Energy flexible manufacturing system two-stage regulation and control method considering demand response priority

Abstract

The invention belongs to the technical field of energy flexible manufacturing system regulation and control, and particularly relates to an energy flexible manufacturing system two-stage regulation and control method considering demand response priority, which comprises the following steps: S1, constructing a comprehensive carbon emission factor model; s2, constructing a demand response priority model; s3, in the day-ahead stage, mixed carbon emission factors of all time periods of the next day are calculated, and a two-level production energy consumption decision-making model is constructed based on the mixed carbon emission factors; solving to obtain a day-ahead plan of production and energy consumption collaboration of the next day; and S4, in the real-time operation stage, when the actual energy supply cannot meet the load demand of the day-ahead plan, calling the priority score of each processing task in the S2, determining an adjustable processing task, constructing a multi-target real-time optimization decision model, and dynamically adjusting the start-stop or processing time period of the adjustable processing task. According to the method, low-carbon, economic and reliable collaborative decision-making of production energy consumption of the energy flexible manufacturing system in an environment with uncertain supply and demand can be realized.

Description

Technical Field

[0001] This invention belongs to the field of energy flexible manufacturing system control technology, and particularly relates to a two-stage control method for energy flexible manufacturing systems that considers demand response priority. Background Technology

[0002] With the deepening of the goals of "carbon peaking and carbon neutrality," the manufacturing industry is accelerating its transformation towards green and low-carbon practices. Against this backdrop, the traditional energy supply structure, dominated by primary energy sources such as thermal power and natural gas, is gradually evolving into a comprehensive energy system that integrates multiple types of clean energy, including photovoltaics, wind power, and energy storage. This transformation significantly enhances the diversity and flexibility of energy use in manufacturing systems, giving rise to a new paradigm: "energy-flexible manufacturing systems." However, renewable energy sources (such as wind and solar power) have inherent intermittency and volatility, and their high penetration rate presents unprecedented uncertainty challenges for energy-flexible manufacturing systems. On the one hand, the output on the supply side is difficult to predict accurately, leading to deviations between energy consumption plans and actual energy supply. On the other hand, production tasks on the manufacturing side have varying time sensitivities, process constraints, and customer requirements, resulting in significant differences in their responsiveness to energy regulation. How to achieve synergistic optimization of energy use's economy, low carbon emissions, and robustness while ensuring production efficiency and delivery fulfillment has become a core research challenge.

[0003] To address the aforementioned challenges, scholars both domestically and internationally have conducted extensive research on multi-timescale energy consumption decision-making in manufacturing systems. For example, some studies focus on the microgrid level, considering energy supply uncertainty and capacity constraints, and employ multi-objective optimization algorithms such as NSGA-II to optimize the allocation of distributed generation resources. Other studies introduce stochastic model predictive control (SMPC) methods, combined with Gaussian mixture models (GMMs) to characterize the probability distribution of renewable energy and load forecasting errors, thus supporting more accurate real-time dispatch decisions. Furthermore, the Chinese invention patent "A Multi-Timescale Dispatch Method and System for Integrated Energy Systems" (CN120409982A) proposes a day-ahead-intraday coordinated electricity-gas integrated energy dispatch framework, achieving a dispatch scheme with minimized operating costs by constructing a multi-timescale optimization model.

[0004] Although the aforementioned studies have made some progress in multi-timescale regulation of energy supply, there are still significant limitations: First, existing methods generally treat the manufacturing side as a passive load, ignoring the heterogeneity of demand response capabilities of different processing tasks and failing to establish a linkage mechanism between task priority and energy regulation strategies; Second, optimization objectives are mostly focused on minimizing economic costs, lacking systematic consideration of environmental constraints such as carbon emissions; Third, carbon emission factors are usually simplified to fixed constants, failing to consider the time-varying characteristics of mixed carbon emission intensity caused by dynamic changes in the energy structure (such as fluctuations in the proportion of wind and solar power output) at different times, making it difficult to truly reflect the actual carbon footprint of the manufacturing system.

[0005] Therefore, how to achieve low-carbon, economical, and reliable collaborative decision-making in energy production and consumption of flexible energy manufacturing systems under uncertain supply and demand environments has become an urgent problem to be solved. Summary of the Invention

[0006] To address the shortcomings of the existing technologies, this invention provides a two-stage control method for energy-flexible manufacturing systems that considers demand response priorities. This method enables low-carbon, economical, and reliable collaborative decision-making for energy consumption in energy-flexible manufacturing systems under uncertain supply and demand conditions.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0008] A two-stage control method for energy-flexible manufacturing systems that considers demand response priorities includes the following steps:

[0009] S1. Construct a comprehensive carbon emission factor model that considers the time-varying nature of energy types. This model is used to calculate the mixed carbon emission factor of the energy flexible manufacturing system based on the output ratio of various energy types and their corresponding unit carbon emission intensity in any given time period, so as to quantify the dynamic carbon emission intensity of the manufacturing system.

[0010] S2. Construct a demand response priority model based on three dimensions: customer needs, product characteristics, and equipment characteristics. This model is used to calculate the demand response priority score for each processing task, thereby quantifying its adjustment priority in real-time demand response.

[0011] S3. In the day-ahead phase, based on the day-ahead forecast data of renewable energy output and the grid time-of-use electricity price information, the mixed carbon emission factor for each time period of the next day is calculated using the comprehensive carbon emission factor model in S1. Based on the mixed carbon emission factor, a two-level production energy consumption decision model is constructed: the first level takes the minimum maximum processing time, the highest equipment load balance, and the minimum time-series carbon emissions as optimization objectives, and generates equipment processing time arrangements and corresponding load curves; the second level, based on the load curve, optimizes the coordinated energy consumption strategy of each energy source with the goal of minimizing energy costs.

[0012] By jointly solving the two-level production energy consumption decision model using a multi-objective optimization algorithm, the day-ahead plan for the coordination of production and energy consumption for the next day is obtained.

[0013] S4. During the real-time operation phase, acquire the actual energy supply data of each energy source and compare it with the day-ahead forecast data. When the actual energy supply cannot meet the load demand of the day-ahead plan, call the priority scores of each processing task in S2 to determine the adjustable processing tasks, and construct a multi-objective real-time optimization decision model with carbon emission management costs, delivery delay penalty costs, and demand response incentive benefits as optimization objectives. Dynamically adjust the start and stop or processing time of the adjustable processing tasks to realize online correction of the day-ahead plan.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] 1. Achieving dynamic and accurate characterization of carbon emission intensity. Unlike existing technologies that use fixed carbon emission factors, this scheme constructs a comprehensive carbon emission factor model that considers the time-varying characteristics of the energy structure. This model can calculate the mixed carbon emission factor in real time based on the actual output ratio of various energy sources (such as wind power and photovoltaics) and their unit carbon emission intensity at different times. This significantly improves the accuracy of carbon emission assessment and provides a scientific basis for low-carbon dispatching.

[0016] 2. Introduce a task-level demand response prioritization mechanism to enhance scheduling flexibility. Existing methods typically treat manufacturing load as an unadjustable or uniformly adjustable whole, ignoring the differences in processing tasks based on customer urgency, product process requirements, and equipment adaptability. This solution constructs a priority scoring model from three dimensions: customer needs, product characteristics, and equipment characteristics. It quantifies the adjustability and adjustment costs of each task, making real-time control more targeted and feasible, and improving the overall system responsiveness while ensuring the fulfillment of critical tasks.

[0017] 3. Constructing a two-stage collaborative optimization framework for day-ahead and real-time operations, balancing planning foresight and operational adaptability. Unlike traditional scheduling methods that focus only on a single time scale (such as day-ahead or real-time only), this scheme designs a two-stage control architecture: the day-ahead stage generates a collaborative energy consumption plan based on forecast information, balancing processing efficiency, load balancing, and minimizing carbon emissions; the real-time stage dynamically adjusts the plan using a priority model based on actual energy supply deviations. This "overall planning + local adjustment" mechanism effectively addresses the uncertainties in renewable energy output and load demand, significantly improving the robustness and practicality of the scheduling scheme.

[0018] 4. Achieving multi-objective synergistic optimization of economy, low carbon emissions, and reliability. Existing research often focuses solely on minimizing energy costs, neglecting the trade-off between carbon emission constraints and production reliability. This scheme simultaneously optimizes processing time, load balancing, and time-series carbon emissions during the day-ahead phase, and comprehensively considers carbon emission management costs, delivery delay penalties, and demand response incentives during the real-time phase. This forms a multi-objective decision-making system covering the entire chain, avoiding suboptimal or even conflicting results that may occur under the single-objective drive of traditional methods.

[0019] In summary, this method enables energy-flexible manufacturing systems to make low-carbon, economical, and reliable collaborative decisions regarding energy consumption in production under uncertain supply and demand environments.

[0020] Preferably, the various energy sources include grid power supply, photovoltaic power generation, and energy storage discharge.

[0021] Preferably, in S1, the integrated carbon emission factor model is:

[0022]

[0023] v∈(grid, pv, ess);

[0024] In the formula, xef t It is a mixture of carbon emission factors; It is the power output of energy v at time t; Let v be the emission factor of energy at time t; grid, pv, and ess represent grid power supply, photovoltaic power generation, and energy storage system, respectively.

[0025] This setup, by incorporating the carbon emission characteristics of diverse energy sources such as photovoltaics, energy storage, and the power grid into a unified calculation framework, provides a scientific basis for subsequent optimization objectives (such as minimizing time-series carbon emissions). Compared to traditional methods that rely solely on a single energy source or ignore the emission reduction potential of clean energy, this model is more conducive to guiding the system to prioritize the use of low-carbon energy and promoting the green transformation of manufacturing processes. Furthermore, due to the volatility of wind and solar power output, traditional fixed carbon factors cannot reflect differences between time periods, potentially leading to high-load production during high-carbon periods. This model, however, can update carbon emission intensity in real time according to changes in energy output, providing crucial input for subsequent day-ahead and real-time regulation, and enhancing the flexibility and adaptability of the entire system in dealing with uncertainties.

[0026] Preferably, in S2, the customer demand is characterized by the urgency of delivery, and the product characteristics and equipment characteristics each include multiple evaluation indicators; the priority score is obtained by weighted fusion of the scores from the three dimensions:

[0027]

[0028] In the formula, PR finPriority scores: PD, PR, MRPR t These are the numerical values ​​for customer needs, product characteristics, and equipment characteristics, respectively; ω PD ω PR , PD, PR, MRPR respectively t The weights, and

[0029] This setup, by constructing a demand response priority model based on multi-dimensional feature weighted fusion, enables precise characterization of the adjustment potential of manufacturing tasks, providing key support for the dynamic scheduling of energy-flexible manufacturing systems under uncertain environments.

[0030] Preferably, in S3, the objective function of the first level is:

[0031] minf1=minC max ;

[0032]

[0033] In the formula, C max The maximum processing time is t; n, n1, and m are the total number of workpieces, process routes, and equipment, respectively; t ijk x represents the processing time of workpiece i on equipment k under process route j; ijk For binary decision variables, xef is 1 when workpiece i is executed on equipment k under process route j, and 0 otherwise; t id The current carbon emission factor; For in t ijk Operating power of processing equipment within the range.

[0034] This setup, by constructing a multi-objective first-level optimization model that integrates processing efficiency, load balancing, and low-carbon operation, achieves deep synergy between production planning and energy use, providing strong support for the efficient, balanced, and low-carbon comprehensive scheduling of energy-flexible manufacturing systems in complex operating environments.

[0035] Preferably, in S3, the objective function of the second level is:

[0036]

[0037] In the formula, T is the total cycle time based on the processing task scheduling; and Let be the energy costs of grid power supply, PV, and ESS at time t, respectively. and These are the time-of-use electricity price of the power grid at time t, the levelized cost of photovoltaic power generation, the levelized cost of energy storage charging, and the levelized cost of energy storage discharging; xeft id This is the carbon emission factor as of today; P t grid P t pv P t ess,ch P t ess,dis These represent the electrical power purchased from the grid at time t, the electrical power output from the photovoltaic power generation system, the charging power of the energy storage system, and the discharging power of the energy storage system, respectively; t grid t pv t ess,ch t ess,dis These are the duration of grid power supply, the duration of photovoltaic power generation, the duration of energy storage system charging, and the duration of energy storage system discharging.

[0038] This approach integrates grid power supply, photovoltaic power generation, and energy storage systems into a unified cost optimization framework. By comprehensively considering the real-time prices and operating characteristics of different energy sources, it enables dynamic adjustments to the energy usage structure. Compared to traditional methods that rely on a single energy source or static allocation, this approach significantly improves the economic efficiency of energy utilization and helps reduce the overall energy expenditure of manufacturing systems.

[0039] Preferably, in S3, the NSGA-II algorithm is used to jointly solve the two-level production energy consumption decision model.

[0040] Preferably, in S4, the objective function of the multi-objective real-time optimization decision model is:

[0041]

[0042] ψ=[P grid ,P pv ,P ess,ch ,P ess,dis ];

[0043]

[0044] In the formula, It is the cost of punishment; It is demand response revenue; It is the cost of carbon management; It is the incentive benefit for participating in demand response; xef t rt It is a real-time carbon emission factor; xef t id The carbon emission factor is as of today; ψ is a control variable that is further adjusted in real time to follow the random fluctuations in clean energy photovoltaic power.

[0045] ΔT is the total scheduling cycle time; Real-time grid power purchase cost; Real-time photovoltaic power generation cost; Cost of real-time energy storage systems; The unit price will be penalized for delays; For delay penalty factor; P t dr For demand response participation power; t dr Duration of demand response; For carbon management unit price; P t rt P represents real-time load power. t id This represents the planned load power for the day. Δt is the real-time control variable; Δt is the real-time optimization time step. P is the control variable for the current day plan; grid P pv P ess ,ch P ess,dis These represent the power consumed by grid power purchase, photovoltaic power generation, energy storage charging, and energy storage discharging, respectively; ψ rt (t+Δt|t) represents the predictive control variable for the future time t+Δt when performing rolling optimization at time t; ψ rt,0 (t) represents the initial control state; N Δt To optimize the number of steps for scrolling; Δu i (t+Δt) represents the load adjustment amount of the i-th adjustable task at time t+Δt.

[0046] This setup integrates grid costs, photovoltaic and energy storage operating costs, delivery delay penalties, carbon management costs, and demand response incentive benefits into a unified optimization framework, forming a multi-dimensional collaborative optimization mechanism. This avoids suboptimal solutions that may arise under a single objective (such as sacrificing delivery for carbon reduction) and achieves a globally optimal balance under complex constraints.

[0047] 2. By introducing real-time carbon emission factor xef t rt And compare it with the current carbon factor xef t id By performing interpolation calculations, the model can quantify the additional carbon emission costs caused by fluctuations in the energy structure. This mechanism prompts the system to prioritize the use of low-carbon energy in real-time and proactively avoid operating during high-carbon periods, thereby achieving precise management and continuous reduction of the carbon footprint.

[0048] 3. By setting a time step Δt and a rolling optimization mechanism, the model can predict and optimize for a future period at each time point, dynamically updating control variables. This "rolling prediction-step-by-step execution" approach effectively addresses the randomness of renewable energy output and improves the timeliness and accuracy of dispatch decisions.

[0049] Preferably, in S4, the Model Predictive Control (MPC) algorithm is used for rolling optimization, and a feedback correction stage is introduced, using the actual multi-energy output vector of the system at the current moment as the initial state for rolling optimization at the next moment:

[0050] ψ rt,0 (t+1)=ψ rt,real (t);

[0051] In the formula, ψ rt,0 (t+1) is the initial control variable for the rolling optimization at the next time step, ψ rt,real (t) represents the actual multi-energy output vector of the system at the current moment, including grid power purchase, photovoltaic power generation, energy storage charging power, and discharge power.

[0052] This setup, by introducing the feedback correction mechanism of Model Predictive Control (MPC), achieves a shift from prediction-driven to state-driven control, significantly enhancing the dynamic response capability and control accuracy of the energy flexible manufacturing system under uncertain environments. Attached Figure Description

[0053] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0054] Figure 1 This is a flowchart of the method;

[0055] Figure 2 This is a schematic diagram of typical photovoltaic power output data and carbon emission factor data in Example 2;

[0056] Figure 3 This is a schematic diagram of the pre-processed Pareto solution set, Gantt chart, and multi-energy output in Example 2;

[0057] Figure 4 This is a schematic diagram of the product demand response matrix constructed in Example 2;

[0058] Figure 5 This is a schematic diagram of the real-time optimized scheduling results constructed in Example 2. Detailed Implementation

[0059] The following detailed explanation illustrates the specific implementation methods:

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment discloses a two-stage control method for an energy-flexible manufacturing system that considers demand response priorities, including the following steps:

[0062] S1. Construct a comprehensive carbon emission factor model that considers the time-varying nature of energy types. This model is used to calculate the mixed carbon emission factor for any given time period based on the output ratio of various energy types and their corresponding unit carbon emission intensity of the energy-flexible manufacturing system. This will quantify the dynamic carbon emission intensity of the manufacturing system.

[0063] The various energy sources mentioned include grid power supply, photovoltaic power generation, and energy storage discharge.

[0064] In practical implementation, the comprehensive carbon emission factor model is as follows:

[0065]

[0066] v∈(grid, pv, ess);

[0067] In the formula, xef t It is a mixture of carbon emission factors; It is the power output of energy v at time t; Let v be the emission factor of energy at time t; grid, pv, and ess represent grid power supply, photovoltaic power generation, and energy storage system, respectively.

[0068] By incorporating the carbon emission characteristics of diverse energy sources such as photovoltaics, energy storage, and the power grid into a unified calculation framework, this model provides a scientific basis for subsequent optimization objectives (such as minimizing time-series carbon emissions). Compared to traditional methods that rely solely on a single energy source or ignore the emission reduction potential of clean energy, this model is more conducive to guiding the system to prioritize the use of low-carbon energy and promoting the green transformation of manufacturing processes. Furthermore, due to the volatility of wind and solar power output, traditional fixed carbon factors cannot reflect differences between time periods, potentially leading to high-load production during high-carbon periods. This model, however, can update carbon emission intensity in real time according to changes in energy output, providing crucial input for subsequent day-ahead and real-time regulation, and enhancing the flexibility and adaptability of the entire system in dealing with uncertainties.

[0069] S2. Construct a demand response priority model based on three dimensions: customer needs, product characteristics, and equipment characteristics. This model is used to calculate the demand response priority score for each processing task, thereby quantifying its adjustment priority in real-time demand response.

[0070] In practice, the customer demand is characterized by the urgency of the delivery date, and the product characteristics and equipment characteristics each include multiple evaluation indicators; the priority score is obtained by weighted fusion of the scores from the three dimensions:

[0071]

[0072] In the formula, PR fin Priority scores: PD, PR, MRPR t These are the numerical values ​​for customer needs, product characteristics, and equipment characteristics, respectively; ω PD ω PR , PD, PR, MRPR respectively t The weights, and

[0073] To better understand the construction process of the demand response prioritization model, the following explanation is provided.

[0074] (1) Customer demand matrix

[0075] Customer demand can be represented by delivery time. Assuming there are K products, customer demand can be constructed by collecting delivery time data for K products.

[0076] PD = (npd1, npd2, npd3, ..., npd) i ,...,npd K ) T ;

[0077] Time-constrained metrics, such as delivery dates, need to be reverse-normalized, as shown below. This is because for time-sensitive data, the shorter the time, the more important it is. The same applies to other time-constrained metrics.

[0078]

[0079] Therefore, the standardized customer demand matrix can be described as:

[0080] PD=(nnpd1,nnpd2,nnpd3,...,nnpd i ,...,nnpd K ) T .

[0081] (2) Product Feature Matrix

[0082] Suppose there are K products, defined as P i = (i = 1, 2, 3, ..., K), each product has N demand evaluation indicators, and the relative performance of the product demand priorities of K products is evaluated to form a K×N product demand priority evaluation matrix:

[0083]

[0084] pr ji Normalized value and and The relationship can be described as follows:

[0085]

[0086] Therefore, the standardized product feature matrix can be described as:

[0087]

[0088] Based on the importance of product performance indicators, the weight of each performance indicator is determined. in The product demand prioritization model is obtained based on the standardized NPPR values:

[0089]

[0090] (3) Equipment feature matrix

[0091] Suppose there are K products, defined as P i = (i = 1, 2, 3, ..., K), to determine the parameters of the equipment that participates in the demand response at time t, where equipment that is currently processing cannot participate. Assume that at time t, the Pth... i Product at M m If there are N evaluation indicators for processing equipment (m=1,2,3,...,j,..,M), then P i The product's equipment operation priority model is as follows, forming a K×N evaluation matrix:

[0092]

[0093] Normalized value and and The relationship can be described as follows:

[0094]

[0095] Therefore, the standardized product feature matrix can be described as follows:

[0096]

[0097] Based on the importance of equipment performance indicators, the weight θ = (θ1, θ2, θ3, ..., θ) of each performance indicator is obtained. N ), where 0 < θ i <1, and According to the standardized NMPR t Numerical data is used to obtain the device operation priority model.

[0098]

[0099] (4) Comprehensive Priority Modeling

[0100]

[0101] In this way, by constructing a demand response priority model based on multi-dimensional feature weighted fusion, the adjustment potential of manufacturing tasks can be accurately characterized, providing key support for the dynamic scheduling of energy flexible manufacturing systems under uncertain environments.

[0102] S3. In the day-ahead phase, based on the day-ahead forecast data of renewable energy output and the grid time-of-use electricity price information, the mixed carbon emission factor for each time period of the next day is calculated using the comprehensive carbon emission factor model in S1. Based on this mixed carbon emission factor, a two-level production energy consumption decision model is constructed: the first level aims to minimize the maximum processing time, maximize the equipment load balance, and minimize the time-series carbon emissions, generating equipment processing time arrangements and corresponding load curves; the second level, based on the load curves, aims to minimize energy costs and optimize the coordinated energy consumption strategies of each energy source; the NSGA-II algorithm is used to jointly solve the two-level production energy consumption decision model to obtain the day-ahead plan for coordinated production and energy consumption for the next day.

[0103] In practice, the day-ahead forecast data of renewable energy output can be obtained by using a forecasting model based on historical renewable energy output data and weather forecast data. This forecasting technology is a mature existing technology and will not be elaborated on here.

[0104] The first level constructs a multi-objective optimization model considering equipment load, carbon emissions, and delivery time. In specific implementation, the objective function of the first level is:

[0105] minf1=minC max ;

[0106]

[0107] In the formula, C max The maximum processing time is t; n, n1, and m are the total number of workpieces, process routes, and equipment, respectively; t ijk x represents the processing time of workpiece i on equipment k under process route j; ijk For binary decision variables, xef is 1 when workpiece i is executed on equipment k under process route j, and 0 otherwise; t id The current carbon emission factor; For in t ijk Operating power of processing equipment within the range.

[0108] The second level constructs an optimization model for energy consumption decisions aimed at minimizing energy costs. In specific implementation, the objective function of the second level is:

[0109]

[0110] In the formula, T is the total cycle time based on the processing task scheduling; and Let be the energy costs of grid power supply, PV, and ESS at time t, respectively. and These are the time-of-use electricity price of the power grid at time t, the levelized cost of photovoltaic power generation, the levelized cost of energy storage charging, and the levelized cost of energy storage discharging; xef t id This is the carbon emission factor as of today; P t grid P t pv P t ess,ch P t ess,dis These represent the electrical power purchased from the grid at time t, the electrical power output from the photovoltaic power generation system, the charging power of the energy storage system, and the discharging power of the energy storage system, respectively; t grid t pv t ess,ch t ess,dis These are the duration of grid power supply, the duration of photovoltaic power generation, the duration of energy storage system charging, and the duration of energy storage system discharging.

[0111] In this way, by constructing a multi-objective first-level optimization model that integrates processing efficiency, load balancing, and low-carbon operation, deep synergy between production planning and energy use is achieved, providing strong support for the efficient, balanced, and low-carbon comprehensive scheduling of energy-flexible manufacturing systems in complex operating environments.

[0112] By integrating grid power supply, photovoltaic power generation, and energy storage systems into a unified cost optimization framework, and comprehensively considering the real-time prices and operating characteristics of different energy sources, the energy usage structure can be dynamically adjusted. Compared to traditional methods that rely on a single energy source or static allocation, this approach significantly improves the economic efficiency of energy utilization and helps reduce the overall energy expenditure of manufacturing systems.

[0113] S4. During the real-time operation phase, acquire the actual energy supply data of each energy source and compare it with the day-ahead forecast data. When the actual energy supply cannot meet the load demand of the day-ahead plan, call the priority scores of each processing task in S2 to determine the adjustable processing tasks, and construct a multi-objective real-time optimization decision model with carbon emission management costs, delivery delay penalty costs, and demand response incentive benefits as optimization objectives. Dynamically adjust the start and stop or processing time of the adjustable processing tasks to realize online correction of the day-ahead plan.

[0114] In practical implementation, the objective function of the multi-objective real-time optimization decision model is:

[0115]

[0116]

[0117] In the formula, It is the cost of punishment; It is demand response revenue; It is the cost of carbon management; It is the incentive benefit for participating in demand response; xef t rt It is a real-time carbon emission factor; xef t id The carbon emission factor is as of today; ψ is a control variable that is further adjusted in real time to follow the random fluctuations in clean energy photovoltaic power.

[0118] ΔT is the total scheduling cycle time; Real-time grid power purchase cost; Real-time photovoltaic power generation cost; Cost of real-time energy storage systems; The unit price will be penalized for delays; For delay penalty factor; P t dr For demand response participation power; t dr Duration of demand response; For carbon management unit price; P t rt P represents real-time load power. t id This represents the planned load power for the day. Δt is the real-time control variable; Δt is the real-time optimization time step. P is the control variable for the current day plan; grid P pv P ess ,ch P ess,dis These represent the power consumed by grid power purchase, photovoltaic power generation, energy storage charging, and energy storage discharging, respectively; ψ rt (t+Δt|t) represents the predictive control variable for the future time t+Δt when performing rolling optimization at time t; ψ rt,0 (t) represents the initial control state; N Δt To optimize the number of steps for scrolling; Δu i (t+Δt) represents the load adjustment amount of the i-th adjustable task at time t+Δt.

[0119] This approach integrates grid costs, photovoltaic and energy storage operating costs, delivery delay penalties, carbon management costs, and demand response incentive benefits into a unified optimization framework, forming a multi-dimensional collaborative optimization mechanism. This avoids suboptimal solutions that might arise under a single objective (such as sacrificing delivery for carbon reduction), achieving a globally optimal balance under complex constraints. Furthermore, it introduces a real-time carbon emission factor, xef. t rt And compare it with the current carbon factor xef t id By performing interpolation calculations, the model can quantify the additional carbon emission costs caused by fluctuations in the energy structure. This mechanism prompts the system to prioritize the use of low-carbon energy in real-time and proactively avoid operating during high-carbon periods, thereby achieving precise management and continuous reduction of the carbon footprint.

[0120] In practice, the Model Predictive Control (MPC) algorithm is used for rolling optimization, and a feedback correction stage is introduced. The actual multi-energy output vector of the system at the current moment is used as the initial state for the rolling optimization at the next moment.

[0121] ψ rt,0 (t+1)=ψ rt,real (t);

[0122] In the formula, ψ rt,0 (t+1) is the initial control variable for the rolling optimization at the next time step, ψ rt,real (t) represents the actual multi-energy output vector of the system at the current moment, including grid power purchase, photovoltaic power generation, energy storage charging power, and discharge power.

[0123] By introducing a feedback correction mechanism of model predictive control (MPC), a shift from prediction-driven to state-driven control was achieved, significantly enhancing the dynamic response capability and control accuracy of the energy flexible manufacturing system under uncertain environments.

[0124] Unlike existing technologies that use fixed carbon emission factors, this solution constructs a comprehensive carbon emission factor model that considers the time-varying characteristics of the energy structure. This model can calculate the mixed carbon emission factor in real time based on the actual output ratio and unit carbon emission intensity of various energy sources (such as wind power and photovoltaics) at different times. This significantly improves the accuracy of carbon emission assessment and provides a scientific basis for low-carbon scheduling. Furthermore, existing methods typically treat manufacturing load as an unadjustable or uniformly adjustable whole, ignoring differences in processing tasks based on customer urgency, product process requirements, and equipment adaptability. This solution constructs a priority scoring model from three dimensions: customer needs, product characteristics, and equipment characteristics. This quantifies the adjustability and adjustment costs of each task, making real-time control more targeted and feasible, improving the overall system response capability while ensuring the fulfillment of critical tasks. Moreover, unlike traditional scheduling methods that only focus on a single time scale (such as only day-ahead or only real-time), this solution designs a two-stage control architecture: the day-ahead stage generates a coordinated energy use plan that balances processing efficiency, load balancing, and minimum carbon emissions based on forecast information; the real-time stage dynamically adjusts the plan using the priority model based on actual energy supply deviations. This "overall planning + local adjustment" mechanism effectively addresses the uncertainty of renewable energy output and load demand, significantly improving the robustness and practicality of the dispatch scheme. Furthermore, existing research often focuses solely on minimizing energy costs, neglecting the trade-off between carbon emission constraints and production reliability. This scheme simultaneously optimizes processing time, load balancing, and time-series carbon emissions during the day-ahead phase, and comprehensively considers carbon emission management costs, delivery delay penalties, and demand response incentives during the real-time phase. This forms a multi-objective decision-making system covering the entire chain, avoiding suboptimal or even conflicting results that may arise from traditional methods driven by a single objective.

[0125] This method enables energy-flexible manufacturing systems to make low-carbon, economical, and reliable collaborative decisions regarding energy consumption in production under uncertain supply and demand environments.

[0126] Example 2

[0127] To better understand this method, the following example is provided.

[0128] The dynamic carbon emission factors of the energy flexible manufacturing system are analyzed, and the time-series carbon emission factors of the energy flexible manufacturing system are determined based on the energy supply characteristics of the energy flexible manufacturing system.

[0129] Depend on Figure 2 It is evident that renewable energy output is significantly affected by weather conditions and exhibits high volatility, demonstrating the inherent uncertainty in production and energy consumption decisions based on clean energy forecast data. Combined with the constructed xef model, xef gradually decreases as clean energy PV output increases, reaching 0.2268 kg CO at night. 2e / kWh, during the day as PV output increases, xef gradually decreases, with a minimum of 0.1444kgCO. 2e / kWh. Therefore, a reasonable allocation of energy composition can effectively reduce carbon emissions. Furthermore, the charging and discharging of energy storage will also change xef, depending on whether the energy storage charging and discharging comes from clean energy or direct power supply.

[0130] Based on the NSGA-II algorithm logic, a two-level multi-objective optimization decision model is formulated to support low-carbon energy use decisions in energy-flexible manufacturing systems.

[0131] In the pre-processing phase, based on the aforementioned xef values, a higher-level scheduling strategy is formulated considering processing time, equipment load, and carbon emissions, resulting in a Pareto optimal frontier solution set, such as... Figure 3 As shown, the shortest processing time to complete the processing task is 23 hours, with carbon emissions of 8157.66 kg CO2e. The longest processing time is 34 hours, with carbon emissions of 7968.96 kg CO2e. The Pareto solution shows that carbon emissions gradually decrease with increasing processing time. Therefore, the shortest processing time is selected as the production operation scheduling scenario, as follows... Figure 3 As shown, a Gantt chart of the production operation is generated, with a processing time of 23 hours and a total energy consumption of 38468 kWh. Based on the production operation load data generated by the upper-level model, and with the goal of minimizing energy costs, a multi-energy supply scheme is generated, such as... Figure 3 As shown in the figure, blue represents grid power supply, yellow represents photovoltaic power supply, green represents energy storage discharge, and brown represents energy storage charging. It can be seen that the predicted PV output offsets part of the direct power supply output during periods of high electricity prices. Furthermore, when PV power generation decreases, during peak electricity price periods, the ESS (Energy Storage System) reduces the consumption and energy cost of direct power supply by discharging, with a total ESS discharge of 1009.58 kWh. The model prioritizes the consumption of clean energy, considering ESS power supply only when clean energy is exhausted. The total energy cost is 24471.01 CNY, and this model achieves minimum energy cost under carbon emission constraints.

[0132] To effectively and quickly analyze the products and processing equipment involved in demand response on the demand side, product demand response indicators, product characteristic indicators, and equipment characteristic indicators are constructed.

[0133] like Figure 4 As shown, a product demand response matrix is ​​constructed, with the customer demand indicator being delivery time (PD); product characteristic indicators being remaining processing time (PPR1), product output value (PPR2), and customer level (PPR3); and equipment characteristic indicators including equipment operating cost (MPR1), equipment energy consumption (MPR2), and equipment responsiveness (MPR3). This addresses the energy supply and demand imbalance caused by errors in clean energy forecasting.

[0134] To effectively and quickly analyze the products and processing equipment involved in demand response on the demand side, a product demand response matrix is ​​constructed. The customer demand indicator is delivery time (PD); product characteristic indicators include remaining processing time (PPR1), product output value (PPR2), and customer level (PPR3); equipment characteristic indicators include equipment operating cost (MPR1), equipment energy consumption (MPR2), and equipment responsiveness (MPR3). Furthermore, the weights of each indicator are obtained using the analytic hierarchy process (AHP), as shown in Tables 1, 2, and 3.

[0135] Table 1 Product Feature Weights

[0136]

[0137] Table 2 Equipment Feature Weights

[0138]

[0139] Table 3 Overall Priority Weights

[0140]

[0141] Based on the relevant data, it can be seen that the actual output of photovoltaic power begins to fall below the day-ahead forecast at 11:00. Therefore, it is necessary to transfer the adjustable production equipment at 11:00. Based on the day-ahead production Gantt chart, the equipment that can participate in demand response are M1 and M4, with a time scale of 11:00 to 13:00. The products they process are products J5-2 and J2-3. The demand response priority matrix for the two is constructed as shown in Table 4. The normalized data is shown in Table 5. Within the interval of 11:00 to 13:00, the processing tasks of equipment M4 need to be transferred to other time periods.

[0142] Table 4. Demand Response Priority Data (Points 11-13)

[0143]

[0144] Table 5. Demand Response Priority Data (Normalized) from Points 11-13

[0145]

[0146] The devices involved in demand response after 1 PM are M1 and M5, with a time scale from 1 PM to 2 PM. The products they process are products J1-3 and J7-2. The demand response priority matrix for both is constructed as shown in Table 6, and the normalized data is shown in Table 7. Within the 1 PM to 2 PM interval, the processing tasks of device M5 need to be transferred to other time periods.

[0147] Table 6. Demand Response Priority Data (Points 13-14)

[0148]

[0149] Table 7. Demand Response Priority Data (Normalized) for Points 13-14

[0150]

[0151] The devices involved in demand response after 2 PM are M4 and M6, with a time scale from 2 PM to 4 PM. The products they process are products J6-3 and J10-3. The demand response priority matrix for both is constructed as shown in Table 8, and the normalized data is shown in Table 9. Within the interval from 2 PM to 6 PM, the processing tasks of device M4 need to be transferred to other time periods.

[0152] Table 8. Demand Response Priority Data (Points 14-16)

[0153]

[0154] Table 9. Demand Response Priority Data (Normalized) from Points 14-16

[0155]

[0156] Based on the actual output data of photovoltaic power, there is no shortage of clean energy supply after 4 PM, which is consistent with the predicted data. Therefore, the optimal energy consumption decision for production can be made subsequently.

[0157] Secondly, considering the optimal management cost under the interaction of energy supply and demand, adjustable production equipment needs to be transferred at 11:00. Based on the production Gantt chart of the day, the equipment that can participate in demand response are M1 and M4, with a time scale of 11:00 to 13:00. The products they process are products J5-2 and J2-3. The demand response priority matrix of the two is constructed as shown in Table 5. The normalized data is shown in Table 4. Within the interval of 11:00 to 13:00, the processing tasks of equipment M4 need to be transferred to other time periods.

[0158] In the real-time phase, a multi-objective real-time optimization decision-making model is constructed, which considers carbon emission management costs, delay penalty costs, and demand response incentives.

[0159] During the real-time optimization phase, the production energy consumption plan is adjusted based on the above four strategies, such as... Figure 5 As shown, to reduce energy costs, energy storage charging and discharging were increased during peak electricity price periods (11:00-12:00 and 14:00-19:00), with energy storage discharge increasing by 350.62 kWh. Without affecting product processing time, maintaining the daily production plan for 23 hours, and without product delays, energy consumption was 39786.42 kWh, carbon emissions were 8585.29 kg CO2e, and carbon management costs were 368.20 CNY.

[0160] This method reveals the impact of source-load fluctuations on dynamic carbon emission factors and constructs a comprehensive carbon emission factor model that considers the time-varying nature of energy types. To quantify the priority of key equipment in energy-flexible manufacturing systems participating in real-time demand response, a demand response prioritization model is constructed from the perspectives of customer demand, product characteristics, and equipment characteristics, supporting multi-stage production energy consumption decisions in energy-flexible manufacturing systems.

[0161] In addition, the present invention proposes a two-stage optimization model for energy-flexible manufacturing systems that considers economics, environment, and delivery time. In the pre-production stage, carbon emissions are calculated based on dynamic carbon emission factor data. In the real-time stage, carbon emission management costs, delay penalty costs, and demand response incentives are considered. Production energy decisions are corrected in a timely manner through real-time supply and demand interaction and dynamic feedback mechanisms. This aims to reduce the deviation caused by the uncertainty of clean energy power generation and load, and improve the stability of production energy use in energy-flexible manufacturing systems.

[0162] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A two-stage control method for energy-flexible manufacturing systems considering demand response priorities, characterized in that: Includes the following steps: S1. Construct a comprehensive carbon emission factor model that considers the time-varying nature of energy types. This model is used to calculate the mixed carbon emission factor of the energy flexible manufacturing system based on the output ratio of various energy types and their corresponding unit carbon emission intensity in any given time period, so as to quantify the dynamic carbon emission intensity of the manufacturing system. S2. Construct a demand response priority model based on three dimensions: customer needs, product characteristics, and equipment characteristics. This model is used to calculate the demand response priority score for each processing task, thereby quantifying its adjustment priority in real-time demand response. S3. In the day-ahead phase, based on the day-ahead forecast data of renewable energy output and the grid time-of-use electricity price information, the mixed carbon emission factor for each time period of the next day is calculated using the comprehensive carbon emission factor model in S1. Based on the mixed carbon emission factor, a two-level production energy consumption decision model is constructed: the first level takes the minimum maximum processing time, the highest equipment load balance, and the minimum time-series carbon emissions as optimization objectives, and generates equipment processing time arrangements and corresponding load curves; the second level, based on the load curve, optimizes the coordinated energy consumption strategy of each energy source with the goal of minimizing energy costs. By jointly solving the two-level production energy consumption decision model using a multi-objective optimization algorithm, the day-ahead plan for the coordination of production and energy consumption for the next day is obtained. S4. During the real-time operation phase, acquire the actual energy supply data of each energy source and compare it with the day-ahead forecast data. When the actual energy supply cannot meet the load demand of the day-ahead plan, call the priority scores of each processing task in S2 to determine the adjustable processing tasks, and construct a multi-objective real-time optimization decision model with carbon emission management costs, delivery delay penalty costs, and demand response incentive benefits as optimization objectives. Dynamically adjust the start and stop or processing time of the adjustable processing tasks to realize online correction of the day-ahead plan.

2. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 1, characterized in that: The various energy sources include grid power supply, photovoltaic power generation, and energy storage discharge.

3. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 2, characterized in that: In S1, the integrated carbon emission factor model is as follows: v∈(grid, pv, ess); In the formula, xef t It is a mixture of carbon emission factors; It is the power output of energy v at time t; Let v be the emission factor of energy at time t; grid, pv, and ess represent grid power supply, photovoltaic power generation, and energy storage system, respectively.

4. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 1, characterized in that: In S2, the customer demand is characterized by the urgency of delivery, and the product characteristics and equipment characteristics each include multiple evaluation indicators; the priority score is obtained by weighted fusion of the scores from the three dimensions: In the formula, PR fin Priority scores: PD, PR, MRPR t These are the numerical values ​​for customer needs, product characteristics, and equipment characteristics, respectively; ω PD ω PR , PD, PR, MRPR respectively t The weight, and 5. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 1, characterized in that: In S3, the objective function of the first level is: min f1=min C max ; In the formula, C max The maximum processing time is t; n, n1, and m are the total number of workpieces, process routes, and equipment, respectively; t ijk x represents the processing time of workpiece i on equipment k under process route j; ijk For binary decision variables, xef is 1 when workpiece i is executed on equipment k under process route j, and 0 otherwise; t id The current carbon emission factor; For in t ijk Operating power of processing equipment within the range.

6. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 5, characterized in that: In S3, the objective function of the second level is: In the formula, T is the total cycle time based on the processing task scheduling; and Let be the energy costs of grid power supply, PV, and ESS at time t, respectively. and These are the time-of-use electricity price of the power grid at time t, the levelized cost of photovoltaic power generation, the levelized cost of energy storage charging, and the levelized cost of energy storage discharging; xef t id This is the carbon emission factor as of today; P t grid P t pv P t ess,ch P t ess,dis These represent the electrical power purchased from the grid at time t, the electrical power output from the photovoltaic power generation system, the charging power of the energy storage system, and the discharging power of the energy storage system, respectively; t grid t pv t ess,ch t ess,dis These are the duration of grid power supply, the duration of photovoltaic power generation, the duration of energy storage system charging, and the duration of energy storage system discharging.

7. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 6, characterized in that: In S4, the objective function of the multi-objective real-time optimization decision model is: Ψ=[P grid ,P pv ,P ess,ch ,P ess,dis ]; In the formula, It is the cost of punishment; It is demand response revenue; It is the cost of carbon management; It is the incentive benefit for participating in demand response; xef t rt It is a real-time carbon emission factor; xef t id The carbon emission factor is as of today; Ψ is the control variable, which is further adjusted in real time to follow the random fluctuations in clean energy photovoltaic power. ΔT is the total scheduling cycle time; Real-time grid power purchase cost; Real-time photovoltaic power generation cost; Cost of real-time energy storage systems; The unit price will be penalized for delays; For delay penalty factor; P t dr Power for demand response participation; t dr Duration of demand response; Carbon management unit price; P t rt P represents real-time load power. t id This represents the planned load power for the day. Δt is the real-time control variable; Δt is the real-time optimization time step. P is the control variable for the current day plan; grid P pv P ess,ch P ess,dis These represent the power consumed by grid power purchase, photovoltaic power generation, energy storage charging, and energy storage discharging, respectively; ψ rt (t+Δt|t) represents the predictive control variable for the future time t+Δt when performing rolling optimization at time t; ψ rt,0 (t) represents the initial control state; N Δt To optimize the number of steps for scrolling; Δu i (t+Δt) represents the load adjustment amount of the i-th adjustable task at time t+Δt.

8. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 7, characterized in that: In S4, the Model Predictive Control (MPC) algorithm is used for rolling optimization, and a feedback correction stage is introduced. The actual multi-energy output vector of the system at the current moment is used as the initial state for the rolling optimization at the next moment. ψ rt,0 (t+1)=ψ rt,real (t); In the formula, ψ rt,0 (t+1) is the initial control variable for the rolling optimization at the next time step, ψ rt,real (t) represents the actual multi-energy output vector of the system at the current moment, including grid power purchase, photovoltaic power generation, energy storage charging power, and discharge power.

9. The two-stage control method for an energy-flexible manufacturing system considering demand response priority as described in claim 1, characterized in that: In S3, the NSGA-II algorithm is used to jointly solve the two-level production energy consumption decision model.

Images