Control method, device and equipment of fruit and vegetable cold chain fresh-keeping equipment and storage medium

Abstract

The embodiment of the invention provides a control method, device and equipment of fruit and vegetable cold-chain fresh-keeping equipment and a storage medium, and is used for the technical field of cold-chain fresh-keeping. The control method of the fruit and vegetable cold chain fresh-keeping equipment comprises the following steps: carrying out space-time alignment and associated matching on a multi-source heterogeneous sensing data stream and a product metadata set to obtain an environment-product state data set, performing shelf life prediction and corruption risk assessment on each product in the environment-product state data set to obtain residual shelf life distribution data and corruption risk data, and performing multi-target optimization decision according to the equipment operation cost data, the residual shelf life distribution data and the corruption risk data to obtain a candidate equipment operation parameter set; and verifying and adjusting the candidate equipment operation parameter set according to the equipment operation and product fresh-keeping safety boundary set to generate an equipment operation instruction. According to the technical scheme, through mechanism prior, Pareto look-up table decision and safety constraint verification, maximum preservation is achieved, and energy consumption and safety are both considered.

Description

Technical Field

[0001] This application relates to the field of cold chain preservation technology, and in particular to a control method, device, equipment, and storage medium for fruit and vegetable cold chain preservation equipment. Background Technology

[0002] Equipment control plays a crucial role in the cold chain preservation of fruits and vegetables, translating an ideal storage and transportation environment into actionable processes. Through the coordinated regulation of subsystems such as temperature, humidity, gas composition, and air distribution, it directly influences the respiration and metabolism, dehydration rate, microbial growth, and chilling injury risk of fruits and vegetables, ultimately determining shelf life and quality maintenance. Simultaneously, precise equipment control can reduce energy consumption, extend equipment lifespan, ensure food safety compliance, and provide verifiable operational records for warehouse scheduling and traceability, achieving a balance between quality assurance and economic and regulatory requirements.

[0003] Currently, the most common equipment control methods are closed-loop control based on thresholds or PID, supplemented by timed / stage-based settings and manual intervention. Many locations use SCADA or cloud platforms for remote monitoring and alarms. Some advanced systems have introduced rule engines or parameter tables based on historical experience to adjust operating strategies. In some high-end applications, model predictive control or offline analysis based on big data is also being explored to assist decision-making. However, in general, these methods are mostly centered on single settings for individual points or entire warehouses, and the control logic is mainly based on steady-state or empirical rules, lacking in-depth modeling of complex spatiotemporal fields, differences between different product categories, and uncertainties.

[0004] It should be noted that the above content is not necessarily prior art, nor is it intended to limit the scope of patent protection of this application. Summary of the Invention

[0005] This application provides a control method, apparatus, equipment, and storage medium for fruit and vegetable cold chain preservation equipment to solve or alleviate one or more of the technical problems mentioned above.

[0006] One aspect of this application provides a control method for a fruit and vegetable cold chain preservation device, the method comprising: Spatiotemporal alignment and correlation matching are performed on the real-time acquired multi-source heterogeneous sensor data stream and the preset product metadata dataset to obtain the environment-product status dataset. Shelf life prediction and spoilage risk assessment are performed on each product in the environmental-product status dataset to obtain the remaining shelf life distribution data and spoilage risk data for each product. Based on the real-time acquired equipment operating cost data, the remaining shelf life distribution data, and the spoilage risk data, multi-objective optimization decision processing is performed to obtain a set of candidate equipment operating parameters. Based on the preset equipment operation and product preservation safety boundary set, the candidate equipment operation parameter set is verified and adjusted to generate equipment operation instructions.

[0007] Optionally, the step of performing spatiotemporal alignment and correlation matching on the real-time acquired multi-source heterogeneous sensor data stream and the preset product metadata dataset to obtain the environment-product status dataset includes: Based on the acquisition timestamps corresponding to each sensor data, the real-time acquired multi-source heterogeneous sensor data streams are time-domain aligned to obtain synchronous sensor data streams. By using a preset spatial coordinate system and based on the acquisition location information corresponding to each sensor data in the synchronous sensor data stream, spatial alignment processing is performed on the synchronous sensor data stream to obtain a spatiotemporally aligned environmental dataset. The environmental dataset is subjected to adaptive Kalman filtering to generate a denoised sensing dataset; The environmental field dataset is obtained by interpolating and compensating the denoised sensing dataset using a thin-plate spline spatial interpolation algorithm. The system performs correlation matching between the preset product metadata dataset and the RFID tags read from each product to generate a product status dataset.

[0008] Optionally, the step of performing shelf-life prediction and spoilage risk assessment on each product in the environment-product status dataset to obtain the remaining shelf-life distribution data and spoilage risk data for each product includes: By using a preset Monte Carlo dropout method, based on the optimal storage parameters corresponding to each product in the product status dataset, the shelf life of each product in the environment-product status dataset is predicted, and a set of predicted remaining shelf life values ​​for each product is obtained. Statistical analysis is performed on the predicted remaining shelf life values ​​to obtain the remaining shelf life distribution data and corresponding confidence values ​​for each product. Based on the spatial location information and confidence value of each product in the product status dataset, a spoilage risk assessment is performed on the remaining shelf life distribution data to generate spoilage risk data for each product.

[0009] Optionally, the step of performing multi-objective optimization decision processing based on real-time acquired equipment operating cost data, remaining shelf-life distribution data, and spoilage risk data to obtain a candidate equipment operating parameter set includes: Based on the remaining shelf life distribution data and the spoilage risk data, calculate the current freshness status index for each product; Based on the preset business priority strategy, determine the weight ratio between the preservation target and the energy consumption target for each product; Using a pre-set Pareto frontier decision query table, a decision matching query is performed based on real-time acquired equipment operating cost data, the current preservation status index, and the weight ratio to obtain a set of candidate equipment operating parameters.

[0010] Optionally, the step of verifying and adjusting the candidate equipment operation parameter set based on a preset equipment operation and product preservation safety boundary set, and generating equipment operation instructions, includes: Step S41: Based on the spatial location information corresponding to each product, perform preservation environment simulation processing on each operating parameter in the candidate equipment operating parameter set to obtain the key feature parameter set corresponding to each product; Step S42: Compare and verify the feature parameters in the key feature parameter set with the boundary thresholds corresponding to the preset equipment operation and product preservation safety boundary set; Step S43: When the feature parameters in the set of key feature parameters exceed the corresponding boundary threshold, the running parameters corresponding to the feature parameters are adjusted and updated according to the preset parameter fine-tuning rules. Repeat steps S41 to S43 until all feature parameters in the key feature parameter set do not exceed the corresponding boundary threshold, and then synthesize instructions according to the updated candidate device operation parameter set using a preset instruction synthesis method to generate device operation instructions.

[0011] Optionally, the method further includes: Based on the device operation instructions, the target device is recorded to obtain a historical operation dataset. Based on the preset safety constraint set and the historical operation dataset, the control parameters of the target device are constrained and optimized to generate a control enhancement and adjustment strategy corresponding to the target device. Based on the control enhancement and adjustment strategy, update the preset Pareto frontier decision lookup table and the preset parameter fine-tuning rules.

[0012] Another aspect of this application provides a control device for a cold chain preservation equipment for fruits and vegetables, the device comprising: The storage sensing module is used to perform spatiotemporal alignment and correlation matching processing on real-time acquired multi-source heterogeneous sensor data streams and preset product metadata datasets to obtain environment-product status datasets. The freshness prediction module is used to predict the shelf life and assess the spoilage risk of each product in the environment-product status dataset, and obtain the remaining shelf life distribution data and spoilage risk data of each product. The control decision module is used to perform multi-objective optimization decision processing based on the real-time acquired equipment operating cost data, the remaining shelf life distribution data, and the spoilage risk data to obtain a set of candidate equipment operating parameters; The adjustment and verification module is used to verify and adjust the candidate equipment operation parameter set according to the preset equipment operation and product preservation safety boundary set, and generate equipment operation instructions.

[0013] Another aspect of this application provides a computer device, including: At least one processor; and A memory that is communicatively connected to the at least one processor; Wherein: the memory stores instructions that can be executed by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0014] Another aspect of this application provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the method described above.

[0015] Another aspect of this application provides a computer program product including a computer program that, when executed by a processor, implements the method described above.

[0016] The embodiments of this application employing the above-described technical solution may have the following advantages: 1. By performing spatiotemporal alignment, adaptive Kalman filtering, and spatial interpolation on multi-source heterogeneous sensor data, high-confidence environmental-product status information can be obtained, enabling accurate perception and location of the microenvironment and stacking risks of fruits and vegetables, thereby significantly reducing misjudgments and losses caused by sensor noise or local anomalies.

[0017] 2. By using the Arrhenius mechanism as a prior and combining it with Bayesian deep learning (Monte Carlo dropout) to output the probability distribution of remaining shelf life, the uncertainty of prediction is quantified and risk-aware decision-making is achieved, thereby improving the accuracy of shelf life prediction and taking more conservative control measures under high uncertainty to reduce the risk of misoperation.

[0018] 3. By constructing Pareto fronts offline and making decisions online through table lookup, supplemented by lightweight CFD agent security verification and constrained reinforcement learning under digital twins to achieve policy evolution, a dynamic balance and continuous adaptive optimization between preservation effect and energy consumption are achieved, thereby maximizing quality preservation and economic benefits while ensuring safety. Attached Figure Description

[0019] The accompanying drawings exemplify embodiments and form part of the specification, serving together with the textual description to explain exemplary implementations of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0020] Figure 1 The schematic diagram illustrates a flow chart of the control method for a fruit and vegetable cold chain preservation device according to Embodiment 1 of this application; Figure 2 This schematic diagram illustrates the functional block diagram of the control device of the fruit and vegetable cold chain preservation equipment according to Embodiment 2 of this application; Figure 3 A schematic diagram of the hardware architecture of a computer device according to Embodiment 3 of this application is shown. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that the descriptions involving "first," "second," etc., in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one feature. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0023] In the description of this application, it should be understood that the numerical labels before the steps do not indicate the order of the steps, but are only used to facilitate the description of this application and to distinguish each step, and therefore should not be construed as a limitation of this application.

[0024] Example 1 like Figure 1 The diagram shows a flowchart of a control method for a fruit and vegetable cold chain preservation device provided in this application embodiment. The control method for the fruit and vegetable cold chain preservation device provided in this application embodiment includes the following steps.

[0025] Step S1: Perform spatiotemporal alignment and correlation matching on the real-time acquired multi-source heterogeneous sensor data stream and the preset product metadata dataset to obtain the environment-product status dataset.

[0026] In the fruit and vegetable cold chain preservation system of this application embodiment, comprehensive digital sensing of the physical space inside the cold chain equipment is required. Multi-source heterogeneous sensor data streams constitute the most fundamental data source of this system. These data streams refer to a time-series data set continuously collected from various sensing devices at different spatial locations and operating principles within the cold chain equipment. The acquisition of these multi-source heterogeneous sensor data streams relies on a carefully deployed sensor network architecture, which consists of two types of core units: one type is high-precision digital sensor nodes fixedly installed on the top, side, and shelf supports of the storage unit. These nodes form a static baseline framework for environmental monitoring, continuously collecting key parameters such as temperature, relative humidity, and carbon dioxide concentration at their location. The other type is mobile sensing units attached to the fruit and vegetable packaging units. These units are typically integrated within active RFID tags and dynamically change position as the goods move within the storage unit, thereby providing data on the microenvironment of the goods themselves. This architecture, which combines fixed and mobile sensors and collects multiple physical quantities simultaneously, naturally gives multi-source heterogeneous sensor data streams heterogeneous characteristics such as wide spatial distribution, multiple sensor types, and inconsistent data formats. Its core function is to comprehensively capture the spatiotemporal distribution details of environmental parameters within the cold chain space, providing raw data support for subsequent intelligent decision-making.

[0027] After obtaining the raw data stream, the primary task is to address its spatiotemporal consistency. This embodiment of the application achieves time-domain alignment by introducing a high-precision time synchronization module as the time base for the entire system. Each sensing unit, when acquiring readings, synchronously records precise timing pulses from this unified clock source, forming a standardized acquisition timestamp. The data processing engine then scans all sensor data within a fixed time window, grouping readings with the same or very close timestamps into the same processing batch. For example, at a specific moment, a fixed temperature sensor on the east wall of the warehouse, a mobile humidity tag on the middle shelf of the west side, and a carbon dioxide detection module in the central area may simultaneously report data. Time-domain alignment integrates these originally independent readings into a synchronous sensor data stream describing the overall environmental state at that moment. This time-domain alignment fundamentally avoids the "time misalignment" phenomenon caused by differences in sensor response speed or communication delays, ensuring that subsequent analysis is based on a snapshot of the real environment at the same moment.

[0028] After unifying the temporal dimension, it is necessary to further construct the intrinsic connections of the data in the spatial dimension. This embodiment achieves this through spatial alignment processing. Specifically, using a pre-defined three-dimensional coordinate system within the cold chain equipment, the spatial coordinates of each fixed sensor node are precisely measured and recorded in the system database during deployment. The position of the mobile sensing unit is dynamically calibrated using its associated RFID reader antenna position or indoor positioning technology based on signal strength. Spatial alignment processing forcibly binds each data point in the synchronous sensing data stream to its spatial coordinates, giving the originally abstract temperature and humidity values ​​explicit spatial attributes, forming a geographically meaningful spatiotemporally aligned environmental dataset.

[0029] Due to inherent measurement errors, electromagnetic interference, and transient environmental disturbances in sensors, raw data inevitably contains noise components. Therefore, this embodiment employs adaptive Kalman filtering for denoising. Specifically, a dynamic prediction model of environmental parameters is established to optimize the measured data. That is, the data sequence of each sensor is continuously tracked, and the possible value of the parameter at the next moment is predicted based on the historical trend of parameter changes. The predicted value is then weighted with the actual measured value. When a sensor experiences a momentary abnormal reading, the adaptive algorithm automatically reduces the weight of that measurement, relying more on the model's predicted value; when the data is stable and reliable, the confidence level of the measured data is increased. This dynamic adjustment mechanism enables the filter to effectively smooth out transient disturbances such as sudden temperature changes caused by the brief opening of the shutter, while preserving the true trend changes, ultimately outputting a denoised sensing dataset with smooth fluctuations and significantly improved reliability.

[0030] Based on obtaining high-quality sampling point data, a continuous environmental field covering the entire cold chain space needs to be constructed using spatial interpolation algorithms. This embodiment employs a thin-plate spline spatial interpolation algorithm, treating discretely distributed sensor points as spatial constraints. By solving a mathematical surface that minimizes bending energy, the surface can accurately pass through the measurements of all known sampling points while maintaining the smoothness and naturalness of the overall surface. Through this interpolation compensation process, the system can generate scientifically reasonable estimates in areas where sensors are not deployed. For example, it can accurately infer the temperature distribution in the intermediate region between two actual temperature measurement points, ultimately forming a uniformly resolved, continuous, and complete dataset of temperature, humidity, and gas concentration fields. The generation of this environmental field dataset allows the system to perceive subtle environmental features such as local hotspots and temperature stratification, providing a panoramic view for precise control.

[0031] Simultaneously, the system executes the product information digitization process in parallel, namely, associating and matching the preset product metadata dataset with RFID tags. The product metadata dataset, serving as a static knowledge base, stores the preservation physiological parameters of various fruits and vegetables, including but not limited to key characteristics such as variety name, optimal storage temperature range, chilling injury critical temperature, suitable humidity range, and respiration rate coefficient. When a fruit and vegetable unit with an RFID tag enters the cold chain environment, the reader automatically captures the unique identifier embedded in the tag and retrieves the corresponding preservation parameters from the product metadata database using this identifier, dynamically creating a product status dataset. This dataset not only records the basic attributes of the product but also continuously receives micro-environmental data from associated mobile sensing units, forming a direct mapping between the individual product and its environment.

[0032] Ultimately, the system deeply integrates the environmental field dataset and the product status dataset to generate a comprehensive data entity: the environment-product status dataset. This dataset achieves cross-domain correlation between the physical environment and biological characteristics, enabling the system to understand complex information such as "goods located in the low-temperature area in the northeast corner of the warehouse are banana varieties sensitive to chilling injury," rather than simply perceiving isolated physical quantities like "the temperature in the northeast corner is 12℃." This deep data fusion provides a complete data foundation for subsequent intelligent analyses such as shelf-life prediction and spoilage risk assessment, encompassing both the spatiotemporal distribution characteristics of the environment and the physiological characteristics of the products. This supports a qualitative leap in the entire fruit and vegetable cold chain preservation system, moving from environmental monitoring to quality management.

[0033] Step S2: Perform shelf life prediction and spoilage risk assessment on each product in the environment-product status dataset to obtain the remaining shelf life distribution data and spoilage risk data for each product.

[0034] In the cold chain preservation system for fruits and vegetables, step S2 of this application transforms the environment-product state dataset into a quantitative prediction of the product's future quality—achieved through shelf-life prediction and spoilage risk assessment. This processing begins with precise modeling of the individual product spoilage dynamics, where a pre-defined Monte Carlo dropout method constitutes the core mechanism of the prediction method. This mechanism is based on a specially designed deep neural network architecture, the key feature of which is that the prediction phase maintains a pattern of randomly disconnecting some neuron connections. This design ensures that each forward propagation is equivalent to computation in a slightly different sub-network. When the environmental data sequence of a specific product is input, the network performs hundreds of independent predictions, each using a different combination of neurons, thereby generating a set of remaining shelf-life prediction values ​​containing multiple predictions. For example, predicting the same box of strawberries in a 2°C environment may yield a set of differentiated prediction results such as [125 hours, 118 hours, 132 hours, 110 hours...]. This difference is not an error, but rather a quantitative expression of the system's prediction uncertainty.

[0035] In the prediction process, the optimal storage parameters for each product in the product status dataset play a crucial role. These parameters include kinetic parameters such as variety-specific activation energy and pre-exponential factors, which are linked to real-time environmental data through the Arrhenius equation. Specifically, the system inputs the temperature sequence from the environmental field data into a trained neural network, while embedding the product's basal metabolic rate as a physical constraint into the network calculation process. This approach allows the prediction model to learn complex nonlinear relationships discovered from historical data while adhering to basic biochemical principles. For example, the prediction for bananas strictly follows the critical chilling injury temperature constraint of no less than 13°C, while the prediction for blueberries may use lower suitable temperature parameters. This differentiated approach ensures the biological rationality of the prediction results.

[0036] After obtaining the predicted remaining shelf life set, the system enters the statistical analysis and processing stage. This process first calculates the mathematical characteristics of the predicted value set, including the mean, variance, and probability distribution. The mean represents the most likely remaining shelf life of the product, while the variance reflects the degree of uncertainty in the prediction. Using non-parametric statistical methods such as kernel density estimation, the system constructs a complete probability distribution curve for the remaining shelf life, which visually displays all possible values ​​and their corresponding probabilities. Simultaneously, the system calculates a confidence score as a quantitative indicator of prediction reliability. This indicator comprehensively considers the dispersion of the predicted value set, the quality of the training data, and the completeness of the current input environmental data. For example, when environmental sensor data is continuous and stable, the confidence score is usually high; while when data is missing or fluctuates drastically, the confidence score will decrease accordingly. This dynamic adjustment mechanism enhances the system's robustness.

[0037] Based on the generated remaining shelf life distribution data, the system further performs a spoilage risk assessment. This process first spatially correlates the product's specific coordinates within the cold chain environment with environmental field data, using spatial location information from the product status dataset. This correlation allows the system to identify individual products under adverse environmental conditions. For example, fruits and vegetables located in areas directly exposed to air vents may face excessive water loss, while goods piled in corners may experience localized temperature increases due to poor ventilation. The assessment algorithm comprehensively considers the statistical characteristics of the remaining shelf life distribution and spatial environmental factors, employing a multi-level scoring system to quantify the spoilage risk.

[0038] The evaluation formula used in the embodiments of this application can be expressed as: .in, C represents the comprehensive corruption risk index (range 0-1); S represents the prediction confidence level (range 0-1); and S represents the average remaining shelf life (in hours). D represents the maximum possible shelf life of this type of product under ideal conditions (in hours); D is the deviation of the environmental parameters of the location from the optimum value. , , Let be the weight coefficients of each influencing factor, and satisfy . + + =1.

[0039] The above assessment methods ensure that the system can respond promptly to low-confidence predictions, short remaining shelf life, and harsh local environments. For example, when a batch of grapes has a low prediction confidence level and is located in a relatively high-temperature area, even if its average remaining shelf life seems sufficient, the system will still assign a higher spoilage risk level, triggering more frequent monitoring and intervention.

[0040] Ultimately, the generated corruption risk data is visualized using heatmaps, with different colored areas intuitively showing the risk distribution within the cold chain space. High-risk areas are highlighted in red, medium-risk areas are transitioned to yellow, and low-risk areas are marked in green. This spatialized risk representation not only helps locate problem areas but also provides a direct basis for developing subsequent differentiated control strategies. The entire process forms a complete closed loop from environmental perception to risk warning, enabling the cold chain management system to shift from passive monitoring to proactive intervention, effectively improving the preservation quality of fruits and vegetables and reducing economic losses.

[0041] Step S3: Based on the real-time acquired equipment operating cost data, the remaining shelf life distribution data, and the spoilage risk data, perform multi-objective optimization decision processing to obtain a set of candidate equipment operating parameters.

[0042] To transform quality monitoring data into control commands, the current freshness status is first precisely quantified. In this embodiment, the calculation of the current freshness status index requires comprehensive consideration of the intrinsic relationship between remaining shelf life distribution data and spoilage risk data. This calculation process employs a probability-weighted comprehensive evaluation method. Specifically, key statistical features are first extracted from the remaining shelf life distribution data, including the distribution mean, standard deviation, and specific quantiles, while also incorporating the spatial risk coefficient and confidence index from the spoilage risk data. These parameters are linearly weighted to form a basic status score, the mathematical expression of which can be described as: Where Q is the current freshness index; μ is the mean of the remaining shelf life distribution; and σ is the standard deviation of the remaining shelf life distribution. is the spatial corruption risk coefficient, with a value range of [0,1]; C is the prediction confidence index, with a value range of [0,1]. to These are the weighting coefficients for each parameter. The above-described method for calculating the freshness status index ensures that the system can simultaneously consider multiple dimensions of factors, including the concentration trend of shelf life, prediction uncertainty, spatial risk distribution, and data reliability. For example, when the average remaining shelf life of a batch of strawberries is high but the prediction confidence is low, the system will automatically lower its status score to avoid making overly optimistic decisions due to misleading single-point data.

[0043] After obtaining the current freshness status index, the system enters the weight ratio determination stage. This stage achieves dynamic target trade-offs through a preset business priority strategy. The business priority strategy is a knowledge base containing multi-dimensional decision rules, comprehensively considering factors such as real-time energy market prices, product value levels, expected storage duration, and customer service level agreements. This strategy uses a combination of rule engines and fuzzy reasoning to transform abstract business requirements into specific weight ratio values. For example, during electricity demand response periods, when the power grid issues an energy-saving instruction, the strategy automatically increases the weight ratio of energy consumption targets; while when handling high-value imported berries, the system prioritizes freshness targets, correspondingly increasing their weight allocation. This dynamic adjustment mechanism allows the cold chain system to flexibly adapt to changes in the external environment and internal business needs, achieving an organic balance between economic benefits and quality assurance.

[0044] After weight configuration, the system initiates the matching and retrieval process for the Pareto front decision query table. This query table is a decision knowledge base pre-built through deep optimization calculations in the offline phase. Its generation process employs a multi-objective evolutionary algorithm to perform massive simulations on historical operating data. In the offline optimization phase, the system explores various possible combinations of equipment operating parameters in tens of thousands of simulation runs by changing operational variables such as the set temperature, humidity control target, and ventilation rate of the refrigeration system, and records the preservation status index and energy consumption cost index corresponding to each combination. After processing such as non-dominated sorting and crowding calculation, the system finally selects the set of parameters that constitute the Pareto optimal boundary. These sets represent the best trade-off solutions achievable under given constraints.

[0045] The actual retrieval process of the query table employs a multidimensional nearest neighbor matching algorithm. The system treats the real-time acquired equipment operating cost data, the current preservation status index, and the weight ratio combination as a multidimensional query vector. It determines the optimal equipment operating parameters by finding the matching point with the minimum Euclidean distance on the Pareto front surface. For example, when the status index corresponding to the query vector is 0.85, the energy consumption weight is 0.6, and the real-time electricity price is at the peak rate, the system will automatically match an operating parameter combination with a set temperature of 2.3℃, relative humidity of 88%, and moderate wind speed. This real-time query mechanism based on pre-calculated results ensures both the scientific nature and global optimality of the decision-making process, while perfectly meeting the stringent requirements of cold chain control for real-time response speed.

[0046] The output candidate equipment operating parameter set of the entire decision-making process contains a complete sequence of equipment control instructions. These instructions are ordered according to execution priority and are labeled with expected performance indicators. The parameter set adopts a hierarchical coding structure: the first layer defines the core temperature control target, the second layer refines the humidity adjustment parameters, the third layer sets the gas composition maintenance range, and the fourth layer plans the equipment operation sequence. This structured parameter organization facilitates item-by-item checking in subsequent verification stages and also enables distributed parsing by the execution units. The final generated candidate equipment operating parameter set not only provides specific control values ​​but also includes progressive adjustment paths to achieve these control targets, ensuring a smooth transition of equipment status and avoiding environmental shocks caused by sudden parameter changes.

[0047] Step S4: Based on the preset equipment operation and product preservation safety boundary set, verify and adjust the candidate equipment operation parameter set to generate equipment operation instructions.

[0048] To ensure the safety and reliability of the operating parameters, it is necessary to verify and adjust the obtained set of operating parameters for candidate equipment. Firstly, this embodiment uses a lightweight computational fluid dynamics surrogate model to simulate the preservation environment of the candidate equipment operating parameter set. At the start of the simulation, the system converts core parameters such as supply air temperature, fan speed, and return air humidity from the candidate equipment operating parameter set into boundary conditions for fluid simulation. Simultaneously, it establishes a three-dimensional spatial grid by combining the spatial location information centrally recorded in the product status data set. The surrogate model uses neural network forward propagation to predict the velocity, temperature, and humidity field distributions of the entire cold chain space within seconds. Its output is condensed into a set of key feature parameters corresponding to each product. This set includes physical quantities that directly affect preservation quality, such as wind speed, minimum contact temperature, relative humidity, and temperature fluctuation amplitude at each product unit's location. For example, for strawberry boxes placed on the top shelf, the system focuses on monitoring whether the surface wind speed is too high, leading to excessive moisture evaporation. For banana boxes stacked in corners, the system focuses on whether the contact temperature is below the chilling injury critical value.

[0049] After completing the environmental simulation, a comparative verification process based on safety boundaries is performed. The system's preset set of equipment operation and product preservation safety boundaries is a structured rule base. This set includes both physical limitations of the equipment, such as the minimum operating frequency of the compressor and the anti-frost temperature of the evaporator, and biological boundaries of the product, such as the critical temperature for chilling injury and the critical humidity for mold growth for various fruits and vegetables. The verification engine compares each feature parameter in the key feature parameter set with its corresponding boundary threshold in real time, employing a multi-level early warning mechanism to identify the degree of parameter deviation. For example, when the simulation shows that the leaf surface temperature of a batch of lettuce is close to the 0°C freezing point, the system generates a level one warning; while when it predicts that the humidity in a certain area is consistently higher than 95%, potentially causing mold growth, a level two warning is triggered. This hierarchical processing allows the system to distinguish the degree of urgency and prioritize handling parameter deviations that may cause irreversible damage.

[0050] When a key feature parameter exceeds its corresponding boundary threshold, the parameter fine-tuning rule is immediately activated. This rule base employs a generative rule architecture, containing fine-tuning logic in the form of condition-action pairs. The fine-tuning process follows the principle of minimal intervention, first analyzing the root cause of the parameter exceeding the boundary, and then specifically adjusting the relevant parameters in the candidate device operating parameter set. For example, when the wind speed on the surface of a grape packaging box near the air outlet exceeds the safety threshold of 3 m / s, the system will not directly shut down the fan, but will lower the wind speed setpoint by 0.2 m / s according to a preset gradient reduction rule; when it is predicted that the temperature in a certain corner is below the critical temperature for chilling injury to kiwifruit, the air supply temperature will be appropriately increased by 0.5℃ and the damper opening will be redistributed. This precise directional adjustment solves both local safety issues and maximizes the overall operating efficiency of the system. The decision logic in the parameter adjustment process in this embodiment can be expressed by the following optimization function: .in, Adjust the vector for the parameters; This represents the current value of the operating parameters for the i-th device. This represents the adjusted value of the operating parameters for the i-th device; This is the predicted value of the j-th key feature parameter after adjustment; Let be the threshold value of the j-th security boundary.

[0051] The entire verification and adjustment process forms a closed-loop iteration. After each parameter update, environmental simulation and safety verification are immediately re-executed until all key characteristic parameters are stable within the safety boundaries. This iterative mechanism ensures that even in complex multi-parameter coupling situations, the system can find a safe and feasible operating point through a step-by-step approximation approach. For example, when adjusting the ventilation parameters in a high-temperature storage area, it may be necessary to go through 3-4 rounds of fine-tuning to simultaneously meet the dual requirements of temperature uniformity and wind speed safety.

[0052] The final parameter set is sent to the instruction synthesis module, which converts the numerical parameters into control instructions executable by the equipment. Instruction synthesis employs a hierarchical coding strategy: the first-level instructions correspond to the refrigeration compressor frequency setting; the second-level instructions control the damper opening distribution; the third-level instructions adjust the humidifier output; and the fourth-level instructions schedule the defrosting cycle. Each instruction includes three elements: target value, rate of change, and transition time, ensuring a smooth transition of the equipment to the new operating state. For example, changes to the temperature setpoint are broken down into multiple steps to avoid thermal shock to fruits and vegetables from sudden temperature changes; humidity adjustment instructions are coordinated with temperature changes to prevent condensation during cooling. The generated equipment operation instruction set is sent to the field controller via an industrial communication protocol, completing the entire transformation from decision-making to execution. Simultaneously, the system records detailed logs of each adjustment, providing data support for subsequent strategy optimization.

[0053] In one optional implementation, to achieve system self-evolution, equipment operation records are transformed into learning data for system intelligent improvement. The equipment operation recording process begins with comprehensive monitoring of the target equipment's execution status. This monitoring covers equipment-level data such as the output power of the refrigeration compressor inverter, the actual value of the fan speed, and the feedback signal of the solenoid valve opening. Simultaneously, it records changes in the temperature field distribution, humidity fluctuation curves, and actual energy consumption data corresponding to equipment operation commands, fed back from environmental sensors. These data streams are aligned and compressed along a unified timeline to form a structured historical operation dataset. This dataset not only includes equipment response characteristic curves but also records the final fruit and vegetable quality maintenance effects achieved under different parameter combinations, forming a complete "decision-execution-result" data chain. For example, the system accurately records the compressor power increase curve, the change in the uniformity of the temperature field within the storage, and the rate of hardness decay in subsequent batches of strawberries after adjusting the supply air temperature from 2℃ to 1.5℃. This correlated data provides rich training samples for subsequent analysis.

[0054] Based on accumulated historical operational datasets, the system initiates a constraint optimization process, which constructs a secure training sandbox within a digital twin environment. The set of safety constraints defines the boundary conditions for the optimization process, including physical limits of the equipment such as the maximum load rate of the compressor and the minimum safe temperature of the evaporator, as well as product preservation limits such as the maximum temperature fluctuation range that various fruits and vegetables can withstand. The optimization algorithm employs a constraint-based reinforcement learning framework, using equipment control parameters as decision variables and exploring and learning with the objective function of maximizing overall operational efficiency. This objective function comprehensively considers the balance between quality retention rate and energy efficiency, seeking the optimal control strategy while satisfying all safety constraints. The learning process explores the optimization space of equipment control parameters through numerous simulation experiments. For example, it was found that for a specific model of refrigeration unit, when maintaining a -0.5℃ operating condition, a phased frequency reduction strategy saves energy and provides more stable temperature operation than continuous high-frequency operation. These findings are summarized by the system into a control reinforcement adjustment strategy, which is stored in the form of a conditional decision tree. Each decision node corresponds to a specific environmental state and equipment state, while the leaf nodes store the validated optimal combination of control parameters.

[0055] The practical application of the enhanced control strategy is reflected in the updating and upgrading process of the system's core knowledge components. The Pareto front decision lookup table is updated using an incremental learning mechanism. The system performs non-dominated sorting of newly discovered high-quality parameter combinations with the existing Pareto solution set, eliminating old solutions dominated by the new ones and retaining all non-dominated solutions to form an extended Pareto front. For example, when the system learns that a certain combination of supply air temperature and humidity can simultaneously improve blueberry freshness and reduce energy consumption under specific external climatic conditions, this combination will be added to the lookup table, enriching decision-making options. Simultaneously, the update of parameter fine-tuning rules is achieved through rule mining algorithms. The system analyzes successful cases in historical adjustment records to summarize more refined adjustment logic. For instance, the original rule might simply require "increasing the supply air temperature when the corner temperature is too low," while the updated rule might be refined to "when the cargo density is detected to be greater than 70% and the corner temperature is too low, prioritize adjusting the damper angle before considering the temperature setpoint."

[0056] The entire optimization and update process employs version control. Newly generated Pareto frontier decision lookup tables and parameter fine-tuning rules are first thoroughly validated in a digital twin environment. After comparative testing confirms that performance improvements have reached the expected threshold, they are then deployed to the online system via a hot-swap mechanism. The system retains the rollback capability of historical versions, ensuring rapid recovery to a stable state in the event of unforeseen circumstances. This continuous optimization mechanism allows the cold chain preservation system to gradually adapt to the long-term effects of equipment performance degradation and climate change, while also rapidly absorbing new knowledge generated during operation, forming a virtuous cycle from experience accumulation to performance improvement. Through continuously refined control strategies, the system ultimately achieves a continuous improvement in operational efficiency by gradually approaching the theoretical optimal energy consumption while ensuring the quality of fruits and vegetables.

[0057] The control method for fruit and vegetable cold chain preservation equipment provided in this embodiment obtains an environment-product status dataset by performing spatiotemporal alignment and correlation matching on multi-source heterogeneous sensor data streams and product metadata datasets. It then performs shelf-life prediction and spoilage risk assessment on each product in the environment-product status dataset to obtain remaining shelf-life distribution data and spoilage risk data. Based on equipment operating cost data, remaining shelf-life distribution data, and spoilage risk data, it performs multi-objective optimization decision-making to obtain a candidate equipment operating parameter set. Finally, it verifies and adjusts the candidate equipment operating parameter set based on the equipment operation and product preservation safety boundary set to generate equipment operating instructions. The technical solution of this application first obtains reliable perception through spatiotemporal fusion and Kalman filtering, then makes decisions based on mechanistic priors, Bayesian prediction, and offline Pareto lookup tables, and finally issues instructions after verification through proxy CFD and safety constraints, thereby maximizing preservation while balancing energy consumption and safety.

[0058] Example 2 like Figure 2 The diagram shown illustrates the functional block diagram of a control device for a fruit and vegetable cold chain preservation equipment according to an embodiment of this application. This device can be divided into one or more program modules. One or more program modules are stored in a storage medium and executed by one or more processors to complete the embodiment of this application. The program module referred to in this embodiment is a series of computer program instruction segments capable of performing a specific function. The following description will specifically introduce the function of each program module in this embodiment. Figure 2 As shown, the control device 1000 for fruit and vegetable cold chain preservation equipment may include: a storage sensing module 1100, a preservation prediction module 1200, a control decision module 1300, an adjustment verification module 1400, and a model optimization module 1500, wherein: The storage sensing module 1100 is used to perform spatiotemporal alignment and correlation matching processing on the real-time acquired multi-source heterogeneous sensor data stream and the preset product metadata dataset to obtain the environment-product status dataset. The freshness prediction module 1200 is used to perform shelf life prediction and spoilage risk assessment on each product in the environment-product status dataset, and obtain the remaining shelf life distribution data and spoilage risk data for each product. The control decision module 1300 is used to perform multi-objective optimization decision processing based on the real-time acquired equipment operating cost data, the remaining shelf life distribution data, and the spoilage risk data to obtain a set of candidate equipment operating parameters; The adjustment and verification module 1400 is used to verify and adjust the candidate equipment operation parameter set according to the preset equipment operation and product preservation safety boundary set, and generate equipment operation instructions.

[0059] As an optional embodiment, the storage sensing module 1100 is specifically used for: Based on the acquisition timestamps corresponding to each sensor data, the real-time acquired multi-source heterogeneous sensor data streams are time-domain aligned to obtain synchronous sensor data streams. By using a preset spatial coordinate system and based on the acquisition location information corresponding to each sensor data in the synchronous sensor data stream, spatial alignment processing is performed on the synchronous sensor data stream to obtain a spatiotemporally aligned environmental dataset. The environmental dataset is subjected to adaptive Kalman filtering to generate a denoised sensing dataset; The environmental field dataset is obtained by interpolating and compensating the denoised sensing dataset using a thin-plate spline spatial interpolation algorithm. The system performs correlation matching between the preset product metadata dataset and the RFID tags read from each product to generate a product status dataset.

[0060] As an optional embodiment, the freshness prediction module 1200 is specifically used for: By using a preset Monte Carlo dropout method, based on the optimal storage parameters corresponding to each product in the product status dataset, the shelf life of each product in the environment-product status dataset is predicted, and a set of predicted remaining shelf life values ​​for each product is obtained. Statistical analysis is performed on the predicted remaining shelf life values ​​to obtain the remaining shelf life distribution data and corresponding confidence values ​​for each product. Based on the spatial location information and confidence value of each product in the product status dataset, a spoilage risk assessment is performed on the remaining shelf life distribution data to generate spoilage risk data for each product.

[0061] As an optional embodiment, the control decision module 1300 is specifically used for: Based on the remaining shelf life distribution data and the spoilage risk data, calculate the current freshness status index for each product; Based on the preset business priority strategy, determine the weight ratio between the preservation target and the energy consumption target for each product; Using a pre-set Pareto frontier decision query table, a decision matching query is performed based on real-time acquired equipment operating cost data, the current preservation status index, and the weight ratio to obtain a set of candidate equipment operating parameters.

[0062] As an optional embodiment, the adjustment verification module 1400 is specifically used for: Step S41: Based on the spatial location information corresponding to each product, perform preservation environment simulation processing on each operating parameter in the candidate equipment operating parameter set to obtain the key feature parameter set corresponding to each product; Step S42: Compare and verify the feature parameters in the key feature parameter set with the boundary thresholds corresponding to the preset equipment operation and product preservation safety boundary set; Step S43: When the feature parameters in the set of key feature parameters exceed the corresponding boundary threshold, the running parameters corresponding to the feature parameters are adjusted and updated according to the preset parameter fine-tuning rules. Repeat steps S41 to S43 until all feature parameters in the key feature parameter set do not exceed the corresponding boundary threshold, and then synthesize instructions according to the updated candidate device operation parameter set using a preset instruction synthesis method to generate device operation instructions.

[0063] As an optional embodiment, the control device 1000 of the fruit and vegetable cold chain preservation equipment further includes a model optimization module 1500, which is specifically used for: Based on the device operation instructions, the target device is recorded to obtain a historical operation dataset. Based on the preset safety constraint set and the historical operation dataset, the control parameters of the target device are constrained and optimized to generate a control enhancement and adjustment strategy corresponding to the target device. Based on the control enhancement and adjustment strategy, update the preset Pareto frontier decision lookup table and the preset parameter fine-tuning rules.

[0064] Example 3 Figure 3 This illustration schematically shows a hardware architecture diagram of a computer device 10000 suitable for implementing a control method for fruit and vegetable cold chain preservation equipment according to Embodiment 3 of this application. In some embodiments, the computer device 10000 may be a terminal device such as a smartphone, wearable device, tablet computer, personal computer, vehicle terminal, game console, virtual device, workbench, digital assistant, set-top box, robot, etc. In other embodiments, the computer device 10000 may be a rack server, blade server, tower server, or cabinet server (including independent servers or server clusters composed of multiple servers), etc. Figure 3 As shown, the computer device 10000 includes, but is not limited to: a memory 10010, a processor 10020, and a network interface 10030 that can communicate and be linked with each other via a system bus. Wherein: The memory 10010 includes at least one type of computer-readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10010 may be an internal storage module of a computer device 10000, such as the hard disk or memory of the computer device 10000. In other embodiments, the memory 10010 may also be an external storage device of the computer device 10000, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 10000. Of course, the memory 10010 may also include both the internal storage module and the external storage device of the computer device 10000. In this embodiment, the memory 10010 is typically used to store the operating system and various application software installed on the computer device 10000, such as the program code for the control method of the fruit and vegetable cold chain preservation equipment. In addition, the memory 10010 can also be used to temporarily store various types of data that have been output or will be output.

[0065] In some embodiments, processor 10020 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other chip. Processor 10020 is typically used to control the overall operation of computer device 10000, such as performing control and processing related to data interaction or communication with computer device 10000. In this embodiment, processor 10020 is used to run program code stored in memory 10010 or process data.

[0066] Network interface 10030 may include a wireless network interface or a wired network interface, which is typically used to establish a communication link between computer device 10000 and other computer devices. For example, network interface 10030 is used to connect computer device 10000 to an external terminal via a network, establishing a data transmission channel and communication link between computer device 10000 and the external terminal. The network may be an intranet, the Internet, Global System for Mobile Communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, or other wireless or wired networks.

[0067] It should be pointed out that, Figure 3 Only computer devices with components 10010-10030 are shown; however, it should be understood that it is not required to implement all of the shown components, and more or fewer components may be implemented instead.

[0068] In this embodiment, the control method of the fruit and vegetable cold chain preservation equipment stored in the memory 10010 can also be divided into one or more program modules and executed by one or more processors (such as processor 10020) to complete the embodiment of this application.

[0069] Obviously, those skilled in the art should understand that the modules or steps of the embodiments of this application described above can be implemented using general-purpose computer devices. They can be centralized on a single computer device or distributed across a network of multiple computer devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computer device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of this application are not limited to any particular combination of hardware and software.

[0070] It should be noted that the above are merely preferred embodiments of this application and do not limit the scope of patent protection of this application. Any equivalent structural or procedural changes made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of this application.

Claims

1. A control method for a fruit and vegetable cold chain preservation equipment, characterized in that, The method includes: Spatiotemporal alignment and correlation matching are performed on the real-time acquired multi-source heterogeneous sensor data stream and the preset product metadata dataset to obtain the environment-product status dataset. Shelf life prediction and spoilage risk assessment are performed on each product in the environmental-product status dataset to obtain the remaining shelf life distribution data and spoilage risk data for each product. Based on the real-time acquired equipment operating cost data, the remaining shelf life distribution data, and the spoilage risk data, multi-objective optimization decision processing is performed to obtain a set of candidate equipment operating parameters. Based on the preset equipment operation and product preservation safety boundary set, the candidate equipment operation parameter set is verified and adjusted to generate equipment operation instructions.

2. The method according to claim 1, wherein the environment-product state dataset includes an environmental field dataset and a product state dataset, characterized in that, The process of performing spatiotemporal alignment and correlation matching on the real-time acquired multi-source heterogeneous sensor data stream and the preset product metadata dataset to obtain the environment-product status dataset includes: Based on the acquisition timestamps corresponding to each sensor data, the real-time acquired multi-source heterogeneous sensor data streams are time-domain aligned to obtain synchronous sensor data streams. By using a preset spatial coordinate system and based on the acquisition location information corresponding to each sensor data in the synchronous sensor data stream, spatial alignment processing is performed on the synchronous sensor data stream to obtain a spatiotemporally aligned environmental dataset. The environmental dataset is subjected to adaptive Kalman filtering to generate a denoised sensing dataset; The environmental field dataset is obtained by interpolating and compensating the denoised sensing dataset using a thin-plate spline spatial interpolation algorithm. The system performs correlation matching between the preset product metadata dataset and the RFID tags read from each product to generate a product status dataset.

3. The method according to claim 2, characterized in that, The process of performing shelf-life prediction and spoilage risk assessment on each product in the environmental-product status dataset to obtain the remaining shelf-life distribution data and spoilage risk data for each product includes: By using a preset Monte Carlo dropout method, based on the optimal storage parameters corresponding to each product in the product status dataset, the shelf life of each product in the environment-product status dataset is predicted, and a set of predicted remaining shelf life values ​​for each product is obtained. Statistical analysis is performed on the predicted remaining shelf life values ​​to obtain the remaining shelf life distribution data and corresponding confidence values ​​for each product. Based on the spatial location information and confidence value of each product in the product status dataset, a spoilage risk assessment is performed on the remaining shelf life distribution data to generate spoilage risk data for each product.

4. The method according to claim 1, characterized in that, The process of performing multi-objective optimization decision-making based on real-time acquired equipment operating cost data, remaining shelf life distribution data, and spoilage risk data to obtain a candidate equipment operating parameter set includes: Based on the remaining shelf life distribution data and the spoilage risk data, calculate the current freshness status index for each product; Based on the preset business priority strategy, determine the weight ratio between the preservation target and the energy consumption target for each product; Using a pre-set Pareto frontier decision query table, a decision matching query is performed based on real-time acquired equipment operating cost data, the current preservation status index, and the weight ratio to obtain a set of candidate equipment operating parameters.

5. The method according to claim 3, characterized in that, The step of verifying and adjusting the candidate equipment operation parameter set based on a preset equipment operation and product preservation safety boundary set, and generating equipment operation instructions, includes: Step S41: Based on the spatial location information corresponding to each product, perform preservation environment simulation processing on each operating parameter in the candidate equipment operating parameter set to obtain the key feature parameter set corresponding to each product; Step S42: Compare and verify the feature parameters in the key feature parameter set with the boundary thresholds corresponding to the preset equipment operation and product preservation safety boundary set; Step S43: When the feature parameters in the set of key feature parameters exceed the corresponding boundary threshold, the running parameters corresponding to the feature parameters are adjusted and updated according to the preset parameter fine-tuning rules. Repeat steps S41 to S43 until all feature parameters in the key feature parameter set do not exceed the corresponding boundary threshold, and then synthesize instructions according to the updated candidate device operation parameter set using a preset instruction synthesis method to generate device operation instructions.

6. The method according to claim 1, characterized in that, The method further includes: Based on the device operation instructions, the target device is recorded to obtain a historical operation dataset. Based on the preset safety constraint set and the historical operation dataset, the control parameters of the target device are constrained and optimized to generate a control enhancement and adjustment strategy corresponding to the target device. Based on the control enhancement and adjustment strategy, update the preset Pareto frontier decision lookup table and the preset parameter fine-tuning rules.

7. A control device for a fruit and vegetable cold chain preservation equipment, applied to the control method of the fruit and vegetable cold chain preservation equipment as described in claim 1, characterized in that, The device includes: The storage sensing module is used to perform spatiotemporal alignment and correlation matching processing on real-time acquired multi-source heterogeneous sensor data streams and preset product metadata datasets to obtain environment-product status datasets. The freshness prediction module is used to predict the shelf life and assess the spoilage risk of each product in the environment-product status dataset, and obtain the remaining shelf life distribution data and spoilage risk data of each product. The control decision module is used to perform multi-objective optimization decision processing based on the real-time acquired equipment operating cost data, the remaining shelf life distribution data, and the spoilage risk data to obtain a set of candidate equipment operating parameters; The adjustment and verification module is used to verify and adjust the candidate equipment operation parameter set according to the preset equipment operation and product preservation safety boundary set, and generate equipment operation instructions.

8. A computer device, characterized in that, include: At least one processor; and A memory communicatively connected to the at least one processor; wherein: The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 6.

Images