Energy optimization decision method, system, device and electronic equipment

Abstract

This invention relates to the field of energy management technology and discloses an energy optimization decision-making method, system, device, and electronic equipment. The method includes: acquiring multi-source heterogeneous data of the current production environment; performing reverse calculations using a pre-constructed digital twin model based on preset product quality qualification standards and multi-source heterogeneous data to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles; constructing a multi-constraint rolling optimization model based on preset constraints and time-varying lower limit constraints on process parameters, with the goal of minimizing comprehensive energy costs; and solving the multi-constraint rolling optimization model at the beginning of each optimization cycle to generate a sequence of optimization control instructions for future optimization cycles, which are then issued and executed. This invention solves the problem of energy optimization decision-making being disconnected from the quality control requirements of production processes, leading to optimization results violating key process constraints and thus affecting the final product quality. It achieves deep collaboration and closed-loop optimization between energy flow and production quality flow.

Description

Technical Field

[0001] This invention relates to the field of energy management technology, specifically to energy optimization decision-making methods, systems, devices, and electronic equipment. Background Technology

[0002] Currently, most battery manufacturers have deployed basic energy management systems to monitor the consumption of energy sources such as electricity, water, and gas. However, these systems generally suffer from the following inherent flaws: 1. Equipment energy consumption data, process parameters, environmental data, and grid price signals are scattered in different independent data subsystems. They are usually divided into energy supply (grid, photovoltaic), energy storage (energy storage system), and energy consumption (production equipment, environmental control) subsystems according to their energy roles, and are managed independently, forming data islands and control islands. They lack effective integration and cannot coordinate the overall situation or carry out multi-objective collaborative optimization. 2. Most data systems only have monitoring and reporting functions, and operate based on preset and fixed rules. They cannot respond in real time to dynamically changing electricity prices, fluctuating photovoltaic output, and emergency production tasks. They lack forward-looking prediction and optimization decision-making capabilities, and scheduling relies on human experience, resulting in delayed decision-making. 3. Most energy optimization decisions often only consider electricity costs, while ignoring their potential impact on core production processes (such as coating drying temperature and dew point in the formation workshop). Abruptly adjusting environmental settings or interrupting equipment in order to reduce electricity costs may lead to a decrease in battery yield and cause huge quality losses that far exceed the electricity cost savings.

[0003] In summary, in existing technologies, energy optimization decisions and quality control requirements of production processes are disconnected, which may lead to optimization results violating key process constraints and thus affecting the quality of the final product. Summary of the Invention

[0004] This invention provides an energy optimization decision-making method, system, device, and electronic device to solve the problem in the prior art where energy optimization decision-making and the quality control requirements of production processes are disconnected, which may lead to optimization results violating key process constraints and thus affecting the quality of the final product.

[0005] In a first aspect, the present invention provides an energy optimization decision-making method, the method comprising: Acquire multi-source heterogeneous data of the current production environment, including energy-side data, production task data, process parameters, and environmental data; Based on preset product quality qualification standards and multi-source heterogeneous data, a pre-built digital twin model is used for reverse calculation to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles. Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the goal of minimizing comprehensive energy cost. At the beginning of each optimization cycle, the multi-constraint rolling optimization model is solved to generate a sequence of optimization control instructions for future optimization cycles, which are then issued and executed.

[0006] This invention provides an energy optimization decision-making method that, by introducing a digital twin model of process quality and performing reverse calculations, dynamically transforms abstract product quality qualification standards into specific, quantifiable, and time-varying lower limit constraints of key process parameters. These constraints are then embedded as hard conditions into a rolling optimization model aimed at minimizing costs. This constructs a closed-loop decision-making system that achieves optimal comprehensive energy costs while strictly ensuring product quality does not exceed the lower limit. This solves the problem in existing technologies where energy optimization decisions and production process quality control requirements are disconnected, potentially leading to optimization results that violate key process constraints and thus affect final product quality. The method achieves deep synergy and closed-loop optimization between energy flow and production quality flow.

[0007] In one alternative implementation, the digital twin model is constructed in the following manner: Acquire historical production batch data, which includes process parameters, environmental parameters, and corresponding product quality data that have occurred. Based on the quality formation mechanism of the engineering process, a mathematical model framework is established with process parameters and environmental parameters as inputs and product quality data as output. Historical production batch data is used to calibrate the parameters to be determined in the mathematical model framework, resulting in a digital twin model of process quality with defined parameters. The digital twin model of process quality is used to describe the quantitative relationship between process parameters, environmental parameters and final product quality indicators.

[0008] The present invention provides an energy optimization decision-making method that integrates engineering mechanisms and historical data to establish a precise and quantifiable dynamic mathematical model between the final product quality, which is difficult to directly observe and control, and measurable process and environmental parameters. This provides key and reliable technical support for achieving the core objective of optimizing energy costs under the hard constraint of ensuring quality.

[0009] In one optional implementation, based on preset product quality acceptance standards and multi-source heterogeneous data, a pre-built digital twin model is used for reverse calculation to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles, including: Extract current real-time process parameters, future environmental parameter prediction sequences, and preset product quality qualification standards from multi-source heterogeneous data; The preset product quality qualification standards, current real-time process parameters, and future environmental parameter prediction sequences are input into the process quality digital twin model, and the lower limit sequence of process parameters that meet the quality requirements in the future optimization cycle is solved by reverse calculation. Based on the sequence of lower limit values ​​of process parameters, time-varying lower limit constraints of process parameters are generated to constrain product quality in future optimization cycles for subsequent energy optimization decisions.

[0010] This invention provides an energy optimization decision-making method that dynamically and accurately transforms static, abstract product quality qualification standards into a sequence of specific, time-varying key process parameter lower limits for a future period by combining a digital twin model of process quality with real-time and predictive data. This ensures at the source of decision-making that any scheduling scheme aimed at energy saving and cost reduction will not sacrifice product quality, fundamentally solving the key contradiction of difficulty in coordinating economic efficiency and quality assurance during the optimization process.

[0011] In one optional implementation, the sequence of lower limit values ​​of process parameters that meet quality requirements within future optimization cycles is solved by reverse calculation, including: The preset product quality qualification standards are transformed into constraint inequalities for the output results of the digital twin model of process quality; For each time period of the future optimization cycle, the constraint inequalities are mathematically solved to obtain the lower limit of the process parameters that meet the quality requirements within the future optimization cycle. Arrange the lower limit values ​​of the process parameters obtained from all time steps in time sequence to generate a sequence of lower limit values ​​of process parameters.

[0012] This invention provides an energy optimization decision-making method that transforms abstract quality standards into precise mathematical inequalities and solves them independently for each future decision moment. This achieves refined and time-varying requirements for process parameters. Furthermore, through serialized output, it provides directly embeddable, time-varying hard constraints for rolling optimization, realizing forward-looking, full-process, and computable quality control.

[0013] In one optional implementation, the preset constraints include: production task constraints, energy supply and demand balance constraints, and energy storage system operation constraints. Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the objective of minimizing the overall energy cost, including: The total energy cost is determined by summing energy purchase cost, carbon emission cost, and equipment depreciation cost, with the optimization objective being to minimize the total energy cost over a future cycle. By integrating production task constraints, energy supply and demand balance constraints, energy storage system operation constraints, and time-varying lower limit constraints of process parameters, a set of constraints is obtained. Based on the optimization objective and constraint set, a multi-constraint rolling optimization model is constructed to solve the problem at the beginning of each cycle. The multi-constraint rolling optimization model is a mixed integer linear programming model.

[0014] This invention provides an energy optimization decision-making method that systematically integrates the time-varying lower limit constraints of process parameters representing quality red lines with the physical, economic, and task constraints of traditional energy systems, constructing a unified mixed-integer linear programming model. Mathematically, this method achieves collaborative modeling and joint solution of quality assurance requirements and multi-dimensional energy and economic objectives (cost, carbon emissions, equipment wear and tear). This allows optimization decisions to simultaneously and automatically satisfy the strict quality boundaries of production processes and all operational limitations of complex energy systems within a unified framework. Therefore, from the model's root, it ensures that the final scheduling scheme, under the premise of technical feasibility and task achievability, is both economically optimal and quality reliable.

[0015] In one alternative implementation, the method further includes: Collect actual process operation data and actual product quality data after the execution of optimized control commands; Based on actual process operation data, a digital twin model is used for forward calculation to obtain predicted values ​​of product quality data; The deviation between the actual product quality data and the predicted product quality data is calculated. When the deviation continues to exceed a preset threshold, an online learning algorithm is triggered to update the initial parameters of the pre-built digital twin model for decision-making in the next optimization cycle.

[0016] This invention provides an energy optimization decision-making method that, by introducing a closed-loop feedback mechanism of execution-monitoring-learning, enables the system to continuously self-optimize. This method dynamically triggers a learning algorithm based on the deviation between actual production data and model predictions, thereby automatically correcting the parameters of the digital twin model and effectively adapting to the uncertainties and time-varying nature of real-world production, such as the slow decline in equipment performance and fluctuations in material properties.

[0017] In one alternative implementation, the online learning algorithm employs recursive least squares, with the update formula as follows: θ_new=θ_old+K×(y_actual-y_pred); Where θ represents the parameters of the digital twin model to be updated, θ_new represents the parameters of the updated digital twin model, θ_old represents the initial parameters of the digital twin model before the update, K represents the algorithm gain matrix, y_actual represents the actual product quality index, and y_pred represents the predicted value of the product quality data.

[0018] Secondly, the present invention provides an energy optimization decision-making system, the system comprising: The perception layer is used to acquire multi-source heterogeneous data of the current production environment, including energy-side data, production task data, process parameters, and environmental data. The decision-making layer, connected to the perception layer, is used to perform reverse calculations based on preset product quality qualification standards and multi-source heterogeneous data, using a pre-built digital twin model to generate time-varying lower limits of process parameters for product quality in future optimization cycles. Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the goal of minimizing comprehensive energy costs. At the beginning of each optimization cycle, the multi-constraint rolling optimization model is solved to generate a sequence of optimization control instructions for future optimization cycles, which are then issued for execution.

[0019] Thirdly, the present invention provides an energy optimization decision-making device, the device comprising: The data acquisition module is used to acquire multi-source heterogeneous data of the current production environment, including energy-side data, production task data, process parameters and environmental data. The reverse calculation module is used to perform reverse calculations based on preset product quality qualification standards and multi-source heterogeneous data, using a pre-built digital twin model, to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles. The optimization model building module is used to construct a multi-constraint rolling optimization model based on preset constraints and time-varying lower limits of process parameters, with the goal of minimizing the overall energy cost. The solver module is used to solve the multi-constraint rolling optimization model at the beginning of each optimization cycle, generate the optimization control instruction sequence for future optimization cycles, and issue it for execution.

[0020] Fourthly, the present invention provides an electronic device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the energy optimization decision-making method of the first aspect or any corresponding embodiment described above.

[0021] Fifthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the energy optimization decision-making method of the first aspect or any corresponding embodiment thereof.

[0022] In a sixth aspect, the present invention provides a computer program product, including computer instructions for causing a computer to execute the energy optimization decision-making method of the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0023] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the first process of the energy optimization decision-making method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a second process for the energy optimization decision-making method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the third process of the energy optimization decision-making method according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the execution layer in the energy optimization decision-making method according to an embodiment of the present invention; Figure 6 This is a structural block diagram of an energy optimization decision-making device according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0027] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0028] As an optional application scenario of this invention, such as Figure 1 As shown, this is the energy optimization decision-making system provided by the present invention. The system includes: a perception layer, a decision-making layer, an execution layer, and a learning and display layer; wherein: The perception layer 110 is used to collect multi-source heterogeneous data on manufacturing process, energy consumption, process quality and environmental status in real time, and to perform edge processing and protocol unification. The decision layer 120, connected to the perception layer, is used to receive and fuse multi-source data, generate inverse constraints based on a pre-built energy-process digital twin model, and use a model predictive control framework to solve multi-objective optimization problems in a rolling manner to generate the optimal control instruction set. The energy-process digital twin model is used to describe the quantitative relationship between process parameters, environmental parameters and final product quality indicators, and the inverse constraint generation is used to transform product quality requirements into dynamic constraints on process parameters in the future time domain. The execution layer 130, connected to the decision layer, receives the optimal control instruction set and converts it into executable instructions for each controlled device. It consists of local controllers from various subsystems, such as the energy storage PCS controller and the oven PLC. They receive current-moment instructions from the decision layer, such as… , And convert it into device-level control signals for execution.

[0029] The learning and presentation layer 140, connected to the decision-making and execution layers, monitors the system's operational status, displays optimization effects, and adaptively updates the key parameters of the digital twin model based on actual operational feedback data. The learning and presentation layer 140 includes an adaptive learning engine and a visualization interface. After each production batch, the learning engine compares the measured final moisture content MC_actual with the predicted moisture content MC_pred. If the error continues to exceed the limit, the digital twin parameter k is updated online using a recursive least squares method. The visualization interface displays real-time load, optimization strategies, cost savings, and quality pass rate dashboards to management personnel.

[0030] The perception layer 110 is responsible for collecting multi-source heterogeneous data across the entire plant. Its components include: smart meters installed at the grid inlet to collect power purchase and sales (P_grid(t)) and electricity consumption; a photovoltaic inverter communication interface to collect real-time photovoltaic output (P_pv(t); an energy storage battery management system to collect the energy storage state of charge (SOC(t)) and maximum power (Pess_max); dew point sensors and temperature / humidity sensors deployed in the workshop to collect ambient dew point temperature (T_dew(t)) and ambient temperature (T_amb(t)); PLCs for key process equipment (such as ovens) to collect their set temperature (T_set(t), actual temperature (T_actual(t)), and operating power (P_oven(t)); and data interfaces with the manufacturing execution system and plant monitoring system to obtain production task scheduling and basic load information. All data is parsed and preliminarily cleaned through the industrial IoT gateway before being uploaded at fixed intervals (e.g., every minute).

[0031] Decision layer 120 is the core computing unit of the system, including: The multi-dimensional data fusion center is used to receive and align multi-source heterogeneous data from the perception layer in a time sequence to form a unified historical-real-time database. The multi-source heterogeneous data includes energy-side data, production task data and key process parameters, environmental parameters and equipment operation data. The energy-process digital twin, comprising a mechanistic model and a data-driven model, is used for forward simulation of process parameters and energy consumption, and for generating inverse constraints, outputting a dynamic lower bound constraint sequence for process parameters. Specifically, when generating inverse constraints, the energy-process digital twin is used for: The system acquires the completed process parameter history, environmental parameter history, and future environmental parameter predictions for the current production batch. Based on the mechanistic model and the final product quality index thresholds that must be met, it reverse-calculates the time-varying lower limits that key process parameters must meet in the remaining production period to ensure quality compliance, generating a dynamic lower limit constraint sequence, i.e., the time-varying lower limit constraints of process parameters to ensure product quality in future optimization cycles. Taking electrode drying as an example, its mechanistic model is: MC_pred=MC_0×exp(-k×Σ(T_set(i) -T_dew(i)) / δ×Δt). The model parameters (k,MC0) are calibrated using historical data. Its core function is to perform reverse constraint generation: based on the quality index MC_spec, the already occurred process data, and the future environmental predictions, it reverse-calculates the lower limit curve T_min(t) that the future oven temperature must meet. MC_pred represents the predicted moisture content, and MC_0 is the initial moisture content.

[0032] The prediction model module is used to predict renewable energy output, energy prices, and base load in the future time domain; it uses time series or machine learning methods to predict photovoltaic output P_pv_pred(t), time-of-use electricity price Price_pred(t), and base load P_base_pred(t) in the future time domain.

[0033] The optimization algorithm engine is used to construct and continuously solve a mixed-integer linear programming model with the goal of minimizing the overall cost, based on the dynamic lower bound constraint sequence, prediction information and the current state of the system, and output the optimal control instruction set including process setpoints and energy storage charging and discharging power.

[0034] The core of the optimization decision-making in this invention is a mixed-integer linear programming model constructed and solved at the beginning of each rolling cycle (e.g., every 15 minutes). Its working principle is: within the "feasible region" jointly defined by dynamic quality constraints, production tasks, and physical system limitations generated by the digital twin model, a set of equipment control command sequences that minimizes the overall energy cost over a future prospective period is sought.

[0035] The specific construction steps are as follows: First, define the decision-making period. Divide the future optimization prospect period, for example, 4 hours, into multiple consecutive time periods, such as 16 15-minute time periods. Number each time period sequentially.

[0036] Secondly, declare the decision variables. Define a series of mathematical variables to represent the actions or states of all controllable devices in each time period, such as the power purchased from the grid in each time period, the power of the energy storage system charging or discharging, and the setpoints of key process equipment.

[0037] Next, we construct the objective function. The overall energy cost that needs to be minimized is expressed as a linear function of all the decision variables mentioned above.

[0038] Next, formalize the constraints. Express all the pre-defined physical limitations, operational requirements, production tasks, and quality constraints as linear equations or inequalities between these decision variables.

[0039] Finally, integration and solution. The objective function and all constraints are combined to form a standard mixed-integer linear programming problem, and a professional mathematical solver is called to perform the calculation, thereby obtaining the optimal control command sequence.

[0040] The Mixed-Integer Linear Programming (MILP) model described in this invention defines the following decision variables to characterize the state or action of each controllable device in different time periods: Among them, continuous variables include: Pbuy ( t ): Indicates a time period t Active power purchased from the power grid, measured in kilowatts (kW). Psell ( t ): Indicates a time period t Active power delivered to the power grid, measured in kilowatts (kW). Pch ( t ): Indicates a time period t The charging power of an energy storage system is measured in kilowatts (kW). Pdis ( t ): Indicates a time period t The discharge power of an energy storage system is measured in kilowatts (kW). SOC ( t ): Indicates a time period t The state of charge of the energy storage system at the end, in percentage (%). Tset ( t ): Indicates a time period t The set temperature of key process equipment (such as ovens) is in degrees Celsius (°C). Poven ( t ): This is an intermediate variable, representing the time period. t Energy consumption of key process equipment (such as ovens), expressed in kilowatts (kW).

[0041] Binary variables (0 / 1 variables) include: uch ( t ): Indicates a time period t Whether the energy storage system is in a charging state; a value of 1 indicates that it is charging. udis ( t ): Indicates a time period t Whether the energy storage system is in a discharging state, a value of 1 indicates that it is discharging.

[0042] These variables together constitute the adjustable dimension in the optimized scheduling. By optimizing their values, the overall energy cost can be minimized.

[0043] The objective function of the MILP model described in this invention aims to minimize the total integrated cost over the next T time periods. C Total. This total cost includes the cost of energy purchases. C ele, carbon emission costsC Carbon and depreciation costs of energy storage devices C The degrade function is composed of three parts, which are summed over time.

[0044] Its mathematical expression is: Minimize the total cost C The total is equal to the sum of the three costs mentioned above for each time period t, multiplied by the duration Δ of each time period. t .

[0045] The specific calculation methods for each part of the cost are as follows: 1. Energy purchase cost (Cele(t)) C ele( t The electricity consumption expenditure consists of two parts: electricity purchase expenses and electricity sales revenue. The calculation formula is as follows: C ele( t )= π buy( t ) P buy( t ) π sell( t ) P sell( t ).

[0046] in, π buy( t The unit price (in yuan / kWh) represents the price of electricity purchased from the grid during time period t.

[0047] P buy( t ) represents the power purchased during time period t (unit: kilowatt).

[0048] π sell( t () represents the unit price of electricity sold to the grid during time period t (unit: yuan / kWh).

[0049] P sell( t ) represents the electricity sales power (unit: kilowatt) during time period t.

[0050] 2. Carbon emission costs C carbon ( t This involves converting the carbon emissions associated with electricity purchases into economic costs. The calculation formula is as follows: C carbon ( t )= λ γ grid P buy( t ).

[0051] in, λ The trading price of a unit of carbon emission rights (e.g., per kilogram of carbon dioxide) (unit: yuan / kilogram of carbon dioxide).

[0052] γ Grid represents the average carbon emission factor of grid electricity, which is the amount of carbon dioxide emissions indirectly generated for every kilowatt-hour of grid electricity consumed (unit: kilograms of carbon dioxide / kilowatt-hour).

[0053] 3. Depreciation cost of energy storage equipment C degrade( t ): Quantify the lifespan loss cost of energy storage devices due to charging and discharging operations. The calculation formula is as follows: C degrade( t )= η ch P ch( t )+ η dis P dis( t ).

[0054] in, η ch and η dis represents the cycle life loss cost coefficient (unit: yuan / kilowatt) corresponding to the unit power of charging and discharging of the energy storage system.

[0055] P ch( t )and P dis( t ) represent the charging and discharging power of energy storage during time period t (unit: kilowatt).

[0056] In summary, this objective function achieves precise optimization of the overall efficiency of system operation by uniformly quantifying economic costs (electricity costs and equipment depreciation) and environmental costs (carbon emissions) into monetary form and jointly minimizing them.

[0057] The MILP model of this invention includes the following key preset constraints: (a) Process quality constraints (dynamic output from the digital twin model): This is the core constraint of the invention, which transforms abstract product quality requirements into hard numerical constraints on key process variables that change over time.

[0058] The mathematical expression of this constraint is: in each time period of the optimization prospect period The set temperature of key process equipment (such as ovens) set( It must be greater than or equal to a corresponding dynamic lower limit value. min( This lower limit value sequence. min( The lower bound is not a fixed value, but rather a sequence of known parameters generated in real time by the decision-making level based on the latest environmental forecasts and product quality standards, using a digital twin model of process quality, before each optimization calculation. For example, in a practical implementation, this dynamic lower bound sequence can be approximated using a piecewise constant approach.

[0059] (b) Energy-process coupling constraints: This constraint establishes a quantitative relationship between process setpoints and equipment energy consumption, serving as a crucial bridge connecting process decisions and energy consumption.

[0060] Its mathematical relationship can be expressed as follows: The process equipment (taking an oven as an example) in... Energy consumption during a given time period oven ), and its set temperature set( ) and ambient temperature amb( There is a linear relationship between them.

[0061] Specifically, energy consumption equals a coefficient. (Unit: kilowatts / degrees Celsius) Multiply by the difference between the set temperature and the ambient temperature, plus a constant term. (Unit: kilowatt). The parameters included... and It was obtained by analyzing the historical operating data of the equipment and fitting it using regression analysis, thereby ensuring that the model can accurately reflect the actual energy consumption characteristics of the specific equipment.

[0062] The MILP model of this invention also includes the following two key operational constraints: (c) Production task constraint: This constraint is used to ensure that the optimized energy scheduling scheme can meet the established production plan and guarantee the completion of production tasks.

[0063] The mathematical expression of this constraint is: throughout the entire optimization prospect period, the total energy provided by key process equipment (such as ovens) must not be less than the total energy required to complete the current production batch.

[0064] Specifically, the device will be used in each time period t Energy consumption P oven t Multiply by the time period length Δt The total energy consumed by the equipment is calculated by summing the values ​​over all time periods. This total energy value must be greater than or equal to a preset minimum energy value required to complete production. E The requirement (in kilowatt-hours) translates the production target into a rigid requirement for system operation from the perspective of total energy consumption.

[0065] (d) System power balance constraint (core physical constraint): This constraint is the core physical law that the entire energy system must abide by, requiring that at any time, the total power consumption of the plant must be in instantaneous balance with the total power supply.

[0066] Its mathematical equation is described as follows: For each time period t Power purchased from the power grid P buy( t Subtract the power sent to the grid P sell( t In addition, the predicted power generation of the photovoltaic system P pvpred( t and the discharge power of the energy storage system P dis( t Their sum must equal the base load forecast value for the same period. P basepred( t Energy consumption of process equipment P oven t and the charging power of the energy storage system P ch( t The sum of ).

[0067] Among them, photovoltaic power generation is predicted. P pvpred( t and basic load forecast P basepred( t The input parameters are derived in advance from weather forecasts and historical data. This equation ensures the real-time balance between power supply and demand, and is the physical basis for the physical feasibility of all optimization schemes.

[0068] In the MILP model of this invention, the operation of the energy storage system must comply with the following physical and operational constraints, which together ensure the safety and rationality of the scheduling scheme's use of energy storage devices: (e) Energy storage system operating constraints: 1. Power limitation: The energy storage system at any given time... t charging power Pch ( t and discharge power Pdis ( tNeither of the values ​​can be negative, and they must not exceed the maximum value allowed by the device. The upper limit of charging power is determined by the binary variable of the charging state. uch ( t ) and maximum charging power Pch The product of max and max determines the upper limit of discharge power; similarly, the upper limit of discharge power is determined by the discharge state variables. udis ( t ) and maximum discharge power Pdis The product of max determines the power. This means that the charging / discharging power can only be greater than zero when the corresponding state variable is "1" (i.e., active state).

[0069] 2. State Mutual Exclusion Constraint: Energy storage systems cannot simultaneously charge and discharge within the same time period. This requirement is guaranteed by an inequality constraint: charging state variables... uch ( t ) and discharge state variables udis ( t The sum of all variables must be less than or equal to 1. This directly introduces 0-1 integer variables, which is one of the core reasons why the model is a mixed integer programming problem.

[0070] 3. Capacity (State of Charge, SOC) Dynamic Constraints: This part includes a state update equation and a range constraint.

[0071] Dynamic equation: time period t t State of charge of stored energy at the end SOC ( t ), which is equal to the state at the previous moment. SOC ( t 1) This includes the change in net stored energy during this period. This change is contributed by charging (charging power multiplied by charging efficiency). ηch (later converted to energy increment) and discharge contribution (discharge power divided by discharge efficiency) ηdis The amount is determined jointly by the energy reduction (later converted to energy savings) and then divided by the rated energy storage capacity. E rated and multiplied by the time period length Δ t get.

[0072] Range limitation: State of charge of energy storage at all times. SOC ( t It must be maintained within the preset safe operating range, that is, not lower than the minimum limit. SOC min and not higher than the maximum limit. SOC Maximize the initial value to ensure battery life and system safety. SOC (0) is for optimizing the actual measured values ​​at the beginning.

[0073] The MILP model of this invention also includes the following two important physical and security constraints: (f) Grid and Transformer Capacity Constraints: This constraint limits the upper limit of power exchange between the plant and the grid, ensuring that the dispatching scheme complies with the hardware and protocol requirements for grid connection. Specifically, it states that at any given time period... t Power purchased from the power grid P buy( t The value must be non-negative and cannot exceed the maximum allowable capacity of the factory's transformer. P transmax; simultaneously, the electricity sold back to the grid. P sell( t It must also be a non-negative value and cannot exceed the maximum return power specified in the grid connection protocol. P feed-inmax.

[0074] (g) Process Equipment Safety Constraints: This constraint sets a static upper limit for safe operation of process equipment to prevent damage or safety issues caused by overload. Specifically, it states that at any given time period... t The set temperature of key process equipment (such as ovens) T set( t The maximum safety setting allowed by the device itself must not be exceeded. T setmax. This constraint is the same as the dynamic lower bound constraint generated by the digital twin model. T set( t )≥ T min( t Together, these constitute a complete safe operating window for process settings.

[0075] For example, for each period (e.g., 15 minutes), a mixed-integer linear programming model with a prospective period (e.g., 4 hours) is constructed. The objective function is to minimize the energy purchase cost (i.e., total electricity cost) and carbon cost. Decision variables include the power purchased and sold, the charging and discharging power and state of energy storage, and the oven set temperature, etc.

[0076] In process manufacturing industries such as lithium batteries, chemicals, and pharmaceuticals, energy purchase costs account for a high proportion of total operating costs, and energy consumption in the production process is closely coupled with product quality (such as purity, moisture content, and yield). Traditional energy management systems mainly focus on load monitoring and simple demand management, independent of production planning and process control, forming information silos. This often leads to neglecting process constraints during energy optimization, potentially compromising product quality; while process control, in order to ensure quality, often adopts conservative and fixed parameter settings, sacrificing energy efficiency.

[0077] While some research in related technologies combines production scheduling with energy management, it primarily focuses on optimization at the planning level, lacking the ability to respond to real-time fluctuations at the second / minute level, and failing to establish a precise and quantifiable dynamic correlation model between "process parameters-energy consumption-quality". Systems typically use fixed safety ranges for process parameters as constraints, unable to dynamically adjust according to real-time environmental changes (such as temperature and humidity) and specific quality targets. This results in low optimization freedom and makes it difficult to fully explore the potential for flexible adjustments such as absorbing renewable energy and utilizing time-of-use pricing while strictly ensuring quality.

[0078] This invention provides an energy optimization decision-making method. Through the MCP (Manufacturing, Consumption, Process) collaborative mechanism, which corresponds to the closed-loop linkage of manufacturing task (M-Mission), energy cost (C-Cost), and process (P-Process), a closed-loop decision-making system is constructed to achieve the optimal comprehensive energy cost under the premise of strictly ensuring that product quality does not exceed the lower limit. This solves the problem in the prior art where energy optimization decision-making and production process quality control requirements are mutually isolated, which may lead to optimization results violating key process constraints and thus affecting the final product quality. This method achieves deep collaboration and closed-loop optimization between energy flow and production quality flow.

[0079] According to an embodiment of the present invention, an embodiment of an energy optimization decision-making method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0080] This embodiment provides an energy optimization decision-making method, which can be used in the aforementioned energy optimization decision-making system. Figure 2 This is a flowchart of an energy optimization decision-making method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain multi-source heterogeneous data of the current production environment. The multi-source heterogeneous data includes energy-side data, production task data, process parameters, and environmental data.

[0081] Specifically, the sensing layer devices continuously collect data from sources, networks, loads, storage, processes, and the environment. The industrial IoT gateway cleans and converts the data according to protocols, uploading it to the multi-dimensional data fusion center at the decision-making layer every one minute. The multi-dimensional data fusion center adds a unified timestamp to all data and stores it, providing a data foundation for subsequent modeling and optimization.

[0082] Step S202: Based on the preset product quality qualification standards and multi-source heterogeneous data, a pre-built digital twin model is used for reverse calculation to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles.

[0083] Specifically, the pre-built digital twin model is based on drying kinetics and mass transfer mechanism, and establishes an energy-process digital twin model with process equipment set temperature and environmental dew point as input and core product quality indicators as output.

[0084] Based on the energy-process digital twin model and the current production status, the decision-making level performs reverse constraint generation, transforming product quality requirements into dynamic process parameter constraints in the future time domain.

[0085] Step S203: Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the goal of minimizing the overall energy cost.

[0086] Specifically, the decision-making level aims to minimize the sum of energy purchase cost, carbon cost and equipment depreciation cost, and establishes a mixed integer linear programming rolling optimization model that includes energy equipment operation constraints, production task constraints and time-varying lower limit constraints of key process parameters generated in step S202.

[0087] Step S204: At the beginning of each optimization cycle, solve the multi-constraint rolling optimization model, generate the optimization control instruction sequence for future optimization cycles, and issue it for execution.

[0088] Specifically, at the beginning of each optimization cycle, the rolling optimization model is solved by combining the latest prediction data and real-time status, and the optimization instruction sequence of energy storage charging and discharging power and adjustable process equipment setpoints for the next cycle is output; and the optimization instruction sequence is sent to the execution layer to execute the optimization instruction sequence.

[0089] The energy optimization decision-making method provided in this embodiment introduces a digital twin model of process quality and performs reverse calculations to dynamically transform abstract product quality qualification standards into specific, quantifiable, and time-varying lower limit constraints of key process parameters. This constraint is then embedded as a hard condition into a rolling optimization model aimed at minimizing costs. This constructs a closed-loop decision-making system that achieves optimal comprehensive energy costs while strictly ensuring that product quality does not exceed the lower limit. This solves the problem in existing technologies where energy optimization decisions and production process quality control requirements are disconnected, leading to optimization results that may violate key process constraints and thus affect the final product quality. It achieves deep synergy and closed-loop optimization of energy flow and production quality flow.

[0090] This embodiment provides an energy optimization decision-making method, which can be used in the aforementioned energy optimization decision-making system. Figure 3This is a flowchart of an energy optimization decision-making method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps: Step S301: Obtain multi-source heterogeneous data from the current production environment. For details, please refer to [link to relevant documentation]. Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0091] Step S302, the digital twin model is constructed in the following manner: Step S3021: Obtain historical production batch data, which includes process parameters, environmental parameters, and corresponding product quality data that have occurred.

[0092] Specifically, process parameters are the sequence of operating setpoints or actual measured values ​​of key adjustable equipment (such as coating machines and ovens) over time during the production process of this batch. For example, the set temperature T_set(t) of the oven at each moment.

[0093] Environmental parameters are the sequence of changes in the workshop environmental conditions during the production of this batch over time. For example, the real-time ambient dew point temperature T_dew(t) in dry air.

[0094] Product quality data refers to the final quality indicators obtained through laboratory testing or online measurement after the production of this batch of products is completed. For example, the measured final moisture content MC_actual of this batch of electrodes.

[0095] Step S3022: Based on the quality formation mechanism of the engineering process, establish a mathematical model framework with process parameters and environmental parameters as inputs and product quality data as outputs.

[0096] Specifically, the physicochemical nature of the target production process is analyzed. For example, in the battery electrode drying process, based on drying kinetics and the principle of heat-mass transfer, the core driving factor affecting the final moisture content is determined to be the difference between the material surface temperature and the ambient dew point temperature. Based on this, a mechanistic equation (such as a differential or difference equation in exponential decay form) is constructed as a mathematical model framework. This equation explicitly defines the functional relationship between input variables (such as the oven set temperature T_set(t) and the dew point temperature T_dew(t)) and the output variable (predicted final moisture content MC_final): MC_final = f(T_set(t), T_dew(t), Θ), where Θ represents a set of undetermined physical parameters (such as the drying rate coefficient k and the initial moisture content MC_0). This framework encodes engineering knowledge into a computable mathematical structure, providing a physically interpretable computational blueprint for quality prediction.

[0097] For example, the mechanism model establishes the relationship between oven energy consumption and temperature and ventilation volume, or the quantitative relationship model of electrode moisture content, based on thermodynamic principles.

[0098] Taking the battery baking process as an example, the mechanism model for the battery baking process is as follows: E_oven(t)=ρ·c_p·V·(dT / dt)+h·A·(T_in-T_out)+P_fan·η_fan (1); Where E_oven(t) is the energy consumption of the oven at time t, ρ is the air density, c_p is the specific heat capacity, V is the cavity volume, T_in and T_out are the internal and external temperatures, h is the heat transfer coefficient, A is the heat transfer area, P_fan is the fan power, and η_fan is the efficiency.

[0099] The mechanism of moisture content after electrode drying can be represented by the following model: MC_final=MC_0×exp(-k×Σ(T_set(t)-T_dew(t)) / δ×Δt)(2); Where MC_final is the predicted final moisture content, MC_0 is the initial moisture content, k is the drying rate coefficient, T_set(t) is the oven set temperature at time t, T_dew(t) is the ambient dew point temperature at time t, δ is the electrode coating thickness, and Δt is the time step.

[0100] Step S3023: The parameters to be determined in the mathematical model framework are calibrated using historical production batch data to obtain a process quality digital twin model with determined parameters; the process quality digital twin model is used to describe the quantitative relationship between process parameters, environmental parameters and final product quality indicators.

[0101] Specifically, the mechanistic framework is combined with historical data from a specific production line, and model instantiation is completed through parameter identification. The process involves collecting sufficient historical production batch data, with each batch containing a complete input sequence {T_set_hist(t), T_dew_hist(t)} and the corresponding final measured quality value MC_actual. This data is then substituted into the model framework f, forming a fitting problem with the parameter set Θ as the optimization variable. Optimization algorithms such as nonlinear least squares are employed to minimize the sum of squared errors between the model's predicted value f(Θ) and the measured value MC_actual, thereby solving for the optimal parameter values ​​that best represent the actual behavior of the production line. Through this calibration, a system containing specific numerical parameters is created. A digital twin model of process quality that can perform positive quality prediction and reverse constraint solution has been constructed.

[0102] Step S303: Based on the preset product quality qualification standards and multi-source heterogeneous data, a pre-built digital twin model is used for reverse calculation to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles.

[0103] Specifically, step S303 includes: Step S3031: Extract the current real-time process parameters, the predicted sequence of future environmental parameters, and the preset product quality qualification standards from the multi-source heterogeneous data.

[0104] Specifically, this step involves targeted filtering and structured organization of the real-time data pool. From the continuously incoming raw data stream (such as sensor readings from the equipment PLC, task status from the MES, and forecast data from the weather forecast interface), three types of key information are identified and extracted in real time: 1) current real-time process parameters reflecting the current production status (such as the current measured temperature of the oven); 2) predicted sequences of future environmental parameters characterizing future external disturbances (such as the predicted dew point temperature for the next few hours); and 3) preset product quality qualification standards as a prerequisite for optimization (such as the upper limit of moisture content). This process transforms massive, heterogeneous data into time-aligned information prepared for subsequent model calculations, providing accurate and complete input conditions for reverse calculations.

[0105] Step S3032: Input the preset product quality qualification standard, the current real-time process parameters and the future environmental parameter prediction sequence into the process quality digital twin model, and solve the process parameter lower limit value sequence that meets the quality requirements in the future optimization cycle through reverse calculation.

[0106] Specifically, this step involves performing a model-based dynamic inverse solution. The extracted information (quality standards, current state, future predictions) is used as input to call the calibrated digital twin model of process quality. The core operation is to solve a mathematical problem in reverse for each time step within the future optimization cycle, based on the quantitative positive relationship between process / environmental parameters and product quality described by the model: Under given (or predicted) environmental conditions, what are the minimum values ​​of the process parameters at the current time step to meet the final product quality standards? By traversing all future time steps and solving this problem, the system outputs a sequence of lower limit values ​​consisting of the minimum allowable values ​​of the process parameters at each time point. Essentially, this decomposes the global, final quality objective into a series of local, process-oriented, and dynamic control boundaries.

[0107] In some optional implementations, step S3032 above includes: Step a1: Transform the preset product quality qualification standard into a constraint inequality on the output of the digital twin model of process quality.

[0108] Step a2: For each time period of the future optimization cycle, solve the constraint inequalities mathematically to obtain the lower limit of the process parameters that meet the quality requirements within the future optimization cycle.

[0109] Step a3: Arrange the lower limit values ​​of the process parameters obtained from all time steps in time sequence to generate a sequence of lower limit values ​​of process parameters.

[0110] For example, to ensure that MC_final ≤ MC_spec (e.g., ≤ 0.5%), the time-varying lower limit constraint T_min(t) that T(t) must satisfy throughout the entire drying cycle is derived in reverse, and the formula is expressed as: T_min(t)=T_dew_point(t)+[-δ×ln(MC_spec / MC_0)] / (k×t_remaining)(3); The specific process of reverse calculation is as follows: Substitute the quality index pass limit MC_spec into MC_final, and solve for the oven set temperature lower limit T_set_min(t) at each future time t under the condition that MC_final≤MC_spec. The drying rate coefficient k and the initial moisture content MC_0 were obtained by fitting the model parameters using the least squares method after collecting the oven temperature history T_set(t), dew point history T_dew(t), and the final measured moisture content MC_measured of the electrode sheets in the historical production batches.

[0111] Step S3033: Based on the sequence of lower limit values ​​of process parameters, generate time-varying lower limit constraints of process parameters for product quality in future optimization cycles to constrain subsequent energy optimization decisions.

[0112] Specifically, the sequence of lower limit values ​​for process parameters is converted into a standardized mathematical constraint form that can be directly recognized and processed by the subsequent optimization engine (e.g., T_set(t) ≥ T_min(t), where T_min(t) is the sequence of lower limit values). This conversion generates time-varying lower limit constraints for process parameters, which clearly define the lower limit of the safe operating range of key adjustable process parameters over time in the upcoming optimization cycle.

[0113] Step S304: Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the objective of minimizing the overall energy cost. For details, please refer to [link to relevant documentation]. Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0114] Step S305: At the beginning of each optimization cycle, solve the multi-constraint rolling optimization model to generate a sequence of optimization control instructions for future optimization cycles, and then issue and execute them. For details, please refer to [link to relevant documentation]. Figure 2Step S204 of the illustrated embodiment will not be described again here.

[0115] The energy optimization decision-making method provided in this embodiment transforms static and abstract product quality qualification standards into a specific and time-varying sequence of lower limit constraints for key process parameters over a future period by combining a digital twin model of process quality with real-time and predictive data. This ensures at the source of decision-making that any scheduling scheme aimed at energy saving and cost reduction will not sacrifice product quality, fundamentally solving the key contradiction that economic efficiency and quality assurance are difficult to coordinate during the optimization process.

[0116] This embodiment provides an energy optimization decision-making method, which can be used in the aforementioned energy optimization decision-making system. Figure 4 This is a flowchart of an energy optimization decision-making method according to an embodiment of the present invention, such as... Figure 4 As shown, the process includes the following steps: Step S401: Obtain multi-source heterogeneous data from the current production environment. For details, please refer to [link to relevant documentation]. Figure 3 Step S301 of the illustrated embodiment will not be described again here.

[0117] Step S402: Based on preset product quality acceptance standards and multi-source heterogeneous data, a pre-built digital twin model is used for reverse calculation to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles. For details, please refer to [link to relevant documentation]. Figure 3 Step S303 of the illustrated embodiment will not be described again here.

[0118] Step S403: Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the goal of minimizing the overall energy cost.

[0119] Specifically, the preset constraints include: production task constraints, energy supply and demand balance constraints, and energy storage system operation constraints. Step S403 includes: Step S4031: Determine the sum of energy purchase cost, carbon emission cost, and equipment depreciation cost as the comprehensive energy cost, with the optimization objective being to minimize the comprehensive energy cost over a future cycle.

[0120] Specifically, the calculation method for comprehensive energy cost is clearly defined, and it is specifically broken down and modeled as a weighted sum or direct sum of three factors: energy purchase cost (based on the product of time-of-use electricity price and purchased power), carbon emission cost (based on the product of grid carbon intensity and purchased power and carbon trading unit price), and equipment depreciation cost (such as an energy storage lifespan loss model based on the number of charge-discharge cycles). Subsequently, minimizing the comprehensive energy cost over a future cycle is set as a single or multi-objective mathematical expression (objective function) for the optimization problem.

[0121] Step S4032 integrates the production task constraints, energy supply and demand balance constraints, energy storage system operation constraints, and time-varying lower limit constraints of process parameters to obtain a set of constraints.

[0122] Specifically, requirements from different dimensions are systematically integrated: production task constraints (ensuring minimum output or total energy consumption), energy supply and demand balance constraints (meeting real-time power balance), energy storage system operation constraints (meeting charging and discharging power and capacity limits), and key time-varying lower limit constraints for process parameters (ensuring quality and safety standards). This integration process encodes these constraints into a set of mathematical equations or inequalities, forming a complete set of constraints.

[0123] Step S4033: Based on the optimization objective and constraint set, construct a multi-constraint rolling optimization model that is solved at the beginning of each cycle. The multi-constraint rolling optimization model is a mixed integer linear programming model.

[0124] Based on the optimization objective and constraint set, an operations research method is used to construct a standard mixed-integer linear programming (MILP) model. The model is characterized by: 1) Rolling: The model is designed to be rebuilt and solved at the beginning of each fixed period based on the latest predictions and state data to achieve closed-loop feedback control; 2) Mixed-integer characteristics: Decision variables include both continuous variables (such as power values) and integer variables (such as the 0 / 1 states of equipment start-up and shutdown), thus accurately describing discrete control actions; 3) Linear structure: The objective function and constraints are expressed as linear relationships between variables, ensuring that the model can be solved quickly and accurately using efficient commercial solvers (such as CPLEX and Gurobi) to achieve global optimum.

[0125] For example, at the current moment, a MILP model is established with a 4-hour outlook period.

[0126] The objective function is MinΣ[Price_pred(t)×P_buy(t)×Δt+λ×P_buy(t)×Δt].

[0127] Constraints include T_set(t)≥T_min(t), oven energy consumption model P_oven(t)=a×(T_set(t)-T_amb(t))+b, production task constraints, power balance constraints and energy storage constraints (such as SOC_min≤SOC(t)≤SOC_max, u_ch(t)+u_dis(t)≤1).

[0128] Step S404: At the beginning of each optimization cycle, solve the multi-constraint rolling optimization model to generate a sequence of optimization control instructions for future optimization cycles, and issue them for execution.

[0129] Specifically, the solution is obtained by MCP collaborative rolling optimization.

[0130] The optimization algorithm engine executes every 15 minutes: it calls the Gurobi solver to solve the model. A successful solution yields the optimal decision trajectory for the next 16 time periods, including... , , .

[0131] The instruction for the first time period in the trajectory (i.e., the current 15 minutes from now). and The output serves as the optimal set of control instructions to be executed.

[0132] like Figure 5 As shown, the execution layer controller receives instructions: the energy storage controller, according to... Control the power flow direction; the oven PLC uses T_set(1) as the new temperature setpoint and starts its internal PID control loop for precise tracking. This execution operation realizes a closed loop from virtual optimization to physical execution.

[0133] Step S405: Collect actual process operation data and actual product quality data after the execution of the optimization control command; based on the actual process operation data, use a digital twin model to perform forward calculation to obtain the predicted value of product quality data; calculate the deviation between the actual product quality data and the predicted value of product quality data; when the deviation continues to exceed the preset threshold, trigger the online learning algorithm to update the initial parameters of the pre-built digital twin model for decision-making in the next optimization cycle.

[0134] Specifically, after a production batch is completed, the actual product quality index MC_actual is obtained from the laboratory; the prediction error e=MC_actual-MC_pred is calculated; if the error continues to exceed the limit, the parameter is updated, and the key parameter k in the digital twin model is corrected using the recursive least squares method. The updated parameter is used for the optimization decision of the next batch.

[0135] In one alternative implementation, the online learning algorithm employs recursive least squares, with the update formula as follows: θ_new=θ_old+K×(y_actual-y_pred)(4); Where θ represents the parameters of the digital twin model to be updated, θ_new represents the parameters of the updated digital twin model, θ_old represents the initial parameters of the digital twin model before the update, K represents the algorithm gain matrix, y_actual represents the actual product quality index, and y_pred represents the predicted value of the product quality data.

[0136] For example, update the drying rate coefficient k in the digital twin. That is, the update formula is k_new = k_old + K × e × ( MC_pred / k), where K is the algorithm gain. The updated k_new will be used for the decision in the next batch.

[0137] The energy optimization decision-making method provided in this embodiment enables the system to continuously self-optimize by introducing a closed-loop feedback mechanism of execution-monitoring-learning. This method dynamically triggers a learning algorithm based on the deviation between actual production data and model predictions, thereby automatically correcting the parameters of the digital twin model and effectively adapting to the uncertainties and time-varying nature of real-world production, such as the slow decline in equipment performance and fluctuations in material properties.

[0138] As one or more specific application embodiments of the present invention, the energy optimization decision-making method provided by the present invention will be further described in detail as follows: First, this embodiment discloses an energy optimization decision-making system based on MCP (Manufacturing, Consumption, Process) collaboration, which includes a perception layer, a decision-making layer, an execution layer, and a learning and display layer.

[0139] The perception layer is responsible for collecting multi-source heterogeneous data across the entire plant. It comprises: smart meters installed at the grid inlet to collect power purchase and sales (P_grid(t)) and electricity consumption; photovoltaic inverter communication interfaces to collect real-time photovoltaic output (P_pv(t); an energy storage battery management system to collect the energy storage state of charge (SOC(t)) and maximum power (P_ess_max); dew point sensors and temperature / humidity sensors deployed in the workshop to collect ambient dew point temperature (T_dew(t)) and ambient temperature (T_amb(t)); PLCs for key process equipment (such as ovens) to collect their set temperature (T_set(t), actual temperature (T_actual(t)), and operating power (P_oven(t)); and data interfaces with the manufacturing execution system and plant monitoring system to obtain production task scheduling and basic load information. All data is parsed and preliminarily cleaned through an industrial IoT gateway before being uploaded at fixed intervals (e.g., every minute).

[0140] The decision-making layer is the core computing unit of the system. It includes: Multidimensional data fusion center: Receives data from the perception layer, performs time-series alignment and associated storage, and forms a unified data pool.

[0141] The energy-process digital twin comprises a mechanistic model and a data-driven model. Taking electrode drying as an example, its mechanistic model is: MC_pred=MC_0×exp(-k×Σ(T_set(i)-T_dew(i)) / δ×Δt). The model parameters (k,MC_0) are calibrated using historical data. Its core function is to perform inverse constraint generation: based on the quality index MC_spec, the process data that has occurred, and future environmental predictions, it calculates the lower limit curve T_min(t) that the future oven temperature must meet.

[0142] Prediction Model Module: Utilizes time series or machine learning methods to predict future photovoltaic power output P_pv_pred(t), time-of-use electricity price Price_pred(t), and base load P_base_pred(t) in the time domain.

[0143] The optimization algorithm engine constructs a mixed-integer linear programming model for a prospective period (e.g., 4 hours) in each cycle (e.g., 15 minutes). The objective function is to minimize the total electricity cost and carbon cost. Decision variables include power purchased and sold, energy storage charging and discharging power and status, and oven set temperature. Key constraints include: dynamic process constraints T_set(t) ≥ T_min(t) provided by the digital twin, production task constraints ΣP_oven(t) ≥ E_required, power balance constraints P_buy(t) + P_pv_pred(t) + P_dis(t) = P_base_pred(t) + P_oven(t) + P_ch(t), and energy storage operation constraints. The engine calls a solver (e.g., Gurobi) to solve the problem and outputs the optimal control instruction set.

[0144] The execution layer consists of local controllers from various subsystems, such as the energy storage PCS controller and the oven PLC. They receive current-moment commands from the decision-making layer (e.g., ...). , ), and convert them into device-level control signals for execution.

[0145] The learning and presentation layer includes an adaptive learning engine and a visualization interface. After each production batch, the learning engine compares the measured moisture content MC_actual with the predicted moisture content MC_pred. If the error continues to exceed the limit, the digital twin parameter k is updated online using a recursive least squares method. The visualization interface displays real-time load, optimization strategies, cost savings, and quality pass rate dashboards to management personnel.

[0146] The energy optimization decision-making method disclosed in this embodiment, applied to the above system, includes the following steps: Step S1: Real-time acquisition and fusion of multi-source data.

[0147] Sensing layer devices continuously collect data from sources, networks, loads, storage, processes, and the environment. Industrial IoT gateways clean and convert the data according to protocols, uploading it to the multi-dimensional data fusion center at the decision-making level every one minute. The center adds a unified timestamp to all data and stores it, providing a data foundation for subsequent modeling and optimization.

[0148] Step S2: Generation of reverse constraints for process quality.

[0149] This is a crucial step connecting process quality and energy decisions. Take, for example, the current batch of electrodes being dried: S21: The digital twin acquires the following quality standards: MC_spec=0.5%, the batch's run time t_elapsed, the oven temperature sequence {T_set_past} and dew point temperature sequence {T_dew_past} that have occurred, and the future dew point prediction sequence {T_dew_future} derived from the weather forecast.

[0150] S22: Based on the mechanistic model MC_pred=MC_0×exp(-k×Σ(T_set(i)-T_dew(i)) / δ×Δt), with MC_pred≤MC_spec as a hard condition, the solution is obtained in reverse. For simplification, the remaining future time is divided into two periods, assuming that the temperature is constant in each period as T_min_seg1 and T_min_seg2. T_min_seg1 and T_min_seg2 are solved using an iterative algorithm (such as the bisection method), thereby generating a piecewise constant future temperature lower limit constraint curve T_min(t).

[0151] S23: Output the dynamic constraint T_set(t)≥T_min(t) to the optimization algorithm engine.

[0152] Step S3: Solve the MCP collaborative rolling optimization.

[0153] The optimization algorithm engine executes every 15 minutes: S31: At the current moment, with a 4-hour outlook period, establish the MILP model. The objective function is MinΣ[Price_pred(t)×P_buy(t)×Δt+λ×P_buy(t)×Δt]. The constraints are integrated from step S2: T_set(t)≥T_min(t), oven energy consumption model P_oven(t)=a×(T_set(t)-T_amb(t))+b, production task constraints, power balance constraints, and energy storage constraints (e.g., SOC_min≤SOC(t)≤SOC_max, u_ch(t)+u_dis(t)≤1).

[0154] S32: Call the Gurobi solver to solve the model. A successful solution yields the optimal decision trajectories for the next 16 time periods, including... , , .

[0155] S33: Instructions for the first time period (i.e., the next 15 minutes) in the trajectory. and The output serves as the optimal set of control instructions to be executed.

[0156] Step S4: Optimize instruction execution: The execution layer controller receives instructions: the energy storage controller, according to... Control the power flow; the oven PLC will As the new temperature setpoint, its internal PID control loop is activated for precise tracking. This step achieves a closed loop from virtual optimization to physical execution.

[0157] Step S5: Feedback learning and visualization.

[0158] S51: After the drying of this batch of electrode sheets is completed and laboratory testing is finished, the system obtains the actual measured final moisture content MC_actual.

[0159] S52: The learning engine calculates the prediction error e = MC_actual - MC_pred. If |e| > 0.1% for three consecutive batches, the model is considered to have bias.

[0160] S53: Trigger adaptive learning by using recursive least squares method, taking the latest batch of data as samples to update the drying rate coefficient k in the digital twin. That is, the update formula is k_new = k_old + K × e × ( MC_pred / k), where K is the algorithm gain. The updated k_new will be used for the decision in the next batch.

[0161] Meanwhile, the visual interface updates in real time, displaying the current load curve, optimization strategies, cumulative energy savings, carbon emission reductions, and batch quality pass rate, providing decision support for managers.

[0162] The construction of the energy-process digital twin model, the working principle of inverse constraint generation, and the steps of specifically writing the output of the energy-process digital twin model into the mathematical expression of the mixed-integer linear programming model are explained in further detail: (a) The steps for constructing the energy-process digital twin model are as follows: Step 1: Preparation and preprocessing of multi-source historical data: Data source: Extracting the complete data chain of all qualified production batches from the factory's historical database over a past period (e.g., 6 months). This includes: Process timing data: Setting parameter curves of key process equipment (such as the set temperature T_set(t) of the oven every minute).

[0163] Environmental time series data: Environmental parameter curves (such as dew point temperature T_dew(t) and ambient temperature T_amb(t)) within the corresponding production period.

[0164] Energy time-series data: Real-time power curves (P_oven(t)) of the corresponding equipment.

[0165] Final quality data: The final quality indicators obtained by laboratory testing after each batch is completed (such as electrode moisture content MC_final).

[0166] Preprocessing: The above data is time-series aligned, outlier removed, and missing values ​​imputed to form a structured dataset, Dataset_hist, for model training.

[0167] Step 2: Establishing the Mechanism Model Framework: Based on physical and chemical laws, a mathematical framework is established to describe the core relationship between process operation and quality evolution.

[0168] Taking electrode drying as an example, its core mechanism is drying kinetics: the drying rate is proportional to the difference between the material surface temperature and the dew point temperature (i.e., the drying driving force).

[0169] The mechanistic model framework established accordingly is as follows: MC_final=MC_0×exp(-k×Σ(T_set(t)-T_dew(t)) / δ×Δt); MC_final: Predicts the final moisture content in the mechanistic model.

[0170] MC_0: Initial moisture content.

[0171] k: Drying rate coefficient, which is a key parameter to be identified that integrates material characteristics, wind speed, and mass transfer efficiency.

[0172] T_set(τ), T_dew(τ): Set temperature and dew point temperature that change over time.

[0173] G: Represents a functional relationship, usually in the form of exponential decay.

[0174] Step 3: Model parameter identification and concretization: Using the data list prepared in step 1, the unknown parameters in the framework of step 2 are calibrated, so that the abstract framework is transformed into a concrete model that can be quantitatively calculated.

[0175] The specific discrete-time model formula is as follows: MC_pred=MC_0×exp(-k×Σ[(T_set(i)-T_dew(i)) / δ]×Δt), i=1, 2,…,N.

[0176] The physical meaning, units, and methods of obtaining each symbol in the formula are disclosed below: MC_pred: The final predicted moisture content (%) calculated by the model, which is the core output of the model; MC_0: Initial moisture content (%) after electrode coating. It can be input via incoming inspection data or used as a parameter identified together with k; k: Drying rate coefficient (m 2 / s or min -1 This is a key process characteristic parameter; it is not fixed but changes slowly with factors such as material formulation and equipment airflow. Its initial value is obtained by fitting historical data. T_set(i): The set temperature (°C) of the oven at the i-th time step. This is a decision variable for the future optimization model; T_dew(i): The dew point temperature (°C) inside the drying chamber at the i-th time step. Measured by a dew point sensor or calculated using temperature and humidity. δ: Electrode coating thickness (m), a known process parameter.

[0177] N: Total number of drying steps.

[0178] Δt: Time step (e.g., 1 minute).

[0179] Parameter identification method: Nonlinear least squares method is adopted. The objective is to minimize the sum of squares of prediction errors across all historical batches, solving for k and MC_0. The objective function is as follows: Where min represents minimization, which is an optimization operation; m represents the summation index, indicating the m-th historical production batch; M represents the total number of historical production batches; This represents the laboratory measured final moisture content of the m-th historical production batch, in % (%). This represents the model-predicted final moisture content of the m-th historical production batch, expressed in %, and is derived from the process data, environmental data, and calculations of the m-th historical production batch. The formula is used to calculate it.

[0180] The initial calibration values ​​of the model parameters are obtained through this optimization solution. and At this point, a digital twin model that can be used for positive quality prediction has been completed.

[0181] Step 4: Supplementing the data-driven model: For some complex relationships that are difficult to describe with precise mechanisms (such as environmental dew point → coating yield), data-driven models are used to supplement them.

[0182] From historical data, feature variables (such as average dew point and dew point fluctuation variance) and quality indicators (yield rate) are extracted, and a correlation model Yield = f(T_dew_features) is obtained by training it using random forest regression or neural network.

[0183] This model can provide another type of quality constraint for optimization problems (such as Yield>=99%), or be used to verify the effective range of mechanistic models.

[0184] (II) The working principle of reverse constraint generation is as follows: Step 1, Input: Quality red line: The ultimate quality indicator MC_spec that must be met (e.g., 0.5%).

[0185] Process status: The elapsed drying time t_elapsed for the current batch, and the actual {T_set_past}, {T_dew_past} sequences that have occurred.

[0186] Future disturbance prediction: Dew point temperature prediction sequence {T_dew_future} for the remaining production time in the future, derived from weather forecasts.

[0187] Current model: Mechanism model with the latest parameters (e.g., k_current).

[0188] Step 2, Processing (Reverse Solving Logic): Problem definition: Find a set of future set temperatures {T_set_future} such that when it is substituted into a forward model along with "past actual data", the calculation result exactly satisfies MC_pred=MC_spec.

[0189] Mathematical description: Solving inequalities (or equations): MC_0×exp(-k_current×{Σ_past(T_set-T_dew)+Σ_future(T_set_future(i)- T_dew_future(i))} / δ×Δt)≤MC_spec; Solution strategy (designed for integration into the optimization model): To simplify and obtain a clear sequence of constraints, a piecewise constant lower bound approximation method is adopted.

[0190] Divide the remaining time into Q time periods (e.g., the step size of the optimization model, Q=16).

[0191] Assume that T_set_future remains constant within each time interval q, denoted as T_min(q).

[0192] The Σ_future(...) term in the above inequality is... Instead, among them, It is the average predicted dew point within time period q. It represents the number of steps taken within a given time period.

[0193] By using numerical iterative algorithms (such as the bisection method), the T_min(q) sequence that makes the equation hold true can be obtained. This sequence is the theoretical minimum temperature trajectory that guarantees the quality will just meet the standard.

[0194] Step 3, Output: Dynamic lower limit constraint sequence: T_min(t), a lower limit value of a process parameter that varies with time.

[0195] Delivered to the optimization model: The sequence is transformed into constraints for the optimization model: T_set(t) ≥ T_min(t). This transforms the uncompromising quality objective into dynamic process variable boundaries that the optimization algorithm can directly handle.

[0196] (III) The specific steps for writing the output of the energy-process digital twin model into the mathematical expression of the mixed-integer linear programming model are as follows: Step 1: Optimize variable definitions: Energy-related variables: P_buy(t) is the power purchased during time period t, P_sell(t) is the power sold during time period t, P_ch(t) is the energy storage charging power during time period t, P_dis(t) is the energy storage discharging power during time period t, and u_ch(t) and u_dis(t) are 0-1 variables representing the charging and discharging states.

[0197] Process-related variables (directly coupled with digital twin): T_set(t): The set temperature of the process equipment (such as an oven) during time period t; this is a key coupling variable. Objective function (example): MinΣ_t[C_ele×(P_buy(t)-P_sell(t))+λ×P_buy(t)+γ_ch×P_ch(t)+γ_dis×P_dis(t)×Δt; Note: Although T_set(t) does not appear directly in the objective function, it indirectly affects P_oven(t) through the "energy consumption coupling equation" below, and then affects P_buy(t) through power balance constraints, ultimately affecting the total cost.

[0198] Step 2, Key Constraints: a. Energy-process coupling constraints (connecting process variables with the energy system): The energy consumption model is P_oven(t) = a × (T_set(t) - T_amb(t)) + b, where a and b are the fitting coefficients.

[0199] This formula illustrates that adjusting the process setting T_set(t) will directly change the energy consumption P_oven(t).

[0200] b. Dynamic process quality constraints (core output from the digital twin model): T_set(t)≥T_min(t),t∈{1,2,...,H}; This is the result of the reverse constraint generation. T_min(t) is a known sequence of input parameters calculated by the decision layer before each rolling optimization, and it is a constant for the optimization model.

[0201] c. System power balance constraints (reflecting energy flow): P_buy(t)-P_sell(t)+P_pv_pred(t)+P_dis(t)=P_base_pred(t)+P_oven(t)+P_ch(t); Here, P_oven(t) is defined by constraint a, so changes in T_set(t) ultimately affect the grid interaction power P_buy(t) / P_sell(t) through this chain.

[0202] Summary of relationships: The digital twin model dynamically calculates T_min(t) -> the optimization model freely adjusts T_set(t) under the constraint T_set(t)≥T_min(t) -> T_set(t) affects energy consumption through P_oven(t)=f(T_set(t)) -> changes in energy consumption affect electricity purchase and sale through power balance, ultimately achieving collaborative decision-making to pursue economic optimization within the quality and safety zone.

[0203] This embodiment also provides an energy optimization decision-making device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0204] This embodiment provides an energy optimization decision-making device, such as... Figure 6 As shown, it includes: The data acquisition module 601 is used to acquire multi-source heterogeneous data in the current production environment.

[0205] The reverse calculation module 602 is used to perform reverse calculations based on preset product quality qualification standards and multi-source heterogeneous data, using a pre-built digital twin model, to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles.

[0206] The optimization model construction module 603 is used to construct a multi-constraint rolling optimization model based on preset constraints and time-varying lower limits of process parameters, with the goal of minimizing the overall energy cost.

[0207] The solver module 604 is used to solve the multi-constraint rolling optimization model at the beginning of each optimization cycle, generate the optimization control instruction sequence for future optimization cycles, and issue it for execution.

[0208] In some alternative implementations, the digital twin model is constructed in the following manner: Acquire historical production batch data, which includes process parameters, environmental parameters, and corresponding product quality data that have occurred. Based on the quality formation mechanism of the engineering process, a mathematical model framework is established with process parameters and environmental parameters as inputs and product quality data as output. Historical production batch data is used to calibrate the parameters to be determined in the mathematical model framework, resulting in a digital twin model of process quality with defined parameters. The digital twin model of process quality is used to describe the quantitative relationship between process parameters, environmental parameters and final product quality indicators.

[0209] In some alternative implementations, the reverse computation module 602 includes: The data extraction unit is used to extract current real-time process parameters, future environmental parameter prediction sequences, and preset product quality qualification standards from multi-source heterogeneous data.

[0210] The reverse calculation unit is used to input the preset product quality qualification standard, the current real-time process parameters, and the predicted sequence of future environmental parameters into the process quality digital twin model, and solve the sequence of lower limit values ​​of process parameters that meet the quality requirements in the future optimization cycle through reverse calculation.

[0211] The process parameter time-varying lower limit constraint generation unit is used to generate process parameter time-varying lower limit constraints for product quality in future optimization cycles based on the process parameter lower limit value sequence, which are used to constrain subsequent energy optimization decisions.

[0212] In some optional implementations, the time-varying lower limit constraint generation unit for process parameters includes: The transformation subunit is used to convert the preset product quality qualification standards into constraint inequalities for the output results of the digital twin model of process quality.

[0213] The inequality solving sub-unit is used to mathematically solve the constraint inequalities for each time step of the future optimization cycle, and to obtain the lower limit of the process parameters that meet the quality requirements within the future optimization cycle.

[0214] The process parameter lower limit value sequence generation sub-unit is used to arrange the process parameter lower limit values ​​obtained from all time steps in time sequence to generate the process parameter lower limit value sequence.

[0215] In some optional implementations, the preset constraints include: production task constraints, energy supply and demand balance constraints, and energy storage system operation constraints; the optimization model construction module 603 includes: The optimization objective determination unit is used to determine the sum of energy purchase cost, carbon emission cost, and equipment depreciation cost as the comprehensive energy cost, with the optimization objective being to minimize the comprehensive energy cost over a future cycle.

[0216] The constraint set construction unit is used to integrate production task constraints, energy supply and demand balance constraints, energy storage system operation constraints, and time-varying lower limit constraints of process parameters to obtain a constraint set.

[0217] The multi-constraint rolling optimization model building unit is used to construct a multi-constraint rolling optimization model that is solved on a rolling basis at the beginning of each cycle, based on the optimization objective and constraint set. The multi-constraint rolling optimization model is a mixed integer linear programming model.

[0218] In some alternative embodiments, the device further includes: The online learning module is used to collect actual process operation data and actual product quality data after the execution of optimization control commands; based on the actual process operation data, a digital twin model is used for forward calculation to obtain the predicted value of product quality data; the deviation between the actual product quality data and the predicted value of product quality data is calculated, and when the deviation continues to exceed a preset threshold, the online learning algorithm is triggered to update the initial parameters of the pre-built digital twin model for decision-making in the next optimization cycle.

[0219] In some optional implementations, the online learning algorithm employs recursive least squares, with the update formula as follows: θ_new=θ_old+K×(y_actual-y_pred); Where θ represents the parameters of the digital twin model to be updated, θ_new represents the parameters of the updated digital twin model, θ_old represents the initial parameters of the digital twin model before the update, K represents the algorithm gain matrix, y_actual represents the actual product quality index, and y_pred represents the predicted value of the product quality data.

[0220] The energy optimization decision-making device provided in this embodiment of the invention can execute the energy optimization decision-making method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0221] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0222] The following is a detailed reference. Figure 7 This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 701, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 702 or a program loaded from memory 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0223] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0224] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 709, or installed from a memory 708, or installed from a ROM 702. When the computer program is executed by the processor 701, it performs the functions defined in the energy optimization decision-making method of the embodiments of the present invention.

[0225] Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0226] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the energy optimization decision-making method shown in the above embodiments is implemented.

[0227] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0228] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. An energy optimization decision-making method, characterized in that, The method includes: Acquire multi-source heterogeneous data of the current production environment, including energy-side data, production task data, process parameters, and environmental data; Based on preset product quality qualification standards and multi-source heterogeneous data, a pre-built digital twin model is used for reverse calculation to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles. Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the goal of minimizing comprehensive energy cost. At the beginning of each optimization cycle, the multi-constraint rolling optimization model is solved to generate a sequence of optimization control instructions for future optimization cycles, which are then issued and executed.

2. The method according to claim 1, characterized in that, The digital twin model is constructed in the following manner: Acquire historical production batch data, which includes process parameters, environmental parameters, and corresponding product quality data that have occurred. Based on the quality formation mechanism of the engineering process, a mathematical model framework is established with the process parameters and environmental parameters as inputs and the product quality data as outputs. The historical production batch data is used to calibrate the undetermined parameters of the mathematical model framework to obtain a process quality digital twin model with determined parameters; the process quality digital twin model is used to describe the quantitative relationship between process parameters, environmental parameters and final product quality indicators.

3. The method according to claim 2, characterized in that, The process, based on preset product quality standards and multi-source heterogeneous data, uses a pre-constructed digital twin model for reverse calculation to generate time-varying lower limit constraints on process parameters for product quality within future optimization cycles, including: Extract current real-time process parameters, future environmental parameter prediction sequences, and preset product quality qualification standards from the multi-source heterogeneous data; The preset product quality qualification standard, the current real-time process parameters, and the predicted sequence of future environmental parameters are input into the process quality digital twin model, and the lower limit sequence of process parameters that meet the quality requirements in the future optimization cycle is solved by reverse calculation. Based on the sequence of lower limit values ​​of the process parameters, time-varying lower limit constraints of process parameters are generated to constrain product quality within future optimization cycles for subsequent energy optimization decisions.

4. The method according to claim 3, characterized in that, The sequence of lower limit values ​​for process parameters that meet quality requirements within future optimization cycles is obtained through reverse calculation, including: The preset product quality qualification standards are transformed into constraint inequalities for the output results of the digital twin model of process quality; For each time period of the future optimization cycle, the constraint inequality is mathematically solved to obtain the lower limit of the process parameters that meet the quality requirements within the future optimization cycle. Arrange the lower limit values ​​of the process parameters obtained from all time steps in time sequence to generate the sequence of lower limit values ​​of the process parameters.

5. The method according to claim 1, characterized in that, The preset constraints include: production task constraints, energy supply and demand balance constraints, and energy storage system operation constraints. The multi-constraint rolling optimization model, based on preset constraints and time-varying lower limits of process parameters, aims to minimize the overall energy cost and includes: The total energy cost is determined by summing energy purchase cost, carbon emission cost, and equipment depreciation cost, with the optimization objective being to minimize the total energy cost over a future cycle. The constraints of production tasks, energy supply and demand balance, energy storage system operation, and time-varying lower limit of process parameters are integrated to obtain a set of constraints. Based on the optimization objective and constraint set, a multi-constraint rolling optimization model is constructed to solve the problem at the beginning of each cycle. The multi-constraint rolling optimization model is a mixed integer linear programming model.

6. The method according to claim 1, characterized in that, The method further includes: Collect actual process operation data and actual product quality data after the execution of optimized control commands; Based on actual process operation data, a digital twin model is used for forward calculation to obtain predicted values ​​of product quality data; The deviation between the actual product quality data and the predicted product quality data is calculated. When the deviation continues to exceed a preset threshold, an online learning algorithm is triggered to update the initial parameters of the pre-built digital twin model for decision-making in the next optimization cycle.

7. The method according to claim 6, characterized in that, The online learning algorithm uses the recursive least squares method, and the update formula is: θ_new=θ_old+K×(y_actual-y_pred); Where θ represents the parameters of the digital twin model to be updated, θ_new represents the parameters of the updated digital twin model, θ_old represents the initial parameters of the digital twin model before the update, K represents the algorithm gain matrix, y_actual represents the actual product quality index, and y_pred represents the predicted value of the product quality data.

8. An energy optimization decision-making system, characterized in that, The system includes: The perception layer is used to acquire multi-source heterogeneous data from the current production environment; The decision layer, connected to the perception layer, is used to perform reverse calculations based on preset product quality qualification standards and multi-source heterogeneous data using a pre-built digital twin model to generate time-varying lower limits of process parameters for product quality in future optimization cycles. Based on preset constraints and time-varying lower limits of process parameters, a multi-constraint rolling optimization model is constructed with the goal of minimizing comprehensive energy costs. At the beginning of each optimization cycle, the multi-constraint rolling optimization model is solved to generate a sequence of optimization control instructions for future optimization cycles, which is then issued for execution.

9. An energy optimization decision-making device, characterized in that, The device includes: The data acquisition module is used to acquire multi-source heterogeneous data of the current production environment, including energy-side data, production task data, process parameters and environmental data. The reverse calculation module is used to perform reverse calculations based on preset product quality qualification standards and multi-source heterogeneous data, using a pre-built digital twin model, to generate time-varying lower limit constraints on process parameters for product quality in future optimization cycles. The optimization model building module is used to construct a multi-constraint rolling optimization model based on preset constraints and time-varying lower limits of process parameters, with the goal of minimizing the overall energy cost. The solver module is used to solve the multi-constraint rolling optimization model at the beginning of each optimization cycle, generate a sequence of optimization control instructions for future optimization cycles, and issue them for execution.

10. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the energy optimization decision-making method of any one of claims 1 to 7 by executing the computer instructions.

Images