A method for intelligent regulation of tobacco stacks based on a three-dimensional shelf microenvironment model

Abstract

The present application relates to the technical field of general control or regulation system, and discloses a kind of intelligent regulation and control method of smoke pile based on stereoscopic shelf microenvironment model, comprising: running an average value control loop to maintain overall stability, while running an equalization control loop, the equalization control loop triggers the direction switching of air flow circulation executor by monitoring the gradient accumulation amount between characteristic zones, and triggers the direction switching of air flow circulation executor based on this; a predictive suppression step is run in parallel, the predictive suppression step actively reduces the output power of temperature and humidity executor when monitoring the instantaneous gradient and its gradient change rate simultaneously exceeds the respective preset threshold, the present application constructs a double-scale control architecture, realizes process equalization on long period using integral logic, while avoiding quality damage risk on instantaneous scale using differential logic, realizes the orthogonal decoupling of equalization and safety.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of based on three-dimensional shelf microenvironment model of smoke pile intelligent regulation method, belong to general control or regulation system technical field. BACKGROUND

[0002] Currently, when regulating objects such as large closed stacked materials, although the basic control mode such as proportional integral derivative (PID) control is widely used, its core task is to ensure that a process variable accurately tracks a set value, but the design essence of such control mode is based on the regulation of lumped parameter system, i.e. the average value of one or a few sensors can represent the overall state of the controlled object by default; however, when this control system based on lumped parameter design is applied to regulate objects such as large closed stacked materials, which essentially have distributed parameter characteristics, its inherent technical limitations will arise, especially in modern automated three-dimensional warehouse designed to solve the defects of traditional flat warehouse, smoke piles are densely stored in high-density, multi-layer three-dimensional shelf locations. This high-density, multi-layer stacking form, i.e. smoke piles are densely stored in different locations of three-dimensional shelves, makes the spatial heterogeneity problem more serious; limited by the physical law of heat and mass transfer, any continuous fluid regulation in a single direction will inevitably lead to spatial heterogeneity within the system, for example, a temperature and humidity gradient is formed between the windward side and the leeward side. The traditional control system can only obtain the average value of this spatial gradient, which cannot reflect the true local state. In the face of such spatial heterogeneity, the traditional control system faces a regulatory constraint: on the one hand, the controller detects the deviation between the average value and the set value, and the regulation means rooted in the lumped parameter logic is to increase the output of the actuator such as fan, trying to force the entire object to be uniform, but this regulation method itself constitutes a direct damage to the material quality in many task processes, and leads to energy waste; on the other hand, trying to build a high-precision process model such as computational fluid dynamics to perform predictive control, it also faces the problems of large amount of calculation, difficult debugging and sensitive model to working condition changes, which leads to its difficulty in wide deployment in actual engineering.

[0003] Not only the hardware facilities exist the form or the structural limitation, the existing control method applied to the micro environment maintenance also exists the deficiency in the logic level, for example, the Chinese invention patent with the publication number CN120495683A discloses a tobacco appearance quality control method based on micro environment intelligent maintenance and related equipment, the method constructs the appearance feature extraction network and the environment feature extraction network, and utilizes the data after the time sequence alignment to carry out the feature extraction and the correlation analysis, finally, the micro environment regulation and control is carried out based on the result of the appearance quality change analysis, however, the core of the scheme is to analyze the relationship between the appearance change feature and the environment change feature through the network model, and then adjust the environment parameter, the regulation and control strategy based on the correlation analysis, although the connection between the appearance and the environment is considered, but its essence depends on the fitting of the model to the historical data, lacks the dynamic, physical level decoupling ability for the endogenous spatial heterogeneity (for example, the temperature and humidity gradient of the windward surface and the leeward surface of the tobacco pile) in the distributed parameter system, once the heat and mass transfer characteristics in the tobacco pile change due to the uneven stacking or the working condition change, the correlation result of the model may be invalid, it is difficult to ensure the balance of the total amount treated in different areas in the long period.

[0004] Therefore, how to provide a brand new control method, which can jump out of the traditional lumped parameter control framework, and realize the process balanced regulation and control of the distributed parameter system in a low overhead and adaptive manner while avoiding quality damage, has become a technical problem to be solved by the present application. SUMMARY

[0005] The present application provides a tobacco pile intelligent regulation and control method based on a three-dimensional shelf micro environment model, which mainly aims to solve the problem that the existing control system cannot realize process balance in a low overhead and adaptive manner when regulating and controlling distributed parameter objects, either by using high intensity adjustment to cause quality damage, or by relying on complex models which are difficult to implement.

[0006] To achieve the above-mentioned purpose, the present application provides a tobacco pile intelligent regulation and control method based on a three-dimensional shelf micro environment model, which is executed by a control system, the control system is connected to at least one temperature and humidity actuator and at least one air flow circulation actuator with switchable air flow direction for regulating and controlling the tobacco pile, and the method comprises:

[0007] Step 101, running an average value control loop for controlling the temperature and humidity actuator according to the average process variable in the tobacco pile;

[0008] Step 102, simultaneously running an equalization control loop, which comprises: step 102.1, acquiring process variables of a first characteristic observation area and a second characteristic observation area in the smoke stack in real time, and calculating an instantaneous gradient between the first characteristic observation area and the second characteristic observation area; step 102.2, calculating a gradient accumulation amount representing the degree of uneven accumulation between the first characteristic observation area and the second characteristic observation area in real time based on the instantaneous gradient; step 102.3, when the airflow circulation executor is in the first airflow direction, monitoring that the gradient accumulation amount reaches a first preset threshold value, triggering the control system to switch the airflow circulation executor to the second airflow direction; step 102.4, when the airflow circulation executor is in the second airflow direction, monitoring that the gradient accumulation amount reaches a second preset threshold value, triggering the control system to switch the airflow circulation executor back to the first airflow direction;

[0009] Step 103, running a predictive suppression step in parallel with steps 101 and 102, which comprises: step 103.1, calculating a gradient change rate in real time based on the instantaneous gradient; step 103.2, when it is monitored that the absolute value of the instantaneous gradient exceeds a preset first instantaneous gradient threshold value, and the absolute value of the gradient change rate simultaneously exceeds a preset second gradient change rate threshold value, determining that the control system enters an instantaneous loss-of-control risk state; step 103.3, during the period when the control system is in the instantaneous loss-of-control risk state, triggering the control system to temporarily override the average value control loop and actively reduce the output power of the temperature and humidity executor until the instantaneous loss-of-control risk state is removed.

[0010] Preferably, the gradient accumulation amount in step 102.2 is obtained by time integration of the instantaneous gradient , wherein is the instantaneous gradient, is the gradient accumulation amount; and the second preset threshold value in step 102.4 is set to zero, for making the cumulative process treatment amounts received by the first characteristic observation area and the second characteristic observation area tend to be equalized at the end of each regulation period of the equalization control loop.

[0011] Preferably, the instantaneous gradient in step 102.1 is the difference between the process variable of the first characteristic observation area and the process variable of the second characteristic observation area; and the gradient change rate in step 103.1 is obtained by performing time differentiation operation on the instantaneous gradient.

[0012] Preferably, the action of actively reducing the output power of the temperature and humidity actuators in step 103.3 is performed independently of the actions of switching the direction of the air flow circulator in steps 102.3 and 102.4; the predictive suppression step further comprises: after the transient runaway risk state is resolved, automatically removing the override of the average value control loop and returning the control of the temperature and humidity actuators to the average value control loop.

[0013] Preferably, the process variables include temperature and humidity; the transient gradient in step 102.1 is a composite gradient formed by a weighted combination of the temperature gradient and the humidity gradient between the first characteristic observation zone and the second characteristic observation zone.

[0014] Preferably, the method further comprises an adaptive gain scheduling step, which comprises: step 601: during the operation of the equilibrium control loop, calculating a process characteristic metric based on the statistical characteristics of the transient gradient in one or more reversal periods; the process characteristic metric is used to represent the real-time physical characteristics of the tobacco column; step 602: presetting a gain scheduling table that maps the process characteristic metric to a set of controller tuning parameters; step 603: automatically adjusting the controller tuning parameters of the average value control loop online according to the gain scheduling table, the controller tuning parameters including the PID gain, the adjustment being used to adapt the control performance of the average value control loop to the real-time physical characteristics of the tobacco column.

[0015] Preferably, the calculation of the process characteristic metric in step 601 specifically comprises: calculating the average absolute value or the root mean square value of the transient gradient in one or more reversal periods; the adjustment rule according to the gain scheduling table in step 603 comprises: comparing the process characteristic metric with a preset characteristic threshold to determine whether the tobacco column is of high inertia characteristics or low inertia characteristics; when it is determined to be of high inertia characteristics, a first set of PID parameters is selected for the average value control loop; when it is determined to be of low inertia characteristics, a second set of PID parameters is selected for the average value control loop, wherein the gain of the first PID parameters is higher than the gain of the second PID parameters.

[0016] Preferably, the method further comprises an adaptive timing feedforward step, which comprises: step 801: measuring a first time period required for the gradient accumulation to reach a first preset threshold from a second preset threshold in the first air flow direction of step 102.3; step 802: measuring a second time period required for the gradient accumulation to return to the second preset threshold from the first preset threshold in the second air flow direction of step 102.4; step 803: taking the first time period and the second time period as feedforward control parameters, and adjusting the operation of the subsequent regulation period of the equilibrium control loop according to the feedforward control parameters.

[0017] Preferably, the method further comprises a diagnostic step, which comprises: step 901: defining an expected physical model that the process variable of the characteristic observation region that just switches from the windward side to the leeward side presents a downward trend; step 902: within a preset time window after the air flow circulation executor switches direction according to step 102.3 or step 102.4, checking the change trend of the process variable of the first characteristic observation region and the second characteristic observation region; step 903: judging whether the change trend conforms to the expected physical model, if the change trend does not conform to the expected physical model, it is determined that there is a fault, and a safety mode is triggered.

[0018] Preferably, the air flow circulation executor is a reversible fan whose rotating direction can be changed by a control system; and the temperature and humidity executor comprises at least one of a heater and a humidifier.

[0019] Compared with the prior art, the present application has the following beneficial effects:

[0020] 1. The method establishes a control logic for converting spatial differences into time series regulation, maintains overall process stability through an average value control loop, and manages local differences through an equalization control loop. The equalization loop calculates the process variable gradient between the characteristic observation regions and the time accumulation thereof in real time, and uses the accumulation as a trigger condition to switch the running direction of the air flow circulation executor. This way separates the uniformity regulation task from the dependence on the intensity of the executor output, so that the system can maintain a low circulation rate, adjust the timing of the air flow direction, and offset the accumulation of instantaneous gradients in a long period of time, avoiding the quality damage or energy consumption problems caused by the traditional control method to pursue uniformity and enhance the output of the executor.

[0021] 2. The control method has self-adaptive ability to the physical time-varying characteristics of the controlled object, and does not depend on the accurate modeling of the internal physical characteristics of the controlled object. Instead, it performs closed-loop feedback on the actual gradient accumulation between the characteristic observation regions. When the controlled object causes the heat and mass transfer characteristics on both sides to be unequal due to internal density unevenness or process evolution, the equalization control loop will automatically adjust the running time ratio in the first air flow direction and the second air flow direction until the effect of the asymmetric timing can make the gradient accumulation return to balance within a complete cycle, thereby achieving automatic compensation for the dynamic changes of the object.

[0022] 3、By parallel running control loops, a control system working in different time scales is constructed. The balance control loop monitors and reverses the accumulated gradient, and realizes the balance management of the total treatment amount in different areas of the controlled object in a long period of minutes or hours. Meanwhile, a predictive suppression step monitors the instantaneous gradient value and its change rate in real time, and judges the system risk in the instantaneous scale of seconds. When the loss of control trend is identified, the output of the temperature and humidity actuator is temporarily intervened and actively suppressed. The two mechanisms work in their respective positions, so that the system maintains long-term cumulative balance while avoiding local damage caused by instantaneous process fluctuations. BRIEF DESCRIPTION OF DRAWINGS

[0023] Fig. 1 The dual-scale collaborative control logic flowchart of the present application;

[0024] Fig. 2 The instantaneous gradient under different control strategies of the present application Evolution comparison chart;

[0025] Fig. 3 The physical architecture and control logic of the present application DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application will be further described in detail below in combination with specific embodiments. It should be understood that the specific embodiments described are only used to explain the present application, but not to limit the protection scope of the present application.

[0027] A method for intelligent regulation of tobacco stacks based on a three-dimensional shelf microenvironment model, which is executed by a control system, such as a programmable logic controller (PLC) or a distributed control system (DCS). The control system is connected to at least one temperature and humidity actuator, such as a heater or humidifier, and at least one airflow circulation actuator with switchable airflow direction, such as a reversible fan that can change its rotation direction through the control system, for regulating the process variables of the tobacco stacks on the hardware link. The control system is logically configured to run three core control steps or loops in parallel, including an average value control loop for maintaining the stability of the overall process variables, an equilibrium control loop for offsetting spatial heterogeneity in the time dimension, and a predictive suppression step for preventing instantaneous process loss of control. These three control steps work together to achieve process equilibrium regulation of distributed parameter systems such as tobacco stacks stored in three-dimensional shelf locations without relying on high-intensity actuator output. When the control system is running, step 101 is first executed, i.e. the average value control loop is run, which undertakes the basic regulation function in the control architecture. To achieve this function, the control system is configured to collect process variable data from multiple sensors deployed inside the tobacco stacks, calculate a weighted average process variable representing the overall state, and use this average value as the process variable (PV) of the main control loop. A standard functional block, such as a proportional-integral-derivative (PID) controller, then compares this PV with a preset process curve set value (SP) and calculates a control output (CV) based on the deviation through a control algorithm. This control output is sent to the temperature and humidity actuator, such as adjusting the power of the heater, with the goal of making the average process variable in the tobacco stacks strictly follow the set value of the process curve throughout the process cycle. While the average value control loop is running, the control system also executes step 102, i.e. runs an equilibrium control loop. The design of this control loop aims to address the problem of spatial heterogeneity inherent in the system that cannot be handled by the average value control alone. To this end, the loop performs the following logical steps: step 102.1, the control system is configured to select at least two physically symmetrical or representative positions for the tobacco stacks stored in the three-dimensional shelf locations, such as selecting the windward side position along the main airflow path in the space as the first characteristic observation area (A area) and the second characteristic observation area (B area), and obtaining the process variables of these two areas in real time in each sampling period, such as temperature and , the controller then performs a mathematical operation to calculate the difference between the two, resulting in an instantaneous gradient , such as .

[0028] To ensure that the first and second characteristic observation areas described in step 102.1 have the maximum characterization capability of the heterogeneity of the internal space of the smokestack, their specific physical locations are determined through a preliminary on-site calibration procedure. This procedure includes: symmetrically arranging a test array containing multiple temporary sensors along the expected main airflow path in the internal space of the smokestack formed by a typical three-dimensional racking location; initiating a calibration program, causing the airflow circulation actuator to run at a calibration speed in the first airflow direction, while the temperature and humidity actuator performs a preset step temperature increase action; during this period, the control system collects process variable data from all sensors in the test array until each measuring point reaches a steady state; the calibration program then analyzes this set of data, identifies the sensor location with the fastest response speed and the first to reach a steady state under this airflow direction, and determines it as the first characteristic observation area; at the same time, it identifies the sensor location with the slowest response speed and the longest lag time, and determines it as the second characteristic observation area. These two determined locations are then fixed as the sensor placement points used to calculate the instantaneous gradient in this control method; step 102.2, to avoid To mitigate the interference of instantaneous fluctuations on control decisions and obtain a control basis that better reflects the true cumulative effect of unevenness, the control system will use the calculated... The signal is input to an integrator function block, which integrator... Performing continuous time integration operations, such as in digital systems... Accumulate, where The sampling period of the controller is used to calculate a gradient accumulation in real time. ;this In the control model, its physical meaning represents the cumulative degree of difference or unevenness in the net heat received by regions A and B since the last equilibrium point. For example, its dimensions can be: seconds or Hours; Steps 102.3 and 102.4 construct a state machine to drive the direction switching of the airflow circulation actuator. This state machine uses It is the only trigger variable; when the control system is running in the first airflow direction (e.g., facing the wind in zone A), The tendency is positive. As the number of cases continues to accumulate, the controller's built-in comparator monitors the situation in real time. The value when Reaching a first preset threshold At this time, the controller immediately triggers a state switch, commanding the airflow circulation actuator to switch to the second airflow direction (e.g., facing the wind in zone B); is an engineering parameter, which is not an arbitrary set value, but determined by off-line process calibration experiments, and it represents the maximum non-uniform accumulation that the tobacco column can tolerate without quality damage; after switching to the second airflow direction, followed by a negative, and then starts to decrease, the controller continues to monitor When decreases and reaches a second preset threshold , the controller triggers the state switching again, switching the airflow circulation actuator back to the first airflow direction; this is preferably set to zero in the control logic, and its control meaning is to ensure that the cumulative process quantities received by the A zone and the B zone are completely equal in the integral sense at the end of each complete control cycle.

[0029] To further optimize the dynamic response of the control system when facing the inherent asymmetry of the controlled object, the method of the present application can also include an adaptive timing feedforward step, which is executed during the operation of the equalization control loop (step 102); specifically, the control system uses its internal timer function to accurately measure the gradient accumulation from the second preset threshold to the first preset threshold required first time period , and then measures the gradient accumulation from the first preset threshold back to the second preset threshold required second time period ; the obtained and values not only reflect the difference in thermal response characteristics of the controlled object under the two airflow directions, but also are used as feedforward control parameters; for example, the control system can use the and values of the last cycle to predict the approximate time point when the airflow direction will switch in the next cycle, and based on this prediction information, a small preset time (e.g. is or 5%) before the switching occurs, the output of the average value control loop (step 101) is fine-tuned (such as slightly reducing or increasing the heater power) to smooth the instantaneous fluctuations of the average process variable that may be caused by the airflow direction switching, or and values can be used to predict the approximate time point when the airflow direction will switch in the next cycle, and based on this prediction information, a small preset timethe ratio of the instantaneous gradient to the filtered gradient change rate as an auxiliary process asymmetry indicator to fine-tune the PID parameters in the adaptive gain scheduling step, so as to optimize the overall control performance in the subsequent control cycle according to the expected operating cycle asymmetry; the control system further runs a predictive suppression step, i.e. step 103, in parallel with the above two control loops, which serves as a monitoring and veto control layer to prevent transient process risks that cannot be covered by the integral logic of step 102, and the step includes: step 103.1, the control system multiplexes the signal calculated in step 102.1 and inputs it into a differentiator functional block, which may, for example, adopt a differential algorithm with low-pass filtering to calculate the gradient change rate in real time characterizes the instantaneous acceleration of the non-uniformity development.

[0030] In step 103.1, to avoid the direct interference of the measurement noise contained in the original instantaneous gradient signal on the differential operation, the calculation of the gradient change rate is performed by a first-order low-pass filtered differential algorithm, which is implemented by a recursive formula wherein is the instantaneous gradient of the current cycle, is the instantaneous gradient of the previous cycle, is the sampling period of the control system, is the filtered gradient change rate calculated in the previous cycle, and is the gradient change rate calculated in the current cycle for the judgment in step 103.2; the value of the filter constant is between 0 and 1, and the specific value is determined by a pre-set calibration procedure, which includes: performing one or more step response tests in the system debugging stage, observing the smoothness and response lag characteristics of the calculated curve, and adjusting the value of to strike a balance until the curve can still reflect the real trend of the gradient change in time while effectively filtering out high-frequency noise interference, at which time the value of is determined as the operating parameter of the algorithm; step 103.2, the control system is built-in a logic judgment unit, which contains two comparators and an AND gate logic, the controller continuously monitors the absolute value of and the absolute value of , when it is monitored that the absolute value of exceeds a preset first instantaneous gradient threshold , and the absolute value of also exceeds a preset second gradient change rate threshold ​​When the output of the logic judging unit becomes true, it is determined that the control system enters the transient risk state; the two thresholds and are also engineering parameters, which are calibrated by process experiments and represent the transient gradient boundary sufficient to cause local quality damage and the trend boundary representing that the gradient is in the risk of uncontrolled climbing; step 103.3, once the system is determined to enter the transient risk state, the logic of the predictive suppression step is triggered immediately, and the execution priority of the step is set to be higher than that of the average value control loop in step 101. The control action of the step is that the controller temporarily and forcibly overrides the control right of the average value control loop to its output and actively sends a suppression signal to the temperature and humidity actuator, for example, forcibly multiplying the PID output of the heater by a suppression factor set to 0.7 or clamping it to a safe upper power limit, so that the system actively reduces the total energy input and avoids the transient risk; it should be noted that the suppression action is independent of the air flow reversal logic in step 102, and realizes the orthogonal decoupling of the two goals of safety and balance in control; the suppression override will remain until the controller monitors that the risk state is removed, for example or the absolute value falls below the safety threshold, at which time the monitoring layer automatically cancels the override and smoothly returns the control right of the temperature and humidity actuator to the average value control loop.

[0031] To further improve the adaptability of the control system to the time-varying characteristics of the controlled object, the method of the application can further include an adaptive gain scheduling step, which aims to solve the problem of decreased control performance of the PID parameters of the average value control loop (step 101) when facing different batches and different physical characteristics (such as high inertia or low inertia) of the controlled object; for this purpose, the control system is configured to multiplex information on the signal during the operation of the balancing control loop (step 102), and after the controller completes one or more complete reversal cycles, the controller calculates statistical characteristics of the data in the cycle, for example, calculates the average absolute value or root mean square value (RMS), and defines the statistical result as a process characteristic measure (PDM), which is used in the control logic to represent the real-time physical characteristics of the controlled object, for example, a high PDM value corresponds to a high-inertia dense stack, and a low PDM value corresponds to a low-inertia loose stack; the control system internally presets a gain scheduling table, which is a lookup table or a function that establishes the PDM value and the controller tuning parameters (i.e. PID gains The mapping relationship between the two is determined through offline calibration experiments. The rules are as follows: when high inertia is identified, a set of first PID parameters with higher gain is selected for the average value control loop to overcome inertia with an aggressive control law; when low inertia is identified, a set of second PID parameters with lower gain is selected to prevent overshoot with a conservative control law. During online system operation, the controller periodically updates the PDM and automatically and online adjusts the PID gain of the average value control loop according to the gain scheduling table, thereby adapting the control performance of the average value control loop to the real-time physical characteristics of the smokestack. To enhance the control system... To enhance system robustness, the method of this invention may further include a diagnostic step, which incorporates an expected physical model. For example, in a characteristic observation area that has just switched from the windward side to the leeward side, the process variables should show a decreasing trend. Within a preset time window after each airflow direction switch, the controller checks the actual changing trends of the process variables in areas A and B. If the controller detects that the actual trend does not conform to the expected physical model, for example, if the temperature in area A is still rising after switching to the leeward side, the controller determines that there is a fault in the sensor or airflow circulation actuator and immediately triggers a safety mode or alarm to prompt maintenance personnel to intervene.

[0032] Example 1: In a specific application, the control system of the present invention is deployed in an automated storage and retrieval system (AS / RS) to perform intelligent control of the tobacco stacks. This AS / RS system avoids the shortcomings of traditional flat warehouses. For example, at a warehouse height of 8.2 meters, a flat warehouse wastes 7 meters of space due to the 1.2-meter stacking height limit and poses a risk of goods being crushed. In contrast, the AS / RS can be designed with multiple layers of racks, such as 5 layers of racks with a 1.5-meter height, increasing storage capacity by more than 4 times. Although this brings economic and management advantages, this high-density, multi-layered storage form is also a typical distributed parameter object for achieving uniform control. Therefore, in this deployment scenario, the control system is deployed to execute a process curve that lasts for several days and includes a rapid heating phase and a long constant holding phase. This process curve requires adjusting the average process variable of the tobacco stack from the initial... Quickly upgrade to and in The temperature was maintained constant for 72 hours; the control system employed a publicly available control method, initiating the average value control loop, the equilibrium control loop, and the predictive suppression step in parallel; during the initial rapid temperature rise phase of the process, the average value control loop detected the average process variable and... Because there is a significant deviation between the setpoints, the PID controller calculates a high-power output command, instructing the temperature and humidity actuator, i.e., the heater, to operate at its maximum output power. Simultaneously, the airflow circulation actuator, in this case a reversible fan, is positioned in the first airflow direction, causing the airflow to move from the first characteristic observation zone to the second characteristic observation zone. Under this condition, due to the high power input of the heater and the unidirectional airflow, the temperature in the first characteristic observation zone rises sharply, while the temperature response in the second characteristic observation zone lags behind, resulting in an instantaneous gradient between the two. and its gradient rate of change All of them increased rapidly in a short period of time.

[0033] exist Cumulative amount The trigger threshold of the balance control loop has not yet been reached. Previously, the logic decision unit of the parallel-running predictive suppression step had detected... The absolute value, in this example, reached It has exceeded its preset first instantaneous gradient threshold. The threshold is set to ,and The absolute value, in this example, reached / minute, which also exceeded its preset second gradient rate of change threshold. The threshold is set to At a rate of / minute, the control system determines that it has entered a state of momentary runaway risk. At this point, the predictive suppression step is triggered, and its control logic temporarily overrides the output of the average value control loop, actively reducing the heater's output power from 100% to a preset safety limit, such as 70%. This action directly suppresses the total energy input rate without changing the airflow direction, thereby avoiding the risk of the temperature in the first characteristic observation zone reaching the quality damage point due to overshoot. As the risk state is resolved, control of the heater is returned to the average value control loop, and the system continues to operate near the 70% power limit until the overall average process variable gradually approaches... The set value is then used to stabilize the PID output of the average value control loop at a certain constant value; however, during the constant holding phase, since the airflow is still in the first airflow direction, a continuous and stable instantaneous gradient is maintained. In this example, The problem persists, causing the first characteristic observation area to continuously receive excessive heat, while the second characteristic observation area receives insufficient heat. At this point, the integrator function block of the balance control loop begins to play its long-period regulatory role, and this loop affects this... of Perform time integration to increase the gradient accumulation. It begins to grow linearly and steadily; after running for approximately 5 hours in the constant-hold phase... The value reached its first preset threshold. This threshold is set to [value] in this application. The hour represents the maximum tolerable net heat difference between the first and second characteristic observation zones; step 102.3 of the equilibrium control loop is triggered, and the control system immediately issues a command to switch the airflow circulation actuator to the second airflow direction, causing the airflow to flow from the second characteristic observation zone to the first characteristic observation zone; under this new airflow direction, the second characteristic observation zone becomes the windward side, and the first characteristic observation zone becomes the leeward side, with an instantaneous gradient It then reverses to a negative value, for example In the second airflow direction, the average value control loop and the predictive suppression step continue to operate in parallel, maintaining... The average process variable stability and instantaneous risk avoidance, while the equilibrium control loop continues to... of Integrating leads to gradient accumulation. from The peak value after one hour begins to decline linearly; after approximately 5 more hours of operation... It decreased and reached its second preset threshold. This threshold is set to At this point, step 102.4 is triggered, and the control system switches the airflow circulation actuator back to the first airflow direction, completing a full control cycle. This control method avoids the risk of local quality damage during the heating phase on a second-level instantaneous scale through a predictive suppression step. At the same time, on a long-term scale of hours through a balance control loop, it uses integral logic and airflow reversal to control the cumulative processing volume received by different regions to be equal throughout the entire cycle. This dual-scale control architecture solves the control conflict between instantaneous safety and long-term cumulative balance in the control system while using only low-speed circulating airflow.

[0034] Example 2: To objectively verify the control method disclosed in this invention, a control test platform was built in distributed parameter system control to compare the control effects of existing technologies in terms of process balance, instantaneous risk avoidance, and system energy consumption. This platform includes a control system, in this example a programmable logic controller (PLC), connected to a power-adjustable heater as a temperature and humidity actuator, and a speed-controlled and reversible fan as an airflow circulation actuator. A sealed test chamber simulating a smokestack was placed in the controlled environment, and temperature sensors for monitoring the first and second characteristic observation areas were deployed symmetrically inside the chamber, with their measurement accuracy set to [value missing]. Simultaneously, multiple other sensors were deployed to calculate the average process variable. Based on this experimental platform, three comparative experiments were set up: Control Group 1, Control Group 2, and the present invention sample group. Control Group 1 simulated the control method in the prior art that relies on high-intensity actuator output. Its control system ran a standard PID algorithm to control the average process variable, and the reversible fan ran continuously at 80% of its rated speed in the first airflow direction. Control Group 2 simulated the control method in the prior art that uses low-intensity actuator output. Its control system also ran a standard PID algorithm, and the reversible fan ran continuously at 20% of its rated speed in the first airflow direction. The present invention sample group was used to verify the technical solution of the present invention. Its control system, according to the method disclosed in the aforementioned specific embodiments, ran the average value control loop, the equalization control loop, and the predictive suppression step in parallel. The base speed of its reversible fan was set to 20% of its rated speed, the running direction was controlled by the equalization control loop, and the temperature and humidity actuator was controlled collaboratively by the average value control loop and the predictive suppression step. The target setpoint (SP) of the average process variable for all experimental groups was set to... The tests were all conducted continuously for 24 hours; for the sample group of this invention, the first preset threshold of the balance control loop was... Set as Hours, second preset threshold Set as Hour, the first instantaneous gradient threshold of the predictive suppression step Set as Second gradient rate of change threshold Set as / minute; the control system records the key performance indicators of each group throughout the test, and the data are summarized in Table 1 below.

[0035] Table 1: Comparison of control methods experimental data.

[0036] ;

[0037] As shown in Table 1, control group 1 operates at high fan speed, with a total system energy consumption of 51.3 kWh and a maximum instantaneous gradient of [data missing]. The control group 1 experienced a 4.2-hour instantaneous risk duration on the windward side, failing to balance energy consumption and process safety. The control group 2 used a low fan speed, resulting in lower total system energy consumption (25.1 kWh) and a lower instantaneous risk duration (0.0 hours). However, due to the consistently unidirectional airflow, its instantaneous gradient (…)… The gradient is continuously accumulated in one direction, resulting in a final accumulated gradient amount of 100%. This indicates that the long-term equilibrium of the process is poor. In the sample group of this invention, the wind turbine base speed is the same as that of control group 2, and the total system energy consumption (25.8 kWh) and instantaneous risk duration (0.0 hours) remain at a low level. Meanwhile, the final gradient accumulation of the sample group of this invention is only... The value is close to zero; this result stems from the fact that the equilibrium control loop of the present invention periodically triggers airflow reversal, resulting in an instantaneous gradient close to zero. Alternating between positive and negative values, the gradient is integrated and canceled out over time.

[0038] Example 3: This example combines Figs. 1 to 3 This paper describes an intelligent control method for tobacco stacks based on a microenvironment model of an automated shelving system. Fig. 1 As shown, the architecture operates three core loops in parallel, including an average control loop that acquires average process variables to maintain overall process variable stability; an equilibrium control loop that acquires process variables from the characteristic observation area to calculate the instantaneous gradient (quantifying the degree of spatial non-uniformity) and further calculates the gradient accumulation (integral) based on this gradient (reflecting the degree of non-uniformity accumulation). When the gradient accumulation reaches a preset threshold (monitoring the first or second threshold condition is met), a direction switching command is triggered to the airflow circulation actuator; and a predictive suppression step that serially calculates the gradient change rate using the instantaneous gradient (characterizing the acceleration of non-uniformity development) and performs a judgment on whether the instantaneous gradient and change rate simultaneously exceed the limit (judging the risk of instantaneous runaway). Once determined to be yes, the instantaneous runaway risk state is determined, and the output power of the temperature and humidity actuator is actively reduced and a coverage command is issued to control the temperature and humidity actuator. In addition, the instantaneous gradient can also be used by an adaptive gain scheduling step (calculating process characteristic metrics, adaptively adjusting PID parameters for online adjustment of PID gain and feeding back to the average control loop).

[0039] like Fig. 2 As shown, this graph uses time (in hours) as the horizontal axis and instantaneous gradient as the horizontal axis. The vertical axis clearly shows the gradient evolution path under different control strategies. The curve of control group 1 (80% speed, unidirectional) (shown by the dashed line in the figure) exhibits a high-level fluctuating instantaneous gradient, while the curve of control group 2 (20% speed, unidirectional) (shown by the dotted line in the figure) has a lower instantaneous gradient but remains positive throughout. In contrast, the curve of the present invention sample group (20% speed, equilibrium reversal) (shown by the solid line in the figure) rises to approximately [value missing] in the initial stage. Then, under the intervention of the equilibrium control loop, it reversed and fell to a negative value, before rising again, exhibiting periodic fluctuations centered on zero throughout the cycle; for example... Fig. 3As shown, the system includes a temperature and humidity actuator and a reversible fan working together on a controlled object comprising two characteristic observation zones, A and B. Sensors are distributed in both zones A and B. The reversible fan can execute airflow circulation in either a first airflow direction or a second airflow direction (reverse direction) according to instructions. The average value control loop maintains overall stability through PID control, and its controlled object is the temperature and humidity actuator. The equalization control loop achieves airflow direction switching through gradient accumulation monitoring, and its controlled object is the reversible fan. The predictive suppression step achieves active power suppression through instantaneous gradient monitoring, and its controlled object is also the temperature and humidity actuator.

[0040] Example 4: To further illustrate the necessity of the gradient accumulation-based feedback-triggered airflow reversal technique in solving the problem of long-period cumulative unevenness, a control group 3 was set up for comparative testing. Control group 3 aims to simulate a control method that only uses fixed time intervals to switch airflow directions, in order to evaluate the lack of the gradient accumulation-based technique of this invention. The effect of feedback-triggered reversal; the control group used the same experimental platform and initial conditions as Example 2, and its control system also ran the average value control loop in parallel to control the average process variable within... It also runs predictive suppression steps to avoid instantaneous risks, and its reversible fan base speed is set to 20% of the rated speed; the only difference from the sample group of this invention is that the airflow circulation actuator switching logic of control group 3 is modified to not monitor gradient accumulation. Instead, it simply triggers reversal based on a fixed time period, specifically by forcibly switching the airflow direction every 5 hours; during the 24-hour test run of control group 3, the working state of the average value control loop and predictive suppression step was similar to that of the sample group of this invention. Since the wind turbine base speed was also 20%, its maximum instantaneous gradient Maintain at The instantaneous risk duration was 0.0 hours, and the total system energy consumption was 25.5 kWh. These indicators are similar to the results of the sample group and control group 2 of this invention. However, since the airflow reversal is based on a fixed 5-hour time interval rather than on the actual gradient accumulation, the results are not satisfactory. Has the equilibrium point been reached? ( Triggered by hours, during the experiment, it was observed that due to slight asymmetries in the heat transfer characteristics inside the test chamber or dynamic changes during the process, at the end of each 5-hour running cycle, The value did not precisely return to zero, but rather exhibited residual bias; this bias gradually accumulated over multiple running cycles, resulting in a final gradient accumulation at the end of the 24-hour experiment. Reached The data from control group 3 were added to Table 1 to obtain Table 2.

[0041] Table 2: Comparison of control methods experimental data.

[0042] ;

[0043] Comparing the data of control group 3 and the sample group of this invention in Table 2, it can be seen that although control group 3 adopted the airflow reversal operation and performed reasonably well in terms of instantaneous risk control and energy consumption, its final gradient accumulation ( ) higher than the sample group of the present invention ( This indicates that using only a fixed-time reversal control method, lacking real-time feedback and adaptive adjustment capabilities regarding the actual uneven accumulation of the system, cannot effectively solve the problem of uneven accumulation caused by inherent asymmetry or dynamic changes in distributed parameter systems during long-term operation. This further demonstrates that the gradient-based cumulative quantity control method in this invention... The technical feature of feedback-triggered airflow reversal is essential for achieving long-term equilibrium in the process.

[0044] Example 5: In the offline calibration stage before the deployment of a control system, to determine the specific algorithm parameters and control thresholds of the control method of the present invention, a representative chimney stack equipped with sensors and actuators or its verified digital twin model is used as the calibration object. The initial state is set as the chimney stack is in a state of normal temperature and humidity to be processed. The control system has the basic functions of data acquisition, calculation and control command output, but the core control parameters, such as thresholds and gains, are to be determined. The purpose of calibration is to set a set of parameters based on the response characteristics of the specific object to ensure that the process runs within the preset limits, achieves the cumulative target, and has specific control performance indicators. To achieve this purpose, a composite gradient is first executed. The weight calibration procedure requires determining the relative weights of temperature and humidity when the process variables involve both, in order to form a composite gradient. A step response test is performed by individually changing the outputs of the temperature and humidity actuators, measuring and recording the temperature gradients generated in the first and second characteristic observation regions. and humidity gradient The response curve is analyzed; combined with process requirements or historical data, the sensitivity of unit temperature gradient and unit humidity gradient to the final product quality is evaluated. If the analysis shows that the effect of temperature gradient is 1.5 times that of humidity gradient, then a temperature weight is set. Humidity weight The composite gradient is calculated accordingly. Next, the gradient rate of change is defined. The specific calculation method, in order to smooth noise interference and obtain a stable rate of change trend, adopts a first-order low-pass filter differential algorithm. The calculation steps are as follows: in each control sampling period... In this example seconds, based on the currently calculated instantaneous gradient Instantaneous gradient of the previous period and gradient rate of change ,pass Calculate the current gradient change rate ,in This is a filter constant, with a value ranging from 0 to 1. It is selected experimentally based on the on-site noise level and response speed requirements. .

[0045] Subsequently, the calibration procedure for the threshold parameter is executed, with the first preset threshold... The purpose of calibration is to determine the maximum allowable accumulation of non-uniformity under unidirectional airflow; from a smaller... Initial value, such as The trial run begins, recording one complete reversal cycle, i.e. Accumulate from 0 Returning to 0, examine the process variable curves in the first and second characteristic observation zones, as well as product quality-related indicators (if measurable); gradually increase... The set value, such as increasing each time. Repeat the trial run and observation until unacceptable signs of quality degradation are observed in the characteristic observation area (such as the triggering of critical indicators for localized over-drying or over-wetting) or the absolute value of the average gradient within the period no longer follows. When increasing and decreasing, select the critical value that leads to quality deterioration. The previous set value, or the inflection point that makes the absolute value of the average gradient tend to stabilize. The value is used as the final calibration result; in this example, it is calibrated as... Second preset threshold Typically set to 0; Predictive suppression threshold and The purpose of calibration is to ensure early intervention when the instantaneous gradient changes drastically; a set of initial thresholds is set, such as... / minute, conduct simulated rapid temperature rise or introduce disturbances in test runs, and observe... and The peak value and extreme process variable values ​​in the characteristic observation area; if the observed extreme process variable values ​​are close to or exceed the quality and safety boundary, and predictive suppression has not yet been triggered, then gradually reduce... and The set value, such as decreasing it each time. and The test is repeated every minute until predictive suppression can be stably triggered before the extreme process variable reaches the safety boundary. The threshold at this point is the calibration result; in this example, the calibration is set to [value missing]. / minute.

[0046] Further, the specific calculation procedure of the process characteristic metric PDM and the establishment procedure of the adaptive gain scheduling table; the calculation method of PDM is as follows: after a complete inversion cycle of the equalization control loop is completed, the control system extracts all the instantaneous gradients recorded in the cycle From Back to , the control system extracts all the instantaneous gradients recorded in the cycle Data points, a total of N, and calculates the root mean square value , wherein is the sampling time point in the cycle; to establish the gain scheduling table, system identification and controller tuning are required; first, prepare or simulate two typical working conditions of the stack: one is high inertia, such as high-density stacking, and the other is low inertia, such as low-density stacking; run the equalization control loop under these two working conditions respectively, and calculate the corresponding characteristic PDM values; measure under high inertia working condition, and under low inertia working condition; according to this, set a characteristic threshold for online differentiation of working conditions; then, for high inertia and low inertia working conditions respectively, use the PID tuning method known in the art, such as the Ziegler-Nichols method based on step response or the relay feedback method, to perform parameter tuning on the PID controller of the average value control loop (step 101) to obtain specific control performance under each working condition, such as minimizing overshoot and regulation time, to obtain two sets of PID parameters: one set is the first PID parameter for high inertia, such as , whose gain is relatively high; the other set is the second PID parameter for low inertia, such as , whose gain is relatively low; finally, construct a gain scheduling table or logical rule: when running online, the control system periodically calculates the PDM value, if , the average value control loop automatically selects the first PID parameter; if , the second PID parameter is automatically selected; by performing the above series of calibration procedures, the core algorithm details of the control method of the present application are clear, the key control parameters and threshold values are obtained, and the setting values have clear physical meaning and engineering basis for specific calibration objects, and the logical rule of adaptive gain scheduling is established, so that the control system can perform specific performance control within the preset limit and achieve the cumulative amount target for different working conditions based on the calibration results in subsequent actual operation, solving the uncertainty in the parameter setting process.

[0047] Example 6: Before deploying the control method of the present invention on a new or heavily maintained smokestack system, a pre-calibration procedure must be performed to determine its key judgment parameters to ensure the effectiveness and reliability of the diagnostic steps. This procedure first involves calibrating the transient response characteristics of the system after airflow reversal, assuming the system is operating stably under a constant condition, such as when the average process variable stabilizes at a certain level. At the same time, at least 10 manually triggered airflow direction switches are repeatedly executed, and the control system continuously records the process variables of the first and second characteristic observation areas after each switch at a sampling frequency of 1Hz. In this case, this is the temperature change curve. The duration of each recording covers at least 3 system response time constants. Based on this set of transient response data, the rate of temperature decrease in the characteristic observation area during the initial stage after each switch to the leeward side is statistically analyzed, for example, from the 10th to the 70th second after the switch. The 5th percentile of the rate distribution is calculated and set as the minimum expected rate of decrease. In this example, we obtain for / second, while simultaneously determining the maximum time required for the statistical temperature decrease trend to stabilize, and adding a 50% safety margin, setting this as the maximum time window. In this example, we obtain For 90 seconds, these two parameters and This was subsequently solidified into the judgment threshold for the diagnostic procedure.

[0048] During online operation of the control system, the diagnostic procedure involves switching the direction of the actuator during each airflow cycle, for example, switching from the first airflow direction to the second airflow direction. At this time, area A becomes the leeward side, and area B becomes the windward side. Within the time window, continuously monitor the rate of change of the process variable in the characteristic observation area that has just switched to the leeward side, which is area A in this example. If the control system determines Failed to achieve the preset goal That is, in The rate of change of multiple consecutive sampling points within the range is lower than or is positive, the control system preliminarily determines that there is a fault; to further distinguish the source of the fault, the control system immediately checks whether the change trend of the process variable of the feature observation area that becomes the windward surface at the same time, in this case, area B, is consistent with the expectation, i.e., whether it presents an upward trend; if the change trend of area B is consistent with the expectation and the change trend of area A is not consistent with the expectation, it is determined that the sensor of area A is faulty; if the change trends of both areas A and B are not consistent with the expectation, for example, area A does not decrease and area B does not increase, it is determined that the airflow circulation actuator fails to successfully reverse or the airflow is abnormal; the control system records this specific diagnostic conclusion together with the relevant data in the log, and then triggers the preset safety mode, the specific action procedures of which include recording fault information, issuing an alarm, disabling the reversing logic of the balancing control loop, locking the airflow circulation actuator in a safe state such as low-speed unidirectional or stopping, and temporarily adjusting the threshold of the predictive suppression step to a more conservative setting until manual intervention confirms that the fault is eliminated.

[0049] Embodiment 7: To further optimize the dynamic response of the control system when facing the inherent asymmetry of the controlled object, the method of the present application can also include an adaptive timing feedforward step, which is executed during the operation of the balancing control loop (step 102); specifically, the control system uses its internal timer function to accurately measure the gradient accumulation amount from the second preset threshold to the first preset threshold in the first airflow direction in each regulation period , and measures the gradient accumulation amount from the first preset threshold to the second preset threshold in the subsequent second airflow direction ; the obtained and values not only reflect the difference in thermal response characteristics of the controlled object in the two airflow directions, but also are used as feedforward control parameters; for example, the control system can use the and values of the last period to predict the approximate time point when the airflow direction is about to switch in the next period, and based on this prediction information, the output of the average value control loop (step 101) is fine-tuned a short preset time , for example or 5% of before the switching occurs, to slightly reduce or increase the heater power, so as to smooth the instantaneous fluctuation of the average process variable that may be caused by the airflow direction switching, or the and The ratio of the two values is used as an auxiliary process asymmetry index to make more refined adjustment of the PID parameters in the adaptive gain scheduling step, so as to optimize the overall control performance of the subsequent regulation period according to the expected operation period asymmetry.

[0050] It is apparent for a person skilled in the art that the present application is not limited to the details of the above exemplary embodiments, but can be implemented in other concrete forms without departing from the spirit or essential characteristics of the present application.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the present application.

Claims

1. A method for intelligent control of tobacco stacks based on a micro-environment model of a three-dimensional shelving system, characterized in that, The method is executed by a control system connected to at least one temperature and humidity actuator for regulating the smoke stack and at least one airflow circulation actuator capable of switching airflow direction. The method includes: Step 101: Run an average value control loop to control the temperature and humidity actuators based on the average process variables within the smoke stack; Step 102: Simultaneously run an equalization control loop, which includes: Step 102.1: Real-time acquisition of process variables for a first characteristic observation area and a second characteristic observation area within the smokestack, and calculation of the instantaneous gradient between the first and second characteristic observation areas; Step 102.2: Based on the instantaneous gradient, real-time calculation of a gradient accumulation amount representing the degree of uneven accumulation between the first and second characteristic observation areas; Step 102.3: When the airflow circulation actuator is in the first airflow direction, if the gradient accumulation amount reaches a first preset threshold, the control system is triggered to switch the airflow circulation actuator to the second airflow direction; Step 102.4: When the airflow circulation actuator is in the second airflow direction, if the gradient accumulation amount reaches a second preset threshold, the control system is triggered to switch the airflow circulation actuator back to the first airflow direction. Step 103, a predictive suppression step is run in parallel with steps 101 and 102. The predictive suppression step includes: Step 103.1, calculating the gradient change rate in real time based on the instantaneous gradient; Step 103.2, when the absolute value of the instantaneous gradient exceeds a preset first instantaneous gradient threshold and the absolute value of the gradient change rate simultaneously exceeds a preset second gradient change rate threshold, the control system is determined to have entered a transient runaway risk state; Step 103.3, during the period when the control system is in the transient runaway risk state, the control system is triggered to temporarily cover the average value control loop and actively reduce the output power of the temperature and humidity actuator until the transient runaway risk state is resolved.

2. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 1, characterized in that, The gradient accumulator in step 102.2 is obtained by integrating the instantaneous gradient over time. And what was obtained, among which For instantaneous gradient, The gradient accumulation amount; and the second preset threshold in step 102.4 is set to zero, so that at the end of each control cycle of the balance control loop, the accumulated process processing amount received by the first feature observation area and the second feature observation area tends to be balanced.

3. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 1, characterized in that, The instantaneous gradient in step 102.1 is the difference between the process variables in the first characteristic observation region and the process variables in the second characteristic observation region; the gradient change rate in step 103.1 is obtained by performing time differentiation on the instantaneous gradient.

4. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 1, characterized in that, The action of actively reducing the output power of the temperature and humidity actuator in step 103.3 is performed independently of the actions of switching the direction of the airflow circulation actuator in steps 102.3 and 102.

4. The predictive suppression step also includes: after the instantaneous runaway risk state is eliminated, automatically removing the coverage of the average value control loop and returning the control of the temperature and humidity actuator to the average value control loop.

5. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 1, characterized in that, The process variables include temperature and humidity; the instantaneous gradient in step 102.1 is a composite gradient formed by weighted composite operation based on the temperature gradient and humidity gradient between the first feature observation area and the second feature observation area.

6. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 1, characterized in that, The method also includes an adaptive gain scheduling step, which includes: Step 601: During the operation of the equalization control loop, a process characteristic metric is calculated based on the statistical characteristics of the instantaneous gradient within one or more reversal cycles; the process characteristic metric is used to characterize the real-time physical characteristics of the tobacco stack; Step 602: A gain scheduling table is preset to map the process characteristic metric to a set of controller tuning parameters; Step 603: According to the gain scheduling table, the controller tuning parameters of the average value control loop are automatically adjusted online, including the PID gain, which is used to make the control performance of the average value control loop adapt to the real-time physical characteristics of the tobacco stack.

7. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 6, characterized in that, Step 601, calculating the process characteristic metric, specifically includes: calculating the average absolute value or root mean square value of the instantaneous gradient over one or more reversal cycles; Step 603, adjusting according to the gain scheduling table, includes: comparing the process characteristic metric with a preset characteristic threshold to determine whether the stack has high inertia or low inertia characteristics; when determined to have high inertia characteristics, selecting a first set of PID parameters for the average value control loop; when determined to have low inertia characteristics, selecting a second set of PID parameters for the average value control loop, wherein the gain of the first PID parameter is higher than the gain of the second PID parameter.

8. The intelligent control method for tobacco stacks based on a micro-environment model of a three-dimensional shelving system according to claim 1, characterized in that, The method further includes an adaptive timing feedforward step, which includes: step 801: measuring the first time period required for the gradient accumulation to reach the first preset threshold from the second preset threshold under the first airflow direction in step 102.3; step 802: measuring the second time period required for the gradient accumulation to return from the first preset threshold to the second preset threshold under the second airflow direction in step 102.4; step 803: using the first time period and the second time period as feedforward control parameters, and adjusting the operation of the subsequent regulation cycle of the equilibrium control loop according to the feedforward control parameters.

9. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 1, characterized in that, The method also includes a diagnostic step, which includes: Step 901: Define an expected physical model, which is: a characteristic observation area that has just switched from the windward side to the leeward side, and whose process variables show a downward trend; Step 902: Within a preset time window after the airflow circulation actuator switches direction according to Step 102.3 or Step 102.4, check the changing trends of the process variables in the first characteristic observation area and the second characteristic observation area; Step 903: Determine whether the changing trend conforms to the expected physical model. If the changing trend does not conform to the expected physical model, it is determined that there is a fault and a safety mode is triggered.

10. The intelligent control method for tobacco stacks based on a three-dimensional shelving microenvironment model according to claim 1, characterized in that, The airflow circulation actuator is a reversible fan whose rotation direction can be changed by the control system; the temperature and humidity actuator includes at least one of a heater and a humidifier.

Images