control module to generate refined control signals based on the nonlinear MPC algorithm are:

[0142] Establish a prediction model for the nonlinear MPC algorithm:

[0143] x2(t+1|t)=f(x2(t),u2(t))

[0144] y²(t)=h(x²(t))

[0145] Among them, x2(t+1|t) represents the predicted state of the system at time t+1 under the condition of the system state and control input known at the current time t; f(x2(t),u2(t)) represents the nonlinear function of state transition, the specific form of which is obtained by actual physical modeling or actual data-driven method, x2(t) is the internal pressure value and pressure change rate of the system at the current time t, u2(t) is the control input vector of the system, including the supply air volume and return air volume; y2(t) is the output vector of the system, which represents the measured output value of the system at the current time t, including the room pressure value; h(x2(t)) represents the nonlinear relationship between output and state;

[0146] Optimize the objective function J f for:

[0147]

[0148] Among them, J2 is the optimization objective function value, and the air supply volume is the target to be minimized; y2(t+k|t) is the predicted air supply volume value at the current time t predicted at the future time t+k; y 2,ref is the target reference value of the air supply volume; Q is the output error weight matrix, which is used to weigh the deviation between the predicted value and the target value of the air supply volume; Δ is the change, Δu2(t+k|t) is the control input change at the prediction moment (the change value of the air supply volume); R is the control input weight matrix, which is used to limit the adjustment range of the air supply volume;

[0149] The nonlinear processing method is numerical optimization, using Newton's method to directly optimize the objective function J2;

[0150] The Newton method prioritizes setting the convergence threshold and establishing the objective function gradient. Use the Hessian matrix to calculate, and finally update the formula of u according to the convergence condition.

[0151] Constraints include flow range constraints and control input range constraints:

[0152] Flow range constraints:

[0153] q s,min ≤q s (t+k|t)≤q s,max

[0154] Where: q s,min is the minimum value of air supply volume; q s,max is the maximum value of air supply volume; q s (t+k|t) is the predicted state of the air supply volume at t+k based on time t;

[0155] Control input range constraints:

[0156] θ s,min ≤θ s (t+k|t)≤θ s,max

[0157] Where: θ s,min is the minimum opening of the air supply valve; θ s,max is the maximum opening of the air supply valve;

[0158] θ s (t+k|t) is the predicted state of the air supply valve opening at t+k based on time t.

[0159] The steps for the room pressure control module to generate the global control signal based on the constrained MPC algorithm are:

[0160] Establish a prediction model with constrained MPC algorithm:

[0161] x1(t+k+1|t)=A1·x1(t+k|t)+B1·u1(t+k|t)+w1(t+k|t)

[0162] y1(t+k|t)=C1·x1(t+k|t)+v1(t+k|t)

[0163] where x1(t+k+1|t) represents the predicted state of the system at time t+k+1 given the system state and control input known at time t, where k = 0, 1, ..., N p -1, N p represents the number of time steps for future prediction; u1(t+k|t), y1(t+k|t), w1(t+k|t), and v1(t+k|t) respectively represent the control input, control output, system noise, and measurement error of the system at time t+k when the system state and control input are known at the current time t.

[0164] The optimization objective function J1 is:

[0165]

[0166] Among them, J1 is the optimization objective function value, which is the target to be minimized. Its value reflects the control performance of the system in the prediction time domain and consists of two parts: the output error term and the weight term of the control input; y1(t+k|t) is the predicted system output room pressure value at t+k when the current time t is known; y 1,ref is the target pressure value; Q is the output error weight matrix, which is used to weigh the deviation between the target pressure and the actual pressure. It is a positive semidefinite matrix. The larger the weight value, the more important the error in this dimension is. Q is adjusted according to the pressure stability requirements in the system design; u1(t+k|t) is the control input vector, which means that when t is known, the control variables at time t+k are predicted, which are the supply air volume and return air volume; R is the control input weight matrix, which is used to limit the range of control input changes. It is a positive definite matrix. The larger the value, the stricter the limit on signal changes.

[0167] Constraints include pressure range constraints, input amplitude constraints, and input change rate constraints:

[0168] Pressure range constraints:

[0169] p min ≤y q (t+k|t)≤p max

[0170] Among them, p min and p max is the allowable room pressure range, indicating the minimum and maximum values ​​respectively;

[0171] Input amplitude constraint:

[0172] q s,min ≤q s (t+k|t)≤q s,max

[0173] q r,min ≤q r (t+k|t)≤q r,max

[0174] Among them, q s,min is the minimum value of air supply volume; q s,max is the maximum value of air supply volume; q s (t+k|t) is the predicted state of the air supply volume at time t+k based on time t; q r,min is the minimum value of return air volume; q r,max is the maximum return air volume; q r(t+k|t) is the predicted state of return air volume at time t+k based on time t;

[0175] Input change rate constraint:

[0176] Δq s,min ≤q s (t+k|t)-q s (t+k-1|t)≤Δq s,max

[0177] Δq r,min ≤q r (t+k|t)-q r (t+k-1|t)≤Δq r,max

[0178] Where Δq s,min The minimum difference between the air supply volume at two predicted moments; Δq s,max The maximum difference between the air supply volume at two predicted moments; Δq r,min The minimum difference in return air volume between the two predicted moments; Δq r,max It is the maximum value of the difference in return air volume between the two predicted moments.

[0179] The actual execution modules include the supply air valve execution module and the return air valve execution module, which are used to control the supply and return air volumes, respectively. The supply air valve execution module receives control signals from the control strategy selection module and the predictive optimization allocation module to adjust the opening of the supply air valve to achieve precise control of the supply air volume. The return air valve execution module adjusts the opening of the return air valve based on the control signals to achieve the return air volume control target, ensuring the overall pressure control of the room is achieved.

[0180] The control system in this embodiment achieves precise control of cleanroom pressure and flow through an automated state-space model, MPC predictive control, and real-time feedback adjustment. Compared to traditional control methods, this invention can dynamically adapt to complex environmental changes and nonlinear interference, maintaining high-precision and stable control of cleanroom pressure and flow under multivariable coupling. This system not only offers high control accuracy but also excellent real-time response capabilities, providing an efficient and reliable solution for rigorous cleanroom environment control.

[0181] The above is a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, various improvements can be made to the present invention without departing from the principles of the present invention, and these improvements should also be considered as the scope of protection of the present invention.

Claims

1. A clean room pressure intelligent control system, characterized by: It includes data processing module, modeling and prediction module, control strategy selection module, prediction and optimization allocation module, control signal output module, actual execution module, physical signal feedback module and real-time feedback adjustment module, among which: The data processing module includes an initialization module and a data acquisition module; the initialization module is used for the initial setup of the system and the acquisition of baseline data; wherein the baseline data acquisition refers to the acquisition of the initial environmental parameters and external factors of the clean room when the system is started, the establishment of the initial state space model and the setting of the initial control parameters; the data acquisition module is used to collect the dynamic environmental data of the clean room in real time, including the supply air volume, return air volume and room pressure; The modeling and prediction module is used to generate a state space model based on the data of the data processing module and dynamically update the state space model through an adaptive algorithm; The control strategy selection module includes a room pressure control module and an air valve flow control module; the room pressure control module, based on a constrained model predictive control algorithm, generates a global control signal for regulating the overall pressure of the room. This control signal directly acts on the supply air valve execution module and the return and exhaust air valve execution module to achieve overall pressure difference control of the room; the air valve flow control module, based on a nonlinear model predictive control algorithm, generates a refined control signal for individually adjusting the opening and flow of the supply air valve, ensuring precise regulation of the air supply volume in the room, thereby optimizing the stability of the room pressure and flow at a micro level; The prediction optimization allocation module is used to receive the control signal from the control strategy selection module and dynamically allocate the control signal to different actuators after optimization; The real-time feedback adjustment module is used to monitor the real-time feedback data in the prediction optimization allocation module and dynamically adjust the control signal of the control strategy selection module, so that the system can quickly respond to changes in the external environment and realize the real-time flow of feedback data and model update; The control signal output module is used to transmit the optimized and allocated control signal to the actual execution module; The actual execution module includes an air supply valve execution module and a return and exhaust air valve execution module; the air supply valve execution module is used to receive the global control signal from the room pressure control module and the refined control signal from the air valve flow control module, adjust the opening of the air supply valve, and realize the air supply control of the clean workshop; the return and exhaust air valve execution module is used to receive the global control signal from the room pressure control module, adjust the opening of the return and exhaust air valve, and realize the control of the overall pressure difference of the room; The physical signal feedback module is used to collect physical feedback data generated after execution, including pressure feedback and flow feedback, and send the feedback data to the data processing module to adjust the control strategy in real time.

2. The intelligent control system for clean room pressure according to claim 1 is characterized by: The specific control steps of the room pressure control module and the air valve flow control module of the control strategy selection module are as follows: The room pressure control module generates a global control signal based on the constrained MPC algorithm, including obtaining the deviation between the current room pressure and the target pressure value, calculating the pressure adjustment amount, generating a global control signal, and sending the signal to the supply air valve execution module and the return air valve execution module for adjusting the supply air volume and exhaust air volume; The air valve flow control module generates a refined control signal based on the nonlinear MPC algorithm, including obtaining the current opening of the air supply valve and the actual flow data, calculating the deviation between the target air supply flow and the actual flow, generating a refined flow control signal, and sending the signal to the air supply valve execution module for adjusting the opening of the air supply valve to accurately control the air supply flow.

3. The intelligent control system for clean room pressure according to claim 2 is characterized in that ,The state space model generated by the modeling and prediction ,module consists of the following two parts: The room pressure control model is used to describe the dynamic characteristics of the overall room pressure and adjust the global control variables of the supply air valve and the return air valve. The state vector, control input, and output vector of the room pressure control model are defined as follows: The state vector x1(t), including the room pressure value p(t) and the pressure change rate Expressed as: Control input u1(t), including the air volume q of the air supply system s (t) and the air volume w of the return air system r (t), which are used to adjust the positive and negative pressures of the room, respectively, and are expressed as: u1(t)=[q s (t)q r (t)] T The output vector y1(t) is the real-time pressure value p(t) of the room, which is used to compare with the target pressure p ref (t) comparison, expressed as: y1(t)=p(t) Equation of state: x1(t+1)=A1x1(t)+B1u1(t)+w1(t) Where: A1 is the system matrix, used to describe the dynamic characteristics of room pressure and its gradient changes; B1 is the input matrix, used to describe the regulatory effect of supply and return air on room pressure; w1(t) is the system noise, including instantaneous pressure fluctuations caused by opening doors and windows and ambient pressure gradient fluctuations caused by starting and stopping equipment fans; Output equation: y1(t)=C1x1(t)+v1(t) Where: C1 is the output matrix used to extract the real-time pressure value of the room; v1(t) is the measurement noise, including the measurement error of the pressure sensor; The damper flow control model is used to describe the dynamic characteristics of the air supply valve. By fine-tuning the opening of the air supply valve, precise control of the air supply flow can be achieved. The state vector, control input, and output vectors of the damper flow control model are defined as follows: The state vector x2(t), including the air volume q of the air supply system s (t) and the opening of the air supply valve θ s (t), expressed as: x2(t)=[w s (t)θ s (t)] T Control input u2(t), represents the adjustment amount Δθ of the air supply valve opening s (t), expressed as: u2(t)=Δθ s (t) Output vector y2(t), representing the actual air supply volume q at the current moment s (t), used to match the target air supply volume q s，ref (t) comparison, expressed as: y2(t)=q s (t) Equation of state: x2(t+1)=A2x2(t)+B2u2(t)+w2(t) Where: A2 is the system matrix, used to describe the dynamic characteristics of the air supply valve; B2 is the input matrix, used to describe the regulating effect of the air supply valve opening on the air supply volume; w2(t) is the system noise, including the nonlinear interference of air supply duct vibration and ambient temperature and humidity fluctuations; Output equation: y2(t)=C2x2(t)+v2(t) Where: C2 is the output matrix used to extract the air supply volume; v2(t) is the measurement noise, including the measurement error of the air volume sensor.

4. The intelligent control system for clean room pressure according to claim 3 is characterized by: The steps for the room pressure control module to generate the global control signal based on the constrained MPC algorithm are: Establish a prediction model with constrained MPC algorithm: x1(t+k+1|t)=A1·x1(t+k|t)+B1·u1(t+k|t)+w1(t+k|t) y1(t+k|t)=C1·x1(t+k|t)+v1(t+k|t) where x1(t+k+1|t) represents the predicted state of the system at time t+k+1 given the system state and control input known at time t, where k = 0, 1, ..., N p -1, N p represents the number of time steps for future prediction; u1(t+k|t), y1(t+k|t), w1(t+k|t) and v1(t+k|t) represent the control input, control output, system noise and measurement error of the system at time t+k, respectively, when the system state and control input are known at the current time t; The optimization objective function J is: Among them, J1 is the optimization objective function value, which is the target to be minimized. Its value reflects the control performance of the system in the prediction time domain and consists of two parts: the output error term and the weight term of the control input; y1(t+k|t) is the predicted system output room pressure value at t+k when the current time t is known; y 1,ref is the target pressure value; Q1 is the output error weight matrix, which is used to weigh the deviation between the target pressure and the actual pressure. It is a positive semidefinite matrix. The larger the weight value, the more important the error in this dimension is. Q1 is adjusted according to the pressure stability requirements in the system design; u1(t+k|t) is the control input vector, which means that when t is known, the control variables at time t+k are predicted, which are the supply air volume and return air volume; R1 is the control input weight matrix, which is used to limit the range of control input changes. It is a positive definite matrix. The larger the value, the stricter the limit on signal changes. Constraints include pressure range constraints, input amplitude constraints, and input change rate constraints: Pressure range constraints: p min ≤y1(t+k|t)≤p max Among them, p min and p max is the allowable room pressure range, indicating the minimum and maximum values ​​respectively; Input amplitude constraint: q s,min ≤q s (t+k|t)≤q s,max q r,min ≤q r (t+k|t)≤q r,max Among them, q s,min is the minimum value of air supply volume; q s,max is the maximum value of air supply volume; q s (t+k|t) is the predicted state of the air supply volume at time t+k based on time t; q r,min is the minimum value of return air volume; q r,max is the maximum value of return air volume; q r (t+k|t) is the predicted state of return air volume at time t+k based on time t; Input change rate constraint: Δq s,min ≤q s (t+k|t)-q s (t+k-1|t)≤Δq s,max Δq r,min ≤q r (t+k|t)-q r (t+k-1|t)≤Δq r,max Where Δq s,min The minimum difference between the air supply volume at two predicted moments; Δq s,max The maximum difference between the air supply volume at two predicted moments; Δq r,min The minimum difference in return air volume between the two predicted moments; Δq r,max It is the maximum value of the difference in return air volume between the two predicted moments.

5. The intelligent control system for clean room pressure according to claim 2 is characterized in that: The steps for the air valve flow control module to generate refined control signals based on the nonlinear MPC algorithm are: Establish a prediction model for the nonlinear MPC algorithm: x2(t+1|t)=f(x2(t),u2(t)) y²(t)=h(x²(t)) Among them, x2(t+1|t) represents the predicted state of the system at time t+1 under the condition of the system state and control input known at the current time t; f(x2(t),u2(t)) represents the nonlinear function of state transition, the specific form of which is obtained by actual physical modeling or actual data-driven method, x2(t) is the internal pressure value and pressure change rate of the system at the current time t, u2(t) is the control input vector of the system, including the supply air volume and return air volume; y2(t) is the output vector of the system, which represents the measured output value of the system at the current time t, including the room pressure value; h(x2(t)) represents the nonlinear relationship between output and state; The optimization objective function J2 is: Among them, J2 is the optimization objective function value, and the air supply volume is the target to be minimized; y2(t+k|t) is the predicted air supply volume value at the current time t predicted at the future time t+k; y 2,ref is the target reference value of the air supply volume; Q2 is the output error weight matrix, which is used to weigh the deviation between the predicted value and the target value of the air supply volume; Δ is the change, Δu2(t+k|t) is the change value of the air supply volume at the prediction moment; R2 is the control input weight matrix, which is used to limit the adjustment range of the air supply volume.

6. The intelligent control system for clean room pressure according to claim 5 is characterized by: Use Newton's method to directly optimize the objective function J2; The Newton method prioritizes setting the convergence threshold and establishing the objective function gradient. Use the Hessian matrix to calculate, and finally update the formula of u according to the convergence condition; Constraints include flow range constraints and control input range constraints: Flow range constraints: q s,min ≤q s (t+k|t)≤q s,max Where: q s,min is the minimum value of air supply volume; q s,max is the maximum value of air supply volume; q s (t+k|t) is the predicted state of the air supply volume at t+k based on time t; Control input range constraints: θ s,min ≤θ s (t+k|t)≤θ s,max Where: θ s,min is the minimum opening of the air supply valve; θ s,max is the maximum opening of the air supply valve; θ s (t+k|t) is the predicted state of the air supply valve opening at t+k based on time t.

7. The intelligent control system for clean room pressure according to claim 3 is characterized by: The prediction optimization allocation module integrates the global and refined control signals to generate the final control instructions, which act on the air supply valve and return and exhaust valve execution modules. The final air valve control signal u final (t) is: u final (t)=u1(t)+u2(t) Among them, u1(t) is the global control signal generated by the pressure control module, and u2(t) is the refined control signal output by the air valve flow control module.

8. The intelligent control system for clean room pressure according to claim 7, characterized in that: The global control signal u1(t) generated by the pressure control module and the refined control signal u2(t) output by the air valve flow control module are respectively: Among them, q s (t) is the global air supply volume; q r (t) is the global return air volume; u2(t)=Δθ s (t) Where Δθ s (t) is the air supply valve opening adjustment amount.

9. The intelligent control system for clean room pressure according to claim 1 is characterized by: The data processing module further includes an external factor input interface for collecting data on external interference factors in the clean room environment, including temperature, humidity and other room pressure differences.

10. A method for intelligent control of room pressure in a clean room, characterized in that: Based on the clean room pressure intelligent control system according to any one of claims 1 to 9, the control method includes the following steps: Step 1: Initialize the system through the data processing module and collect the baseline data of the clean room; Step 2: Generate the initial state space model using the modeling and prediction module, and update the model by acquiring real-time data through the data acquisition module; Step 3: The room pressure control module generates a global control signal based on the constrained MPC algorithm. Specifically, it obtains the deviation between the current room pressure and the target pressure, calculates the adjustment amount and generates a global control signal, and sends the signal to the supply air valve execution module and the return air exhaust valve execution module. Step 4: The air valve flow control module generates a refined control signal based on the nonlinear MPC algorithm. Specifically, it obtains the current opening and actual flow of the air supply valve, calculates the flow deviation and generates a refined control signal, which is sent to the air supply valve execution module. Step 5: The prediction optimization allocation module optimizes the control signal and allocates it to the corresponding actuator; Step 6: The control signal output module transmits the optimized control signal to the actual execution module; Step 7: The air supply valve execution module and the return and exhaust air valve execution module adjust the valve opening according to the control signal to adjust the supply air volume and return air volume; Step 8: The physical signal feedback module collects the feedback data of the room pressure and flow after execution and sends it to the real-time feedback adjustment module to adjust the control strategy according to the real-time feedback; Step 9: Return to step 2 for loop control to achieve adaptive control of the pressure and flow in the clean room.

Images