Multivariable cooperative control method and system for fruit and vegetable warehouse preservation

Abstract

The invention discloses a multivariable cooperative control method and system for fruit and vegetable warehouse fresh-keeping, and belongs to the technical field of fruit and vegetable fresh-keeping, and the method comprises the steps: setting a reference target range of each fresh-keeping parameter, carrying out the correction through combining with a physical characteristic parameter, so as to generate a target initial vector, constructing a coupling incidence matrix, and carrying out the self-updating; obtaining an actual parameter sequence, calculating an absolute deviation to screen main adjustment parameters, synchronously setting a dynamic reset window, and generating a main adjustment instruction; calculating a predicted disturbance variable based on a planned adjustment variable of the main adjustment parameter and the coupling incidence matrix, and generating a hierarchical compensation instruction; an instruction feedback data set is obtained, and when a feedback parameter value enters the dynamic reset window for the first time, first-level real-time stability verification is carried out to judge whether a main adjustment parameter is preliminarily stable or not; and once all the main adjustment parameters are preliminarily stabilized, second-level collaborative stability judgment is carried out, so that the fresh-keeping mode is switched, unnecessary adjustment is reduced, and the fluctuation risk is reduced.

Description

Technical Field

[0001] This invention relates to a multivariate collaborative control method and system for fruit and vegetable warehouse preservation, belonging to the field of fruit and vegetable preservation technology. Background Technology

[0002] Fruits and vegetables, rich in water and sugar, are prone to dehydration, rotting, and other quality deterioration if not stored properly. As the fruit and vegetable industry expands, post-harvest preservation has become crucial for maintaining quality and reducing losses. Warehouses, as the core centralized storage and preservation facilities, directly impact industry efficiency. Multiple variables within warehouses, such as temperature, humidity, oxygen, and carbon dioxide concentrations, affect the respiration and metabolism of fruits and vegetables and the growth of microorganisms, thus determining the preservation period and quality.

[0003] Chinese Patent No. CN115454179B discloses a fruit and vegetable warehouse preservation control system and method, including: a data processing module that compares and analyzes the environmental dynamic balance coefficient with a first threshold, and outputs the current state of preservation parameters in the K fruit and vegetable warehouse; an abnormal state analysis module that analyzes the preservation parameters of the fruit and vegetable warehouse according to abnormal instructions, obtains the abnormal ratio and discrete value of the preservation parameters, compares and analyzes the abnormal ratio with a second threshold, compares and analyzes the discrete value with a third threshold, and marks the abnormal state as occasional or frequent abnormal according to the comparison and analysis results. When the abnormal state is marked as occasional, a correction instruction is generated; when the abnormal state is marked as frequent, a maintenance instruction is generated; and a parameter adjustment module that corrects the preservation parameters of the fruit and vegetable warehouse.

[0004] Although existing technologies can ensure the preservation effect of fruits and vegetables in warehouses, they do not consider the precise reset and stability verification after parameter correction. After adjusting the preservation parameters, there is no follow-up monitoring and verification to ensure that the parameters are stably maintained at the target initial value. This can easily lead to situations where fluctuations after parameter adjustment go unnoticed, affecting the stability of the preservation environment. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a multivariate collaborative control method and system for fruit and vegetable warehouse preservation. By generating a dynamic target vector adapted to the fruit and vegetable and the storage environment, quantifying parameter coupling interference and collaboratively compensating for it, setting a dynamic reset window to avoid over-adjustment, and performing real-time and collaborative dual stability verification, the method solves the problem of undetected parameter fluctuations after adjustment, thus ensuring a stable preservation environment.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Multivariate collaborative control methods for fruit and vegetable warehouse preservation include:

[0008] Set the baseline target range for each preservation parameter, and make corrections based on physical characteristic parameters to generate the target initial vector, construct the coupling correlation matrix, and perform self-updating;

[0009] Obtain the actual parameter sequence, combine it with the target initial vector, calculate the absolute deviation, filter the main adjustment parameters, and simultaneously set the dynamic reset window;

[0010] Based on the coupling correlation matrix, the planned adjustment amount and predicted disturbance amount of the main adjustment parameter are calculated to generate a coordinated control instruction.

[0011] Obtain the instruction feedback dataset. When the feedback parameter value enters the dynamic reset window for the first time, perform the first level of real-time stability verification to determine whether the main adjustment parameter has initially stabilized.

[0012] Once all the main control parameters have initially stabilized, a second-level collaborative stability determination is performed to determine whether the fruit and vegetable warehouse is in a collaboratively stable state, so as to switch the preservation mode.

[0013] Specifically, the steps for generating the target initial vector include:

[0014] Obtain the attribute sequence and expected storage time of the fruits and vegetables to be stored, combine them with physical characteristic parameters, generate an initial basic sequence, and obtain the benchmark target range of each preservation parameter;

[0015] Based on historical preservation data, we screened correction factors that affect the preservation adaptation effect.

[0016] For each correction factor, calculate the correction weight and correction coefficient to obtain the target correction value;

[0017] Calculate the baseline median to obtain the corrected target value and generate the target initial vector;

[0018] Based on the preservation patterns of fruits and vegetables and the estimated storage duration, storage stages are defined;

[0019] Set storage cycle switching trigger conditions, use actual storage duration as the criterion, automatically trigger stage switching, and modify the target initial vector in combination with changes in the physiological state of fruits and vegetables.

[0020] Specifically, the steps for constructing the coupling correlation matrix include:

[0021] Test scenarios are divided based on warehouse type, fruit and vegetable combination, and parameter adjustment type.

[0022] For each test scenario, adjust the preservation parameters to the target initial vector, divide the gradient parameters and the disturbed parameters, set the adjustment gradient, record the disturbance characteristic parameters, and generate a coupled dataset;

[0023] Using the adjustment amount as the independent variable and the steady-state disturbance amount as the dependent variable, multiple linear regression is used to fit the disturbance coefficient, and a symbolic label for the disturbance direction is set to construct a preliminary coupling correlation matrix.

[0024] Obtain the interference delay coefficient, correct the interference coefficient, and for mixed category scenarios, perform weighted fusion of the interference coefficients of single category scenarios based on category quality weights to generate a coupling correlation matrix;

[0025] Multiple trigger conditions are set up. Each time an update is performed, historical data is extracted. With the goal of minimizing the sum of squared errors between the predicted and measured interference, the least squares method is used to iteratively correct the coupling correlation matrix.

[0026] Specifically, the steps for screening the main control parameters include:

[0027] Obtain the actual parameter sequence and calculate the real-time deviation and absolute deviation;

[0028] Set a basic threshold, and then make secondary adjustments based on the fruit and vegetable attribute sequence and the current storage stage to generate a deviation threshold;

[0029] For each preservation parameter, real-time deviation data for five consecutive monitoring cycles are extracted to form a deviation sequence;

[0030] Linear fitting is used to calculate the rate of change of deviation in order to determine the deviation trend of each preservation parameter;

[0031] If the real-time deviation and the rate of change of deviation have the same sign, it is determined that there is a trend of deviation deterioration; if they have opposite signs, it is determined that there is a trend of deviation convergence.

[0032] If the absolute deviation is greater than the deviation threshold, it is determined to be the main control parameter; if the absolute deviation is not greater than the deviation threshold but there is a trend of deviation deterioration, it is determined to be the main control parameter.

[0033] If the absolute deviation is not greater than the deviation threshold and there is a trend of deviation convergence, it is judged as a compliant parameter.

[0034] Specifically, the steps to set up a dynamic reset window include:

[0035] Retrieve the target adjustment value of the main adjustment parameter in the current storage stage and set the base window width. ;

[0036] Based on the absolute deviation and the deviation threshold, the deviation ratio of the main control parameter is calculated, the deviation ratio levels are divided, and the deviation correction coefficient of the main control parameter is obtained.

[0037] Obtain the sensitivity correction coefficient, stage correction coefficient, and environmental correction coefficient, and generate the four-dimensional correction coefficient;

[0038] Obtain the dynamic window width and calculate the initial boundary constraints;

[0039] Set basic limit ratio Based on the initial boundary constraints, the actual width is calculated. To determine the validity of the window; if Preserve the initial boundary constraints;

[0040] like Forced adjustments are performed to generate a forced window, which is then verified in conjunction with the benchmark target range; finally, a dynamic reset window for the main adjustment parameters is generated.

[0041] Specifically, the steps for generating coordinated control instructions include:

[0042] Obtain the actual value of the main control parameter, calculate the reset median based on the median of the dynamic reset window, and calculate the planned adjustment amount of the main control parameter with the goal of driving the main control parameter to the reset median. ;

[0043] Obtain the interference coefficients of the corresponding disturbed parameters and calculate the predicted interference for each disturbed parameter. ;

[0044] Obtain the rate of change of the corresponding disturbed parameter to perform trend correction on the predicted disturbance amount;

[0045] If the rate of change of the deviation has the same sign as the predicted disturbance, the predicted disturbance is corrected as follows: ;

[0046] If the rate of change of the deviation and the predicted disturbance have opposite signs, the predicted disturbance should be corrected as follows: ;

[0047] Based on the predicted intervention amount of each disturbed parameter, the disturbance deviation threshold of the corresponding disturbed parameter is obtained, and the disturbance impact ratio is calculated to initially classify the disturbance priority.

[0048] For mixed-category scenarios, obtain the fruit and vegetable quality ratio and sensitivity weight, calculate the mixed weight, and adjust the interference priority of the initial division;

[0049] Obtain the compensation efficiency coefficient and calculate the compensation amount for each disturbed parameter.

[0050] Specifically, the steps for generating coordinated control instructions also include:

[0051] The planned adjustment amount is used as the main adjustment amount of the main adjustment command, and the main adjustment amount is divided into a fast approach segment and a stable convergence segment according to a preset ratio. The compensation amount of each disturbed parameter is obtained, and a hierarchical compensation command is generated according to the disturbance priority.

[0052] Establish a parameter-actuator association table to clarify the actuators and operating parameters corresponding to each instruction. Based on the parameter-actuator association table, analyze whether there are conflicts between the actuators corresponding to the main control instruction and the graded compensation instruction, and make conflict adjustments.

[0053] The system obtains the response lag time and its own response delay, uses the minimum value as the response benchmark, and calculates the trigger time of the graded compensation command in combination with the trigger time of the main control command, ultimately generating a coordinated control command.

[0054] Specifically, the steps for real-time stability verification include:

[0055] The actual values ​​of various preservation parameters are collected to form an actual control sequence. At the same time, the operating status data of each actuator are collected to form an instruction feedback dataset, and the collected data are preprocessed.

[0056] Calculate the real-time adjustment deviation and compensation deviation for each monitoring cycle;

[0057] Set the criteria for determining when the main control parameters first stabilize and enter the window, so as to trigger real-time stability verification;

[0058] Set a stability determination period and monitor the deviation data within the stability determination period;

[0059] If the absolute value of the real-time adjustment deviation is less than the fine-tuning threshold and the absolute value of the compensation deviation is less than the coordinated stability threshold, the main control parameter is determined to be initially stable; otherwise, it is determined to be not initially stable and real-time fine-tuning is triggered.

[0060] If there are multiple main control parameters, perform real-time stability verification one by one. The real-time stability verification is passed only when all main control parameters are initially stable.

[0061] Specifically, the steps for determining cooperative stability include:

[0062] Obtain a stable parameter sequence, calculate the coefficient of variation for each preservation parameter, and perform a consistency check; if the coefficient of variation is greater than the consistency check threshold, re-collect the stable parameter sequence.

[0063] For each preservation parameter, calculate its own coordination difference; for each set of master control parameters and disturbed parameters, calculate the coupling residual difference, thereby generating a coordination difference matrix.

[0064] Based on the fruit and vegetable attribute sequence, the current storage stage, and the physical characteristics of the warehouse, a two-level judgment threshold is set, including a collaborative stability threshold and a coupling residual threshold.

[0065] Based on the collaborative difference matrix and the secondary decision threshold, a dual stability determination is performed, including single-parameter collaborative stability and multi-parameter collaborative stability.

[0066] If both stability criteria are met, the system is considered to be in a state of coordinated stability. If only single-parameter coordinated stability is met but multi-parameter coordinated stability is not met, then coupling residual fine-tuning is triggered. Otherwise, the regulation and real-time stability verification are re-executed.

[0067] A multivariate collaborative control system for fruit and vegetable warehouse preservation includes: a monitoring module, a collaborative control module, and a verification module;

[0068] The monitoring module is used to acquire fruit and vegetable attribute sequences and warehouse physical characteristic parameters, generate target initial vectors, collect the actual parameter sequences of each preservation parameter in real time, determine parameter trends, and classify main adjustment parameters and compliance parameters.

[0069] The coordinated control module is used to collect coupled datasets, calculate interference coefficients, update interference coefficients by combining interference delay correction and mixed category weighted fusion, construct a coupled correlation matrix, set a dynamic reset window, calculate planned adjustment amount and predicted interference amount, and generate coordinated control instructions.

[0070] The verification module is used to monitor the main control parameters in real time. Once the main control parameters enter the dynamic reset window for the first time, real-time stability verification is performed by adjusting the deviation and compensating the deviation to determine whether the main control parameters are initially stable. Only when all main control parameters are initially stable is a collaborative stability determination performed.

[0071] The beneficial effects of this invention are:

[0072] By integrating fruit and vegetable attributes with warehouse physical characteristics, a target initial vector is generated, ensuring that the target values ​​of preservation parameters align with the physiological needs of fruits and vegetables and the storage environment from the outset. This reduces the risk of pseudo-stability caused by baseline deviations. A coupling correlation matrix is ​​constructed based on the control variable method, and through self-updating using multiple linear regression and least squares, the coupling interference between parameters is accurately quantified. The generated coordinated control instructions can simultaneously predict the impact of the main control parameter on other parameters and provide tiered compensation, suppressing multivariate control conflicts and avoiding cascading fluctuations caused by single parameter adjustments. During the control process, a reasonable space is reserved for parameter fluctuations through a dynamic reset window, and the window width is dynamically adjusted and boundary constraints are applied using a four-dimensional correction coefficient, avoiding the pursuit of absolute values. Over-adjustment of the target value ensures that the actual parameter value is driven to a safe range. Real-time stability verification continuously monitors the deviation sequence after the parameter first enters the window. Preliminary stability is determined by fine-tuning threshold and co-stability threshold to avoid misjudgment of instantaneous fluctuations. Co-stability determination is performed after all main adjustment parameters have stabilized. The co-stability status of multiple parameters is verified by coefficient of variation analysis and co-stability difference matrix to ensure that the relative fluctuations between parameters are strictly controlled, forming a double guarantee. This achieves precise control of the entire chain from parameter initialization to continuous monitoring after adjustment, effectively solving the problem of undetected fluctuations after adjustment in existing technologies. It significantly improves the long-term stability of the preservation environment and provides a reliable parameter control solution for fruit and vegetable storage. Attached Figure Description

[0073] Figure 1 A schematic diagram of a multivariate collaborative control method for preserving fresh fruits and vegetables in a warehouse;

[0074] Figure 2 This is a flowchart illustrating the process of selecting the main control parameters in this invention;

[0075] Figure 3 A flowchart for generating coordinated control instructions for this invention;

[0076] Figure 4 This is a flowchart of the real-time stability verification and collaborative stability determination of the present invention. Detailed Implementation

[0077] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0078] Example 1:

[0079] refer to Figures 1 to 4 As shown in the figure, this embodiment introduces a multivariate collaborative control method for fruit and vegetable warehouse preservation, including the following steps:

[0080] Obtain basic information about fruits and vegetables to be stored in the fruit and vegetable warehouse, such as the type of fruit and vegetable and the maturity level. At the same time, obtain the physical characteristic parameters of the fruit and vegetable warehouse, such as the actual volume of the warehouse, the layout of the ventilation ducts, and the upper limit of the ventilation volume. Based on the basic information and the basic theory of preservation, set the benchmark target range for each preservation parameter. Combine the physical characteristic parameters to correct the benchmark target range to generate the target initial vector.

[0081] Due to the unavoidable coupling interference between various preservation parameters, such as the decrease in humidity due to increased temperature and the increase in carbon dioxide concentration due to decreased oxygen concentration, multiple sets of coupled datasets are obtained based on the control variable method. Multiple linear regression is used to fit the datasets and calculate the interference coefficients of each preservation parameter to construct a coupling correlation matrix. The matrix update period is set, and based on historical response data, including adjustment event data, including gradient parameters, adjustment amounts, adjustment times, and actual changes, the least squares method is used to self-update the coupling correlation matrix to improve the accuracy of coupling interference prediction.

[0082] By deploying a distributed sensor network, the actual parameter sequences of each preservation parameter are collected in real time. Combined with the target initial vector, the absolute deviation of each preservation parameter is calculated. Combined with the detection threshold, it is determined whether the preservation parameter needs to be adjusted, so as to distinguish the main adjustment parameter and the compliance parameter. Once the main adjustment parameter is detected, the preservation mode in the fruit and vegetable warehouse is switched from the monitoring and maintenance mode to the adjustment mode, and the adjustment command is generated for the main adjustment parameter. At the same time, in order to avoid over-adjustment due to the pursuit of absolute target value during the adjustment process and to reserve a reasonable space for parameter fluctuation, a dynamic reset window is set so that the actual value of the main adjustment parameter is driven into the dynamic reset window through the adjustment command, thereby reducing the risk of fluctuation during the adjustment process.

[0083] The system calls the self-updated coupling correlation matrix to calculate the planned adjustment amount of the current main control parameter, thereby generating the main control instruction. It calculates the predicted interference amount of the current planned adjustment amount on other disturbed parameters and generates a graded compensation instruction to suppress negative interference between multiple variables. Based on the main control instruction and the graded compensation instruction, it generates a coordinated control instruction and sends it to the corresponding actuator to trigger the actuator to operate according to the adjustment amplitude and adjustment rate set by the instruction. At the same time, it continuously acquires the instruction feedback dataset of the sensor network. When the feedback parameter value enters the dynamic reset window for the first time, it performs the first level of real-time stability verification to determine whether the main control parameter has initially stabilized.

[0084] Once all main control parameters have initially stabilized, a second-level collaborative stability determination is performed. The actual parameter sequence of the preservation parameters after real-time stability verification is obtained as the stable parameter sequence. The collaborative difference is calculated in combination with the target initial vector. The collaborative stability threshold is used to determine whether the fruit and vegetable warehouse is in a collaboratively stable state. This allows the preservation mode to be switched from the control mode to the monitoring and maintenance mode, which operates with lower power consumption but continues to monitor, reducing unnecessary adjustments and lowering the risk of fluctuations.

[0085] Specifically, the steps for generating the target initial vector include:

[0086] Based on the fruits and vegetables to be stored, the system obtains the fruit and vegetable category, maturity level, respiration type, and ethylene release level to generate an attribute sequence for the fruits and vegetables to be stored. Simultaneously, it obtains the estimated storage time of the fruits and vegetables to be stored and, combined with physical characteristic parameters, generates an initial basic sequence. Specifically, the system uses manual visual recognition combined with category traceability information, such as the origin and variety descriptions provided by suppliers, to clarify the fruit and vegetable category. The maturity level of the fruits and vegetables is obtained through sugar content testing, hardness testing, and soluble solids content testing. The respiration type is determined based on fundamental theories in fruit and vegetable preservation and category characteristics, such as aerobic respiration-dominant or anaerobic respiration-sensitive types. Referring to industry classification standards for ethylene release from fruits and vegetables, and combining category characteristics and maturity, the ethylene release level is initially determined.

[0087] Based on fundamental theories, industry standards, and offline experimental data in the field of fruit and vegetable preservation, the baseline target ranges for various preservation parameters have been preliminarily determined, including the baseline ranges for temperature, humidity, and gas concentration. Among these, offline experimental data were obtained through experiments on the tolerance ranges of preservation parameters for different maturity levels of the same variety. The temperature baseline range is based on the low-temperature tolerance of fruits and vegetables, the humidity baseline range is based on the sensitivity of fruits and vegetables to transpiration and water loss, and the gas concentration baseline range is based on the respiration intensity.

[0088] Based on historical preservation data, we screened correction factors that affect the preservation adaptation effect, including: constructing a candidate factor set to store all candidate correction factors to be screened, such as warehouse insulation coefficient, sealing performance, and ethylene release level; calculating the adaptation degree of preservation parameters based on whether the preservation effect of fruits and vegetables meets the target value of preservation parameters, which is used as the preservation adaptation effect; for each preservation parameter, we established a correlation analysis dimension that includes the correspondence between candidate correction factors and preservation adaptation effect to avoid cross-parameter interference; using correlation analysis, we initially screened out preliminary candidate correction factors related to the preservation adaptation effect; using the preservation adaptation effect as the dependent variable and the preliminary candidate correction factors as independent variables, we conducted regression analysis, and determined the final correction factors that affect the preservation adaptation effect through regression coefficients.

[0089] For each correction factor, the correction weight is calculated using the analytic hierarchy process (AHP). Standard values ​​are obtained through industry experience and offline experimental verification. At the same time, the actual detection values ​​in the initial base sequence are read. The correction coefficient is obtained by the degree of deviation between the actual detection values ​​and the standard values. The target correction value is obtained by weighted summation. Based on the benchmark target range, the benchmark median of the upper and lower limits is calculated. The correction target value is calculated by multiplying the benchmark median and the target correction value. The correction target values ​​of each preservation parameter are integrated to generate the target initial vector.

[0090] Based on the preservation patterns of fruits and vegetables and the expected storage duration, the system divides storage into three stages: initial storage, middle storage, and late storage. It also clarifies the duration rules for each stage. In the warehouse management system, a preset storage cycle switching trigger condition is established, using the actual storage duration as the determining factor to automatically trigger stage switching. Combined with changes in the physiological state of fruits and vegetables, the target initial vector is corrected. A correction rule library based on fundamental theories and offline experiments is invoked for each stage to calculate the correction magnitude of each preservation parameter, updating the target initial vector and generating a dynamic target initial vector for the corresponding stage. This ensures that the target values ​​of preservation parameters not only meet the physiological needs of fruits and vegetables but also adapt to the warehouse environment, reducing the probability of post-adjustment fluctuations from the source and avoiding pseudo-stability or pseudo-fluctuations caused by baseline deviations.

[0091] Specifically, the steps for constructing the coupling correlation matrix include:

[0092] Test scenarios were divided based on warehouse type, fruit and vegetable mix, and parameter adjustment type. Warehouse types included enclosed and ventilated warehouses; fruit and vegetable mixes included single-category and mixed-category warehouses; and parameter adjustment types included... kind, The number of preservation parameters, that is, for the first... Preservation parameters ,correspond Adjust the scene, ;

[0093] For each test scenario, all preservation parameters are adjusted to their target initial values ​​based on the target initial vector. Then, using the controlled variable method, a single preservation parameter is used... Set the gradient parameters. adjustment amount Gradient parameters are obtained by combining the initial value of the target. Adjusting the gradient to adjust the gradient parameters Adjustments are made and the disturbed parameters are compensated in real time. , To prevent fluctuations and ensure stability at the target initial value, avoiding the introduction of additional interference, real-time response data of all preservation parameters are acquired based on a preset sampling frequency under each adjustment gradient. Interference characteristic parameters are recorded, including interference occurrence time, interference peak value, and interference duration. Each scenario experiment is repeated three times. Valid experimental data are selected and integrated from all scenarios to construct a structured coupled dataset, including scenario identifier, gradient parameters, adjustment amount, disturbed parameters, and interference characteristic parameters. Among these, the disturbed parameters... To divide the gradient parameter All other preservation parameters, the time of interference is Start adjusting to The time of the first deviation from the target initial value, the peak interference value is The maximum deviation value under the corresponding adjustment gradient, and the duration of the disturbance are: The time from the initial deviation to the return to stability;

[0094] Adjustment amount Independent variable, steady-state disturbance Using the variable as the dependent variable, multiple linear regression is used to fit the interference coefficients between various preservation parameters. Where is the gradient parameter. After adjusting to the adjustment gradient, when the disturbed parameter During the stable phase, the average value during the stable phase is taken as the disturbed parameter. The steady-state value, obtained through the disturbed parameter The difference between the steady-state value and the initial target value is used to obtain the steady-state disturbance. The stage in which the fluctuation amplitude does not exceed the detection accuracy threshold within 5 consecutive sampling periods is the stable stage.

[0095] For the interference coefficient, set a symbolic marker for the interference direction; if the gradient parameter... Adjusting the disturbance parameters If the interference is positive, the interference coefficient is positive; if the interference is negative, the interference coefficient is negative. Based on the interference coefficient with the sign, a preliminary coupling correlation matrix is ​​constructed, where the diagonal elements are 1, so that the influence coefficient of self-regulation on itself is 1.

[0096] Obtain gradient parameters based on the coupled dataset. Adjusted Disturbance Parameter The time interval between the occurrence of interference peaks is used as the interference delay factor. The interference coefficient is corrected using an exponential function. For mixed-category scenarios, category quality weights are allocated according to the quality proportion of each category. The interference coefficients for single-category scenarios are weighted and fused to ensure that the matrix adapts to the common needs of mixed storage in actual warehousing. The interference coefficients after latency correction are integrated with the weighted fused interference coefficients of mixed categories to generate a coupling correlation matrix. The expression for correcting the interference coefficients is shown below:

[0097]

[0098] In the formula, This is the corrected interference coefficient. The base delay time;

[0099] Multiple triggering conditions are set, including regular periodic updates, special time-triggered updates, and fluctuation-triggered updates. During each update, historical data corresponding to the triggering condition is extracted, including real-time response data of gradient parameters, adjustment amounts, adjustment times, and preservation parameters. With the goal of minimizing the sum of squared errors between the predicted and measured interference amounts, the least squares method is used to iteratively correct the coupling correlation matrix to update the interference coefficients. The predicted interference amount is the product of the interference coefficient and the adjustment amount.

[0100] Specifically, the steps to determine whether preservation parameters need adjustment include:

[0101] Obtain the actual parameter sequence, call the target initial vector for the current stage, and calculate the real-time deviation and absolute deviation for any preservation parameter;

[0102] Based on the core characteristics of each preservation parameter, such as control precision requirements, equipment adjustment capabilities, and basic tolerance of fruits and vegetables, a basic threshold is set for each preservation parameter. Based on the fruit and vegetable attribute sequence and the current storage stage, the basic threshold is adjusted a second time to generate the final deviation threshold.

[0103] For each preservation parameter, real-time deviation data from five consecutive monitoring cycles prior to the current moment are extracted to form a deviation sequence. Linear fitting is then used to calculate the rate of change of deviation in order to determine the deviation trend of each preservation parameter.

[0104] If the real-time deviation and the rate of change of deviation have the same sign, such as if the real-time deviation is positive and continues to increase, it indicates that the deviation is moving away from the target value, and the corresponding preservation parameter is judged to have a deteriorating deviation trend; if the real-time deviation and the rate of change of deviation have opposite signs, such as if the real-time deviation is positive but continues to decrease, it indicates that the deviation is moving towards the target value, and the corresponding preservation parameter is judged to have a convergence deviation trend; in this embodiment, when the real-time deviation or the rate of change of deviation is 0, it is assumed that they have the same sign.

[0105] The main control parameters are selected by using deviation thresholds and deviation trends. For any preservation parameter, if the absolute deviation is greater than the deviation threshold, it indicates that the parameter has exceeded the limit and needs immediate adjustment; the corresponding preservation parameter is then identified as the main control parameter. If the absolute deviation is not greater than the deviation threshold but there is a trend of deterioration, it indicates that the parameter is about to exceed the limit and needs to be adjusted in advance; the corresponding preservation parameter is then identified as the main control parameter to avoid exceeding the limit. If the absolute deviation is not greater than the deviation threshold and there is a trend of deviation convergence, it indicates that the parameter is compliant and will remain compliant during the subsequent preservation process; no adjustment is needed, and it is identified as a compliant parameter.

[0106] If at least one primary control parameter exists, the preservation mode of the fruit and vegetable warehouse will be switched from monitoring and maintenance mode to control mode, and a list of primary control parameters will be generated to provide a target for the generation of control instructions. If all preservation parameters are compliant, the current monitoring and maintenance mode will be maintained until a primary control parameter or storage phase switch occurs. In the monitoring and maintenance mode, the warehouse data and status will be continuously collected and the control equipment will not be activated. In the control mode, the control equipment will be activated and control instructions will be generated and executed.

[0107] Specifically, the steps to set up a dynamic reset window include:

[0108] Based on the target initial vector, obtain the target adjustment value of the master adjustment parameter in the current storage stage. To ensure the window size is fully adapted to the current physiological needs and storage cycle of fruits and vegetables, and based on the characteristics of the main control parameters, such as the sensitivity of parameter adjustment, the adjustment accuracy of storage equipment, and the impact of parameter fluctuations on preservation, a basic window width is set. At this point, the center of the base window needs to be aligned with the target adjustment value; that is, the initial range of the base window is... This ensures that the window always revolves around the adjustment target value of the current stage, avoiding window offset that could lead to deviation in the adjustment direction;

[0109] Obtain the absolute deviation of the master control parameter and the corresponding deviation threshold The deviation ratio is calculated by the ratio of the absolute deviation to the deviation threshold. The deviation ratio is divided into levels by the second-level deviation ratio threshold, including basic deviation, medium deviation, and upper limit deviation. Each deviation ratio level corresponds to a deviation correction coefficient, thereby obtaining the deviation correction coefficient of the main control parameter.

[0110] Based on the fruit and vegetable attribute sequence, fruits and vegetables are divided into three sensitivity levels to obtain the sensitivity correction coefficient of the fruits and vegetables to be stored. Combined with the storage stage division, the physiological state of fruits and vegetables in different cycles is matched to obtain the stage correction coefficient of the fruits and vegetables to be stored. Based on real-time data of warehouse physical characteristics, the stability of the warehouse environment is determined to obtain the environmental correction coefficient corresponding to the current warehouse environment. Finally, four maintenance correction coefficients are generated.

[0111] The dynamic window width is obtained by multiplying the four maintenance correction coefficients by the base window width. Based on the baseline target range and adjusted target values, calculate the initial boundary constraints, including the upper limit of constraints. Lower bound constraint To avoid adjusting parameters to an unsafe range, the expression is as follows:

[0112]

[0113]

[0114] In the formula, , These are the lower and upper limits of the benchmark target range, respectively.

[0115] Set basic limit ratio The actual width of the window after constraints is calculated by using the difference between the upper and lower constraints. To determine the validity of the window; if At this point, the actual width is valid, preventing frequent adjustments and preserving the initial boundary constraints, thus generating a dynamic reset window for the main adjustment parameters; if If the actual width is too narrow, even a small fluctuation in the parameter will trigger adjustment, which can easily lead to frequent adjustments. In this case, a forced adjustment is performed. A forced window is generated based on the initial range of the base window. The forced window is then checked again to see if it exceeds the reference target range. If it does, the intersection of the forced window and the reference target range is taken as the final upper and lower limits to ensure safety. If it does not exceed, the forced window is used directly, thereby generating a dynamic reset window for the main adjustment parameter. This clarifies that the goal of the adjustment command is to drive the actual value of the parameter into the dynamic reset window and keep it stable.

[0116] Specifically, the steps for generating coordinated regulation instructions include:

[0117] Obtain the actual value of the main control parameter, calculate the reset median based on the median of the dynamic reset window, aiming to drive the main control parameter to the reset median, and calculate the planned adjustment amount of the main control parameter by the difference between the reset median and the actual value. ;

[0118] The latest self-updated coupling correlation matrix is ​​invoked, and the disturbance coefficients of the corresponding disturbed parameters are obtained using the master adjustment parameter as the gradient parameter. The predicted disturbance amount of each disturbed parameter is calculated by multiplying the disturbance coefficients by the planned adjustment amount. ;

[0119] Obtain the rate of change of the corresponding disturbed parameter's deviation to perform trend correction on the predicted disturbance. If the rate of change of deviation has the same sign as the predicted disturbance, the trends of the disturbance and the deviation of the disturbed parameter itself are superimposed, amplifying the risk. In this case, the predicted disturbance is amplified to proactively address the superimposed risk, and the predicted disturbance is corrected as follows: If the rate of change of the deviation is opposite in sign to the predicted disturbance, the deviation trends of the disturbance and the disturbed parameter cancel each other out, reducing the predicted disturbance to avoid overcompensation. In this case, the predicted disturbance is corrected as follows: In this embodiment, when the rate of change of deviation or the amount of predicted interference is 0, the default sign is the same; the expression is as follows:

[0120]

[0121]

[0122] In the formula, The rate of change of the deviation of the disturbed parameter. As the reference rate;

[0123] Based on the predicted intervention amount of each disturbed parameter, the disturbance deviation threshold of the corresponding disturbed parameter is obtained. By the ratio of the predicted disturbance amount to the disturbance deviation threshold of the corresponding disturbed parameter, the disturbance impact ratio is calculated to initially classify the disturbance priority, including high priority, medium priority, and low priority. For mixed category scenarios, the quality ratio of fruits and vegetables is obtained. At the same time, based on the sensitivity of the disturbed parameters, sensitivity weights are assigned to various fruits and vegetables. By weighting the quality ratio and sensitivity weights, the mixed weight is obtained. The initially classified disturbance priority is adjusted by the mixed weight to ensure that the priority matches the actual preservation needs.

[0124] Based on the interference priority corresponding to the predicted interference amount, the compensation efficiency coefficient corresponding to each priority is called. With the goal of completely canceling the predicted interference, the corrected predicted interference amount is multiplied by the compensation efficiency coefficient, and then the opposite number is taken to ensure that the compensation direction is opposite to the interference direction, so as to obtain the compensation amount of each disturbed parameter. The compensation strength is adapted to the corresponding interference priority.

[0125] The planned adjustment amount is used as the main adjustment amount of the main adjustment command. The main adjustment amount is divided into a fast approach segment and a stable convergence segment according to a preset ratio. The compensation amount of each disturbance parameter is obtained. According to the disturbance priority, a hierarchical compensation command is generated. The high priority is executed synchronously with the main adjustment command. A dual control system combining feedforward and feedback is adopted. Before execution, the actuator parameters are preset according to the compensation amount. During execution, the changes of the disturbance parameters are monitored in real time and the compensation intensity is dynamically adjusted. The medium priority is executed with a 10-second delay to avoid equipment operation conflicts with the main adjustment command. For example, the refrigeration and humidification equipment are prevented from starting at high power at the same time. The execution progress of the main adjustment command is bound during the execution process. The low priority is executed with a 30-second delay and a time-sharing compensation mode is adopted to avoid small disturbance compensation from causing new fluctuations.

[0126] Establish a parameter-actuator association table to clarify the actuators and operating parameters corresponding to each command. Based on the parameter-actuator association table, analyze whether there is a conflict between the actuators corresponding to the main regulation command and the graded compensation command. If there is a slight conflict, such as energy consumption superposition but without affecting the regulation effect, maintain the command execution order and reduce the operating power of the secondary actuators. If there is a severe conflict, such as the regulation effects cancel each other out, re-plan the execution order and execute them in order of priority.

[0127] Obtain the interference delay coefficient in the coupled dataset, determine the response lag time of the main control parameter to each disturbed parameter, and simultaneously extract the self-response delay of each actuator. Use the minimum of the response lag time and its own response delay as the response benchmark. Based on the difference between the trigger time of the main control command and the response benchmark, calculate the trigger time of the corresponding disturbed parameter in the graded compensation command, and correct the trigger time of the graded compensation command to ensure that the compensation effect and the interference effect of the main control command appear synchronously, offsetting the lag effect. Finally, generate the coordinated control command, including the main control segment command, the graded compensation command, the corrected trigger time, and the operating parameters of the actuator.

[0128] Specifically, the steps for real-time stability verification include:

[0129] The coordinated control instruction set is sent to the warehouse intelligent control terminal via industrial Ethernet. After the terminal parses the instruction, it drives the corresponding actuator to start the control action. The instruction trigger timestamp is recorded simultaneously to facilitate subsequent delay verification and traceability. During the execution process, the actual control values ​​of each preservation parameter are collected at a preset sampling frequency to form the actual control sequence. At the same time, the operating status data of each actuator, such as power, running time, and fault codes, are collected to form an instruction feedback dataset. The collected data is preprocessed, including outlier removal, noise reduction, and data standardization.

[0130] For the main control parameter, the reset median is used to approach the target. The real-time control deviation for each monitoring cycle is calculated by adjusting the difference between the actual value and the reset median to quantify the degree of execution of the main control command. For the disturbed parameter, the control target value is used as the stability benchmark. The compensation deviation for each monitoring cycle is calculated by adjusting the difference between the actual value and the control target value to quantify the effect of the compensation command on the cancellation of coupled interference. The real-time control deviation and the compensation deviation of each disturbed parameter are sorted according to the monitoring cycle to form the corresponding deviation sequence, including the control deviation sequence and the compensation deviation sequence.

[0131] The actual value of the main control parameter is monitored in real time. When the actual value of the control parameter is within the dynamic reset window for one consecutive sampling period, it is determined that the main control parameter has entered the window stably for the first time. At this time, the first level of real-time stability verification is automatically triggered to avoid false triggering caused by instantaneous entry into the window.

[0132] A stability determination period is set. Based on the dynamic reset window and the disturbance deviation threshold of the corresponding disturbed parameter, two types of determination thresholds are set, including fine-tuning threshold and collaborative stability threshold. The fine-tuning threshold is a preset fine-tuning ratio of the width of the dynamic reset window to control the stability of the main adjustment parameter in the central region of the dynamic reset window. The collaborative stability threshold is the stability ratio of the disturbance deviation threshold of the corresponding disturbed parameter to ensure that the compensation effect of the disturbed parameter meets the standard.

[0133] After triggering the real-time stability check, continuously monitor the deviation data within the stability determination period; if the absolute value of the real-time adjustment deviation is less than the fine-tuning threshold and the absolute value of the compensation deviation is less than the collaborative stability threshold, then the main control parameter is determined to be initially stable; otherwise, it is determined to be not initially stable, and real-time fine-tuning is triggered.

[0134] For the initial stable judgment results, trend analysis is performed based on the deviation sequence, including: adjusting the adjustment amplitude of the stable convergence segment based on the real-time adjustment deviation of the main adjustment parameter, reducing the adjustment amplitude if the deviation is positive and reducing the adjustment amplitude if it is negative; adjusting the strength of the corresponding compensation command based on the compensation deviation of the disturbance parameter, increasing the compensation amount if the compensation deviation and the predicted disturbance amount are of the same sign, and decreasing the compensation amount if they are of different signs; and re-entering the stability verification process after fine-tuning.

[0135] If there are multiple main control parameters, the above real-time stability verification must be performed one by one. The first-level real-time stability verification is passed only when all main control parameters are marked as initially stable; otherwise, the control and fine-tuning will continue until all main control parameters are initially stable or an abnormal alarm is triggered. If the verification fails within 10 consecutive stability judgment periods, an abnormal alarm will be triggered.

[0136] Specifically, the steps for determining cooperative stability include:

[0137] When the real-time stability verification is passed, all main control parameters are marked as initially stable. Real-time control values ​​are continuously collected for 5 sampling periods to form a stable parameter sequence. The consistency of the stable parameter sequence of each preservation parameter is checked. Based on the ratio of the sequence standard deviation to the mean, the coefficient of variation of each preservation parameter is calculated.

[0138] If the coefficient of variation is greater than the consistency check threshold, the sequence of the corresponding preservation parameter is too volatile and has potential instability. In this case, five stable parameter sequences are collected again for five sampling periods, and the coefficient of variation calculation and verification process is repeated until the coefficient of variation of all preservation parameters is not greater than the consistency check threshold, and finally a qualified stable parameter sequence is generated.

[0139] For each preservation parameter, the self-coordination difference is obtained based on the difference between the average value of the stable parameter sequence and the corresponding target value. For each set of master control parameters and disturbed parameters, the coupling residual difference is calculated by the difference between the self-coordination difference of the master control parameters and the corresponding disturbed parameters. The self-coordination difference and the coupling residual difference are arranged in a row and column structure to construct a multi-parameter coordination difference matrix.

[0140] Based on the fruit and vegetable attribute sequence, current storage stage, and warehouse physical characteristic parameters, a two-level judgment threshold is set, including a collaborative stability threshold and a coupling residual threshold. In the collaborative stability threshold, the main adjustment parameter is taken as 15% of the dynamic reset window width to strictly control the fluctuation of the main adjustment parameter, and the disturbed parameter and compliant parameter are taken as 70% of the corresponding disturbed deviation threshold to adapt to the collaborative stability requirements. The coupling residual threshold is taken as 50% of the average of the deviation thresholds of all preservation parameters to strictly control residual coupling interference and avoid excessive relative fluctuations between parameters.

[0141] Based on the collaborative difference matrix and the secondary judgment threshold, dual stability judgment is performed, including single-parameter collaborative stability and multi-parameter collaborative stability. The judgment criterion for single-parameter collaborative stability is that the absolute value of the self-collaborative difference of all preservation parameters is not greater than the collaborative stability threshold. The judgment criterion for multi-parameter collaborative stability is that the absolute value of the coupling residual difference between the left and right main adjustment parameters and the disturbed parameter is not greater than the coupling residual threshold.

[0142] If both stability criteria are met simultaneously, the system is considered to be in a state of coordinated stability. If only single-parameter coordinated stability is met but multi-parameter coordinated stability is not, then coupling residual fine-tuning is triggered, and the compensation command for the disturbed parameter is adjusted based on the coupling residual difference until the coupling residual difference reaches the target. Otherwise, the control and real-time stability verification are re-executed.

[0143] Example 2:

[0144] Another embodiment of the present invention provides a multivariate collaborative control system for fruit and vegetable warehouse preservation, comprising: a monitoring module, a collaborative control module, and a verification module;

[0145] The monitoring module is used to acquire the fruit and vegetable attribute sequence and warehouse physical characteristic parameters. Based on preservation theory, industry standards and offline experimental data, it generates a target initial vector containing updates for each storage stage. At the same time, it collects the actual parameter sequence of each preservation parameter in real time through a distributed sensor network, calculates the real-time deviation and absolute deviation, and determines the parameter trend, including deterioration and convergence trends, by combining the deviation change rate and deviation threshold. It divides the main adjustment parameters and compliance parameters, automatically switches the preservation mode, and continuously monitors the parameter status and storage duration to trigger the storage stage switch in order to identify parameter anomalies in advance, avoid parameter exceeding the standard, reduce ineffective adjustment, and adapt to the needs of the entire fruit and vegetable storage cycle.

[0146] The collaborative control module is used to collect coupled datasets, fit interference coefficients through multiple linear regression, and update interference coefficients by combining interference delay correction and mixed category weighted fusion. It constructs a coupled correlation matrix and uses historical response data to achieve matrix self-update through the least squares method. It sets a dynamic reset window, calculates planned adjustment amount and predicted interference amount, generates collaborative control instructions, and issues them to the corresponding execution agencies to predict and suppress coupling interference between multiple parameters, avoid over-adjustment and frequent adjustment, ensure that the adjustment instructions are adapted to the physiological needs of fruits and vegetables and the warehouse environment, and improve the scientificity and effectiveness of control.

[0147] The verification module is used to monitor the main control parameters in real time. Once the main control parameters enter the dynamic reset window for the first time, real-time stability verification is performed by checking the adjustment deviation and compensation deviation to determine whether the main control parameters are initially stable. Only when all main control parameters are initially stable is a collaborative stability judgment performed. Stable parameter sequences are continuously collected and the consistency of the coefficient of variation is checked to construct a collaborative difference matrix containing its own collaborative difference and coupling residual difference. Differentiated collaborative stability thresholds and coupling residual thresholds are dynamically generated for dual collaborative stability judgment. Adaptive fine-tuning is triggered for coupling residual interference. If the judgment fails, the adjustment is readjusted to ensure that the main control parameters accurately approach the target range and are initially stable. This achieves the upgrade from single-parameter stability to multi-parameter collaborative stability, eliminates residual coupling interference, reduces the risk of parameter fluctuations, and ensures the preservation effect of fruits and vegetables.

[0148] Working principle and effects:

[0149] By integrating the attributes of the fruits and vegetables to be stored with the physical characteristics of the warehouse, a dynamic initial target vector is generated through correction factor screening and weight calculation. This approach aligns with the physiological needs of fruits and vegetables and the storage environment from the outset, reducing the risk of pseudo-stability caused by baseline deviations. A self-updating coupled correlation matrix is ​​constructed based on the control variable method to accurately quantify coupling interference between parameters. Collaborative control commands are used to adjust the main control parameters and provide tiered compensation for disturbed parameters, suppressing cascading fluctuations caused by multi-variable adjustments. Distributed sensors monitor the actual parameter sequence in real time, classifying the main control parameters and compliant parameters. After switching to control mode, a dynamic reset window is used to reserve space for parameter fluctuations, avoiding over-adjustment. After adjustment, real-time stability verification determines that the main control parameters are initially stable. Collaborative stability verification then verifies the collaborative state of multiple parameters. Combined with deviation fine-tuning and coupling residual correction, this solves the problem of undetected fluctuations after adjustment in existing technologies, achieving long-term stability of the preservation environment and ensuring the quality of stored fruits and vegetables.

[0150] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A multivariate collaborative control method for fruit and vegetable warehouse preservation, characterized in that, include: Set the baseline target range for each preservation parameter, and make corrections based on physical characteristic parameters to generate the target initial vector, construct the coupling correlation matrix, and perform self-updating; Obtain the actual parameter sequence, combine it with the target initial vector, calculate the absolute deviation, filter the main adjustment parameters, and simultaneously set the dynamic reset window; Based on the coupling correlation matrix, the planned adjustment amount and predicted disturbance amount of the main adjustment parameter are calculated to generate a coordinated control instruction. Obtain the instruction feedback dataset. When the feedback parameter value enters the dynamic reset window for the first time, perform the first level of real-time stability verification to determine whether the main adjustment parameter has initially stabilized. Once all the main control parameters have initially stabilized, a second-level collaborative stability determination is performed to determine whether the fruit and vegetable warehouse is in a collaboratively stable state, so as to switch the preservation mode.

2. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 1, characterized in that, The steps for generating the target initial vector include: Obtain the attribute sequence and expected storage time of the fruits and vegetables to be stored, combine them with physical characteristic parameters, generate an initial basic sequence, and obtain the benchmark target range of each preservation parameter; Based on historical preservation data, we screened correction factors that affect the preservation adaptation effect. For each correction factor, calculate the correction weight and correction coefficient to obtain the target correction value; Calculate the baseline median to obtain the corrected target value and generate the target initial vector; Based on the preservation patterns of fruits and vegetables and the estimated storage duration, storage stages are defined; Set storage cycle switching trigger conditions, use actual storage duration as the criterion, automatically trigger stage switching, and modify the target initial vector in combination with changes in the physiological state of fruits and vegetables.

3. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 2, characterized in that, The steps to construct the coupling correlation matrix include: Test scenarios are divided based on warehouse type, fruit and vegetable combination, and parameter adjustment type. For each test scenario, adjust the preservation parameters to the target initial vector, divide the gradient parameters and the disturbed parameters, set the adjustment gradient, record the disturbance characteristic parameters, and generate a coupled dataset; Using the adjustment amount as the independent variable and the steady-state disturbance amount as the dependent variable, multiple linear regression is used to fit the disturbance coefficient, and a symbolic label for the disturbance direction is set to construct a preliminary coupling correlation matrix. Obtain the interference delay coefficient, correct the interference coefficient, and for mixed category scenarios, perform weighted fusion of the interference coefficients of single category scenarios based on category quality weights to generate a coupling correlation matrix; Multiple trigger conditions are set up. Each time an update is performed, historical data is extracted. With the goal of minimizing the sum of squared errors between the predicted and measured interference, the least squares method is used to iteratively correct the coupling correlation matrix.

4. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 3, characterized in that, The steps for selecting master control parameters include: Obtain the actual parameter sequence and calculate the real-time deviation and absolute deviation; Set a basic threshold, and then make secondary adjustments based on the fruit and vegetable attribute sequence and the current storage stage to generate a deviation threshold; For each preservation parameter, real-time deviation data for five consecutive monitoring cycles are extracted to form a deviation sequence; Linear fitting is used to calculate the rate of change of deviation in order to determine the deviation trend of each preservation parameter; If the real-time deviation and the rate of change of deviation have the same sign, it is determined that there is a trend of deviation deterioration; if they have opposite signs, it is determined that there is a trend of deviation convergence. If the absolute deviation is greater than the deviation threshold, it is determined to be the main control parameter; if the absolute deviation is not greater than the deviation threshold but there is a trend of deviation deterioration, it is determined to be the main control parameter. If the absolute deviation is not greater than the deviation threshold and there is a trend of deviation convergence, it is judged as a compliant parameter.

5. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 4, characterized in that, The steps to set up a dynamic reset window include: Retrieve the target adjustment value of the main adjustment parameter in the current storage stage and set the base window width. ; Based on the absolute deviation and the deviation threshold, the deviation ratio of the main control parameter is calculated, the deviation ratio levels are divided, and the deviation correction coefficient of the main control parameter is obtained. Obtain the sensitivity correction coefficient, stage correction coefficient, and environmental correction coefficient, and generate the four-dimensional correction coefficient; Obtain the dynamic window width and calculate the initial boundary constraints; Set basic limit ratio Based on the initial boundary constraints, the actual width is calculated. To determine the validity of the window; if Preserve the initial boundary constraints; like Forced adjustments are performed to generate a forced window, which is then verified in conjunction with the benchmark target range; finally, a dynamic reset window for the main adjustment parameters is generated.

6. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 5, characterized in that, The steps for generating coordinated control instructions include: Obtain the actual value of the main control parameter, calculate the reset median based on the median of the dynamic reset window, and calculate the planned adjustment amount of the main control parameter with the goal of driving the main control parameter to the reset median. ; Obtain the interference coefficients of the corresponding disturbed parameters and calculate the predicted interference for each disturbed parameter. ; Obtain the rate of change of the corresponding disturbed parameter to perform trend correction on the predicted disturbance amount; If the rate of change of the deviation has the same sign as the predicted disturbance, the predicted disturbance is corrected as follows: ; If the rate of change of the deviation and the predicted disturbance have opposite signs, the predicted disturbance should be corrected as follows: ; Based on the predicted intervention amount of each disturbed parameter, the disturbance deviation threshold of the corresponding disturbed parameter is obtained, and the disturbance impact ratio is calculated to initially classify the disturbance priority. For mixed-category scenarios, obtain the fruit and vegetable quality ratio and sensitivity weight, calculate the mixed weight, and adjust the interference priority of the initial division; Obtain the compensation efficiency coefficient and calculate the compensation amount for each disturbed parameter.

7. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 6, characterized in that, The steps for generating coordinated control instructions also include: The planned adjustment amount is used as the main adjustment amount of the main adjustment command, and the main adjustment amount is divided into a fast approach segment and a stable convergence segment according to a preset ratio. The compensation amount of each disturbed parameter is obtained, and a hierarchical compensation command is generated according to the disturbance priority. Establish a parameter-actuator association table to clarify the actuators and operating parameters corresponding to each instruction. Based on the parameter-actuator association table, analyze whether there are conflicts between the actuators corresponding to the main control instruction and the graded compensation instruction, and make conflict adjustments. The system obtains the response lag time and its own response delay, uses the minimum value as the response benchmark, and calculates the trigger time of the graded compensation command in combination with the trigger time of the main control command, ultimately generating a coordinated control command.

8. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 7, characterized in that, The steps for real-time stability verification include: The actual values ​​of various preservation parameters are collected to form an actual control sequence. At the same time, the operating status data of each actuator are collected to form an instruction feedback dataset, and the collected data are preprocessed. Calculate the real-time adjustment deviation and compensation deviation for each monitoring cycle; Set the criteria for determining when the main control parameters first stabilize and enter the window, so as to trigger real-time stability verification; Set a stability determination period and monitor the deviation data within the stability determination period; If the absolute value of the real-time adjustment deviation is less than the fine-tuning threshold and the absolute value of the compensation deviation is less than the coordinated stability threshold, the main control parameter is determined to be initially stable; otherwise, it is determined to be not initially stable and real-time fine-tuning is triggered. If there are multiple main control parameters, perform real-time stability verification one by one. The real-time stability verification is passed only when all main control parameters are initially stable.

9. The multivariate collaborative control method for fruit and vegetable warehouse preservation according to claim 8, characterized in that, The steps for determining cooperative stability include: Obtain a stable parameter sequence, calculate the coefficient of variation for each preservation parameter, and perform a consistency check; if the coefficient of variation is greater than the consistency check threshold, re-collect the stable parameter sequence. For each preservation parameter, calculate its own coordination difference; for each set of master control parameters and disturbed parameters, calculate the coupling residual difference, thereby generating a coordination difference matrix. Based on the fruit and vegetable attribute sequence, the current storage stage, and the physical characteristics of the warehouse, a two-level judgment threshold is set, including a collaborative stability threshold and a coupling residual threshold. Based on the collaborative difference matrix and the secondary decision threshold, a dual stability determination is performed, including single-parameter collaborative stability and multi-parameter collaborative stability. If both stability criteria are met, the system is considered to be in a state of coordinated stability. If only single-parameter coordinated stability is met but multi-parameter coordinated stability is not met, then coupling residual fine-tuning is triggered. Otherwise, the regulation and real-time stability verification are re-executed.

10. A multivariable collaborative control system for fruit and vegetable warehouse preservation, used to implement the multivariable collaborative control method for fruit and vegetable warehouse preservation as described in any one of claims 1-9, characterized in that, include: Monitoring module, coordinated control module, and verification module; The monitoring module is used to acquire fruit and vegetable attribute sequences and warehouse physical characteristic parameters, generate target initial vectors, collect the actual parameter sequences of each preservation parameter in real time, determine parameter trends, and classify main adjustment parameters and compliance parameters. The coordinated control module is used to collect coupled datasets, calculate interference coefficients, update interference coefficients by combining interference delay correction and mixed category weighted fusion, construct a coupled correlation matrix, set a dynamic reset window, calculate planned adjustment amount and predicted interference amount, and generate coordinated control instructions. The verification module is used to monitor the main control parameters in real time. Once the main control parameters enter the dynamic reset window for the first time, real-time stability verification is performed by adjusting the deviation and compensating the deviation to determine whether the main control parameters are initially stable. Only when all main control parameters are initially stable is a collaborative stability determination performed.

Images