A multi-variable self-adaptive decoupling multi-factor environment collaborative control system and method

Abstract

The present application belongs to the field of environmental simulation and automatic control technology, and particularly relates to a multi-variable self-adaptive decoupling multi-factor environmental collaborative control system and method, wherein the method comprises the following steps: S1, a monitoring module acquires real-time multi-source measurable disturbance data sets of sunlight radiation, specimen heating and snowfall phase change heat absorption; S2, adaptive matching is performed according to spatial thermal inertia drift caused by changes in physical properties of specimen loading; S3, a updated thermodynamic state space model is combined with the multi-source measurable disturbance data sets to perform feedforward dynamic prediction, and is substituted into a quadratic programming optimization objective function containing a spatial gradient uniformity penalty term; S4, an optimal multi-variable control increment sequence is acquired by solving the quadratic programming optimization objective function on line, synchronous driving instructions are generated, and are issued to multi-source executing mechanisms in real time to perform closed-loop regulation and control.

Description

Technical Field

[0001] This invention belongs to the field of environmental simulation and automatic control technology, and more specifically, it relates to a multivariable adaptive decoupling multi-factor environmental collaborative control system and method. Background Technology

[0002] In the field of modern high-end equipment manufacturing, ultra-large environmental test chambers are core infrastructure for verifying the reliability of complex industrial products such as aerospace vehicles, rail transit trains, heavy armored vehicles, and large-scale new energy battery storage stations under extreme weather conditions. These test chambers are enormous, typically ranging from hundreds to thousands of cubic meters, and require highly integrated simulations of extreme cold and heat, extreme humidity environments, full-spectrum solar radiation, and complex multi-factor meteorological conditions such as heavy snowfall and icing.

[0003] Due to the extremely large experimental space, the aerodynamic and thermodynamic distribution exhibits significant non-uniformity and enormous system thermal inertia. Even more challenging is the deep nonlinear physical coupling between various environmental control factors, which presents significant technical limitations for traditional multi-loop independent PID controllers when dealing with such controlled objects. Since PID, as a purely error-based feedback control strategy, is a typical post-event correction mechanism, it only reacts when the macroscopic temperature and humidity within the cabin have deviated significantly, easily causing unacceptable drastic dynamic fluctuations and overshoot when facing objects with large time lags.

[0004] Secondly, traditional control lacks mathematical decoupling mechanisms for the cross-coupling of physical quantities. For example, forced cooling of the surface of the cooling coil causes the moisture in the air to condense and precipitate rapidly, significantly reducing the absolute humidity content. This forces the humidity control system to activate a powerful humidifier, and the release of latent heat causes the temperature to rise again, making the system prone to falling into a vicious cycle of energy consumption oscillation. Moreover, traditional feedback control cannot detect sudden strong disturbances in advance. For example, the high radiant heat of the solar simulation array, the sudden dynamic heat generation of ultra-large specimens during the test, and the huge phase change heat absorption process when the spray network of the snow and ice covering device is turned on all constitute strong disturbance sources that conventional systems cannot suppress in advance.

[0005] Even though some existing systems have introduced static matrix decoupling technology, the overall airflow field and thermodynamic characteristics inside the chamber frequently drift with the test mission due to the huge differences in the volume, material heat capacity and wind resistance area of ​​the samples loaded in each ultra-large test chamber. The fixed parameter model is very easy to fail.

[0006] Therefore, in view of this, we will study and improve the existing structure and its shortcomings, and provide a multivariable adaptive decoupling multi-factor environmental collaborative control system and method, in order to achieve a more practical value. Summary of the Invention

[0007] Based on the problems mentioned in the background art above, the present invention provides a multivariable adaptive decoupling multi-factor environmental collaborative control system and method.

[0008] On one hand, the technical solution adopted by this invention is as follows: a multivariable adaptive decoupling multi-factor environmental collaborative control system, including an environmental simulation chamber, wherein an actuator and a sensing mechanism are installed inside the environmental simulation chamber; the actuator includes a variable frequency chiller, a silicon controlled rectifier proportional heater, an ultrasonic humidifier, a rotary dehumidifier, a variable frequency fan array, a snow and ice spray network, and a sunlight simulation lamp array, which are respectively used to adjust the temperature, humidity, wind speed flow field, phase change load, and light radiation intensity inside the environmental simulation chamber; the sensing mechanism includes multiple sets of temperature and humidity sensors, wind speed sensors, and light sensors installed at different spatial nodes inside the environmental simulation chamber, which are used to collect real-time global environmental feedback data and extract spatial gradient difference characteristics inside the chamber; a monitoring module is used to read the dynamic heat generation power data of the sample in real time through an industrial communication bus, and combine it with the operating setpoint of the sunlight simulation lamp array and the snow and ice spray network. The phase change endothermic power is used to construct a multi-source measurable disturbance dataset; a central controller is electrically and communicatively connected to the sensing mechanism, monitoring module, and actuator respectively. The central controller integrates: a model prediction and decoupling module, which is used to calculate the multivariable control increment to eliminate the nonlinear coupling effect between temperature, humidity, and wind speed cycles based on the thermodynamic state space model with spatial gradient characteristics and the multi-source measurable disturbance dataset obtained by the monitoring module as the feedforward compensation input matrix; an adaptive parameter correction module, which is used to identify and update the system state transition matrix and control input matrix of the thermodynamic state space model online in real time based on the real-time historical environmental feedback data collected by the sensing mechanism and using a recursive least squares algorithm with a dynamic forgetting factor; and a cooperative instruction generation module, which is used to convert the multivariable control increment into decoupled synchronous drive instructions and send them to the various underlying drive ends of the multi-source actuator.

[0009] Furthermore, the thermodynamic state-space model is a discretized model, and the mapping relationship between its variables is defined as follows: the system state variable at the next discrete moment consists of the sum of three parts: the first part is the product of the system state variable at the current moment and the dynamically updated time-varying system state transition matrix; the second part is the product of the system control input at the current moment and the dynamically updated time-varying control input matrix; the third part is the product of the measurable disturbance at the current moment and the time-invariant disturbance input matrix; the measured environmental response output variable at the current moment is obtained by multiplying the system state variable at the current moment and the output matrix. The system state variables include the average absolute temperature, average absolute humidity, and average convective wind speed within the environmental simulation chamber; the system control inputs include the net heating power commands of the variable frequency chiller and the silicon controlled rectifier proportional heater, the net humidification command of the ultrasonic humidifier and the rotary dehumidifier, and the frequency control command of the variable frequency fan array; the measurable disturbances include the real-time radiation power of the sunlight simulation lamp array, the real-time heat generation power of the sample, and the latent heat disturbance power of the snow and ice spray network; the measured environmental response output variable is the measured environmental response vector after fusion processing by the distributed sensor network.

[0010] The specific mathematical expression is limited to:

[0011]

[0012]

[0013] Among them, state variables Includes mean absolute temperature, mean absolute humidity, and mean convective wind speed within the environmental simulation chamber; control inputs. This includes net heating power commands for the variable frequency chiller and the silicon controlled rectifier proportional heater, net humidification commands for the ultrasonic humidifier and the rotary dehumidifier, and frequency control commands for the variable frequency fan array; measurable disturbances. Includes real-time radiant power of the simulated sunlight light array, real-time heat generation power of the sample, and latent heat disturbance power of the snow and ice spray network; output This is the measured environment response vector after fusion processing by the distributed sensor network. and For the online identification parameter vector Real-time driven update of time-varying system matrix, The perturbation input matrix is ​​time-invariant.

[0014] Furthermore, the model prediction and decoupling module calculates the rolling optimization quadratic objective function on which the multivariable control increment is based, which consists of the sum of penalty terms in the following three dimensions: The first term is the quadratic form of the deviation between the predicted system environmental response caused by the feedforward compensation input and the set desired environmental trajectory sequence in the prediction time domain, and is weighted and summed using the output tracking error penalty weight matrix; the second term is the quadratic form of the system control increment in the control time domain, and is weighted and summed using the control increment suppression weight matrix; the third term is the quadratic form of the spatial uniformity gradient established based on the difference between the highest and lowest temperature measurement points extracted by the distributed sensor network in the prediction time domain, and is weighted and summed using the spatial equilibrium adjustment coefficient and the gradient penalty weight matrix. Specifically defined as:

[0015]

[0016] Where, N p For prediction in the time domain, N c To control the time domain, r(k+i) is the set desired environmental trajectory sequence; It includes the system environment response prediction value caused by the feedforward compensation input; Q is the output tracking error penalty weight matrix, and R is the control increment suppression weight matrix; This is a spatial uniformity penalty term established based on the difference between the highest and lowest temperature measurement points extracted from the distributed sensor network. S is the spatial equilibrium adjustment coefficient, and S is the gradient penalty weight matrix.

[0017] Furthermore, the collaborative instruction generation module maps the multivariable control increment into physical execution instructions based on the inverse matrix decoupling control law. In the operation logic of the decoupling matrix, the off-diagonal elements representing the interaction between the cooling / heating power regulation loop and the humidification / dehumidification regulation loop are dynamically configured as mutual cancellation coefficients to directly suppress the dehumidification effect caused by cooling and the latent heat cooling interference caused by forced humidification at the underlying hardware execution end.

[0018] Furthermore, the recursive least squares algorithm with a dynamic forgetting factor executed by the adaptive parameter correction module has the following online parameter update iteration steps: Calculate the prediction error at the current moment, which is the difference between the measured environmental response vector at the current moment and the product of the regression vector of the system input and output observation data and the system parameter vector identified at the previous moment; calculate the Kalman gain matrix at the current moment, which is the product of the error covariance matrix at the previous moment and the regression vector at the current moment, divided by the dynamic forgetting factor and the sum of the quadratic terms formed by the regression vector and the error covariance matrix at the previous moment; update the system parameter vector at the current moment, which is obtained by adding the product of the system parameter vector at the previous moment and the prediction error at the current moment. The system error covariance matrix at the current moment is updated by subtracting the product of the current Kalman gain matrix and the transpose of the regression vector from the identity matrix, multiplying it by the error covariance matrix at the previous moment, and then dividing the whole matrix by the dynamic forgetting factor at the current moment. The system parameter vector contains dynamically updated elements of the state transition matrix and the control input matrix. The system input-output observation data regression vector is composed of a preset order of system historical input variables and historical output variables. When the monitoring module detects a sudden high-power heat load fluctuation in the sample, or when the distributed sensor network detects a step change in the heat capacity of the cabin space, the central controller adaptively reduces the value of the dynamic forgetting factor to accelerate the removal of the constraint weights of outdated historical observation data on the current model parameter update. Specifically as follows:

[0019] Calculate the prediction error at time k :

[0020]

[0021] Calculate the Kalman gain matrix :

[0022]

[0023] Update system parameter vector :

[0024]

[0025] Update the system error covariance matrix P(k):

[0026]

[0027] in, The parameter vector identified at time k contains a matrix. and Dynamically updated elements; The system input and output observation data regression vector is composed of a combination of system historical input variables and historical output variables of a preset order; The Kalman gain matrix; Let I be the error covariance matrix, and let I be the identity matrix; This is a dynamic forgetting factor; when the monitoring module detects a sudden high-power heat load fluctuation in the sample, or when the distributed sensor network detects a step change in the heat capacity of the cabin space, the central controller adaptively reduces... The value of is chosen to accelerate the removal of the constraint weights imposed by outdated historical observation data on the current model parameter updates.

[0028] Secondly, the technical solution adopted by this invention is as follows: a multivariable adaptive decoupling multi-factor environmental collaborative control method, comprising the following steps: S1, real-time acquisition of multi-node temperature, humidity, and wind speed data within a large space through a sensing mechanism, and extraction of spatial gradient difference features; simultaneously, acquisition of real-time multi-source measurable disturbance datasets of solar radiation, sample heating, and snowfall phase change heat absorption by a monitoring module; S2, using a recursive least squares algorithm with a dynamic forgetting factor, online identification and updating of the time-varying state transition matrix and control input matrix of the thermodynamic state-space model based on historical time window input and output data, in order to adaptively... The spatial thermal inertia drift caused by changes in the physical properties of the sample loading should be matched; S3, the updated thermodynamic state space model is combined with the multi-source measurable disturbance dataset to perform feedforward dynamic prediction, and it is substituted into the quadratic programming optimization objective function containing the spatial gradient uniformity penalty term; S4, the optimal multivariable control increment sequence is obtained by solving the quadratic programming optimization objective function online, and the decoupling control law with mutual inverse cancellation of off-diagonal elements is used to generate synchronous drive commands for cooling / heating, humidification / dehumidification and variable frequency fan that mathematically eliminate mutual interference, and is sent to the multi-source actuators in real time for closed-loop control.

[0029] Furthermore, the feedforward dynamic prediction process in step S3 includes: coupling the radiation set power data of the sunlight simulation lamp array with the preset absorptivity parameters of the outer surface material of the sample to construct a transfer function between illumination and transient temperature rise of the solid surface; and using this transfer function to inject the predicted secondary heat source of solid surface radiation as a heat compensation vector into the perturbation channel B of the thermodynamic state space model in advance. d In d(k).

[0030] Thirdly, the technical solution adopted by this invention is as follows: A multivariable adaptive decoupling multi-factor environmental collaborative device, implemented using a modular programmable logic controller or industrial computer hardware architecture, specifically includes: an interface unit for receiving environmental feedback signals from sensing mechanisms, multi-source measurable disturbance datasets, and environmental setting trajectories issued by the operation terminal; a calculation unit for outputting a predicted value using the thermodynamic state-space model locked in the previous control cycle, comparing it with the currently measured distributed environmental feedback fusion value, and generating a predictive deviation vector representing the model; an identification unit for calculating the time-varying Kalman gain based on the predicted deviation vector by introducing a dynamically decaying forgetting factor mechanism, and updating the parameter matrix representing the fluid dynamics and thermodynamics characteristics of the cabin online; and a collaborative control law reconstruction and solution unit for substituting the updated parameter matrix into the predictive control domain containing feedforward disturbance compensation terms and spatial gradient penalty terms, using a quadratic programming solver to solve the optimal control sequence in real time, and outputting multivariable decoupling control commands to the underlying drive interface after decoupling mapping.

[0031] Fourthly, the present invention also discloses a computer-readable storage medium having a computer program stored thereon, the computer program being used to cause a computer to perform the above-described method.

[0032] The beneficial effects of this invention are:

[0033] 1. By using the radiant heat of the simulated sunlight lamp array, the latent heat of phase change from snowfall and icing, and the dynamic heat generation of the test sample as feedforward inputs, the bottleneck of traditional PID feedback lag is overcome. As soon as the disturbance source is triggered but before it causes a substantial shift in the macroscopic temperature and humidity inside the cabin, the underlying actuators have already made precise hedging and compensation actions under the scheduling of the algorithm, which greatly suppresses dynamic fluctuations and shortens the steady-state establishment time.

[0034] 2. By eliminating loop crossover effects at the mathematical level through an online optimized decoupling feedback matrix, the coupling conflict between cooling and humidification operations is resolved, avoiding ineffective energy waste. Simultaneously, the introduction of an algorithm with a dynamic forgetting factor enables the system to learn online and converge to the optimal mathematical model, mitigating the failure risk of fixed-parameter models to some extent. Attached Figure Description

[0035] The present invention can be further illustrated by the non-limiting embodiments given in the accompanying drawings;

[0036] Figure 1 This is a schematic diagram of the hardware structure of the control system of the present invention;

[0037] Figure 2 This is a flowchart illustrating the control method of the present invention. Detailed Implementation

[0038] like Figure 1 As shown, a multivariable adaptive decoupled multi-factor environmental collaborative control system includes: an environmental simulation chamber, a monitoring module, a central controller, a model prediction and decoupling module, an adaptive parameter correction module, and a collaborative command generation module, wherein the environmental simulation chamber is equipped with actuators and sensing mechanisms.

[0039] To simulate extremely complex weather conditions within a vast space, this system integrates multi-source actuators at its terminals. To meet the bidirectional adjustment and precise quantization requirements of decoupled control, the system's output capacity is designed with strict symmetry: the temperature dimension includes a variable frequency chiller and a silicon controlled rectifier (SCR) proportional heater, which respectively provide precise cooling load and steplessly adjustable heat output; the humidity dimension includes an ultrasonic humidifier and a rotary dehumidifier to provide pure water vaporization and adsorption removal of absolute moisture from the air. Simultaneously, to address convective heat transfer in the large space, a fan array driven by a frequency converter is also configured. Furthermore, a xenon lamp solar simulation array with independent power tuning interfaces and a snow and ice spray network are deployed on the roof and perimeter of the cabin.

[0040] As the sensing layer, the sensing mechanism deploys sensor nodes in a three-dimensional grid distribution within a chamber with a volume of over 1,000 cubic meters. These nodes are located near the sample surface, dead corners of the chamber, and airflow stagnation points. The system includes multiple sets of temperature and humidity sensors, wind speed sensors, and light sensors, which are used to collect real-time environmental feedback data and extract spatial gradient difference characteristics within the chamber.

[0041] The monitoring module can be deployed in the equipment electrical control cabinet to read the dynamic heat generation power data of the sample in real time through the industrial communication bus, and combine it with the operating set value of the sunlight simulation lamp array and the phase change heat absorption power of the snow and ice spray network to construct a multi-source measurable disturbance dataset.

[0042] The central controller is electrically and communicatively connected to the sensing mechanism, monitoring module, and actuator, and integrates the following internal components:

[0043] The model prediction and decoupling module is used to calculate the multivariable control increment to eliminate the nonlinear coupling effect between temperature, humidity and wind speed cycles, based on the thermodynamic state space model with spatial gradient characteristics and the multi-source measurable disturbance dataset obtained by the monitoring module as the feedforward compensation input matrix.

[0044] This invention abstracts the ultra-large space environment into a discretized linear time-varying state-space equation, the core discretized equation of which is expressed as:

[0045]

[0046] In this model, a vector of state variables is defined. .in The average absolute temperature inside the cabin obtained from distributed sensor fusion; This represents the average absolute moisture content. The reason for using absolute moisture content instead of relative humidity here is that relative humidity is a non-linear function of temperature in thermodynamics. Directly modeling with relative humidity would lead to severe non-linear coupling distortion. Using absolute moisture content can effectively decouple latent heat and sensible heat in mathematical form. The average convective wind speed inside the cabin.

[0047] The control input vector is defined as . Characterizes the net cooling / heating power command of the underlying controller after pulse width modulation or analog-to-analog conversion; Net humidification / dehumidification mass flow rate command; This refers to the frequency command for the wind turbine array. The generalized measurable disturbance vector is defined as... . To simulate the radiant power of sunlight, The dynamic heating power of the sample is obtained directly through the communication bus. The enormous latent heat of phase change absorbed when spraying supercooled water or ice crystals onto a snow-covered and icy pipeline network. Disturbance transfer matrix. These physical energy levels are precisely mapped as differential increments to the pre-impact on the cabin state.

[0048] State transition matrix in the equation With control input matrix This includes physical constants such as equivalent heat capacity, air mass flow rate, and flow resistance for a very large space. Because the tanks or wind turbines tested inside the cabin are extremely large, they occupy a significant amount of space and drastically alter the airflow circulation cross-sectional area and overall heat capacity within the cabin. and It must be a parameter set that changes over time. Evolution.

[0049] The rolling optimization quadratic objective function upon which the model prediction and decoupling module calculates the multivariable control increment is defined as:

[0050]

[0051] This objective function incorporates three dimensions of collaborative optimization:

[0052] First, the first summation represents the summation in the prediction time domain N. p Internally, the system predicts the response. The desired environmental trajectory r must be closely tracked. Because at this time... The derivation has been completed in the feedforward channel. The radiation setting power P of the sunlight simulation light array was incorporated into the design. solar and heat absorption by snowfall Q snowThe algorithm can predict the future trend of temperature and humidity soaring or plummeting before the error is actually reflected in the sensor, thus achieving disturbance-resistant feedforward prediction.

[0053] Secondly, the second summation uses the suppression weight matrix R to adjust the control increment. It punishes drastic fluctuations, ensuring smooth operation of high-power electrical appliances such as heavy-duty compressors and variable frequency fans, extending their service life and saving energy.

[0054] Finally, the third point addresses the persistent problem of thermal stratification in ultra-large spaces by introducing spatial temperature gradient differences. This parameter represents the difference between the highest and lowest temperatures measured in the distributed network. When the quadratic programming solver searches for the optimal control increment sequence, the system is forced not only to meet the average temperature and humidity standards, but also to adjust the control increment of the variable frequency fan. By adjusting the opening of the cooling and heating valves in different spatial locations, the temperature can be forcibly reduced. .

[0055] An adaptive parameter correction module is used to identify and update the system state transition matrix and control input matrix of the thermodynamic state space model online in real time based on the real-time historical environmental feedback data collected by the sensing mechanism and a recursive least squares algorithm with a dynamic forgetting factor.

[0056] To endow the control system with strong self-learning and evolutionary capabilities, the adaptive parameter correction module in this embodiment employs an algorithm with a dynamic forgetting factor, defining a system data regression vector containing the current state feedback and control input as follows: The system unknown parameter vector that needs to be identified at the current moment is: The iterative calculation logic for online parameter updates is as follows:

[0057] Calculate the prediction error at time k :

[0058]

[0059] Calculate the Kalman gain matrix :

[0060]

[0061] Update system parameter vector :

[0062]

[0063] Update the system error covariance matrix P(k):

[0064]

[0065] The core breakthrough of the above algorithm lies in the dynamic forgetting factor. The introduction of ) . During the conventional large-volume stable heat dissipation test phase, the physical boundaries within the cabin are fixed, and the central controller will (k) is set to a value close to 1, which is equivalent to giving historical observation data a very high weight, ensuring the smoothness of filtering and robust anti-interference ability of the model under small sensor noise.

[0066] However, when the control system detects a sudden full-load operation of the high-power generator under test via the monitoring module, or when the hatch is opened and a new sample is placed in, causing a step reconstruction of the system's thermophysical boundaries, clinging to old historical data will lead to control stagnation. At this point, the central controller will automatically... (k) drops sharply to around 0.85. In the covariance matrix update formula, due to dividing by a smaller... As matrix P(k) rapidly expands, the Kalman gain K(k) increases dramatically. This mathematical action forces the algorithm to forget outdated historical data that has lost its reference value, causing the parameter vector to... It instantly acquires high sensitivity, rapidly evolves and converges to new matrix parameters that reflect the new environmental load state.

[0067] The collaborative instruction generation module is used to convert the multivariate control increment into decoupled synchronous drive instructions and send them to the underlying drive ends of the multi-source actuator.

[0068] In the operational logic of this decoupling matrix, the off-diagonal elements representing the interaction between the cooling / heating power regulation loop and the humidification / dehumidification regulation loop are set as mutual cancellation coefficients. This means that when the central controller issues a large cooling command due to cooling demand, the underlying matrix operation will automatically generate a parallel humidification compensation command and send it to the ultrasonic humidifier. The compensation amount precisely corresponds to the absolute moisture content lost by the cooling coil due to condensation and water separation. This proactive compensation based on mathematical mechanism reduces the control oscillations and energy consumption caused by physical coupling between the system's temperature and humidity subsystems, achieving high-precision, high-stability, and ultra-large-scale environmental simulation test performance.

[0069] The present invention has been described in detail above. The specific embodiments are provided only to help understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make various improvements and modifications to the present invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A multivariable adaptive decoupling multi-factor environmental collaborative control system, characterized in that: include: An environmental simulation chamber, wherein actuators and sensing mechanisms are installed inside the environmental simulation chamber; The actuators include a variable frequency chiller, a silicon controlled rectifier proportional heater, an ultrasonic humidifier, a rotary dehumidifier, a variable frequency fan array, a snow and ice spray network, and a sunlight simulation lamp array, which are used to adjust the temperature, humidity, wind speed flow field, phase change load, and light radiation intensity inside the environmental simulation chamber, respectively. The sensing mechanism includes multiple sets of temperature and humidity sensors, wind speed sensors and light sensors installed at different spatial nodes in the environmental simulation chamber, which are used to collect real-time environmental feedback data and extract spatial gradient difference features within the chamber. The monitoring module is used to read the dynamic heat generation power data of the test sample in real time through the industrial communication bus, and combine it with the operating set value of the sunlight simulation lamp array and the phase change heat absorption power of the snow and ice spray network to construct a multi-source measurable disturbance dataset. The central controller is electrically and communicatively connected to the sensing mechanism, monitoring module, and actuator, respectively. The central controller integrates the following: The model prediction and decoupling module is used to calculate the multivariable control increment to eliminate the nonlinear coupling effect between temperature, humidity and wind speed cycles, based on the thermodynamic state space model with spatial gradient characteristics and the multi-source measurable disturbance dataset obtained by the monitoring module as the feedforward compensation input matrix. An adaptive parameter correction module is used to identify and update the system state transition matrix and control input matrix of the thermodynamic state space model online in real time based on the real-time historical environmental feedback data collected by the sensing mechanism and a recursive least squares algorithm with a dynamic forgetting factor. The collaborative instruction generation module is used to convert the multivariate control increment into decoupled synchronous drive instructions and send them to the various underlying drive ends of the actuator. The thermodynamic state-space model is a discretized model, and the mapping relationship between its variables is defined as follows: The system state variable at the next discrete moment consists of the sum of three parts: the first part is the product of the system state variable at the current moment and the dynamically updated time-varying system state transition matrix; the second part is the product of the system control input at the current moment and the dynamically updated time-varying control input matrix; and the third part is the product of the measurable disturbance at the current moment and the time-invariant disturbance input matrix. The measured environmental response output variable at the current moment is obtained by multiplying the system state variable at the current moment with the output matrix; The system state variables include the average absolute temperature, average absolute humidity and average convective wind speed inside the environmental simulation chamber. The system control inputs include the net heating power command of the variable frequency chiller and the silicon controlled rectifier proportional heater, the net humidification command of the ultrasonic humidifier and the rotary dehumidifier, and the frequency control command of the variable frequency fan array. The measurable disturbances include the real-time radiation power of the simulated sunlight array, the real-time heat generation power of the sample, and the latent heat disturbance power of the snow and ice spray network. The measured environment response output variable is the measured environment response vector after fusion processing by the distributed sensor network.

2. The multivariable adaptive decoupling multi-factor environmental collaborative control system according to claim 1, characterized in that: The model prediction and decoupling module calculates the rolling optimization quadratic objective function upon which the multivariate control increment is based, which consists of the sum of penalty terms in the following three dimensions: The first term is a quadratic form of the deviation between the predicted system environmental response caused by the feedforward compensation input and the set desired environmental trajectory sequence in the prediction time domain, and is weighted and summed using the output tracking error penalty weight matrix; The second term is the quadratic form of the system control increment in the control time domain, and is weighted and summed using the control increment suppression weight matrix; The third term is a quadratic form of the spatial uniformity gradient established based on the difference between the highest and lowest temperature measurement points extracted by the distributed sensor network within the prediction time domain, and is weighted and summed using the spatial equilibrium adjustment coefficient and the gradient penalty weight matrix.

3. The multivariable adaptive decoupling multi-factor environmental collaborative control system according to claim 2, characterized in that: The collaborative instruction generation module maps the multivariable control increment into physical execution instructions based on the inverse matrix decoupling control law. In the operation logic of the decoupling matrix, the off-diagonal elements representing the interaction between the cooling / heating power regulation loop and the humidification / dehumidification regulation loop are dynamically configured as mutual cancellation coefficients to directly suppress the dehumidification effect caused by cooling and the latent heat cooling interference caused by forced humidification at the underlying hardware execution end.

4. The multivariable adaptive decoupling multi-factor environmental collaborative control system according to claim 1, characterized in that: The recursive least squares algorithm with a dynamic forgetting factor executed by the adaptive parameter correction module has the following online parameter update iteration steps: The prediction error at the current moment is calculated as the difference between the measured environmental response vector at the current moment and the product of the regression vector of the system input and output observation data and the system parameter vector identified at the previous moment. Calculate the Kalman gain matrix at the current time step. Its value is the product of the error covariance matrix at the previous time step and the regression vector at the current time step, divided by the dynamic forgetting factor and the sum of the two terms formed by the regression vector and the error covariance matrix at the previous time step. The system parameter vector at the current time step is updated by adding the product of the system parameter vector at the previous time step and the Kalman gain matrix at the current time step and the prediction error at the current time step. To update the system error covariance matrix at the current time, subtract the product of the current Kalman gain matrix and the transpose of the regression vector from the identity matrix, multiply by the error covariance matrix at the previous time, and divide the whole matrix by the dynamic forgetting factor at the current time. The system parameter vector includes dynamically updated elements of the state transition matrix and the control input matrix; the system input-output observation data regression vector is composed of a combination of historical input variables and historical output variables of a preset order. When the monitoring module detects a sudden high-power heat load fluctuation in the sample, or when the distributed sensor network detects a step change in the heat capacity of the cabin space, the central controller adaptively reduces the value of the dynamic forgetting factor to accelerate the removal of the constraint weights of outdated historical observation data on the current model parameter update.

5. A multivariable adaptive decoupling multi-factor environmental cooperative control method, applied to the control system according to any one of claims 1 to 4, characterized in that: Includes the following steps: S1 collects real-time temperature, humidity and wind speed data of multiple nodes in a large space through the sensing mechanism, and extracts spatial gradient difference features; simultaneously, the monitoring module acquires real-time multi-source measurable perturbation datasets of solar radiation, sample heating and snowfall phase change heat absorption. S2 employs a recursive least squares algorithm with a dynamic forgetting factor to identify and update the time-varying state transition matrix and control input matrix of the thermodynamic state-space model online based on the input and output data of the historical time window, so as to adaptively match the spatial thermal inertia drift caused by changes in the physical properties of the sample loading. S3, combine the updated thermodynamic state-space model with the multi-source measurable perturbation dataset to perform feedforward dynamic prediction, and substitute it into the quadratic programming optimization objective function that includes a spatial gradient uniformity penalty term; S4. The optimal multivariable control increment sequence is obtained by solving the quadratic programming objective function online. Through the decoupling control law with mutual inverse cancellation of off-diagonal elements, synchronous drive commands for cooling / heating, humidification / dehumidification and variable frequency fan that eliminate mutual interference are generated and sent to the multi-source actuators in real time for closed-loop control.

6. The multivariable adaptive decoupling multi-factor environmental collaborative control method according to claim 5, characterized in that: The feed-forward dynamic prediction process in the step S3 includes: coupling the radiation set power data of the sunlight simulation lamp array with the preset absorption rate parameter of the outer surface material of the test product, constructing an illumination and solid surface transient temperature rise transfer function, through which a predicted solid surface radiation secondary heat source is injected in advance into the disturbance channel B of the thermodynamic state space model as a heat compensation vector d in d(k).

7. A multivariable adaptive decoupling multi-factor environmental coordination device, used in the control system described in any one of claims 1-4, characterized in that: Implemented using modular programmable logic controllers or industrial computer hardware architectures, specifically including: The interface unit is used to receive environmental feedback signals from the sensing mechanism, multi-source measurable disturbance datasets, and environmental setting trajectories sent by the operation terminal. The calculation unit is used to output the predicted value using the thermodynamic state-space model locked in the previous control cycle, compare it with the currently measured distributed environmental feedback fusion value, and generate a vector representing the prediction deviation of the model. The identification unit is used to calculate the time-varying Kalman gain based on the predicted deviation vector by introducing a dynamically decaying forgetting factor mechanism, and to update the parameter matrix characterizing the fluid dynamics and thermodynamics of the cabin online. The cooperative control law reconstruction and solution unit is used to substitute the updated parameter matrix into the predictive control domain containing feedforward disturbance compensation terms and spatial gradient penalty terms, use a quadratic programming solver to solve the optimal control sequence in real time, and output multivariable decoupled control commands to the underlying drive interface after decoupling mapping.

8. A computer-readable storage medium, characterized in that: It stores a computer program thereon, the computer program being used to cause the computer to perform the method of any one of claims 5-6.

Images