Ecological ice production process self-optimization method and system oriented to dynamic environment parameters

Abstract

The application discloses an ecological ice preparation process self-optimization method and system oriented to dynamic environment parameters, and particularly relates to the technical field of intelligent control of industrial processes, preparation of environmental protection materials and optimization of industrial production management. The method collects environment, raw material, equipment operation and production management related data in real time through a multi-source sensor, inputs a digital twin model fusing physical mechanism and historical management data, and simulates output of production management index prediction results such as ice body microstructure, ecological function achievement degree, energy consumption, cost, efficiency and the like. An intelligent agent of reinforcement learning oriented to maximization of production management benefits generates optimal process parameters, and controls equipment to realize process and management target adaptation. After single batch preparation, the model and the intelligent agent are iteratively optimized in combination with measured data of finished products and production management data. The application solves the problems of poor adaptation of traditional processes to dynamic conditions and disconnection with production management, and improves product consistency, production management decision precision and comprehensive benefits.

Description

Technical Field

[0001] This invention relates to the fields of intelligent control of industrial processes, preparation of environmentally friendly materials, and optimization of industrial production management. More specifically, this invention relates to a self-optimization method and system for ecological ice preparation processes oriented towards dynamic environmental parameters. Background Technology

[0002] Ecological ice, as an environmental material with specific pore structure and function, is widely used in fields such as ecological restoration, water purification and temperature control. Its functional effectiveness is highly dependent on the microstructure of the ice body, which is in turn affected by the complex coupling of multiple parameters such as raw water quality, ambient temperature and humidity, and freezing process.

[0003] Currently, the preparation process of ecological ice is mostly based on fixed formulas and static process parameters, lacking the ability to adapt to dynamically changing environmental and raw material conditions. Although some studies have attempted to optimize the process, they are usually limited to offline analysis, single-objective optimization, or local adjustments based on simple rules. They have failed to achieve closed-loop control that is guided by the final ecological function and production management goals, perceives the environment in real time, and performs global self-optimization. Production management goals include energy consumption control, cost control, efficiency improvement, and resource utilization optimization, resulting in insufficient production scheduling flexibility, high control costs, waste of resources, low production efficiency, difficulty in ensuring product batch consistency, and a serious disconnect between process optimization and production management decisions.

[0004] Therefore, in view of the above situation, the present invention provides a self-optimization method and system for ecological ice preparation process oriented towards dynamic environmental parameters. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, this invention provides a self-optimization method and system for ecological ice preparation processes oriented towards dynamic environmental parameters, in order to solve the problems mentioned in the background art.

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

[0007] On the one hand, this invention provides a self-optimization method for ecological ice preparation process oriented towards dynamic environmental parameters, specifically including the following steps:

[0008] S1. Real-time collection of dynamic environmental parameters, raw material parameters, and production process operation data during the preparation of ecological ice, and collection of basic production management data, including raw material procurement cost data, equipment depreciation data, and energy consumption metering data, to support production management indicator calculation and process optimization decisions;

[0009] S2. Input the collected parameters into the pre-trained digital twin model. The digital twin model integrates the physical mechanism of the production process with historical management data. The historical management data includes historical energy consumption records, efficiency statistics, cost accounting data, and resource consumption data. It dynamically simulates the ice growth process and outputs the ice microstructure, ecological function target achievement degree, and production management indicator prediction results under the corresponding process conditions. The production management indicators include at least two of the following: energy consumption control targets, production efficiency targets, cost control targets, and resource utilization rate targets. The production management indicator prediction results include estimated energy consumption, estimated production cycle, estimated unit product cost, and estimated raw material utilization rate.

[0010] S3. Compare the prediction results with the preset ecological function targets and production management indicators, calculate the target deviation, and dynamically generate the optimal set of process parameters that match the current parameter conditions through the reinforcement learning agent based on the deviation. The reinforcement learning agent takes maximizing production management benefits as the core decision-making orientation, while taking into account the ecological function achievement rate and the management indicator achievement rate.

[0011] S4. Based on the optimal process parameter set, adjust the operating parameters of the ecological ice preparation equipment to achieve coordinated adaptation between process execution and production management goals, ensuring that process adjustments directly serve management needs such as energy consumption reduction, efficiency improvement, and cost saving.

[0012] S5. After a single batch of preparation is completed, obtain the actual microstructure data of the finished ice and the production process management data. The production process management data includes single batch energy consumption statistics, production time, raw material consumption cost, equipment operation loss data, product qualification rate, and actual resource utilization rate. Based on this data, the digital twin model and reinforcement learning agent are jointly iteratively optimized to continuously improve the accuracy of production management decisions, the accuracy of cost accounting, and the adaptability of process and management objectives.

[0013] Preferably, in step S1, the dynamic environmental parameters and raw material parameters include multiple parameters such as ambient temperature, ambient humidity, raw material water flow rate, raw material water pH value, raw material water turbidity, specific pollutant concentration, refrigerant temperature, stirrer speed, solution subcooling, raw material procurement cost related parameters, and raw material quality grade parameters.

[0014] Preferably, in step S2, the digital twin model is constructed by integrating a physical mechanism model and a data-driven model, wherein:

[0015] The physical mechanism model is constructed based on the thermodynamics of ice-water phase transition and the kinetics of solute migration, providing a basic simulation framework for the ice growth process.

[0016] The data-driven model is trained using machine learning based on historical process data, corresponding ice detection results, and historical production management data. Historical production management data includes energy consumption data, cost ledgers, efficiency records, and resource consumption statistics.

[0017] Specifically, the fusion is manifested in that the data-driven model is used to compensate for the simulation error of the physical mechanism model under complex and variable preparation boundary conditions, while optimizing the prediction accuracy of production management indicators and the accuracy of cost accounting.

[0018] Preferably, in step S3, the reinforcement learning agent uses the real-time status of the collected parameters and the progress of achieving the current production management indicators, including energy consumption compliance rate, efficiency compliance rate, and cost control progress, as input status. It uses adjustable process parameters as its action space and the comprehensive consistency between the prediction results output by the digital twin model and the ecological function goals and production management indicators as a reward signal to make online decisions. The adjustable process parameters include cooling rate curves, heat preservation stage parameters, stirring modes, and external intervention parameters. The production management indicators include at least one of energy consumption control targets, production efficiency targets, cost control targets, resource utilization targets, and equipment maintenance cost control targets.

[0019] Preferably, in step S2, the ecological function target includes at least one of the target porosity distribution, the target pollutant load, and the target cooling rate, and the prediction results of the ice microstructure and ecological function target include at least one of the ice porosity, average grain size, pollutant distribution uniformity, and theoretical cooling curve.

[0020] Preferably, in step S5, acquiring the actual microstructure data and production process management data of the finished ice specifically includes: scanning the prepared ice sample through an online microscopic image acquisition system, and obtaining the actual pore structure and ice crystal morphology data based on image analysis technology; acquiring production management data such as single-batch production energy consumption, production time, raw material consumption, raw material loss rate, equipment operation and maintenance costs, product qualification rate, and unit product cost through a production management data acquisition unit, forming a complete production management data ledger.

[0021] This invention also provides a self-optimization system for ecological ice preparation processes oriented towards dynamic environmental parameters, used to implement the above-mentioned method. The system includes:

[0022] The parameter sensing module is used to collect dynamic environmental parameters, raw material parameters, production equipment operation data and basic production management data in real time during the preparation of ecological ice. The basic production management data includes raw material procurement cost data, equipment depreciation data and energy consumption metering data, thus constructing a data collection system for the entire production process.

[0023] The digital twin and prediction module is used to carry the digital twin model and, based on the data from the parameter sensing module, outputs prediction results of the ice body microstructure, the achievement of ecological function indicators, and the achievement of production management indicators. The prediction results of the achievement of production management indicators include predicted values ​​of energy consumption, cost, efficiency, and resource utilization, providing data support for production decisions.

[0024] The intelligent decision-making and optimization module is used to run a reinforcement learning agent. Based on the difference between the prediction results and the preset ecological function goals and production management goals, it generates an optimal set of process parameters that takes into account both ecological function and production management benefits. Production management benefits include energy consumption reduction, cost saving and efficiency improvement.

[0025] The process execution control module is used to drive the preparation equipment to operate according to the optimal set of process parameters, so as to achieve precise matching between process execution and production management goals, and ensure that management goals are effectively achieved through process adjustments.

[0026] The online model evolution module is used to update and optimize the parameters of the models in the digital twin and prediction module and the intelligent decision-making and optimization module based on the actual microstructure data of the finished ice and the production process management data after a single batch is prepared. This continuously improves the production management decision-making capabilities, cost accounting accuracy, and the pertinence of process optimization.

[0027] Preferably, the parameter sensing module includes a multi-source sensor network deployed in the raw material pipeline, preparation reaction unit, perimeter environment, production equipment, and management data acquisition nodes. The management data acquisition nodes include energy consumption metering nodes, cost statistics nodes, and resource consumption monitoring nodes. The process execution control module includes a distributed controller connected to the refrigeration unit, stirring unit, and external intervention device. The controller executes control actions according to the optimal process parameter set to ensure that the production process strictly conforms to the management objectives such as energy consumption and cost.

[0028] Preferably, the digital twin and prediction module includes a model fusion unit, which is used to fuse a phase transition dynamics model based on physical mechanisms with a machine learning prediction model trained based on historical process data, ice detection results, and historical production management data, including energy consumption, efficiency, cost, and resource consumption data. The former provides the simulation basis, while the latter optimizes the prediction accuracy of production management indicators, the accuracy of cost accounting, and the adaptability to complex production scenarios.

[0029] Preferably, the system also includes a human-computer interaction and target management module, which allows users to set and adjust ecological function targets and production management indicators. The production management indicators include indicator weights, target thresholds, cost control ranges, and energy consumption limits. The system also displays the production management optimization process, cost accounting details, energy consumption analysis reports, efficiency improvement curves, simulation prediction results, system operating status, and production management data statistical reports in real time. The production management data statistical reports include comparative analysis of single-batch and multi-batch management data.

[0030] The technical effects and advantages of this invention are as follows:

[0031] 1. This invention captures real-time environmental and raw material dynamic changes, production equipment operation data, and basic production management data through a parameter sensing module. It utilizes a digital twin and prediction module for online simulation and combines reinforcement learning with an intelligent decision-making and optimization module to generate optimal process parameters that balance ecological functions and production management goals in real time. This enables the preparation process to automatically and accurately adapt to current production conditions, fundamentally overcoming the shortcomings of traditional fixed process parameters that cannot respond to environmental fluctuations, have low production control efficiency, high costs, insufficient resource utilization, and a disconnect between management decisions and process execution. It ensures that the microstructure and functional indicators of ecological ice remain stable and meet standards under different operating conditions, while optimizing production energy consumption, efficiency, and costs. This greatly improves the robustness of the production system, batch consistency of products, production management decision-making level, and cost accounting accuracy.

[0032] 2. This invention constructs an optimization paradigm guided by both ecological function goals and production management indicators. The digital twin model integrates physical mechanisms with a data-driven model containing historical management data, and establishes a high-precision dynamic input parameter, adjustable process parameter, terminal ice body function, and production management benefits. The production management benefits include the predictive relationship of energy consumption, cost, efficiency, and resource utilization. The reinforcement learning agent makes decisions based on this predictive relationship, so that the optimization process simultaneously serves application needs and production management benefits, solving the problems of traditional optimization methods having a single objective and being disconnected from production management.

[0033] 3. This invention constructs a complete closed loop of production-inspection-data feedback-model update through an online model evolution module. It utilizes real microscopic data of finished ice and production process management data, including energy consumption, cost, efficiency, and resource consumption data, to continuously iterate and optimize the digital twin model and reinforcement learning strategy. This allows the system's prediction and decision-making capabilities to evolve continuously with the accumulation of production data, effectively addressing challenges such as raw material characteristic drift and changes in production conditions. Long-term operation can significantly reduce reliance on expert experience, while continuously optimizing production management efficiency, achieving synergistic and continuous optimization of process and production management, and reducing control costs and resource waste. Attached Figure Description

[0034] Figure 1This is a flowchart illustrating the overall steps of the present invention.

[0035] Figure 2 This is a system diagram of the present invention.

[0036] The attached diagram is labeled as follows: 1. Parameter sensing module; 2. Digital twin and prediction module; 3. Intelligent decision-making and optimization module; 4. Process execution control module; 5. Online model evolution module; 6. Human-computer interaction and target management module; 7. Multi-source sensor network; 8. Distributed controller; 9. Model fusion unit. Detailed Implementation

[0037] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Example 1

[0038] This invention provides a self-optimization method and system for ecological ice preparation processes oriented towards dynamic environmental parameters;

[0039] In this embodiment, as shown in the appendix Figure 2 As shown, the system configuration of the present invention is as follows:

[0040] The parameter sensing module 1 includes a multi-source sensor network 7 deployed in the raw material inlet pipeline, the preparation reaction unit (such as a reactor or a freezing chamber), the perimeter environment, production equipment, and management data acquisition nodes. The management data acquisition nodes include a multi-source sensor network 7 on energy consumption meters, raw material cost statistics terminals, and resource consumption monitors. The multi-source sensor network 7 specifically includes temperature sensors, humidity sensors, flow meters, pH meters, conductivity meters, turbidity meters, specific pollutant concentration detectors, production operation status sensors and cost-related sensors, energy consumption metering sensors, and resource consumption sensors, etc., for real-time acquisition of production process operation data and basic production management data such as ambient temperature, ambient humidity, raw material water flow rate, pH value, conductivity, turbidity, pollutant concentration, refrigerant temperature, stirrer speed, solution subcooling degree, production equipment operating power, operating time, raw material procurement cost related data, single batch energy consumption data, and raw material consumption data.

[0041] Digital Twin and Prediction Module 2: The core of this module is a pre-trained digital twin model, which is built by integrating a physical mechanism model and a data-driven model. The physical mechanism model is based on the basic thermodynamic laws of ice-water phase transition (such as energy conservation) and the kinetic equations of solute (contaminant) migration at the freezing front, and is used to describe the macroscopic physical processes of ice growth.

[0042] The data-driven model is based on a large amount of historical process data (i.e., historical data collected by parameter sensing module 1), corresponding ice body detection results (such as porosity and grain size measurement values), and historical production management data. The historical production management data includes historical batch energy consumption records, production cycle statistics, cost accounting ledgers, and resource utilization reports. It is trained through machine learning algorithms (such as deep neural networks) to capture complex nonlinear relationships and correlation patterns related to production management that are difficult to describe precisely by the mechanism model.

[0043] The integration of the two is specifically manifested as follows: the physical mechanism model provides the basic framework and interpretability of the simulation, while the data-driven model is responsible for compensating for the simulation error of the physical mechanism model under complex and variable specific preparation boundary conditions (such as the presence of specific pollutants and non-uniform temperature fields), while optimizing the prediction accuracy of production management indicators and the accuracy of cost accounting. Production management indicators include energy consumption, cost, efficiency, and resource utilization.

[0044] After receiving real-time parameters, this module can dynamically simulate the predicted microstructure (such as porosity, average grain size, and uniformity of contaminant distribution), functional indicators (such as theoretical cooling curve), and production management indicators under the current conditions if the predetermined process is followed. The production management indicators include estimated energy consumption, estimated production time, estimated unit product cost, and estimated raw material utilization rate.

[0045] The intelligent decision-making and optimization module 3 has a built-in reinforcement learning agent. This agent takes the real-time data (including environmental and raw material parameters, production process operation data and current production management indicator achievement progress) collected by the parameter perception module 1 as its input state, and takes a series of adjustable process parameters (consisting of action space, such as cooling rate curve, temperature and duration of heat preservation stage, start and stop mode and speed of stirrer, whether to apply external field intervention such as ultrasound / electromagnetism and its intensity / frequency) as its executable actions.

[0046] The goal of its decision-making is to maximize a reward signal, which is calculated by the comprehensive fit between the prediction results output by the digital twin and prediction module 2 and the user's preset ecological function goals (e.g., target porosity distribution, target pollutant load, target cooling rate) and production management indicators, including energy consumption control targets, production efficiency targets, cost control targets, and resource utilization targets.

[0047] Through continuous interaction and trial and error with this digital twin environment (without actual production), the agent learns an optimal strategy and can instantly make an online decision on the current optimal set of process parameters based on any real-time state, ensuring that the adjustment of process parameters directly serves to maximize the benefits of production management.

[0048] The process execution control module 4 includes a distributed controller 8 that is connected to external intervention devices such as refrigeration units, stirring units, ultrasonic generators or electromagnetic generators. It receives the optimal set of process parameters from the intelligent decision-making and optimization module 3 and converts them into specific control commands to drive the precise actions of each actuator, thereby regulating the operation of the ecological ice preparation equipment and ensuring that the process execution is precisely matched with production management goals such as energy consumption, cost, and efficiency.

[0049] Model Online Evolution Module 5 is key to enabling the system's self-evolution. After each batch of ecological ice is prepared, the online microscopic image acquisition system (e.g., an automatic scanning device equipped with a high-definition microscope and a cold stage) will quickly scan the finished ice sample and obtain the actual pore structure and ice crystal morphology data of the ice body through image analysis technology. At the same time, the production management data acquisition unit summarizes production process management data such as single batch production energy consumption, production time, raw material consumption, raw material loss rate, equipment operation and maintenance costs, product qualification rate, unit product cost, and actual resource utilization rate.

[0050] The online model evolution module 5 compares these measured data, production process management data and the prediction results of the digital twin model in this round, calculates the prediction error, and uses this error data to fine-tune the model parameters in the digital twin and prediction module 2. At the same time, it updates the reinforcement learning strategy in the intelligent decision-making and optimization module 3, so that the entire system can continuously learn from production practice and continuously improve the accuracy of production management decisions and cost accounting.

[0051] The Human-Computer Interaction and Target Management Module 6 provides a graphical interface for process engineers to set, adjust, and switch different ecological function targets and production management indicators. Production management indicators include the weight of each indicator, the threshold for achieving the target, the cost control range, and the upper limit of energy consumption. It also monitors the entire production management optimization process in real time, allows viewing digital twin simulation animations, prediction reports, the operating status of each piece of equipment, and production management data statistical reports. The production management data statistical reports include energy consumption analysis, efficiency calculation, cost details, and comparison of data from multiple batches, realizing the combination of human management of advanced strategies and autonomous system optimization.

[0052] As attached Figure 1 As shown, this embodiment of the invention also provides a self-optimization method for ecological ice preparation process oriented towards dynamic environmental parameters. Based on the above system, it specifically includes the following steps:

[0053] S1. Utilize parameter sensing module 1 to collect dynamic environmental parameters, raw material parameters, production process operation data, and basic production management data in real time during the preparation of ecological ice.

[0054] S2. Input the collected parameters into the pre-trained digital twin model in the digital twin and prediction module 2. The digital twin model dynamically simulates the ice growth process according to the input parameters and outputs the prediction results of the ice microstructure, ecological function targets and production management indicators under the corresponding process conditions.

[0055] S3, the intelligent decision-making and optimization module 3, compares the prediction results with the preset ecological function goals and production management indicators, and based on the comparison results, dynamically generates the optimal set of process parameters that match the current parameter conditions through reinforcement learning agents.

[0056] S4, the process execution control module 4, adjusts the operating parameters of the ecological ice preparation equipment according to the optimal process parameter set to achieve coordinated adaptation between process execution and production management objectives;

[0057] S5. After a single batch of preparation is completed, the actual microstructure data of the finished ice is obtained through an online microscopic image acquisition system, and the production process management data is obtained through a production management data acquisition unit. Based on these two types of data, the digital twin model and the reinforcement learning agent are jointly iteratively optimized to continuously improve the accuracy of production management decisions. Example 2

[0058] This embodiment is basically the same as Embodiment 1, except that the reward function of the reinforcement learning agent is specifically designed to synergistically optimize ice quality and production management efficiency, achieving the dual optimization of ecological function and production management goals, and the reward signal... The calculation formula is:

[0059] ;

[0060] The specific meanings of each parameter are as follows:

[0061] : indicates the first The predicted values ​​of the ecological function indicators (output by the digital twin model, corresponding to the predicted results of specific indicators such as ice porosity, pollutant distribution uniformity, and theoretical cooling rate);

[0062] : indicates the first The preset target values ​​for each ecological function indicator (set by the user through human-computer interaction and the target management module, such as target porosity distribution, target pollutant load, target cooling rate, etc.).

[0063] : indicates the first The user-defined weight coefficients corresponding to each ecological function indicator have values ​​ranging from [0,1] and satisfy the following conditions: This is used to distinguish the importance and priority of different ecological function indicators in practical applications;

[0064] : Indicates the total number of ecological function indicators optimized ( (Flexible settings based on actual application scenarios)

[0065] : Represents the adjustment coefficient of the ecological function conformity optimization term, with a value range of [1,5], used to adjust the weight of ecological function optimization in the overall reward;

[0066] : Represents the estimated energy consumption corresponding to the current set of process parameters (calculated by simulation of production equipment operating parameters using a digital twin model, unit: kWh);

[0067] : Indicates the preset baseline energy consumption (based on the actual energy consumption statistics of the best historical process batches, or set by the user according to production requirements, unit: kWh);

[0068] : Represents the adjustment coefficient of the energy consumption optimization term, with a value range of [1,4], used to strengthen the guidance of energy consumption control;

[0069] : Represents the estimated unit product cost corresponding to the current set of process parameters (calculated by simulation using a digital twin model combined with data on raw material consumption, equipment wear and tear, energy consumption, etc., unit: yuan);

[0070] : Represents the preset benchmark unit product cost (derived from the actual cost statistics of the historical best process batch, or set by the user according to production requirements, unit: yuan);

[0071] : Represents the adjustment coefficient of the cost optimization item, with a value range of [1,4], used to strengthen the guidance of cost control;

[0072] : Represents the estimated resource utilization rate corresponding to the current set of process parameters (calculated by simulation using a digital twin model combined with data such as raw material consumption and product output, unit: %).

[0073] : Represents the preset baseline resource utilization rate (based on the actual utilization rate statistics of the historical best process batch, or set by the user according to production requirements, unit: %).

[0074] : Represents the adjustment coefficient of the resource utilization optimization term, with a value range of [1,3], used to strengthen the guidance of resource conservation.

[0075] The core logic of this reward function design is: by multi-dimensionally weighted and integrated ecological function compliance rate, energy consumption reduction rate, cost saving rate, and resource utilization improvement rate, the reinforcement learning agent is guided to simultaneously consider the functional quality of ice body and production management efficiency in the decision-making process, so as to achieve multi-objective collaborative optimization of "functional compliance, energy consumption reduction, cost saving, and resource efficiency", and completely solve the defects of traditional processes that only focus on a single objective and ignore production management efficiency.

[0076] pass , , and The value allocation can flexibly balance the optimization priorities in different scenarios: for example, in scenarios where the core requirement is functional effectiveness, such as ecological restoration, the following settings can be configured: =5、 =1、 =1、 =1, prioritizing ecological functions; in large-scale production scenarios, it can be set =1、 =4、 =4、 =3, prioritize optimizing production management efficiency.

[0077] In summary, the system of the present invention first senses the dynamic parameters of the environment and raw materials, the operating data of production equipment and the basic data of production management in real time, and inputs them into a digital twin that integrates physical mechanisms and data-driven models containing historical management data, so as to predict the microstructure, functional indicators and the degree of achievement of production management indicators of ice body in seconds through simulation.

[0078] Subsequently, the reinforcement learning agent compares the prediction results with the preset ecological function goals and production management indicators, and based on this difference, explores and decides online in the digital twin environment the optimal set of process parameters that best matches the current conditions, and sends it to the execution agency to control production, ensuring that the process adjustment and production management goals are accurately matched.

[0079] Finally, by using actual testing data of finished ice and production process management data, the digital twin model and reinforcement learning strategy are jointly iteratively optimized to form a complete self-evolving closed loop of perception-simulation-decision-execution-learning, thereby achieving synergistic optimization of process and production management.

[0080] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A self-optimization method for ecological ice preparation process oriented towards dynamic environmental parameters, characterized in that: Specifically, the following steps are included: S1. Real-time collection of dynamic environmental parameters, raw material parameters, and production process operation data during the preparation of ecological ice. The data is used to support the calculation of production management indicators and process optimization decisions. S2. The collected parameters are input into a pre-trained digital twin model. The digital twin model is constructed by fusing a physical mechanism model and a data-driven model. The physical mechanism model is based on the thermodynamics of ice-water phase transition and the kinetics of solute migration. The data-driven model is trained by machine learning based on historical process data, corresponding ice body detection results, and historical production management data. The fusion is manifested in the data-driven model compensating for the simulation error of the physical mechanism model under complex and variable preparation boundary conditions, and optimizing the prediction accuracy of production management indicators and the accuracy of cost accounting. The digital twin model dynamically simulates the ice body growth process and outputs the prediction results of ice body microstructure, ecological function target achievement, and production management indicators under corresponding process conditions. The production management indicators include at least two of the following: energy consumption control targets, production efficiency targets, cost control targets, and resource utilization rate targets. S3. Compare the prediction results with the preset ecological function goals and production management indicators, calculate the target deviation, and dynamically generate the optimal set of process parameters that match the current parameter conditions through the reinforcement learning agent based on the deviation. The input state of the reinforcement learning agent is the real-time status of the collected parameters and the progress of the current production management indicators. The action space is the adjustable process parameters. The reward signal is the comprehensive consistency between the prediction results output by the digital twin model and the ecological function goals and production management indicators. The reinforcement learning agent takes maximizing production management benefits as the core decision-making orientation. S4. Based on the optimal set of process parameters, adjust the operating parameters of the ecological ice preparation equipment to achieve coordinated adaptation between process execution and production management objectives; S5. After a single batch of preparation is completed, the actual microstructure data and production process management data of the finished ice are obtained. The actual microstructure data is obtained by scanning the prepared ice sample through an online microscopic image acquisition system and obtaining the actual pore structure and ice crystal morphology data based on image analysis technology. Based on this data, the digital twin model and reinforcement learning agent are jointly iteratively optimized to continuously improve the accuracy of production management decisions and process adaptability.

2. The self-optimization method for ecological ice preparation process oriented towards dynamic environmental parameters according to claim 1, characterized in that: In step S1, the dynamic environmental parameters and raw material parameters include multiple parameters such as ambient temperature, ambient humidity, raw material water flow rate, raw material water pH value, raw material water conductivity, raw material water turbidity, specific pollutant concentration, refrigerant temperature, stirrer speed, solution subcooling, raw material procurement cost related parameters, and raw material quality grade parameters.

3. The self-optimization method for ecological ice preparation process oriented towards dynamic environmental parameters according to claim 1, characterized in that: In step S2, the ecological function target includes at least one of target porosity distribution, target pollutant load, and target cooling rate. The prediction results of the ice microstructure and ecological function target include at least one of ice porosity, average grain size, pollutant distribution uniformity, and theoretical cooling curve. The prediction results of the production management indicators include at least two of the following: estimated energy consumption, estimated production cycle, estimated unit product cost, estimated raw material utilization rate, and estimated equipment wear cost.

4. The self-optimization method for ecological ice preparation process oriented towards dynamic environmental parameters according to claim 1, characterized in that: In step S5, the actual microstructure data and production process management data of the finished ice are obtained. This also includes obtaining production management data such as energy consumption, production time, raw material consumption, raw material loss rate, equipment operation and maintenance costs, product qualification rate and unit product cost through the production management data acquisition unit.

5. A self-optimizing system for ecological ice preparation process oriented towards dynamic environmental parameters, used to implement the method described in claim 4, characterized in that: The system includes: The parameter sensing module (1) is used to collect dynamic environmental parameters, raw material parameters, production equipment operation data and production management basic data including raw material procurement cost data, equipment depreciation data and energy consumption metering data in real time during the preparation of ecological ice, and to build a data collection system for the entire production process. The digital twin and prediction module (2) is used to carry the digital twin model and output the prediction results of the achievement of ice microstructure, ecological function indicators and production management indicators including energy consumption, cost, efficiency and resource utilization rate based on the data of the parameter perception module (1), so as to provide data support for production decision-making. The intelligent decision-making and optimization module (3) is used to run a reinforcement learning agent and generate an optimal set of process parameters that takes into account both ecological function and production management benefits, including energy consumption reduction, cost saving and efficiency improvement, based on the difference between the prediction results and the preset ecological function goals and production management goals. The process execution control module (4) is used to drive the preparation equipment to operate according to the optimal set of process parameters, so as to achieve precise matching between process execution and production management objectives. The online model evolution module (5) is used to update and optimize the parameters of the models in the digital twin and prediction module (2) and the intelligent decision-making and optimization module (3) based on the actual microstructure data of the finished ice and the production process management data after a single batch of preparation is completed, so as to continuously improve the production management decision-making ability, cost accounting accuracy and process optimization targeting.

6. The system according to claim 5, characterized in that: The parameter sensing module (1) includes a multi-source sensor network (7) deployed on the raw material pipeline, preparation reaction unit, perimeter environment, production equipment and management data acquisition nodes such as energy consumption metering nodes, cost statistics nodes and resource consumption monitoring nodes. The process execution control module (4) includes a distributed controller (8) connected to the refrigeration unit, stirring unit and external intervention device. The distributed controller (8) executes control actions according to the optimal process parameter set to ensure that the production process meets the management objectives of energy consumption and cost.

7. The system according to claim 5, characterized in that: The digital twin and prediction module (2) includes a model fusion unit (9), which is used to fuse a phase change dynamics model based on physical mechanisms with a machine learning prediction model trained on historical process data, ice detection results and production management historical data including energy consumption, efficiency, cost and resource consumption data. The former provides the simulation basis, while the latter optimizes the prediction accuracy of production management indicators, the accuracy of cost accounting and the adaptability to complex scenarios.

8. The system according to claim 5, characterized in that: The system also includes a human-computer interaction and target management module (6), which allows users to set and adjust ecological function targets and production management indicators including indicator weights, target thresholds, cost control ranges, and energy consumption upper limits. It also displays in real time the production management optimization process, cost accounting details, energy consumption analysis reports, simulation prediction results, system operating status, and production management data statistical reports including single-batch / multi-batch management data comparison analysis.

Images