Method, device and equipment for optimizing carbon emission in vegetable producing area treatment process
By employing sensitivity analysis and deep reinforcement learning methods, key parameters in the vegetable production process are identified and optimized, addressing the problem of poor adaptability of carbon emission optimization methods in vegetable production processes and achieving efficient carbon emission optimization under complex operating conditions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-25
- Publication Date
- 2026-04-07
AI Technical Summary
Existing technologies for optimizing carbon emissions in vegetable production site processing have poor adaptability and low optimization efficiency, making it difficult to quickly and accurately identify key emission reduction parameters and output the optimal operation plan under complex and variable operating conditions.
By employing sensitivity analysis and deep reinforcement learning methods, this study acquires parameters from multiple stages of the vegetable production process, performs sensitivity analysis, selects adjustable parameters that are above a set threshold as the optimization parameter set, constructs a deep reinforcement learning environment, and performs iterative optimization based on a deep deterministic policy gradient algorithm to output the optimal combination of optimization parameters and carbon emission reduction scheme with the highest reward value.
It enables dynamic and intelligent optimization of carbon emissions under complex and variable operating conditions, significantly improving the targeting, efficiency and rationality of optimization, and providing reliable technical support for low-carbon agricultural production.
Smart Images
Figure CN121809809A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of carbon emission optimization, and in particular to a method, apparatus and equipment for optimizing carbon emissions in vegetable production site processing. Background Technology
[0002] With the deepening of the goals of "carbon peaking and carbon neutrality," energy conservation and emission reduction in the agricultural sector have become a crucial link in achieving the national dual-carbon goals. Vegetables, as an important component of fresh agricultural products, involve multiple stages in their on-site processing, including harvesting, sorting, pre-cooling, storage, and packaging, involving significant energy consumption and equipment operation. This makes them a potential key area for carbon emissions in the agricultural product cold chain logistics system. In recent years, to ensure vegetable quality and reduce spoilage rates, my country has accelerated the construction of cold chain facilities for the "first mile" of vegetable production. However, this has been accompanied by a significant increase in energy demand and carbon emissions in the on-site processing stage. Therefore, how to effectively identify and optimize carbon emissions during the on-site processing stage while ensuring vegetable quality and distribution efficiency has become a key technological requirement for promoting the green and low-carbon transformation of agriculture and implementing the whole-chain emission reduction goals.
[0003] Currently, in the field of cold chain carbon emission research, some literature has proposed methods for calculating carbon emissions from vegetable production site processing, which can quantitatively assess carbon emissions based on specific parameters. However, existing methods are mostly limited to static calculations and local analyses, lacking a dynamic optimization method that can adapt to different regions, vegetable types, and processing conditions. Due to the complexity of vegetable production site processing, the diversity of equipment types, and the high variability of operating parameters, traditional methods struggle to quickly and accurately identify key emission reduction parameters and output optimal operating schemes under multiple constraints, resulting in problems such as low optimization efficiency, poor adaptability, and weak operability in practical applications. Summary of the Invention
[0004] This invention provides a method, apparatus, and equipment for optimizing carbon emissions during vegetable production site processing, which solves the problems of poor adaptability and low optimization efficiency in existing methods for optimizing carbon emissions during vegetable production site processing, and realizes rapid identification and dynamic optimization of key carbon emission parameters under complex and variable working conditions.
[0005] This invention provides a method for optimizing carbon emissions during vegetable processing at the production site, comprising the following steps: The parameters of multiple stages in the vegetable production and processing process are obtained, and the carbon emissions are calculated based on a preset carbon emission calculation model. Sensitivity analysis was performed on the multiple parameters of the process, and the results of the sensitivity analysis were determined. Based on the sensitivity analysis results, parameters with sensitivity higher than the set threshold and adjustable in actual operation are selected as the set of optimization parameters. A deep reinforcement learning environment is constructed based on the optimized parameter set, and a carbon emission optimization algorithm is run based on the deep reinforcement learning environment to iteratively optimize the multi-dimensional optimization objectives in the vegetable production area processing process; the multi-dimensional optimization objectives include the carbon emission amount, quality preservation rate, and storage period. Based on the set of optimization parameters, the optimal combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme are output; the reward value is the optimization objective of the carbon emission optimization algorithm.
[0006] According to the present invention, a method for optimizing carbon emissions in vegetable production site processing includes obtaining parameters of multiple stages in the vegetable production site processing process and obtaining carbon emissions based on a preset carbon emission calculation model. Specifically, this method includes: obtaining physical parameters and operational parameters related to the vegetable production site processing stages; inputting the physical parameters and operational parameters into the preset carbon emission calculation model; and outputting a quantitative result characterizing the total carbon emissions of the vegetable production site processing process from the carbon emission calculation model. The preset carbon emission calculation model is a parameterized carbon emission calculation model constructed based on model parameters determined from historical data.
[0007] According to the present invention, a method for optimizing carbon emissions in a vegetable production process includes performing sensitivity analysis on multiple parameters to determine the sensitivity analysis results. Specifically, this includes: sequentially applying perturbation increments to individual parameters while keeping the remaining parameters constant; recording the change in carbon emissions output by the carbon emission calculation model after each perturbation; using the ratio of the change in carbon emissions to the perturbation increment as the sensitivity index for each parameter to form a sensitivity ranking list; marking parameters in the sensitivity ranking list that exceed a preset threshold as key sensitive parameters; and outputting the key sensitive parameters and their corresponding sensitivity indices as the sensitivity analysis results.
[0008] According to the present invention, a method for optimizing carbon emissions in vegetable production site processing is provided. The step of selecting parameters with sensitivity higher than a set threshold and adjustable in actual operation as an optimization parameter set based on the sensitivity analysis results specifically includes: screening parameters with sensitivity indices higher than a set threshold from the sensitivity analysis results to form a high-sensitivity parameter set; determining whether each parameter in the high-sensitivity parameter set belongs to a preset operable parameter type; and selecting parameters belonging to the operable parameter type as the optimization parameter set.
[0009] According to the present invention, a method for optimizing carbon emissions in vegetable origin processing includes constructing a deep reinforcement learning environment based on the optimization parameter set, and iteratively optimizing multi-dimensional optimization objectives in the vegetable origin processing based on the deep reinforcement learning environment. Specifically, this includes: defining a discrete-continuous action space using the adjustable range of the optimization parameter set; defining a state space using the current state value of the optimization parameter set; and defining a reward function using the negative value or reduction of carbon emissions output by the carbon emission calculation model, thus completing the construction of the deep reinforcement learning environment. The reward function is determined based on key features, with carbon emission minimization, quality preservation rate maximization, and storage period minimization as optimization objectives. The key features simultaneously satisfy the low-carbon requirements of cold chain logistics, the actual needs of growers, and the essential characteristics of vegetable origin processing. A deep deterministic policy gradient algorithm is employed. Gradient (DDPG) interacts and learns in the deep reinforcement learning environment to generate action policies with the goal of maximizing cumulative rewards; the action policies are then used as new parameter combinations to input into a carbon emission optimization algorithm for multiple iterations until the agent meets the convergence condition or reaches a preset number of iterations; the agent is the carrier that interacts with the deep reinforcement learning environment.
[0010] According to the present invention, a method for optimizing carbon emissions in a vegetable production process includes the following steps: First, based on the set of optimization parameters, outputting the optimal combination of optimization parameters with the highest reward value and the corresponding carbon reduction optimization scheme. Specifically, this includes: after the iterative optimization process is completed, selecting the parameter combination with the highest reward value from all iterations as the optimal combination of optimization parameters; generating a carbon reduction optimization scheme based on the optimal combination of optimization parameters; the carbon reduction optimization scheme including specific parameter adjustment suggestions; and outputting the optimal combination of optimization parameters and the carbon reduction optimization scheme.
[0011] The present invention also provides an optimization device for carbon emissions in the vegetable production site processing, comprising the following modules: The carbon emission calculation module is used to obtain parameters from multiple stages of the vegetable production and processing process, and to obtain the carbon emission amount based on the preset carbon emission calculation model. The sensitivity analysis module is used to perform sensitivity analysis on the multiple process parameters and determine the sensitivity analysis results; The parameter selection optimization module is used to select parameters with sensitivity higher than a set threshold and adjustable in actual operation as the optimization parameter set based on the sensitivity analysis results. An iterative optimization module is used to construct a deep reinforcement learning environment based on the optimization parameter set, and to iteratively optimize the multi-dimensional optimization objectives in the vegetable production site processing process based on the deep reinforcement learning environment; the multi-dimensional optimization objectives include carbon emissions, quality preservation rate, and storage period. The scheme output module is used to output the best combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme based on the optimization parameter set; the reward value is the optimization target of the carbon emission optimization algorithm.
[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the carbon emission optimization method for vegetable production site processing as described above.
[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for optimizing carbon emissions in the vegetable production site processing as described above.
[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for optimizing carbon emissions in the vegetable production site processing as described above.
[0015] This invention provides a method, apparatus, and equipment for optimizing carbon emissions in vegetable production site processing, offering the following advantages: By introducing a sensitivity analysis mechanism, it can accurately identify key parameters that significantly impact carbon emissions and are practically adjustable from numerous parameters, effectively focusing on optimization objectives and overcoming the low optimization efficiency caused by numerous parameters and unclear priorities in traditional methods. Furthermore, by constructing a deep reinforcement learning environment and employing an iterative optimization strategy, the scheme can adaptively explore carbon emission performance under different parameter combinations. With minimizing carbon emissions, maximizing quality preservation rate, and minimizing storage period as optimization objectives, a reward function is set, ultimately enabling the agent to automatically output the optimal parameter combination and emission reduction scheme with the highest reward value. This not only significantly improves the targeting, efficiency, and rationality of carbon emission optimization but also achieves dynamic and intelligent optimization of vegetable production site processing under complex and variable conditions, providing reliable technical support for low-carbon agricultural production. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating the optimized method for reducing carbon emissions during vegetable processing at the production site, provided by the present invention.
[0018] Figure 2This is a pseudocode diagram of the carbon emission optimization procedure for on-site processing provided by the present invention.
[0019] Figure 3 This is a schematic diagram of the structure of the optimization device for carbon emissions in the vegetable production site processing provided by the present invention.
[0020] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0022] The following is combined Figures 1-4 The embodiments of the present invention are described in detail.
[0023] Figure 1 This is a flowchart illustrating the optimized method for carbon emissions during vegetable processing at the production site provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps: S110. Obtain parameters from multiple stages of the vegetable production process and calculate carbon emissions based on a pre-set carbon emission calculation model.
[0024] According to the present invention, a method for optimizing carbon emissions in vegetable production site processing is provided. This method involves acquiring parameters from multiple stages of the vegetable production site processing process and obtaining carbon emissions based on a preset carbon emission calculation model. Specifically, the method includes: acquiring physical and operational parameters related to the vegetable production site processing stages; inputting the physical and operational parameters into the preset carbon emission calculation model; and outputting a quantitative result representing the total carbon emissions from the vegetable production site processing process from the carbon emission calculation model. The preset carbon emission calculation model is a parameterized carbon emission calculation model constructed based on model parameters determined from historical data.
[0025] Specifically, let's take the processing of broccoli in a certain vegetable production area as an example.
[0026] Through on-site surveys and equipment monitoring, physical and operational parameters were obtained, including diesel consumption during harvesting, electricity consumption during sorting, refrigerant type and operating time during pre-cooling, type and average temperature of cold storage, and amount of plastic film used during packaging. These specific parameter values were then input into a pre-set parameterized carbon emission calculation model. This model has been pre-loaded with key parameters such as carbon emission factors for various energy sources and global warming potential values for various refrigerants, based on extensive historical data. After receiving the input parameters, the model performs calculations using its built-in logic; the calculation formula can be expressed as: The final output is a clear total carbon emission amount measured in carbon dioxide equivalent (kg CO2e), such as "the total carbon emissions at the production site for processing one ton of broccoli are 15.8 kg CO2e".
[0027] By systematically collecting measured parameters covering all processing stages and using a pre-calibrated parametric model based on historical data for calculation, the carbon emissions from vegetable production site processing were accurately quantified. This method effectively overcomes the problems of large errors and low reliability caused by traditional estimation methods relying on experience and partial data. It provides a reliable and objective data foundation for subsequent sensitivity analysis and optimization, thereby ensuring the scientific nature and accuracy of the entire optimization process.
[0028] S120. Perform sensitivity analysis on multiple parameters and determine the sensitivity analysis results.
[0029] According to the present invention, a method for optimizing carbon emissions in vegetable production site processing is provided. Sensitivity analysis is performed on multiple parameters to determine the results. Specifically, the method includes: sequentially applying perturbation increments to individual parameters while keeping the other parameters constant, recording the change in carbon emissions output by the carbon emission calculation model after each perturbation; using the ratio of the change in carbon emissions to the perturbation increment as the sensitivity index for each parameter to form a sensitivity ranking list; marking parameters in the sensitivity ranking list that exceed a preset threshold as key sensitive parameters, and outputting the key sensitive parameters and their corresponding sensitivity indices as the sensitivity analysis results.
[0030] Specifically, taking the aforementioned broccoli processing as an example, key parameters include precooling time, storage temperature, and packaging material usage. During sensitivity analysis, a perturbation increment is first applied to the "precooling time" parameter, for example, increasing it by 10% (0.2 hours) from the original value of 2 hours, while keeping all other parameters, such as storage temperature and packaging material usage, unchanged. This new set of parameters is then input into the carbon emission calculation model, and the change in carbon emissions is recorded. Assuming the output value increases from 15.8 kgCO2e to 16.1 kgCO2e, the change is +0.3 kgCO2e. Subsequently, the sensitivity index of this parameter is calculated: 0.3 / 0.2 = 1.5 (kg CO2e / hour). This process is repeated for all parameters, such as "storage temperature" and "packaging material usage." Finally, the sensitivity indices of all parameters are summarized and sorted, for example, resulting in a list: [Precooling time: 1.5, Storage temperature: 1.2, Packaging material usage: 0.3, …]. If the preset sensitivity threshold is 0.5, then "pre-cooling time" and "storage temperature" will be marked as key sensitive parameters, and this list will be output as the analysis results.
[0031] By systematically applying perturbations and quantifying comparisons, the abstract impact of each parameter on total carbon emissions is transformed into concrete and comparable sensitivity indicators. This allows for the objective and accurate identification of the key parameters that have the greatest impact on the final carbon emission results, thus focusing the complex optimization problem on a few core levers. This provides a clear and reliable scientific basis for the precise allocation of subsequent optimization resources and the formulation of efficient optimization strategies.
[0032] S130. Based on the sensitivity analysis results, select parameters whose sensitivity is higher than the set threshold and can be adjusted in actual operation as the set of optimization parameters.
[0033] According to the present invention, a method for optimizing carbon emissions in vegetable production site processing is provided. Based on the sensitivity analysis results, parameters with sensitivity higher than a set threshold and adjustable in actual operation are selected as the optimization parameter set. Specifically, the method includes: screening parameters with sensitivity indices higher than a set threshold from the sensitivity analysis results to form a high-sensitivity parameter set; determining whether each parameter in the high-sensitivity parameter set belongs to a preset operable parameter type; and selecting the parameters belonging to the operable parameter type as the optimization parameter set.
[0034] Specifically, the sensitivity analysis results output a list of key sensitive parameters, for example: [Precooling time: 1.5, Storage temperature: 1.2, Packaging material usage: 0.3, Harvesting method: 0.1]. First, a sensitivity threshold of 0.5 is set, from which "precooling time," "storage temperature," and "packaging material usage" are selected to form a set of highly sensitive parameters. Then, the system or operator makes a judgment based on a preset list of "operable parameter types," which defines the categories of parameters that can be adjusted in actual production. For example, "precooling time" and "storage temperature" are equipment operating settings and can be changed by adjusting the control strategy, therefore they are judged as operable parameters; while "packaging material usage" may not be changeable in the short term due to fixed product specifications and customer requirements, and is therefore judged as an inoperable parameter; as for "harvesting method," although its sensitivity is below the threshold and it is not included, even if it were included, if it is determined by a fixed agronomic process, it may also be judged as inoperable. Ultimately, the system officially selected "pre-cooling time" and "storage temperature," two parameters that are both highly sensitive and operable, as the set of optimization parameters for this optimization.
[0035] By combining theoretically high-sensitivity parameters with adjustability in actual production, the optimization objective is ensured to not only have a significant impact but also possess practical feasibility. This effectively avoids wasting optimization resources on parameters that are theoretically effective but practically unchangeable, allowing subsequent deep reinforcement learning optimization to focus on truly controllable key variables, thereby significantly improving the practical value and success rate of the entire optimization scheme.
[0036] S140. Construct a deep reinforcement learning environment based on the optimized parameter set, run the carbon emission optimization algorithm based on the deep reinforcement learning environment, and iteratively optimize the multi-dimensional optimization objectives in the vegetable production process; the multi-dimensional optimization objectives include carbon emissions, quality preservation rate and storage period.
[0037] According to the present invention, a method for optimizing carbon emissions in vegetable production site processing is provided. A deep reinforcement learning environment is constructed based on an optimization parameter set. This environment iteratively optimizes multi-dimensional optimization objectives in the vegetable production site processing process. Specifically, the method includes: defining a discrete-continuous action space based on the adjustable range of the optimization parameter set; defining a state space based on the current state value of the optimization parameter set; and defining a reward function based on the negative value or reduction of carbon emissions output by the carbon emission calculation model, thus completing the construction of the deep reinforcement learning environment. The reward function is determined based on key features, with the optimization objectives being the minimization of carbon emissions, the maximization of quality preservation rate, and the minimization of storage period. The key features simultaneously satisfy the low-carbon requirements of cold chain logistics, the actual needs of growers, and the essential characteristics of vegetable production site processing. A deep deterministic policy gradient algorithm is used for interactive learning within the deep reinforcement learning environment, aiming to maximize cumulative rewards and generate action strategies. These action strategies are then input as new parameter combinations into the carbon emission optimization algorithm for multiple iterations until the agent meets the convergence condition or reaches the preset number of iterations. The agent serves as the carrier interacting with the deep reinforcement learning environment.
[0038] Specifically, taking the optimization parameter sets of "pre-cooling time" and "storage temperature" as an example, a deep reinforcement learning environment can be built using the Visual Studio Code platform and Python language. The adjustable range of "pre-cooling time" is defined as 1.5 to 3.0 hours, and the adjustable range of "storage temperature" is defined as 1 to 5℃. All possible combinations of the two constitute the action space. The specific values of the two at a certain moment (2 hours, 2℃) are defined as the state space. The carbon emissions, vegetable quality retention rate, and origin storage time output by the carbon emission calculation model are defined as complex reward functions for multi-objective optimization. The goal is to simultaneously meet the actual needs of low carbon emissions, high quality retention, and efficient circulation at the origin. The carbon emissions are directly used as the negative value as the reward value. The vegetable quality retention rate and origin storage time adopt the idea of course learning, setting thresholds to correspond to reward values. For example, a quality retention rate greater than 95% will receive 15 reward values, and greater than 90% will receive 10 reward values; an origin storage time less than 0.5 days will receive 15 reward values, and an origin storage time greater than 0.5 days but less than 1 day will receive 10 reward values. Subsequently, a deep deterministic policy gradient algorithm is used to enable the agent to explore and learn in this environment.
[0039] For example, an action strategy is first generated, with parameters adjusted to (2.2 hours, 3℃). This combination is then input into the carbon emission calculation model and related quality and efficiency evaluation modules, resulting in the following output: carbon emissions of 16.5 kg CO2e, vegetable quality retention rate of 93%, and on-site storage time of 0.6 days. Subsequently, the system calculates the total reward value based on a preset multi-objective reward function: the negative value of carbon emissions (-16.5), plus a reward of 10 for the quality retention rate >90%, plus another 10 for the storage time being between 0.5 and 1 day, resulting in a total reward of 3.5. The agent adjusts its strategy based on this moderate reward value, and in the next attempt (2.5 hours, 4℃), a new output is obtained: carbon emissions of 15.2 kg CO2e, quality retention rate of 96%, and storage time of 0.4 days. The new total reward value is: carbon emission reward (-15.2), plus a reward of 15 for quality retention rate >95%, plus a reward of 15 for storage time <0.5 days, resulting in a significant increase in the total reward to 14.8. Through multiple rounds of trial and error iteration, the agent is incentivized to find a parameter combination that simultaneously achieves low carbon emissions, high quality, and high efficiency (i.e., a parameter combination that balances carbon emissions, vegetable quality retention rate, and on-site storage time) until it finds the globally optimal solution with the highest total reward value.
[0040] By formalizing the complex multi-parameter optimization problem into a standard deep reinforcement learning task, advanced intelligent algorithms such as deep deterministic policy gradient algorithms can autonomously and efficiently search within a vast parameter space. This method avoids the limitations of traditional optimization methods that are prone to getting trapped in local optima or relying on expert experience. It achieves automated, adaptive global optimization of carbon emission targets under multiple constraints, significantly improving the efficiency and effectiveness of finding the optimal emission reduction scheme.
[0041] S150. Based on the set of optimized parameters, output the optimal combination of optimized parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme; the reward value is the optimization objective of the carbon emission optimization algorithm.
[0042] According to the present invention, a method for optimizing carbon emissions in vegetable production site processing is provided. Based on an optimization parameter set, the method outputs the optimal combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme. Specifically, the method includes: after the iterative optimization process is completed, selecting the parameter combination with the highest reward value from all iterations as the optimal combination of optimization parameters; generating a carbon emission reduction optimization scheme based on the optimal combination of optimization parameters; the carbon emission reduction optimization scheme includes specific parameter adjustment suggestions; and outputting the optimal combination of optimization parameters and the carbon emission reduction optimization scheme.
[0043] Specifically, after the DDPG algorithm completes all 10,000 preset iterations of training, the system iterates through and compares the total reward values recorded in all iterations. Assuming that from the historical data of the 7524th iteration, the parameter combination that maximizes the total reward value to 25.5 is selected, namely "pre-cooling time: 2.8 hours, storage temperature: 3.5℃", the corresponding model output is: carbon emissions 12.1 kg CO2e, vegetable quality retention rate 97%, and on-site storage time 0.4 days. This combination is determined to be the optimal parameter combination.
[0044] Based on this optimal parameter combination, the system automatically generates a structured carbon reduction optimization plan. This plan not only includes specific parameter adjustment suggestions, such as "It is recommended to extend the pre-cooling cycle duration from the current 2.0 hours to 2.8 hours, and adjust the average temperature of the storage cycle from the current 2.0℃ to 3.5℃," but also clarifies the expected comprehensive benefits of this plan: "According to model calculations, implementing this plan can reduce carbon emissions to 12.1 kg CO2e while increasing the vegetable quality retention rate to 97% and shortening the on-site storage time to 0.4 days, achieving the comprehensive optimization goals of energy saving, quality preservation, and efficiency improvement." The system outputs this optimal parameter combination containing specific values, the executable textual adjustment suggestions, and its comprehensive benefit analysis as the final optimization conclusion.
[0045] By automating the screening of historical best solutions and transforming them into clear and actionable guidelines, the complex results of algorithmic iterations are directly translated into specific production adjustment instructions. This process significantly lowers the barrier to entry for the solutions, enabling frontline production personnel to implement carbon reduction measures without needing to understand the underlying algorithms. This significantly improves the practicality and efficiency of the optimization results, ensuring that emission reduction targets can be effectively achieved in actual production.
[0046] The following pseudocode fully describes the software system framework for implementing the technical solution of this invention. It transforms theoretical optimization methods into executable computational tasks. Its core lies in autonomously finding the optimal parameter combination that minimizes the output value of the carbon emission calculation model in a simulated deep reinforcement learning environment using the DDPG deep reinforcement learning algorithm.
[0047] like Figure 2 The following is the pseudocode for the on-site processing carbon emission optimization procedure: initialization: Set random seed (42) Define the carbon emission model parameter class CarbonModel: Initialize fixed parameters (Ks, Kf, s, L, Kc, Wf, Ke, Δt, Cp, Kw, Kn, Wb,m) Define a list of precooling methods (water cooling, vacuum cooling, differential pressure precooling, cold storage cooling). Define a list of pre-cooling refrigerant types (R22, R134a, R404A, R407C, CO2). Define a list of storage methods (general cold storage, controlled atmosphere storage, low-temperature storage, intelligent cold storage). Define a list of storage refrigerant types (R22, R134a, R404A, R407C, CO2) Define parameter adjustment range Define the environment class VegetableEnv: Initialization: Receive carbon emission model Define the state dimension (4) and the action dimension (9). Define discrete parameter index Implement parameter sampling function Implement the state reset function Implement the action execution function Implement the reward calculation function Define the Actor network: Three-layer fully connected neural network Using the ReLU activation function and Tanh output Define the Critic network: Receive state and action as input Three-layer fully connected neural network Output Q value Define the experience replay buffer: Implement experience storage function Implement random sampling function Define the OU noise class: Achieve noise generation and attenuation functions Define the DDPG intelligent agent class: Initialize the Actor and Critic networks and the target network. Define optimizer and hyperparameters Implement the action selection function Implement network update function Implement experience playback function Implement exploration rate decay function Main program: 1. Initialize carbon emission models and environment 2. Perform a random search to obtain the initial optimal solution. For num_samples iterations: Randomly generate parameter combinations Calculate carbon emissions Record the optimal solution 3. Initialize the DDPG agent and load the initial optimal solution. 4. Training the DDPG agent For each training cycle (episodes): Reset Environment For each time step: Choose an action (with exploration) Perform actions and receive rewards Store experience in replay buffer Update network parameters Update the global optimal solution Decay Exploration Rate Record training metrics 5. Evaluate the final results Multi-step evaluation using a trained agent. Record emission changes and optimal parameters Visual evaluation results 6. Output optimization results Comparing the results of random search and DDPG Displaying the carbon emission decomposition at each stage Show the optimal parameter combination 7. Visualize the training process Plot the reward change curve Plotting carbon emission optimization curves Plot the loss function curve Plot the exploration rate decay curve Finish The following describes the device for optimizing carbon emissions in the vegetable production process provided by the present invention. The device for optimizing carbon emissions in the vegetable production process described below and the method for optimizing carbon emissions in the vegetable production process described above can be referred to in correspondence.
[0048] like Figure 3 The image shows an optimization device for carbon emissions in vegetable production site processing provided by the present invention, comprising: The carbon emission calculation module 310 is used to obtain parameters of multiple stages in the vegetable production and processing process, and to obtain the carbon emission amount based on the preset carbon emission calculation model. Sensitivity analysis module 320 is used to perform sensitivity analysis on multiple parameters and determine the sensitivity analysis results; The optimization parameter selection module 330 is used to select parameters with sensitivity higher than a set threshold and adjustable in actual operation as the optimization parameter set based on the sensitivity analysis results. The iterative optimization module 340 is used to construct a deep reinforcement learning environment based on the optimization parameter set, and to iteratively optimize the multi-dimensional optimization objectives in the vegetable production site processing process based on the deep reinforcement learning environment; the multi-dimensional optimization objectives include carbon emissions, quality preservation rate and storage period. The solution output module 350 is used to output the best combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme based on the optimization parameter set; the reward value is the optimization objective of the carbon emission optimization algorithm.
[0049] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other through the communications bus 440. The processor 410 can call logic instructions in the memory 430 to execute an optimization method for carbon emissions in the vegetable production site processing process. This method includes: acquiring multiple parameters in the vegetable production site processing process and obtaining carbon emissions based on a preset carbon emission calculation model; performing sensitivity analysis on the multiple parameters to determine the sensitivity analysis results; selecting parameters with sensitivity higher than a set threshold and adjustable in actual operation as an optimization parameter set based on the sensitivity analysis results; constructing a deep reinforcement learning environment based on the optimization parameter set; running a carbon emission optimization algorithm based on the deep reinforcement learning environment to iteratively optimize multi-dimensional optimization objectives in the vegetable production site processing process; the multi-dimensional optimization objectives include carbon emissions, quality preservation rate, and storage period; and outputting the optimal combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme based on the optimization parameter set; the reward value is the optimization objective of the carbon emission optimization algorithm.
[0050] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0051] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the optimization method for carbon emissions in the vegetable production site processing process provided by the above methods. The method includes: acquiring multiple parameters in the vegetable production site processing process and obtaining carbon emissions based on a preset carbon emission calculation model; performing sensitivity analysis on the multiple parameters to determine the sensitivity analysis results; selecting parameters with sensitivity higher than a set threshold and adjustable in actual operation as an optimization parameter set based on the sensitivity analysis results; constructing a deep reinforcement learning environment based on the optimization parameter set; running a carbon emission optimization algorithm based on the deep reinforcement learning environment to iteratively optimize the multi-dimensional optimization objectives in the vegetable production site processing process; the multi-dimensional optimization objectives include carbon emissions, quality preservation rate, and storage period; and outputting the best combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme based on the optimization parameter set; the reward value is the optimization objective of the carbon emission optimization algorithm.
[0052] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements an optimization method for carbon emissions in the vegetable production site processing process provided by the above methods. This method includes: acquiring multiple parameters in the vegetable production site processing process and obtaining carbon emissions based on a preset carbon emission calculation model; performing sensitivity analysis on the multiple parameters to determine the sensitivity analysis results; selecting parameters with sensitivity higher than a set threshold and adjustable in actual operation as an optimization parameter set based on the sensitivity analysis results; constructing a deep reinforcement learning environment based on the optimization parameter set; running a carbon emission optimization algorithm based on the deep reinforcement learning environment to iteratively optimize multi-dimensional optimization objectives in the vegetable production site processing process; the multi-dimensional optimization objectives include carbon emissions, quality preservation rate, and storage period; and outputting the optimal combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme based on the optimization parameter set; the reward value is the optimization objective of the carbon emission optimization algorithm.
[0053] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0054] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.
[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for optimizing carbon emissions during vegetable production site processing, characterized in that, include: The parameters of multiple stages in the vegetable production and processing process are obtained, and the carbon emissions are calculated based on a preset carbon emission calculation model. Sensitivity analysis was performed on the multiple parameters of the process, and the results of the sensitivity analysis were determined. Based on the sensitivity analysis results, parameters with sensitivity higher than the set threshold and adjustable in actual operation are selected as the set of optimization parameters. A deep reinforcement learning environment is constructed based on the optimized parameter set, and a carbon emission optimization algorithm is run based on the deep reinforcement learning environment to iteratively optimize the multi-dimensional optimization objectives in the vegetable production area processing process; the multi-dimensional optimization objectives include the carbon emission amount, quality preservation rate, and storage period. Based on the set of optimization parameters, the optimal combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme are output; the reward value is the optimization objective of the carbon emission optimization algorithm.
2. The method for optimizing carbon emissions during vegetable production site processing according to claim 1, characterized in that, The acquisition of parameters from multiple stages of the vegetable production and processing process, and the calculation of carbon emissions based on a pre-set carbon emission calculation model, specifically includes: Obtain physical and operational parameters related to vegetable processing at the production site; The physical parameters and the operational parameters are input into the preset carbon emission calculation model; The carbon emission calculation model outputs a quantitative result characterizing the total carbon emissions from the vegetable production site processing process; The preset carbon emission calculation model is a parameterized carbon emission calculation model constructed based on model parameters determined from historical data.
3. The method for optimizing carbon emissions during vegetable production site processing according to claim 1, characterized in that, The sensitivity analysis of the multiple process parameters, and the determination of the sensitivity analysis results, specifically include: The perturbation increment is applied sequentially to a single parameter among the multiple parameters, while keeping the other parameters unchanged, and the change in carbon emissions output by the carbon emission calculation model after each perturbation is recorded. The ratio of the change in carbon emissions to the increment of disturbance is used as the sensitivity index for each parameter, forming a sensitivity ranking list. The parameters in the sensitivity ranking list that are higher than a preset threshold are marked as key sensitive parameters, and the key sensitive parameters and their corresponding sensitivity indices are output as the sensitivity analysis results.
4. The method for optimizing carbon emissions during vegetable production site processing according to claim 1, characterized in that, The step of selecting parameters with sensitivity higher than a set threshold and adjustable in actual operation as the optimization parameter set based on the sensitivity analysis results specifically includes: From the sensitivity analysis results, parameters with sensitivity indices higher than a set threshold are selected to form a set of high-sensitivity parameters; For each parameter in the set of highly sensitive parameters, determine whether it belongs to a preset operable parameter type; The parameters belonging to the operable parameter type are selected as the optimization parameter set.
5. The method for optimizing carbon emissions during vegetable production site processing according to claim 1, characterized in that, The process of constructing a deep reinforcement learning environment based on the optimized parameter set, and iteratively optimizing the multi-dimensional optimization objectives in the vegetable production site processing based on the deep reinforcement learning environment, specifically includes: A discrete-continuous action space is defined by the adjustable range of the optimization parameter set, a state space is defined by the current state value of the optimization parameter set, and a reward function is defined by the negative value or reduction of carbon emissions output by the carbon emission calculation model, thus completing the construction of the deep reinforcement learning environment. The reward function is determined based on key features, with the optimization objectives being the minimization of carbon emissions, the maximization of quality preservation rate, and the minimization of storage period. The key features simultaneously satisfy the low-carbon requirements of cold chain logistics, the actual needs of growers, and the essential characteristics of vegetable production site processing. A deep deterministic policy gradient algorithm is used to perform interactive learning in the deep reinforcement learning environment, with the goal of maximizing cumulative reward, to generate action policies; The action strategy is input as a new parameter combination into the carbon emission optimization algorithm and iterated multiple times until the agent meets the convergence condition or reaches the preset number of iterations; the agent is the carrier that interacts with the deep reinforcement learning environment.
6. The method for optimizing carbon emissions during vegetable production site processing according to claim 1, characterized in that, The process of outputting the optimal combination of optimization parameters that yields the highest reward value and the corresponding carbon emission reduction optimization scheme based on the optimized parameter set specifically includes: After the iterative optimization process is completed, the parameter combination with the highest reward value is selected from all iterations as the optimal optimization parameter combination; Based on the optimal combination of parameters, a carbon emission reduction optimization scheme is generated; the carbon emission reduction optimization scheme includes specific parameter adjustment suggestions. Output the optimal combination of parameters and the carbon emission reduction optimization scheme.
7. An optimization device for carbon emissions in vegetable production site processing, characterized in that, include: The carbon emission calculation module is used to obtain parameters from multiple stages of the vegetable production and processing process, and to obtain the carbon emission amount based on the preset carbon emission calculation model. The sensitivity analysis module is used to perform sensitivity analysis on the multiple process parameters and determine the sensitivity analysis results; The parameter selection optimization module is used to select parameters with sensitivity higher than a set threshold and adjustable in actual operation as the optimization parameter set based on the sensitivity analysis results. The iterative optimization module is used to construct a deep reinforcement learning environment based on the optimization parameter set, and to iteratively optimize the multi-dimensional optimization objectives in the vegetable production area processing process based on the deep reinforcement learning environment. The multidimensional optimization objectives include carbon emissions, quality preservation rate, and storage period; The scheme output module is used to output the best combination of optimization parameters with the highest reward value and the corresponding carbon emission reduction optimization scheme based on the optimization parameter set; the reward value is the optimization target of the carbon emission optimization algorithm.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for optimizing carbon emissions in the vegetable production site processing process as described in any one of claims 1 to 6.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for optimizing carbon emissions in the vegetable production site processing process as described in any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for optimizing carbon emissions in the vegetable production site processing process as described in any one of claims 1 to 6.