Heat treatment intelligent production scheduling planning method and system

Abstract

The invention relates to the technical field of production scheduling, and particularly discloses a heat treatment intelligent production scheduling planning method and system, and the method comprises the steps: obtaining to-be-scheduled order data, equipment capability data and energy price data; constructing a multi-objective optimization model based on the acquired data, wherein the objective function of the model aims at minimizing the comprehensive cost of production scheduling; solving the multi-objective optimization model under the condition of meeting process constraint conditions, and generating a production scheduling scheme; executing production based on the production scheduling scheme, and triggering a dynamic rescheduling process when production disturbance occurs; according to the method, a specific optimization mechanism is introduced into a hybrid optimization strategy, a modeling method based on real historical data and a nonlinear mathematical form is constructed, deviation caused by a traditional linear model or empirical estimation is avoided, and it is ensured that in heat treatment intelligent production scheduling planning, the efficiency is greatly improved. The comprehensive optimal production scheduling scheme considering the production efficiency, the delivery punctuality and the energy cost can be quickly and accurately generated.

Description

Technical Field

[0001] This invention relates to the field of production scheduling technology, specifically to a method and system for intelligent production scheduling in heat treatment. Background Technology

[0002] As a core component of the manufacturing industry, heat treatment technology is widely used in key fields such as machinery manufacturing, automotive industry, and aerospace. Its production process relies on high-energy-consuming equipment such as high-temperature industrial furnaces for long-term continuous operation. The process parameters are strictly required, the order types are diverse, and the production environment is dynamic and ever-changing. The core challenge for enterprises is how to scientifically formulate production plans to ensure on-time product delivery and meet complex process constraints, while effectively balancing multiple objectives such as production efficiency, order fulfillment rate, and energy cost control.

[0003] Traditional production scheduling techniques are mainly based on manual experience or simple rule-driven scheduling strategies. These methods are often limited to a single optimization dimension, such as pursuing only the shortest completion time, and cannot achieve coordinated optimization of multiple objectives such as production time, order delay penalties, and energy expenditure. In addition, the production site frequently encounters disturbances such as order changes, sudden equipment failures, or process parameter adjustments. Traditional static production scheduling schemes lack effective real-time response mechanisms. When disturbances occur, enterprises usually rely on manual intervention to make adjustments. This process is inefficient and cannot guarantee the overall optimality of the adjusted scheme. It is very easy to cause production plan interruptions or resource waste, which seriously affects production continuity and economic benefits.

[0004] Therefore, there is an urgent need to develop a production scheduling technology framework that can deeply integrate multi-source production data, accurately model the nonlinear energy consumption characteristics of equipment, comprehensively optimize multi-dimensional objectives, and possess intelligent dynamic response capabilities. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent scheduling of heat treatment production to solve the problems mentioned in the background art.

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

[0007] A method for intelligent scheduling and planning of heat treatment processes, the method comprising:

[0008] Obtain data on pending production orders, equipment capacity, and energy prices;

[0009] A multi-objective optimization model is constructed based on pending production order data, equipment capacity data, and energy price data. The objective function of the model aims to minimize the comprehensive cost of production scheduling, which is defined as the weighted sum of the first, second, and third terms. The first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-hour energy consumption cost, which is calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information.

[0010] The multi-objective optimization model is solved under the condition of satisfying process constraints to generate a production scheduling plan;

[0011] Production is executed based on the production scheduling plan, and a dynamic rescheduling process is triggered when production disturbances occur.

[0012] As a further aspect of the present invention: the calculation process for the total peak-period energy consumption cost includes:

[0013] The average power is calculated based on the processing load of each furnace and the nonlinear power function of the corresponding equipment.

[0014] The actual processing time for each furnace batch during the predicted peak electricity price period is determined based on the production scheduling plan;

[0015] The average power, the predicted electricity price for the corresponding time period, and the actual processing time are multiplied and summed to obtain the total peak-period energy consumption cost.

[0016] The fitting and calibration process of the nonlinear power function is dynamic: the system collects load energy consumption data in actual production in real time based on a preset frequency, and periodically updates the parameters of the nonlinear power function to keep the energy consumption model synchronized with the actual equipment status.

[0017] As a further aspect of the present invention: the nonlinear energy consumption model is a quadratic or higher-order function obtained by fitting historical energy consumption data, used to characterize the nonlinear relationship between equipment power and processing load.

[0018] As a further aspect of the present invention: the objective function adopts a normalized weighted sum form, which is expressed as the normalized weighted sum of the first term, the maximum completion time, the second term, the total delay penalty, and the third term, the total peak-period energy consumption cost.

[0019] As a further aspect of the present invention: the solution process of the multi-objective optimization model adopts a hybrid optimization strategy that integrates heuristic rules and metaheuristic algorithms, specifically including:

[0020] Generate an initial solution that satisfies the key constraints;

[0021] A multi-objective genetic algorithm is used for global optimization;

[0022] The solution is optimized based on a local search algorithm;

[0023] The multi-objective genetic algorithm employs an elite retention strategy and introduces domain-knowledge-based crossover and mutation operators during the evolution process. The local search algorithm is a simulated annealing algorithm, used to perform a more accurate search on the non-dominated solution set obtained by the genetic algorithm.

[0024] As a further aspect of the present invention: the dynamic rearrangement process includes:

[0025] A tiered strategy is adopted based on the type and scope of the disturbance:

[0026] For local disturbances, the constraint propagation algorithm is applied for repair.

[0027] For global disturbances, the construction and solution process of the multi-objective optimization model is executed cyclically within a compressed time window.

[0028] The present invention also provides a heat treatment intelligent scheduling planning system, the system comprising:

[0029] The data acquisition module is used to acquire data on pending production orders, equipment capacity, and energy prices.

[0030] The cost optimization module is used to construct a multi-objective optimization model based on pending production order data, equipment capacity data, and energy price data. The objective function of the model aims to minimize the comprehensive cost of production scheduling, which is defined as the weighted sum of the first, second, and third terms. The first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-period energy consumption cost, which is calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information.

[0031] The model solving module is used to solve the multi-objective optimization model under the condition of satisfying process constraints and generate a production scheduling plan.

[0032] The dynamic rescheduling module is used to execute production based on the production scheduling plan and to trigger the dynamic rescheduling process when production disturbances occur.

[0033] As a further aspect of the present invention: the calculation process for the total peak-period energy consumption cost includes:

[0034] The average power is calculated based on the processing load of each furnace and the nonlinear power function of the corresponding equipment.

[0035] The actual processing time for each furnace batch during the predicted peak electricity price period is determined based on the production scheduling plan;

[0036] The average power, the predicted electricity price for the corresponding time period, and the actual processing time are multiplied and summed to obtain the total peak-period energy consumption cost.

[0037] The fitting and calibration process of the nonlinear power function is dynamic: the system collects load energy consumption data in actual production in real time based on a preset frequency, and periodically updates the parameters of the nonlinear power function to keep the energy consumption model synchronized with the actual equipment status.

[0038] As a further aspect of the present invention: the nonlinear energy consumption model is a quadratic or higher-order function obtained by fitting historical energy consumption data, used to characterize the nonlinear relationship between equipment power and processing load.

[0039] As a further aspect of the present invention: the objective function adopts a normalized weighted sum form, which is expressed as the normalized weighted sum of the first term, the maximum completion time, the second term, the total delay penalty, and the third term, the total peak-period energy consumption cost.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention introduces a specific optimization mechanism into the hybrid optimization strategy and constructs a modeling method based on real historical data and nonlinear mathematical forms, which avoids the deviations caused by traditional linear models or empirical estimations. This enables more accurate prediction of the actual energy consumption of equipment under different processing loads when calculating the total peak energy consumption cost, and ensures that in the intelligent scheduling planning of heat treatment, a comprehensive optimal scheduling scheme that takes into account production efficiency, on-time delivery and energy cost can be generated quickly and accurately. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0042] Figure 1 This is a flowchart of a smart scheduling method for heat treatment.

[0043] Figure 2 This is a block diagram of the composition structure of a heat treatment intelligent scheduling system. Detailed Implementation

[0044] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0045] Figure 1 This is a flowchart of a heat treatment intelligent scheduling planning method. In this embodiment of the invention, a heat treatment intelligent scheduling planning method includes:

[0046] Step S100: Obtain pending production order data, equipment capacity data, and energy price data;

[0047] Step S200: Construct a multi-objective optimization model based on pending production order data, equipment capacity data, and energy price data. The objective function of the model aims to minimize the comprehensive cost of production scheduling, which is defined as the weighted sum of the first, second, and third terms. Wherein, the first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-period energy consumption cost, the value of which is calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information.

[0048] Step S300: Solve the multi-objective optimization model under the condition of satisfying process constraints to generate a production scheduling plan;

[0049] Step S400: Execute production based on the production scheduling plan, and trigger a dynamic rescheduling process when production disturbances occur.

[0050] In one embodiment of the technical solution of this invention, during the data acquisition phase, pending production order data, equipment capacity data, and energy price data are acquired. For example, pending production order data can be manually entered into the system or exported from an Enterprise Resource Planning (ERP) system via a data interface; equipment capacity data can be manually configured by technicians, such as equipment model, maximum load, and available time; energy price data can be obtained from the official website of the power supplier or from historical bills. Alternatively, the system can be configured to automatically synchronize pending production order data periodically from the Production Management System (MES) via a preset data interface. Equipment capacity data can be stored in an equipment management database and directly read by the system. Energy price data can be subscribed from an external data service provider or predicted through historical data analysis.

[0051] Secondly, a multi-objective optimization model is constructed based on the acquired data. The objective function of this model aims to minimize the overall cost of production scheduling, which is defined as the weighted sum of the maximum completion time, the total delay penalty, and the total peak-period energy consumption cost. Specifically, when constructing the multi-objective optimization model, a linear weighted method can be used to combine the three objectives of maximum completion time, total delay penalty, and total peak-period energy consumption cost. For example, a fixed weight coefficient is set for each objective, and then the values ​​of each objective function are multiplied by their corresponding weights and summed to form a single overall cost objective function. For the calculation of the total peak-period energy consumption cost, an energy consumption lookup table can be pre-established, which records the average power consumption of equipment under different processing loads. During the production scheduling process, based on the processing load of each batch in the production plan, the corresponding average power is looked up from the lookup table, and combined with the predicted time-of-use electricity price and processing time, the peak-period energy consumption cost is estimated. The nonlinear energy consumption model can be a simple piecewise function used to approximate the relationship between equipment power and load.

[0052] Next, a multi-objective optimization model is solved under the condition of satisfying process constraints to generate a production scheduling plan. When solving the multi-objective optimization model, traditional mathematical programming methods, such as the branch and bound method or the simplex method, can be used to find the optimal solution while satisfying all process constraints (such as furnace temperature profile, holding time, cooling rate, etc.). For smaller-scale production scheduling problems, a production scheduling plan can also be generated by enumerating partially feasible solutions and evaluating them. As another solution approach, some basic heuristic rules, such as "earliest delivery date first" or "shortest processing time first," can be used, combined with some simple local search strategies, to gradually build and improve the production scheduling plan until all process constraints are satisfied and an acceptable overall cost level is reached.

[0053] Finally, production is executed based on the generated production schedule, and dynamic rescheduling is triggered when production disturbances occur. During production execution, adjustments to the production schedule can be manually triggered when production disturbances occur, such as equipment failure, emergency order insertion, or raw material shortages. For example, operators can judge the scope of the disturbance based on experience, then manually modify the affected processes or furnaces, and rearrange subsequent production tasks. As another implementation method, the system can set a threshold, and automatically issue an alarm when the deviation between the actual production progress and the production schedule exceeds the threshold. At this time, a simplified rescheduling process can be initiated, for example, adjusting only the local area affected by the disturbance, or re-entering all unfinished tasks into the optimization model for a complete re-optimization.

[0054] The above solution integrates multi-source data, constructs a multi-objective optimization model, and introduces nonlinear energy consumption calculation. This enables heat treatment scheduling to accurately control peak-hour energy consumption costs while taking into account the maximum completion time and total delay penalty. In addition, a dynamic rescheduling mechanism is introduced, which can effectively cope with production disturbances and avoid the shortcomings of traditional methods in multi-objective balance, energy consumption estimation, and dynamic response, thereby improving the overall efficiency and adaptability of scheduling.

[0055] As a preferred embodiment of the technical solution of the present invention, the calculation process of the total peak-hour energy consumption cost includes:

[0056] The average power is calculated based on the processing load of each furnace and the nonlinear power function of the corresponding equipment.

[0057] The actual processing time for each furnace batch during the predicted peak electricity price period is determined based on the production scheduling plan;

[0058] The average power, the predicted electricity price for the corresponding time period, and the actual processing time are multiplied and summed to obtain the total peak-period energy consumption cost.

[0059] The fitting and calibration process of the nonlinear power function is dynamic: the system collects load energy consumption data in actual production in real time based on a preset frequency, and periodically updates the parameters of the nonlinear power function to keep the energy consumption model synchronized with the actual equipment status.

[0060] The phrase "based on the processing load of each furnace" refers to the comprehensive reflection of the amount of material, type of material, process requirements, and other production parameters affecting energy consumption in each furnace (i.e., one heat treatment task) during the heat treatment production process. Processing load is a key factor affecting equipment energy consumption, and its quantification methods can include, but are not limited to: precise measurement through real-time monitoring of data such as the weight, volume, and temperature change curves of materials inside the furnace using sensors; or estimation based on information such as the workpiece model, quantity, and processing cycle preset in the production plan, combined with historical production data.

[0061] The phrase "calculating the average power using the corresponding equipment's nonlinear power function" refers to employing a mathematical model to describe the nonlinear relationship between the power output of the heat treatment equipment and the processing load, thereby calculating the average power under a specific processing load. This nonlinear power function can more accurately reflect the differences in energy efficiency of the equipment under different load conditions, avoiding errors caused by traditional linear models or fixed power assumptions. For example, this function can be a quadratic or higher-order polynomial function obtained by regression analysis fitting of historical power data of the equipment under different loads; or it can be a complex nonlinear model constructed based on physical mechanism models or machine learning methods (such as neural networks) to more precisely capture the energy consumption characteristics of the equipment.

[0062] The phrase "determining the actual processing time of each heat treatment furnace during the predicted peak electricity price period based on the production scheduling plan" refers to accurately identifying and quantifying the actual operating time of each heat treatment furnace during the predicted peak electricity price period, based on the production scheduling plan generated in step S3 above. The production scheduling plan details the start and end times of each furnace. By overlaying these time periods with the high electricity price time intervals published by the power supplier or predicted through historical data analysis on the time axis, the duration of the overlapping portion can be accurately calculated, which is the actual processing time.

[0063] The phrase "multiplying and summing the average power, the predicted electricity price for the corresponding time period, and the actual processing time to obtain the total peak-period energy consumption cost" refers to multiplying the average power of each furnace batch calculated above by the actual processing time of that furnace batch during the peak period of the predicted electricity price, and then multiplying by the predicted electricity price corresponding to that peak period to obtain the energy consumption cost of that furnace batch during that peak period. Subsequently, the energy consumption costs generated by all furnace batches during all peak periods of the predicted electricity price are summed to finally obtain the total peak-period energy consumption cost for the entire production scheduling cycle.

[0064] Through the above technical solution, this invention can accurately calculate the total peak-period energy consumption cost in the third term of the multi-objective optimization model, solving the problem of inaccurate energy consumption estimation in traditional methods. Specifically, by considering the processing load of each furnace and the nonlinear power function of the equipment, it can accurately reflect the actual power consumption of the equipment under different loads, avoiding errors caused by fixed energy consumption coefficients or linear models. Simultaneously, by combining the production scheduling plan and the predicted peak electricity price periods, it can accurately determine the actual operating time of each furnace during high electricity price periods, ensuring that energy consumption cost calculation focuses on the high electricity price periods that have the greatest impact on total cost. Finally, by multiplying and summing the average power, predicted electricity price, and actual processing time, highly accurate and reliable energy consumption cost data is provided for the aforementioned multi-objective optimization model. This precise cost quantification allows the optimization model to more effectively balance production efficiency, delivery time, and energy costs, thereby generating better production scheduling plans, guiding production to proactively avoid or reduce energy consumption during high electricity price periods, achieving significant energy-saving and cost-reduction effects, and improving the overall economic and environmental benefits of heat treatment production.

[0065] The fitting and calibration process of the nonlinear power function is dynamic: the system continuously collects load-energy consumption data from actual production and periodically updates the parameters of the nonlinear power function to keep the energy consumption model synchronized with the actual equipment status.

[0066] It should be noted that the fitting and calibration process of the nonlinear power function in this invention is dynamic. This ensures that the energy consumption model can reflect the actual operating status of the equipment in real-time or near real-time, rather than relying on fixed historical data. This dynamism means that the process can adaptively adjust according to changes in factors such as the equipment's operating environment, wear and tear, and maintenance conditions. One implementation is an event-driven system design, where the fitting and calibration process is triggered when a significant change in equipment performance is detected (e.g., the deviation between energy consumption and model predictions exceeds a preset threshold). Another implementation is to integrate the fitting and calibration process into the equipment's routine maintenance or production cycle, for example, automatically executing it after each major overhaul or after the completion of a specific batch of production.

[0067] Meanwhile, the system continuously collects load-energy consumption data from actual production. This technical feature is the foundation for dynamic fitting and calibration. By continuously acquiring equipment operating data under actual production conditions, it provides accurate and reliable input for updating the energy consumption model. Load data can include parameters affecting equipment energy consumption such as furnace temperature, processing time, material type, and batch size; energy consumption data directly reflects the actual power consumption of the equipment. One implementation method is to install various sensors (such as temperature sensors, current sensors, voltage sensors, flow meters, etc.) on the heat treatment equipment, combined with a data acquisition unit and industrial control system, to acquire and store this data in real time. Another implementation method is to utilize Internet of Things (IoT) technology to connect the equipment to a cloud platform, and use edge computing devices to preprocess and transmit the raw data, achieving remote and continuous data acquisition.

[0068] Based on this, the parameters of the nonlinear power function are periodically updated. This technical feature is a key operation for achieving the dynamic nature of the energy consumption model. By periodically recalculating or adjusting the coefficients of the nonlinear power function, the model can adapt to long-term drift or short-term fluctuations in equipment performance. This periodic update balances model accuracy and computational resource consumption. One implementation is to set a fixed time interval (e.g., daily, weekly, or monthly) for parameter updates. At the end of each update cycle, the system refits the nonlinear power function using the latest collected load-energy consumption data, for example, using statistical methods such as least squares or regression analysis. Another implementation is to adopt a performance-based update strategy, such as triggering parameter updates when the model's prediction error exceeds a certain threshold for several consecutive cycles, or when the cumulative operating time of the equipment reaches a preset value. The update process can employ adaptive filtering algorithms or online learning algorithms to gradually adjust the model parameters.

[0069] Ultimately, through the aforementioned dynamic fitting and calibration process, the energy consumption model can be synchronized with the actual equipment status. This synchronization means the model can capture the impact of factors such as equipment aging, maintenance, process adjustments, and changes in ambient temperature on energy consumption characteristics. This allows the production scheduling system to base its energy consumption predictions and optimizations on equipment performance data that most closely reflects reality.

[0070] As a preferred embodiment of the technical solution of the present invention, the nonlinear energy consumption model is a quadratic or higher-order function obtained by fitting historical energy consumption data, which is used to characterize the nonlinear relationship between equipment power and processing load.

[0071] In one embodiment of the technical solution of this invention, the nonlinear power function is a mathematical model used to describe the nonlinear relationship between the power consumption of heat treatment equipment and its processing load. This function can capture the complex energy consumption characteristics of the equipment under different load conditions. For example, when the processing load increases, power consumption may show an accelerating or decelerating growth trend, rather than a simple linear proportional relationship. Its role is to provide a more realistic energy consumption prediction tool, laying the foundation for subsequent energy cost calculations. For example, the function can be a polynomial function, such as a quadratic function, a cubic function, or more complex exponential or logarithmic functions, to adapt to the specific energy consumption curves of different equipment.

[0072] The historical energy consumption data refers to the set of data continuously collected and recorded by the system during the actual operation of the heat treatment equipment, including information about the equipment's processing load (e.g., the type, quantity, weight, and temperature curve of materials inside the furnace) and its corresponding actual electrical energy consumption (e.g., instantaneous power and cumulative electricity consumption measured by meters or sensors). This data reflects the equipment's energy consumption behavior in a real production environment and serves as the empirical basis for building an accurate energy consumption model. For example, this data can originate from the equipment's SCADA system, energy management system, or dedicated data acquisition module, and typically includes timestamps, load parameters, and energy consumption readings.

[0073] The fitting process refers to the process of finding a mathematical function (in this case, a nonlinear power function) that best approximates or describes a set of discrete historical energy consumption data points using statistical or mathematical methods. This process aims to determine the specific parameters of the function so that its curve passes through or approximates all data points as closely as possible, thus enabling the function to represent the average energy consumption behavior of the equipment under different processing loads. Common fitting methods include least squares, maximum likelihood estimation, or various regression analysis techniques, using optimization algorithms to minimize the error between model predictions and actual observations.

[0074] The quadratic or higher-order function is a type of polynomial function, where the quadratic function is of the form: ( For power, (where a, b, and c are coefficients), higher-order functions contain load terms of higher powers. Quadratic or higher-order functions are chosen as the specific form of nonlinear power functions because they possess sufficient mathematical flexibility to effectively capture the complex nonlinear characteristics of power consumption in heat treatment equipment as a function of processing load, such as curve bending and inflection points. For example, when energy consumption changes gradually under low load but rises sharply under high load, quadratic or higher-order functions can better simulate this trend and provide a more accurate description compared to linear functions.

[0075] The nonlinear relationship characterizing the power of the equipment as a function of the processing load refers to the fact that the nonlinear power function obtained through the above fitting can accurately and quantitatively reflect the nonlinear variation law of the instantaneous power consumption or energy consumption per unit time of the heat treatment equipment under different processing load conditions. This nonlinear relationship may originate from the characteristics of the heating elements of the equipment, the variation of heat loss with temperature and load, and the differences in energy consumption at different process stages. For example, when the load in the furnace increases from no load to half load, the power may rise rapidly, while when it increases from half load to full load, the rate of increase in power may slow down. This complex behavior can be accurately characterized by a nonlinear function.

[0076] As a preferred embodiment of the technical solution of the present invention, the objective function adopts a normalized weighted sum form, which is expressed as the weighted sum of the first term maximum completion time, the second term total delay penalty, and the third term total peak period energy consumption cost after normalization.

[0077] In one example of the technical solution of this invention, normalization refers to transforming data with different dimensions or numerical ranges to a unified scale to eliminate the differences between them and ensure that each objective function can participate in the calculation fairly in multi-objective optimization. Specifically, normalization can be implemented in various ways. For example, the min-max normalization method can be used to linearly transform the original data to the range [0,1]. The calculation formula is: Normalized value = (Original value - Minimum value) / (Maximum value - Minimum value). In this way, regardless of the dimensions and numerical range of the original data, its normalized value will fall between 0 and 1.

[0078] By employing the aforementioned technical solution, when constructing a multi-objective optimization model, the first objective—maximum completion time—the second—total delay penalty—and the third—total peak-period energy consumption cost—are normalized. This effectively resolves the issue of uneven weight distribution caused by differences in dimensions and numerical ranges among these three objectives. Specifically, before weighted summation of each objective, they are normalized to ensure they are on a comparable, uniform scale. This ensures that when minimizing the overall cost of production scheduling, each objective contributes fairly to the total cost according to its assigned weight, avoiding situations where some objectives are overemphasized or ignored due to differences in numerical values. Therefore, this invention generates a more balanced, comprehensive, and practically guiding production scheduling scheme. This scheme not only effectively controls the maximum completion time and total delay penalty but also precisely manages peak-period energy consumption costs, thereby achieving a better overall balance between production efficiency, order on-time performance, and energy cost control.

[0079] As a preferred embodiment of the technical solution of the present invention, the solution process of the multi-objective optimization model adopts a hybrid optimization strategy that integrates heuristic rules and metaheuristic algorithms, specifically including:

[0080] Generate an initial solution that satisfies the key constraints;

[0081] A multi-objective genetic algorithm is used for global optimization;

[0082] The solution is optimized based on a local search algorithm;

[0083] The multi-objective genetic algorithm employs an elite retention strategy and introduces domain-knowledge-based crossover and mutation operators during the evolution process. The local search algorithm is a simulated annealing algorithm, used to perform a more accurate search on the non-dominated solution set obtained by the genetic algorithm.

[0084] In one embodiment of the technical solution of this invention, the hybrid optimization strategy combining heuristic rules and metaheuristic algorithms refers to a solution method that combines two or more different types of optimization algorithms to leverage their respective advantages and overcome the limitations of a single algorithm. Heuristic rules are typically based on domain knowledge and experience, enabling rapid generation of feasible solutions; metaheuristic algorithms, on the other hand, possess strong global search capabilities, allowing them to escape local optima. This strategy aims to combine the ability to rapidly generate feasible solutions with powerful global search capabilities, thereby improving the efficiency and quality of solutions in complex optimization problems. For example, a sequential hybrid approach can be adopted, where heuristic rules are first used to generate an initial solution, followed by optimization using a metaheuristic algorithm.

[0085] In generating initial solutions that satisfy key constraints, the role is to provide a starting point that already meets the core limitations of the problem before the optimization algorithm begins its iterations. This avoids the algorithm starting its search from completely random or infeasible states, thereby significantly reducing invalid computations, accelerating the convergence process, and improving the quality of the final solution. Specifically, an initial production scheduling plan can be constructed through pre-defined priority rules, greedy algorithms, or scheduling logic based on expert experience to ensure that key constraints such as equipment capacity, process sequence, and delivery time are met.

[0086] When using a multi-objective genetic algorithm for global optimization, this algorithm is a metaheuristic based on natural selection and genetic mechanisms, specifically designed to solve optimization problems with multiple conflicting objective functions. It searches for a set of non-dominated solutions in the solution space by simulating selection, crossover, and mutation operations in biological evolution. The algorithm's global search capability allows it to explore a broad solution space, effectively avoiding getting trapped in local optima, and handling trade-offs between multiple objectives, thereby finding a diverse and high-quality Pareto optimal solution set. For example, classic multi-objective genetic algorithm frameworks such as NSGA-II or SPEA2 can be used, maintaining population diversity through non-dominated sorting and crowding distance calculation.

[0087] When using a local search algorithm to fine-tune a solution, this algorithm is an iterative improvement algorithm that starts from the current solution and searches for a better solution in its neighborhood. If a better solution is found, the algorithm moves to that solution and repeats the process until no better solution can be found or a stopping condition is met. This algorithm excels at deep exploration in the vicinity of the current solution, enabling fine-tuning and optimization of solutions obtained by global optimization algorithms, thereby improving the accuracy and local optima of the solution and compensating for the shortcomings of global optimization algorithms in detailed optimization. For example, simulated annealing can be used, which introduces a probability of accepting a worse solution to escape local optima, and gradually reduces the acceptance probability in the later stages of the search to converge to a high-quality solution.

[0088] In a preferred embodiment of the present invention, the multi-objective genetic algorithm in the hybrid optimization strategy adopts an elite retention strategy and introduces domain-knowledge-based crossover and mutation operators during the evolution process to accelerate the search for feasible solutions and improve the quality of the solutions; the local search algorithm is a simulated annealing algorithm, which is used to perform deep exploration in the vicinity of the non-dominated solution set obtained by the genetic algorithm.

[0089] Specifically, the multi-objective genetic algorithm employs an elite preservation strategy. This strategy aims to ensure that the best individuals (i.e., elite individuals) in each generation can be directly inherited by the next generation, unaffected by crossover, mutation, or other operations. This avoids losing discovered excellent solutions during evolution, accelerates algorithm convergence, and improves the quality of the final solution. For example, after each generation is generated, the K individuals with the highest fitness in the current population can be directly copied to the next generation, replacing the K individuals with the lowest fitness. Alternatively, an independent elite archive can be maintained, storing all non-dominated solutions discovered in previous generations. After each iteration, the non-dominated solutions in the current population are merged and filtered with the solutions in the archive, the archive is updated, and during the generation of the next generation, some elite individuals are selected from the archive to be added to the new population.

[0090] Simultaneously, domain-knowledge-based specialized crossover and mutation operators are introduced during the evolutionary process. Crossover and mutation operators are core operations in genetic algorithms used to generate new individuals (offspring). Domain-knowledge-based specialized operators mean that the design of these operations fully considers the specific constraints, objective function characteristics, and process rules of the heat treatment scheduling problem, rather than using general, random operators. This helps guide the search process to explore the feasible solution space more effectively, reduce invalid searches, improve the algorithm's search efficiency and solution quality, and make it more in line with actual production needs. For example, a "process block crossover" operator can be designed to select one or more complete process blocks (e.g., all heat treatment processes for an order) from two parent scheduling schemes, and then exchange these process blocks while ensuring that the exchange does not violate key constraints such as equipment capacity and process sequence. For example, a "time window adjustment mutation" operator can be designed to make small random adjustments to the start time or end time of a certain furnace or order without violating its earliest start time, latest finish time, and process dependencies with other furnaces / orders, or to transfer it from one device to another compatible device, in order to explore a better production schedule or equipment allocation.

[0091] Furthermore, the local search algorithm is simulated annealing. Simulated annealing is a heuristic search algorithm that avoids getting trapped in local optima by accepting solutions worse than the current solution with a certain probability during the search process, thus possessing strong global search capabilities. This simulated annealing algorithm is used for deep exploration in the vicinity of the non-dominated solution set obtained by the genetic algorithm. A "non-dominated solution set" refers to a set in a multi-objective optimization problem where no single solution is better than any other solution in the set across all objectives. Deep exploration means that the local search algorithm, through its mechanism (such as the probability of accepting inferior solutions in simulated annealing), can escape local optima and perform a detailed search in a broader neighborhood space. For example, the initial solution of the simulated annealing algorithm can be randomly selected from the non-dominated solution set obtained by the genetic algorithm. Then, starting from this solution, neighborhood solutions are iteratively generated and selected according to the acceptance criterion, gradually optimizing the solution. Alternatively, each solution in the non-dominated solution set obtained by the genetic algorithm can be used as the initial solution for an independent local search in the simulated annealing algorithm, thereby refining the entire non-dominated solution set.

[0092] like Figure 2 As shown, a heat treatment intelligent scheduling planning system, applied to the above-mentioned heat treatment intelligent scheduling planning method, includes:

[0093] The data acquisition module is configured to acquire data on pending production orders, equipment capacity, and energy prices.

[0094] The model building module is configured to build a multi-objective optimization model based on the data. The objective function of the model aims to minimize the overall cost of production scheduling, which is defined as the weighted sum of the first, second, and third terms. The first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-period energy consumption cost, the value of which is calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information.

[0095] The optimization solution module is configured to solve the multi-objective optimization model under the condition of satisfying process constraints and generate a production scheduling plan.

[0096] The execution and dynamic scheduling module is configured to execute production based on the production scheduling plan and trigger dynamic rescheduling when production disturbances occur.

[0097] In this embodiment, the data acquisition module is configured to acquire pending production order data, equipment capacity data, and energy price data. Specifically, this module is responsible for collecting, organizing, and preprocessing the basic information required for production scheduling from different data sources. For example, it can interface with existing information systems such as Enterprise Resource Planning (ERP) systems and Manufacturing Execution Systems (MES) to achieve automatic synchronization of data such as order information, equipment status, and process parameters.

[0098] The model building module is configured to construct a multi-objective optimization model based on the data. The objective function of the model aims to minimize the overall cost of production scheduling, defined as the weighted sum of the first, second, and third terms; where the first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-period energy consumption cost, calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information. This module abstracts the actual heat treatment production scheduling problem into a mathematical model for calculation and optimization. For example, this module can use mathematical programming language to define variables, objective functions, and constraints, using the maximum completion time, total delay penalty, and total peak-period energy consumption cost as optimization objectives. The calculation of the total peak-period energy consumption cost is particularly crucial. It calculates the average power of each furnace based on its processing load using the corresponding equipment's nonlinear power function, determines the actual processing time of each furnace during the predicted peak electricity price period according to the production scheduling plan, and finally multiplies and sums the average power, the predicted electricity price for the corresponding period, and the actual processing time. This nonlinear power function can be used to characterize the nonlinear relationship between equipment power and processing load by fitting a quadratic or higher-order function to historical energy consumption data. Furthermore, the objective function Z can be expressed as a normalized weighted sum, representing the normalized weighted sum of the first term (maximum completion time), the second term (total delay penalty), and the third term (total peak-period energy cost), thus balancing objectives with different dimensions.

[0099] The optimization solution module is configured to solve the multi-objective optimization model under the condition of satisfying process constraints and generate a production scheduling plan. This module is the core of intelligent production scheduling. It receives the mathematical model generated by the model building module and uses advanced algorithms to find the optimal or near-optimal production scheduling plan that satisfies all constraints such as production processes and equipment capabilities. For example, this module can use a hybrid optimization strategy that combines heuristic rules and metaheuristic algorithms to solve the problem. First, it generates an initial solution that satisfies the key constraints, then uses a multi-objective genetic algorithm for global optimization, and finally uses a local search algorithm to fine-tune the solution. Alternatively, this module can directly call commercial optimization solvers or open-source solvers to solve complex mixed integer programming problems.

[0100] The execution and dynamic scheduling module is configured to execute production based on the production scheduling plan and trigger dynamic rescheduling when production disturbances occur. This module is responsible for distributing the production scheduling plan generated by the optimization solution module to the production site and monitoring the production process in real time. When disturbances such as order changes, equipment failures, or material shortages occur during production, this module can quickly identify and trigger the corresponding dynamic rescheduling mechanism. For example, depending on the type and scope of the disturbance, this module can adopt a hierarchical strategy: for local disturbances, a constraint propagation algorithm is applied for rapid repair; for global major disturbances, the optimization process of model building and optimization solution is re-executed within a compressed time window. Alternatively, this module can also quickly adjust by matching the corresponding contingency plan according to the disturbance type using a preset contingency plan library.

[0101] Through the aforementioned system, this invention effectively addresses the challenges of data acquisition, model building, optimization, and dynamic response in heat treatment production scheduling. The data acquisition module ensures comprehensive and real-time decision-making, avoiding decision-making biases caused by missing or delayed data in traditional methods. The model building module, by introducing nonlinear energy consumption models of the equipment and time-of-use electricity price prediction information, achieves accurate estimation and optimization of energy costs, overcoming the inaccuracy of traditional linear models. This significantly reduces energy consumption and production costs while ensuring production efficiency and delivery time.

[0102] As a preferred embodiment of the technical solution of the present invention, the dynamic rearrangement process includes:

[0103] A tiered strategy is adopted based on the type and scope of the disturbance:

[0104] For local disturbances, the constraint propagation algorithm is applied for repair.

[0105] For global disturbances, the construction and solution process of the multi-objective optimization model is executed cyclically within a compressed time window.

[0106] In one embodiment of the technical solution of this invention, the dynamic rescheduling process in step S4 is defined. A layered strategy is adopted based on the type and scope of the disturbance: for local disturbances, a constraint propagation algorithm is applied for rapid repair; for significant global disturbances, the optimization processes of steps S2 and S3 are re-executed within a compressed time window. In this embodiment, "dynamic rescheduling" refers to adjusting the existing production schedule to adapt to new production conditions when unplanned events (such as equipment failure, urgent orders, material shortages, etc.) occur during the production process. This adjustment can be local, affecting only tasks directly related to the disturbance; or it can be global, requiring a complete replanning of the entire production schedule.

[0107] The aforementioned "layered strategy based on the type and scope of disturbances" aims to select the most appropriate response mechanism according to the nature and impact of the disturbance event. Disturbance types can include equipment failure, order changes, abnormal process parameters, and material supply delays. The scope of the disturbance can be assessed based on indicators such as the number of affected devices, the number of orders, and the production cycle length. For example, disturbances affecting a single device or a few tasks are considered localized disturbances, while disturbances affecting multiple production lines or critical bottleneck devices are considered global, significant disturbances. This layered strategy avoids overly complex handling of minor disturbances while ensuring sufficient optimization for major disturbances.

[0108] For "local disturbances," this invention employs a "constraint propagation algorithm" for rapid repair. The constraint propagation algorithm is a commonly used technique in constraint satisfaction problems; it narrows the search space by identifying and eliminating inconsistent variable values. In production scheduling repair, when a local disturbance occurs, such as a brief equipment failure, the constraint propagation algorithm can quickly identify the affected tasks and their subsequent dependent tasks, and make local adjustments based on existing constraints (such as process sequence, equipment availability, delivery dates, etc.). For example, it can transfer affected tasks to backup equipment or adjust task start / end times without recalculating the entire production schedule. For instance, the AC-3 algorithm or the FC (Forward Checking) algorithm can be used to maintain constraint consistency and quickly find locally feasible repair solutions.

[0109] For "globally significant disturbances," this invention "re-executes the optimization process of steps S2 and S3 within a compressed time window." Globally significant disturbances typically refer to events that have a wide and profound impact on the entire production system, such as prolonged downtime of critical equipment, a surge of urgent orders, or disruptions in the supply of major raw materials. In such cases, simple local adjustments may not be effective in solving the problem and could even lead to suboptimal solutions or chain reactions. Therefore, it is necessary to re-execute steps S2 (building a multi-objective optimization model) and S3 (solving the multi-objective optimization model) to generate a completely new production scheduling plan. The "compressed time window" means completing the model construction and solution within a limited and typically short time to ensure rapid response to production changes. This can be achieved by employing more efficient solution algorithms, simplifying model parameters, or utilizing pre-computed knowledge bases. For example, during the solution process, a fast iterative version of heuristic or metaheuristic algorithms (such as genetic algorithms or simulated annealing algorithms) can be used, or parallel computing resources can be used to accelerate the solution process.

[0110] Figure 2 This is a block diagram of the composition structure of a heat treatment intelligent scheduling planning system. In this embodiment of the invention, a heat treatment intelligent scheduling planning system 10 includes:

[0111] Data acquisition module 11 is used to acquire data on pending production orders, equipment capacity data, and energy price data;

[0112] The cost optimization module 12 is used to construct a multi-objective optimization model based on pending production order data, equipment capacity data, and energy price data. The objective function of the model aims to minimize the comprehensive cost of production scheduling, which is defined as the weighted sum of the first, second, and third terms. The first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-period energy consumption cost, the value of which is calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information.

[0113] Model solving module 13 is used to solve the multi-objective optimization model under the condition of satisfying process constraints and generate a production scheduling plan;

[0114] The dynamic rescheduling module 14 is used to execute production based on the production scheduling plan and to trigger the dynamic rescheduling process when production disturbances occur.

[0115] Furthermore, the calculation process for the total peak-hour energy consumption cost includes:

[0116] The average power is calculated based on the processing load of each furnace and the nonlinear power function of the corresponding equipment.

[0117] The actual processing time for each furnace batch during the predicted peak electricity price period is determined based on the production scheduling plan;

[0118] The average power, the predicted electricity price for the corresponding time period, and the actual processing time are multiplied and summed to obtain the total peak-period energy consumption cost.

[0119] The fitting and calibration process of the nonlinear power function is dynamic: the system collects load energy consumption data in actual production in real time based on a preset frequency, and periodically updates the parameters of the nonlinear power function to keep the energy consumption model synchronized with the actual equipment status.

[0120] Specifically, the nonlinear energy consumption model is a quadratic or higher-order function obtained by fitting historical energy consumption data, used to characterize the nonlinear relationship between equipment power and processing load.

[0121] Furthermore, the objective function is expressed in a normalized weighted sum form as the normalized weighted sum of the first term, the maximum completion time, the second term, the total delay penalty, and the third term, the total peak-period energy consumption cost.

[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent scheduling and planning of heat treatment processes, characterized in that, The method includes: Obtain data on pending production orders, equipment capacity, and energy prices; A multi-objective optimization model is constructed based on pending production order data, equipment capacity data, and energy price data. The objective function of the model aims to minimize the comprehensive cost of production scheduling, which is defined as the weighted sum of the first, second, and third terms. The first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-hour energy consumption cost, which is calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information. The multi-objective optimization model is solved under the condition of satisfying process constraints to generate a production scheduling plan; Production is executed based on the production scheduling plan, and a dynamic rescheduling process is triggered when production disturbances occur.

2. The intelligent scheduling method for heat treatment according to claim 1, characterized in that, The calculation process for the total peak-hour energy consumption cost includes: The average power is calculated based on the processing load of each furnace and the nonlinear power function of the corresponding equipment. The actual processing time for each furnace batch during the predicted peak electricity price period is determined based on the production scheduling plan; The average power, the predicted electricity price for the corresponding time period, and the actual processing time are multiplied and summed to obtain the total peak-period energy consumption cost. The fitting and calibration process of the nonlinear power function is dynamic: the system collects load energy consumption data in actual production in real time based on a preset frequency, and periodically updates the parameters of the nonlinear power function to keep the energy consumption model synchronized with the actual equipment status.

3. The intelligent scheduling method for heat treatment according to claim 2, characterized in that, The nonlinear energy consumption model is a quadratic or higher-order function obtained by fitting historical energy consumption data, used to characterize the nonlinear relationship between equipment power and processing load.

4. The intelligent scheduling method for heat treatment according to claim 1, characterized in that, The objective function is expressed in a normalized weighted sum form as the normalized weighted sum of the first term, the maximum completion time, the second term, the total delay penalty, and the third term, the total peak-period energy consumption cost.

5. The intelligent scheduling method for heat treatment according to claim 1, characterized in that, The solution process of the multi-objective optimization model adopts a hybrid optimization strategy that combines heuristic rules and metaheuristic algorithms, specifically including: Generate an initial solution that satisfies the key constraints; A multi-objective genetic algorithm is used for global optimization; The solution is optimized based on a local search algorithm; The multi-objective genetic algorithm employs an elite retention strategy and introduces domain-knowledge-based crossover and mutation operators during the evolution process. The local search algorithm is a simulated annealing algorithm, used to perform a more accurate search on the non-dominated solution set obtained by the genetic algorithm.

6. The intelligent scheduling method for heat treatment according to claim 1, characterized in that, The dynamic rearrangement process includes: A tiered strategy is adopted based on the type and scope of the disturbance: For local disturbances, the constraint propagation algorithm is applied for repair. For global disturbances, the construction and solution process of the multi-objective optimization model is executed cyclically within a compressed time window.

7. A heat treatment intelligent scheduling system, characterized in that, The system includes: The data acquisition module is used to acquire data on pending production orders, equipment capacity, and energy prices. The cost optimization module is used to construct a multi-objective optimization model based on pending production order data, equipment capacity data, and energy price data. The objective function of the model aims to minimize the comprehensive cost of production scheduling, which is defined as the weighted sum of the first, second, and third terms. The first term is the maximum completion time; the second term is the total delay penalty; and the third term is the total peak-period energy consumption cost, which is calculated based on the equipment's nonlinear energy consumption model and time-of-use electricity price prediction information. The model solving module is used to solve the multi-objective optimization model under the condition of satisfying process constraints and generate a production scheduling plan. The dynamic rescheduling module is used to execute production based on the production scheduling plan and to trigger the dynamic rescheduling process when production disturbances occur.

8. The intelligent scheduling system for heat treatment according to claim 7, characterized in that, The calculation process for the total peak-hour energy consumption cost includes: The average power is calculated based on the processing load of each furnace and the nonlinear power function of the corresponding equipment. The actual processing time for each furnace batch during the predicted peak electricity price period is determined based on the production scheduling plan; The average power, the predicted electricity price for the corresponding time period, and the actual processing time are multiplied and summed to obtain the total peak-period energy consumption cost. The fitting and calibration process of the nonlinear power function is dynamic: the system collects load energy consumption data in actual production in real time based on a preset frequency, and periodically updates the parameters of the nonlinear power function to keep the energy consumption model synchronized with the actual equipment status.

9. The intelligent scheduling system for heat treatment according to claim 8, characterized in that, The nonlinear energy consumption model is a quadratic or higher-order function obtained by fitting historical energy consumption data, used to characterize the nonlinear relationship between equipment power and processing load.

10. The intelligent scheduling system for heat treatment according to claim 7, characterized in that, The objective function is expressed in a normalized weighted sum form as the normalized weighted sum of the first term, the maximum completion time, the second term, the total delay penalty, and the third term, the total peak-period energy consumption cost.

Images