Dynamic energy consumption monitoring and energy-saving and carbon-reducing regulation and control method for final assembly construction production link

Abstract

The invention provides an energy consumption dynamic monitoring and energy-saving and carbon-reducing regulation and control method for a final assembly construction production link. The method comprises the following steps: S1, acquiring energy consumption related data of each process in an industrial production cycle; s2, constructing an energy consumption prediction model which comprises a mechanical power energy consumption prediction model and a thermal process modification energy consumption prediction model; S3, inputting the data acquired in the step S1 into the energy consumption prediction model constructed in the step S2 to obtain related energy consumption prediction data; s4, converting the energy consumption prediction data into carbon emission, and calculating based on the carbon intensity coefficient of the energy type to obtain total carbon emission; and S5, inputting the total carbon emission into a multi-objective genetic algorithm to generate a Pareto optimal solution set, and obtaining an optimal combination of energy consumption and carbon emission. According to the method, the energy consumption cost, the carbon emission intensity and the production efficiency are planned as a whole through the multi-objective optimization algorithm, the global optimal decision under the equipment cooperation and dynamic working conditions is realized, and the complex requirements of multi-variety small-batch production are effectively met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy consumption dynamic monitoring, and particularly relates to a method for energy consumption dynamic monitoring and energy saving and carbon reduction regulation in a final assembly manufacturing process. BACKGROUND

[0002] Final assembly manufacturing is a core link of complex equipment manufacturing, involving multi-process collaborative operation such as mechanical assembly, electrical integration and system debugging, and has the characteristics of high energy consumption, concentrated carbon emission and large dynamic fluctuation. Traditional methods are mostly based on historical data for static energy consumption management, which is difficult to adapt to dynamic needs in the multi-variety and small-batch production mode, and lacks fine control of carbon emissions. Although the existing patent CN117193212A introduces online modeling technology, it does not combine carbon emission calculation and multi-device collaborative optimization. Therefore, an intelligent method is needed that can realize dynamic perception of energy consumption, accurate accounting of carbon footprint and multi-target collaborative regulation. SUMMARY

[0003] The present application aims to at least solve the technical problems existing in the prior art, and particularly innovatively provides a method for energy consumption dynamic monitoring and energy saving and carbon reduction regulation in a final assembly manufacturing process.

[0004] In order to achieve the above purpose of the present application, the present application provides a method for energy consumption dynamic monitoring and energy saving and carbon reduction regulation in a final assembly manufacturing process, comprising the following steps: S1, collecting data related to energy consumption in each process in an industrial production cycle; S2, constructing an energy consumption prediction model, the energy consumption prediction model comprising a mechanical power energy consumption prediction model and a heat process improvement energy consumption prediction model: S3, inputting the data collected in step S1 into the energy consumption prediction model constructed in step S2 to obtain relevant energy consumption prediction data; S4, converting the energy consumption prediction data into carbon emissions, and calculating the total carbon emissions based on the carbon intensity coefficient of the energy type; S5, inputting the total carbon emissions into a multi-objective genetic algorithm to generate a Pareto optimal solution set, and obtaining the best combination of energy consumption and carbon emissions.

[0005] Preferably, the expression of the mechanical power energy consumption prediction model is: ; wherein, is the total mechanical power energy consumption; is the total number of processes; is the number of independent devices involved in the i-th process; represents the comprehensive efficiency coefficient of device j in process i. Total duration of the i-th process; Pbasej,i represents the base standby power of equipment j in process i; Pdynamicj,i represents the dynamic load power increment of equipment j in process i; Pactivej,i represents the active state function of equipment j in process i; when the equipment is in the processing state, when the equipment is in the processing state, the equipment is in the standby / idle state, ; Ptotalj represents the total standby energy consumption and the total processing energy consumption.

[0006] Preferably, The calculation formula of Ptotalj is: ; Wherein, Pmotor represents the motor efficiency; Pprocess represents the process matching efficiency; Paging represents the aging attenuation factor.

[0007] Preferably, Ptotalj is calculated by the following formula: ; Wherein Ptotal represents the total standby energy consumption; Pprocess represents the processing energy consumption; , Pstartj,i and Pendj,i respectively represent the actual processing start and end time of equipment j in process i.

[0008] Preferably, the expression of the thermal process modification energy consumption prediction model is: ; Wherein, Ptotal represents the total thermal process modification energy consumption; Ntotal represents the total number of processes; Nj,i represents the number of equipment of the i-th process; Qj,i represents the sensible heat energy consumption of equipment j in process i; Rj,i represents the radiation heat loss of equipment j in process i.

[0009] Preferably, The specific calculation formula of is as follows: ; Wherein, represents the constant-pressure specific heat capacity of the equipment j in the process i; represents the mass of the material processed by the equipment j in a single process in the process i; represents the target temperature rise absolute value of the equipment j in the process i; represents the limit temperature of the equipment j in the process i, taken ; represents the reference temperature of the equipment j in the process i, taken .

[0010] Preferably, The specific calculation formula of is as follows: ; Wherein, represents the comprehensive emissivity of the equipment j in the process i; is the Stefan-Boltzmann constant; represents the effective radiation area of the equipment j in the process i; represents the operating average temperature of the equipment j in the process i; represents the ambient temperature of the equipment j in the process i; represents the effective heating time of the equipment j in the process i.

[0011] Preferably, the energy consumption prediction data is converted into carbon emissions, and the total carbon emissions are calculated based on the carbon intensity coefficient of the energy type, which is obtained by the following calculation formula: ; Wherein, represents the total carbon emissions; represents the mechanical power energy consumption at time t; represents the regional power grid carbon emission factor at time t; represents the thermal process improvement energy consumption at time t; represents the calorific value of natural gas, which is used to convert the thermal process improvement energy consumption into natural gas consumption.

[0012] represents the carbon emission factor of natural gas.

[0013] Preferably, the expression for inputting the total carbon emission into the multi-objective genetic algorithm to generate a Pareto optimal solution set is: ; is a minimum value symbol; represents the weight coefficient of the energy consumption target; represents the total energy consumption, including mechanical power energy consumption and heat process improvement energy consumption; represents the historical maximum energy consumption; represents the weight coefficient of the carbon emission target; represents the total carbon emission; historical maximum carbon emission; s.t. is a constraint condition symbol; represents the production cycle time; maximum allowable production cycle time; product quality index; minimum allowable product quality index.

[0014] Preferably, S6, dynamic feedback and compensation, is further included: S6-1, according to the real-time power of mechanical power deviation from the prediction value of the twin model , dynamically adjust the equipment parameters to make the energy consumption error converge within a threshold value, while considering the dynamic accounting result of carbon emission to optimize the adjustment strategy; S6-2, by optimizing the process enabling sequence and time, minimizing the comprehensive target of total assembly process idle energy consumption and carbon emission, while meeting the process dependency relationship and resource constraints.

[0015] In summary, due to the adoption of the above technical solutions, the present application can monitor multi-process energy consumption and carbon emission data in real time, dynamically compensate for energy efficiency deviation caused by production fluctuations, break through the limitations of traditional static management; establish a full life cycle carbon footprint tracking model, accurately associate materials, processes and carbon emissions, and solve the problem of lack of fine management and control; through a multi-objective optimization algorithm, the energy consumption cost, carbon emission intensity and production efficiency are coordinated, global optimal decision-making under device collaboration and dynamic working conditions is realized, and the complex needs of multi-variety and small-batch production are effectively adapted.

[0016] Additional aspects and advantages of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following and the accompanying drawings or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0017] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, taken in conjunction with the accompanying drawings, in which: Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0018] Embodiments of the present application are described below in detail, examples of which are shown in the accompanying drawings, in which the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0019] The method builds a dynamic energy consumption-carbon emission model of the whole assembly process by fusing digital twin technology with Internet of Things monitoring, combines real-time data feedback with multi-objective optimization algorithm, and realizes double precise regulation and control of energy consumption and carbon emission. The specific steps are as follows: Step one: digital twin modeling and energy consumption prediction Based on the assembly process card, a digital twin model is built to simulate the energy consumption characteristics of each process (such as mechanical assembly, welding, painting, and system testing) and predict the initial energy consumption value.

[0020] Through three-dimensional modeling technology (using SolidWorks and Unity3D to collaboratively build 1:1 physical mapping), the physical entities (equipment / materials / environment) of the assembly workshop are converted into digital twins, integrating three types of data collected by Internet of Things sensors in real time (equipment operating parameters: current / voltage / speed; environmental parameters: temperature / humidity / cleanliness; process parameters: material properties / processing precision requirements), and building a dynamic energy consumption simulation engine based on finite element analysis (FEA). The digital twin model and the data acquisition system of step S1 interact in real time through the OPC UA protocol, with a sensor sampling frequency of 1 kHz to ensure dynamic response, providing millisecond-level time series data foundation for the energy consumption prediction model of step S2, reducing the prediction error to ±3%.

[0021] (1) Mechanical power energy consumption: The mechanical work consumption of the driving equipment during the assembly process is calculated by the formula:

[0022] Wherein, is the total mechanical power energy consumption; is the total number of processes; Number of independent devices involved in the ith process; Total duration of the ith process; Base standby power of device j in process i; Dynamic load power increment of device j in process i; Activation state function of device j in process i; when the device is in the processing state, when the device is in the processing state, the device is in standby / idle state, Comprehensive efficiency coefficient of device j in process i: wherein, motor efficiency; process matching efficiency (considering cooling / lubrication system loss); aging attenuation factor; The power composition of the process device is decomposed by using a time slicing algorithm (100 ms / slice), the activation state (processing / standby / idle) of the device is judged in real time by a state machine model, and the energy consumption accumulation calculation in the time dimension is realized by combining the energy efficiency curve of the device out of the factory with the real-time efficiency attenuation coefficient (which is dynamically corrected by the optimization algorithm in step S5). The sub-process energy consumption data (grain size up to device level) output by this model is directly used as the basic data source for carbon emission accounting in step S4, and at the same time, the sensor sampling weight in step S1 is corrected through a feedback channel, and the accuracy of process-level energy saving potential analysis is improved in typical scenarios.

[0023] By connecting motor efficiency (real-time query based on IEC 60034 standard curve), process matching efficiency (fusion of cooling system flow sensor and lubrication pressure data), and aging factor (Weibull distribution model is used to fit the device failure curve), a device full life cycle energy efficiency evaluation model is constructed. The calculation results are used to correct the mechanical power energy consumption prediction value in step S2 (correction coefficient range 0.85-1.12) on the one hand, and as a device health constraint condition of step S5 multi-objective optimization algorithm, guiding the optimization of device maintenance cycle, which can improve the comprehensive efficiency of the device and prolong the maintenance interval in typical scenarios.

[0024] total standby energy and total processing energy; ​​

[0025] wherein represents the standby total energy consumption; represents the processing energy consumption; , respectively represent the actual processing start and end time of the equipment j in the process i; (2) Thermal process modification energy consumption The effective energy consumption of heating links such as welding, coating drying, etc. adopts an improved enthalpy change model: Based on the thermodynamic enthalpy change theory (ΔH = mcΔT), the sensible heat transfer data collected by the infrared temperature sensor and the radiation heat loss calculation model (Monte Carlo method is used to simulate the heat radiation path) are combined to establish a temperature-energy consumption dynamic response model. Real-time interaction with the infrared thermal imager (sampling frequency 25 Hz) in the Internet of Things system in step S1, the accuracy is improved by 15% compared with the traditional heat balance model, which can capture dynamic characteristics such as welding arc heat loss (error ≤ ± 5%), coating drying chamber temperature field distribution (spatial resolution 0.1℃) in real time, and provide thermal process special correction parameters for energy consumption prediction in step S3.

[0026]

[0027] wherein, represents the total energy consumption of the thermal process modification; represents the total number of processes; represents the number of equipment of the i-th process; represents the sensible heat energy consumption of the equipment j in the process i; The specific calculation formula is as follows: , wherein, represents the constant-pressure specific heat capacity (unit: J / (kg·K)) of the material processed by the equipment j in the process i; represents the mass of the material processed by the equipment j in the process i at a time (unit: kg); represents the target temperature rise absolute value of the equipment j in the process i; represents the limit temperature of the equipment j in the process i, which is taken as ; represents the reference temperature of the equipment j in the process i, which is taken as ; represents the radiant heat loss of equipment j in process i; The specific calculation formula is as follows:

[0028] wherein, represents the comprehensive emissivity of equipment j in process i; this parameter is affected by the surface oxide layer, such as the decrease in emissivity after heat treatment of aluminum alloy.

[0029] is the Stefan-Boltzmann constant, which physically represents the total energy radiated by a black body per unit surface area per unit time to a hemispherical space, and the ratio of the fourth power of the thermodynamic temperature; .

[0030] represents the effective radiation area of equipment j in process i; represents the operating average temperature of equipment j in process i; represents the ambient temperature of equipment j in process i; represents the effective heating time of equipment j in process i; related to the process curve (such as annealing holding time).

[0031] The theoretical sensible heat requirement is calculated by the product of the material specific heat capacity database (integrating 1000+ industrial material parameters) and the real-time mass data of the weighing sensor, combined with the target temperature rise feedback of the infrared temperature measuring instrument and the temperature threshold correction coefficient (based on PID closed-loop control dynamic adjustment). The output value is used as the core parameter of the heat process improvement energy consumption model in step S2, and is transmitted to the optimization algorithm in step S5 through the OPC server to provide a theoretical benchmark value for heating equipment power regulation, avoiding energy waste caused by over-heating, reducing heat loss in the aluminum alloy welding scene, and reducing the calculation deviation of carbon emissions in step S4.

[0032] Based on the Stefan-Boltzmann law (P=εσA(T 4 -T0 4) by an infrared spectrometer, the effective radiation area (A) and temperature field distribution (T is the equipment temperature, and T0 is the ambient temperature) obtained by a thermal imager, and the radiation heat loss in different process stages is calculated by the finite difference method. The calculation results are used to correct the radiation heat loss parameter (correction factor 0.92-1.08) of the thermal process performance improvement energy consumption model in step S2, and are transmitted to the carbon emission calculation model in step S4 through the data bus, thereby providing a quantitative basis for the optimization of the heating equipment insulation layer and the adjustment of the process parameters, reducing the radiation heat loss in the coating drying process, and reducing the calculation error of the carbon emission in step S4.

[0033] Step two: dynamic accounting of carbon emissions The energy consumption data is converted into carbon emissions, and the carbon intensity coefficient is calculated based on the energy type (electricity, gas, etc.):

[0034] wherein, represents the total carbon emissions (unit: kilogram of carbon dioxide equivalent, kgCO2e).

[0035] ∑ is the summation symbol, indicating the accumulation of carbon emissions at different times or different energy types.

[0036] represents the mechanical power energy consumption at time t (unit: kilowatt-hour, kWh), which is assumed to be entirely from electricity consumption.

[0037] represents the regional power grid carbon emission factor at time t (unit: kgCO2e / kWh).

[0038] represents the thermal process performance improvement energy consumption at time t (unit: joule, J).

[0039] represents the calorific value of natural gas (unit: J / m 3 ), which is used to convert the thermal process performance improvement energy consumption into natural gas consumption.

[0040] represents the carbon emission factor of natural gas (unit: kgCO2e / m 3 ).

[0041] Based on the theory of life cycle assessment (LCA) (ISO 14040 standard framework), a "energy consumption-carbon emission" mapping model is constructed. Through the API interface, the real-time carbon intensity coefficient (updated every 15 minutes) is obtained from the regional power grid dispatching center. The mechanical power energy consumption (electricity) and the heat process improvement energy consumption (gas) classified data output in step S2 are fused to establish a dynamic accounting engine coupled with time and space. Real-time interaction with the Internet of Things system in step S1 (collection frequency 1 Hz) is carried out. After separating the energy consumption data by energy type (electricity / gas), the corresponding carbon intensity coefficient (electricity from regional power grid real-time data, gas using GB / T 32151.2-2015 benchmark value) is multiplied, realizing minute-level dynamic measurement of carbon emissions, improving the accuracy of traditional batch accounting, and providing second-level carbon emission data input for multi-objective optimization in step S5.

[0042] The mechanical power energy consumption (electricity, unit kWh) and the heat process improvement energy consumption (gas, unit J) output in step S2 are separated by energy flow decomposition algorithm (EFA). The electricity part is multiplied by the real-time regional power grid carbon intensity factor (kgCO2e / kWh) obtained in step S1. The gas part is converted by heat value (natural gas heat value 36MJ / m 3 ) and multiplied by the gas carbon emission factor (2.16kgCO2e / m 3 ). The total carbon emission of the whole process is accumulated by time-weighted average method. The accounting results are stored in the blockchain notarization system (meeting the carbon footprint traceability requirements) on the one hand, and are fed back to the energy consumption prediction model in step S3 to correct the deviation (feedback period 500ms) on the other hand, providing a quantitative benchmark for carbon emissions for double-objective optimization. In the multi-energy structure scenario, the carbon contribution ratio of electricity and gas can be accurately distinguished (typical scenario electricity proportion 65%-75%), supporting the dynamic adjustment of weight coefficients in step S5.

[0043] Step three: double-objective optimization of energy consumption and carbon emission A multi-objective genetic algorithm (NSGA-II) is used to generate a Pareto optimal solution set to balance energy consumption and carbon emission:

[0044] For minimization, it means finding the solution that minimizes the objective function; The weight coefficient of the energy consumption target is represented; The total energy consumption (unit: J) includes mechanical power energy consumption and heat process improvement energy consumption; The historical maximum energy consumption is represented; The weight coefficient of the carbon emission target is represented; Total carbon emission (unit: kgCO2e); Historical maximum carbon emission (unit: kgCO2e); s.t. is constraint symbol, constraints include upper limit of production cycle and quality requirements.

[0045] Production cycle time; Allowed maximum production cycle time; Product quality index; Allowed minimum product quality index.

[0046] A double-objective optimization framework is constructed using the NSGA-II algorithm, taking the total energy consumption (mechanical + thermal process) output in step S2 and the total carbon emission calculated in step S4 as optimization objectives, maintaining population diversity through fast nondominated sorting and crowding distance, and integrating three types of constraints (production cycle ≤ T_max from the process card data in step S1, product pass rate ≥ 99.5% from the quality detection system, and equipment health degree ≥ 0.85 from the aging factor model in step S2). The fitness function is dynamically adjusted in real time during the algorithm iteration process by calling the real-time data (device load rate, material properties) of the Internet of Things in step S1, and the crossover probability (0.7-0.9) and mutation probability (0.01-0.05) are adaptively adjusted according to the energy consumption prediction bias (from step S3). Under the premise of ensuring product pass rate (≥ 99.5%), the Pareto optimal balance of energy consumption and carbon emission is achieved, and the optimization efficiency is improved by 50% compared with traditional single-objective algorithms. The generated solution set directly drives the dynamic control strategy in step S5.

[0047] The optimization priority of energy consumption and carbon emission is adjusted by dynamic weight coefficients (ω1, ω2, satisfying ω1+ω2=1), which are corrected by fuzzy logic controller according to external factors (carbon tariff policy triggers ω2 to increase to 0.6-0.7, energy price peak triggers ω1 to increase to 0.6-0.7), combined with the production cycle upper limit (T≤T_max) in step S1 and the equipment efficiency constraint (η≥η_min) in step S2 to build the optimization function. The energy consumption term (E / E_max) of the objective function is normalized by the predicted energy consumption data in step S2, and the carbon emission term (C / C_max) is normalized by the accounting data in step S4. The generated Pareto frontier (8-12 optimization schemes) is displayed through a visual interface, supporting decision-makers to select control strategies according to real-time working conditions, and passing the preferred scheme parameters (such as device power, process timing) to the adaptive PID controller in step S5. In typical scenarios, the decision response time is shortened to 30 seconds.

[0048] The generated Pareto optimal solution set represents a series of optimal combinations of energy consumption and carbon emission under different trade-offs. These solution sets form a frontier, which is the Pareto frontier. Each solution on the frontier represents a possible optimization scheme. Decision-makers can choose the most suitable solution as the optimization target according to the specific needs of the enterprise (such as energy cost, environmental policy, etc.).

[0049] Step four: dynamic feedback and compensation control The dynamic feedback and compensation control mechanism mainly includes intra-process compensation and inter-process coordination.

[0050] A three-stage closed-loop system of "real-time monitoring-deviation analysis-dynamic control" is constructed. Real-time energy consumption data is collected by step S1 Internet of Things sensors (1 kHz sampling rate), input into step S2 digital twin model to calculate theoretical prediction values, and after deviation analysis module (using Kalman filter for denoising), drive intra-process adaptive PID control (adjust device power) and inter-process timing optimization (adjust process order). The real-time monitoring module interacts with the data acquisition system in step S1 through MQTT protocol, the deviation analysis data is fed back to step S2 to modify the energy consumption prediction model (modification period 200 ms), and the dynamic control instruction receives the Pareto optimal parameters output by step S5 multi-objective optimization algorithm, so that the production process energy consumption fluctuation is controlled within ±5%, the carbon emission peak is reduced by 22%, and the response speed reaches seconds level (≤3s).

[0051] (1) Intra-process compensation: adaptive PID control According to the real-time power of mechanical power and the predicted value of the twin model The dynamic adjustment of equipment parameters (such as motor speed, heating power) is made to enable the energy consumption error to converge within a threshold (±5%) while considering the dynamic accounting results of carbon emissions to optimize the adjustment strategy.

[0052] A Mamdani-type fuzzy logic controller (input: energy consumption deviation e(t), carbon emission deviation ΔC(t)=C_real-C_target, output: PID coefficient correction) is adopted to establish a two-dimensional fuzzy rule base (5x5 rule matrix), and the deviation of the carbon emission dynamic accounting result (C_real) from the target value (C_target from step S5 optimization scheme) is taken as the feedback variable to real-time correct the proportional (Kp), integral (Ki), and differential (Kd) coefficients. The carbon emission weight coefficient of the fuzzy logic rule base is dynamically updated by the optimization algorithm of step S5 (the weight of ΔC(t) is increased to 0.6 when the carbon price is high), which reduces the overshoot of the traditional PID control by 25% and shortens the steady-state error convergence time to 15 seconds, achieving an energy saving rate of 8%-10% in motor-driven equipment, while reducing the carbon emission accounting deviation of step S4 to ±2%.

[0053] (1.1) Deviation calculation Calculate the deviation between real-time power and predicted power:

[0054] The real-time power P_real (sampling frequency 1 kHz) is collected by the power sensor (accuracy ±0.1 kW) of step S1 Internet of Things system, and the difference between P_pred (update frequency 500 ms) output by step S2 digital twin model is calculated. The sliding average filter (window size 5 sampling points) is used to eliminate high-frequency noise, and the deviation between actual energy consumption and theoretical value is quantified. The deviation signal e(t) is used as the core input parameter of the adaptive PID control in this step (the polarity determines the power adjustment direction: positive deviation reduces power, negative deviation increases power), and on the other hand, it corrects the efficiency coefficient of step S2 energy consumption prediction model through the feedback channel (correction formula: η_new=η_old×(1-0.1×|e(t)| / P_pred)), and the model prediction accuracy is improved by 18% in typical scenarios.

[0055] (1.2) Adaptive PID output According to the deviation calculation control quantity u ( t ), a time-varying gain coefficient 、 、 :

[0056] The weights of the proportional, integral, and derivative components are dynamically adjusted using time-varying gain coefficients (Kp(t), Ki(t), Kd(t)). These gain coefficients are corrected in real-time by the fuzzy logic controller based on dual deviations (energy consumption deviation e(t) from the comparison between steps S1 and S2, and carbon emission deviation ΔC(t) from the calculation result of step S4). The correction formula is Kp(t) = α × γ × δ (α is the base coefficient, γ is the energy consumption error factor, and δ is the carbon emission factor). The calculation of the control quantity u(t) integrates the weight coefficients (ω1, ω2) output from the multi-objective optimization algorithm in step S5. When the carbon emission weight ω2 > 0.5, Kd(t) is automatically increased to accelerate the carbon emission response. The generated control quantity guides equipment power adjustment, causing the energy consumption deviation to converge to the ±5% threshold within 15 seconds. In typical scenarios, the dynamic response speed is improved by 30%, while providing accurate input for equipment parameter correction in step S106.

[0057] in, The gain coefficient is a time-varying value, dynamically adjusted using fuzzy logic, with the adjustment rules taking into account the dynamic calculation results of carbon emissions. Feedback.

[0058] (1.3) Adjustment rules for time-varying gain coefficient The gain coefficient is dynamically adjusted using fuzzy logic, taking into account the dynamic carbon emission accounting results. Feedback:

[0059] Adjust the error factor ( Based on the energy consumption prediction deviation e(t) in step S2, the PI algorithm calculates γ = 1 + 0.2 × |e(t)| / P_pred) and carbon emission adjustment factors (δ, ζ, ι, calculated based on the carbon emission deviation ΔC(t) in step S4: δ = 1 + 0.3 × max(ΔC(t) / C_target, 0)). These are coupled through a fuzzy logic rule base (containing 25 if-then rules). When the actual carbon emission exceeds the target value (C_target comes from the optimized solution set in step S5), the weights of the carbon emission factors are automatically increased (δ, ζ, ι with an upper limit of 1.5). Output function: To achieve a coordinated response of the PID controller to the dual targets of "energy consumption-carbon emission," the carbon emission adjustment sensitivity can be improved by 40% during periods of high carbon price (>80 yuan / ton). Simultaneously, the adjustment results are fed back to the optimization algorithm in step S5 via the data bus to update the Pareto optimal solution set.

[0060] in, α , β , θ All are basic gain coefficients; All are error-related adjustment factors; δ ,ζ 、 ι are carbon emission related adjustment factors; is the carbon emission target value at time t .

[0061] (1.4) Equipment parameter correction Adjust the equipment power according to the control variable u ( t ):

[0062] For example: the predicted power of a welding equipment is = 50 kW, the real-time monitoring value is = 54 kW, the deviation is = +4 kW. If = 0.8, the corrected power is: = 50 + 0.8 x 4 = 53.2 kW.

[0063] Based on the linear superposition model of the control variable u(t) output by step S104 and the predicted power P_pred of the twin model in step S2 (P_corr = P_pred + K x u(t), K is the proportional coefficient dynamically adjusted by the optimization algorithm in step S5), a "deviation-correction" dynamic mapping relationship is established (the mapping accuracy is 0.01 kW). The corrected power P_corr is fed back to the Internet of Things system in step S1 to adjust the equipment output, and the actual power data P_real is returned to step S4 to update the carbon emission accounting results, so that the deviation between the actual power of the equipment and the predicted value of the twin model is controlled within ±2%, the single-process energy consumption fluctuation is reduced by 18%, the power adjustment accuracy of typical welding process is 0.1 kW, and the real-time performance of carbon emission accounting is improved to seconds.

[0064] (2) Inter-process coordination: timing optimization model The timing optimization model optimizes the sequence and time of process activation to minimize the comprehensive target of total assembly process idle energy consumption and carbon emission, while meeting the process dependency relationship and resource constraints. Thus, the objective function is:

[0065] wherein, are the weight coefficients of energy consumption and carbon emission; represents the idle energy consumption of process i ; represents the total carbon emission; , represents the parameter of whether process i is activated.

[0066] Using 0-1 integer programming theory, based on the process activation status ( (1 indicates enabled, 0 indicates disabled) are used as decision variables, and weighted by coefficients. (Dynamically output by the multi-objective optimization algorithm in step S5, ω∈[0,1]) Balance idle energy consumption (from the standby data of the device monitored by the Internet of Things in step S1) and carbon emission target (from the dynamic calculation result in step S4). The constraints of the decision variables are integrated with the process energy consumption matrix output by the energy consumption prediction model in step S2. The integer programming problem is solved by the branch and bound method. The process sequence is optimized to reduce the idle waiting time of the final assembly process by 45 minutes / batch and reduce carbon emissions in the multi-device collaborative scenario. At the same time, the optimized process sequence is fed back to step S5 to update the Pareto optimal solution set.

[0067] Based on the process dependency matrix generated by the digital twin model in step S2, a genetic algorithm (GA) is used to traverse all possible process combinations (population size 200, 50 generations). Combined with the carbon emission accounting data in step S4, non-dominated solutions that meet resource constraints (equipment load rate ≤ 90% from step S1) are selected. The smaller the objective function value, the better the overall optimization effect. The generated non-dominated solutions serve as the input subset of the multi-objective optimization algorithm in step S5. After secondary optimization by the NSGA-II algorithm, the final decision scheme is formed, providing an executable process sequencing scheme for production scheduling. This allows decision-makers to flexibly choose between "fast delivery" (high energy consumption, ω1=0.7) and "low-carbon production" (low energy consumption, ω2=0.7). In typical scenarios, the scheduling scheme generation time is shortened to 2 minutes.

[0068] The constraints on the above objective function are: (a) Process timing constraints:

[0069] in, Indicate process i Equipment preparation time.

[0070] This indicates the maximum permissible production cycle time.

[0071] The processing time of each step (real-time PLC data from the IoT system in step S1, with an accuracy of ±1 second) and the equipment preparation time (simulation results of equipment switching based on the digital twin model in step S2) are summed up, and a linear programming (LP) model is used to ensure that the total cycle time does not exceed the upper limit. (Dynamically set by the multi-objective optimization algorithm in step S5 based on the order delivery cycle). The calculation results of the timing constraints are fed back to the optimization algorithm in step S5 in real time. When the predicted total cycle exceeds the limit, the process splitting or parallel processing strategy is automatically triggered to prevent equipment conflicts caused by process overlap, ensure the stability of the production rhythm, and control the cycle fluctuation of a typical production line within ±3%. At the same time, it provides cycle constraint parameters for the timing optimization model in step S110.

[0072] (b) Energy consumption upper limit constraint:

[0073] in, Indicate process i Total energy consumption (including mechanical power energy consumption and thermal process modification energy consumption).

[0074] This indicates the maximum permissible total energy consumption.

[0075] The total energy consumption of each process (mechanical power energy consumption + thermal process modification energy consumption, from the output of the energy consumption prediction model of step S2, at the granularity of each process) is summed and then compared with a threshold. (Dynamically generated by the multi-objective optimization algorithm in step S5 based on the carbon quota policy) In comparison, a logistic regression (LR) model is used to predict energy consumption trends 0.5 steps in advance, forcibly intercepting processes that exceed energy consumption limits. The interception signal is sent in real time to the IoT system in step S1 via the OPC UA protocol, triggering the equipment to reduce power (refer to the PID control parameters in step S105) to avoid total energy consumption exceeding the limit due to a surge in energy consumption in local processes. Energy consumption anomalies in the motor assembly process can be warned in advance, and the interception record is fed back to step S4 to correct the carbon emission accounting model.

[0076] (c) Carbon emission ceiling constraints

[0077] in, This indicates the maximum permissible carbon emissions.

[0078] The real-time carbon emissions (sampling frequency 1Hz) dynamically calculated in step S4 are compared with the threshold. (Based on the process-level thresholds of the enterprise's annual carbon quota decomposition, updated monthly by the optimization algorithm in step S5) Dynamic comparison is performed. When the threshold is exceeded for three consecutive sampling periods, inter-process collaborative optimization (adjusting the process sequence or equipment load allocation) is triggered through the time-series optimization model in step S110. The collaborative optimization instruction calls the scheme with the highest carbon emission weight (ω2=0.7) in the Pareto optimal solution set of step S5 to ensure that the production process complies with the carbon quota policy requirements. At the same time, the optimization results are sent back to step S2 to update the carbon emission prediction parameters of the digital twin model.

[0079] Example 1: Application of Algorithm for Energy Consumption Optimization in Automobile Chassis Assembly Production Line 1. Basic parameter settings: Number of processes =5 (Welding, Painting, Mechanical Assembly, Inspection, Packaging); Electricity carbon intensity factor =0.58kgCO2e / kWh, fuel gas carbon intensity factor =2.1kgCO2e / m³ 3 ; Maximum production cycle =240 minutes, maximum total energy consumption =8000kWh; 2. Digital twin modeling and energy consumption prediction: Mechanical power energy consumption: Welding robots =15kW (standby) + 30kW (dynamic loading), overall efficiency coefficient =0.82, predicted energy consumption =4200 kWh; Thermal process energy consumption: Specific heat capacity of materials in the coating drying chamber =1200J / (kg·K), mass =500kg, target temperature rise =180K, predicted heat energy consumption = 1.2 × 10 8 J; 3. Carbon emission calculation: Carbon emissions from mechanical power = 4200 kWh × 0.58 kg CO2e / kWh = 2436 kg CO2e; Carbon emissions from thermal processes = (1.2 × 10⁻⁶) 8 J / 3.6×10 6 J / kWh) × 2.1kgCO2e / kWh = 70kgCO2e, total carbon emissions =2506kgCO2e; 4. Dual-objective optimization: Set weighting coefficients =0.4 (energy consumption) =0.6 (carbon emissions), the NSGA-II algorithm generates the Pareto optimal solution: energy consumption 7200kWh, carbon emissions 2100kgCO2e; 5. Dynamic control effect: In-process compensation: The real-time power of the welding equipment is 54kW, the predicted power is 50kW, the deviation is e(t)=4kW, the PID control quantity is u(t)=3.2kW, the corrected power is 53.2kW, and the energy consumption of a single process is reduced. Inter-process synergy: The optimized process sequence is "mechanical assembly → welding → painting → detection → packaging", which reduces the energy consumption by 280 kWh and shortens the production cycle by 22 minutes; 6. Implementation results: total energy consumption is reduced by 10%, carbon emissions are reduced by 16%, production efficiency is improved by 9%, and the requirements of GB / T23331 energy management system are met.

[0080] Although the embodiments of the present application have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for dynamic monitoring and energy-saving and carbon-reducing regulation of energy consumption in the production process of general assembly construction, characterized in that, The method comprises the following steps: S1, collecting data related to energy consumption of each process in the industrial production cycle; S2, constructing an energy consumption prediction model, wherein the energy consumption prediction model comprises a mechanical power energy consumption prediction model and a thermal process improvement energy consumption prediction model; S3, inputting the data collected in step S1 into the energy consumption prediction model constructed in step S2 to obtain relevant energy consumption prediction data; S4, converting the energy consumption prediction data into carbon emissions, and calculating the total carbon emissions based on the carbon intensity coefficient of the energy type; S5, inputting the total carbon emissions into a multi-objective genetic algorithm to generate a Pareto optimal solution set, and obtaining the best combination of energy consumption and carbon emissions.

2. The method according to claim 1, wherein the method is characterized by, The expression of the mechanical power energy consumption prediction model is: ； wherein, Pmech is the total mechanical power consumption; N is the total number of procedures; the number of independent devices involved in the ith process step; represents the overall efficiency coefficient of the equipment j in the process i; Ttotai is the total duration of the i-th process; Pbasej,i represents the base standby power of device j in process i; represents the dynamic incremental power load of equipment j in process i; a function representing the activation state of device j in process i; represents the total energy consumption in standby and in processing.

3. The method according to claim 2, wherein the method is characterized by, The calculation formula is: ； wherein represents the motor efficiency; represents the process matching efficiency; denotes the aging attenuation factor.

4. The method according to claim 2, wherein, The calculation is made by the following formula: ； wherein represents the total energy consumption in standby mode; represents the processing energy consumption; , respectively represent the actual processing start and end times of equipment j in process i.

5. The method according to claim 1, wherein the method is characterized by, The expression of the thermal process improvement energy consumption prediction model is: ； wherein, represents the total energy consumption of the thermal process modification; represents the total number of processes; represents the number of devices of the i-th process; Gj,i represents the sensible heat energy consumption of device j in process i; represents the radiative heat loss of device j in process i.

6. The method according to claim 5, wherein the method is characterized by, The specific calculation formula is as follows: ； wherein, Cp, represents the constant pressure specific heat capacity of the material being processed by device j in process i; represents the mass of material processed by device j in one pass through process i; represents the target temperature rise absolute value of the apparatus j in the process i; represents the limit temperature of the device j in the process i; represents the reference temperature of the device j in the process i.

7. The method according to claim 5, wherein the method is characterized by, The specific calculation formula is as follows: ； wherein εj,i represents the overall emissivity of the device j in the process i; S is the Stefan-Boltzmann constant; represents the effective radiation area of device j in process i; represents the average temperature of the operation of the device j in the process i; Tj represents the ambient temperature of the device j in the process i; represents the effective heating time of the device j in the process i.

8. The method according to claim 1, characterized in that, The total carbon emissions are obtained by converting the energy consumption prediction data into carbon emissions and calculating based on the carbon intensity coefficient of the energy type, and the calculation formula is as follows: ； wherein, represents the total carbon emissions; Pm(t) represents the mechanical power consumption at time t; represents the regional grid carbon emission factor at time t; represents the thermal process improvement energy consumption at time t; Btu / scf represents the heating value of natural gas; represents the carbon emission factor for natural gas.

9. The method according to claim 1, wherein the method is characterized by, The expression of inputting the total carbon emissions into a multi-objective genetic algorithm to generate a Pareto optimal solution set is: ； is the minimum value symbol; a weight coefficient representing an energy consumption target; represents the total energy consumption; represents the historical maximum energy consumption; a weight coefficient representing a carbon emission target; represents the total carbon emissions; The largest carbon emissions in history; s.t. is a constraint condition symbol; represents the production cycle time; maximum production cycle time allowed; Product quality indicators; Minimum product quality indicators allowed.

10. The method according to claim 1, wherein the method is characterized by, Further comprising S6, dynamic feedback and compensation: S6-1, real-time power according to mechanical power deviation from the twin model prediction value dynamic adjustment of device parameters to enable energy consumption error to converge within a threshold value, while considering dynamic accounting results of carbon emissions to optimize adjustment strategies; S6-2, minimizing the comprehensive target of total assembly process idle energy consumption and carbon emissions by optimizing the process enabling sequence and time, while meeting the process dependency relationship and resource constraints.

Images