Line resource scheduling management system based on energy efficiency data

Abstract

The present application relates to the technical field of intelligent manufacturing and production scheduling, in particular to a production line resource scheduling management system based on energy efficiency data, comprising: a data acquisition step: deploying sensing end to capture transient changes to obtain original operation data; a state quantization step: analyzing energy efficiency nonlinear decay law, and constructing an energy efficiency decay dynamics model in the form of a mixed state differential equation set; a drift prediction step: predicting energy efficiency time-varying nonlinear drift trend, and identifying energy efficiency critical low value points; a sequence generation step: generating candidate job sequences using a hybrid generation strategy, and calculating energy consumption entropy of the job sequences to evaluate rhythm matching degree; a global optimization step: combining drift trend and energy consumption entropy to calculate global energy efficiency-flux coupling ratio, and screening optimal job sequences; the present application can actively avoid equipment thermal saturation or precision drift, and reduce hidden energy consumption and equipment loss.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and production scheduling technology, specifically a production line resource scheduling and management system based on energy efficiency data. Background Technology

[0002] In high-precision machining scenarios with continuous variable loads, production line equipment exhibits complex energy consumption fluctuations when performing differentiated tasks over a long period of time. Existing resource scheduling solutions generally treat equipment energy efficiency as a static constant, estimating energy consumption and scheduling tasks based solely on equipment rated parameters or a single historical average, lacking the ability to dynamically perceive the evolution of the equipment's physical state.

[0003] This static processing method ignores the fatigue and recovery mechanism of equipment as a physical entity under continuous operation. It cannot capture the nonlinear energy efficiency decay and recovery hysteresis effect caused by thermal saturation, accuracy drift or mechanical wear. Existing systems have difficulty accurately identifying the critical low point of energy efficiency and often forcibly assign high load tasks before the equipment has recovered to a steady state. This leads to cumulative errors in energy consumption estimation, resulting in increased implicit energy consumption, excessive equipment wear and yield reduction due to thermal deformation. Furthermore, it cannot avoid the mechanical shock risk caused by drastic power fluctuations in the work sequence.

[0004] Therefore, how to construct a dynamic model of energy efficiency degradation that conforms to physical characteristics based on real-time operating data, accurately predict the time-varying nonlinear drift trend of energy efficiency, and generate a globally optimal resource allocation scheme that can actively avoid energy efficiency inflection points and take into account output throughput and equipment health status has become an urgent technical problem to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a production line resource scheduling and management system based on energy efficiency data. Specifically, the technical solution of this invention includes:

[0006] The data acquisition module is configured to collect real-time operating status data of production line equipment, and obtain raw operating data including load intensity, instantaneous power and environmental parameters.

[0007] The state quantification module is configured to analyze the energy efficiency change characteristics of the equipment under continuous operation based on the original operating data, extract the energy efficiency decay coefficient and energy efficiency recovery lag time, and construct an energy efficiency decay dynamic model based on the energy efficiency decay coefficient and energy efficiency recovery lag time.

[0008] The drift prediction module is configured to predict the time-varying nonlinear drift trend of equipment energy efficiency under different operating durations based on the energy efficiency decay dynamic model and combined with the load requirements of the production tasks to be scheduled, and identify the critical low point of energy efficiency index below the preset threshold.

[0009] The sequence generation module is configured to generate multiple candidate job sequences based on production delivery constraints, and calculate the job sequence energy consumption entropy corresponding to each candidate job sequence, which characterizes the energy consumption fluctuation characteristics, in order to evaluate the matching degree between the task sequence and the equipment energy efficiency rhythm.

[0010] The global optimization module is configured to combine the time-varying nonlinear drift trend of energy efficiency with the energy consumption entropy of the job sequence, calculate the global energy efficiency-flux coupling ratio that characterizes the comprehensive benefits of output and energy consumption, and select the optimal job sequence from the candidate job sequences based on the global energy efficiency-flux coupling ratio to generate a resource allocation scheme that can avoid the critical low point of energy efficiency.

[0011] Preferably, the method for extracting the energy efficiency degradation coefficient includes:

[0012] Select a segment of historical operating data of the equipment under a specific continuous load intensity;

[0013] Calculate the rate of increase of energy consumption per unit output per unit time within this segment;

[0014] The rate of rise is used as the energy efficiency attenuation coefficient under the corresponding load intensity;

[0015] Construct a mapping library between load intensity and energy efficiency degradation coefficient;

[0016] Based on the preset load intensity of the current production tasks to be scheduled, the corresponding energy efficiency attenuation coefficient is matched from the mapping relationship library to characterize the rate at which the equipment enters the low energy efficiency zone.

[0017] Preferably, the method for extracting the energy efficiency recovery lag time includes:

[0018] Identify the moment when a device switches from a high-load state to a low-load or standby state;

[0019] The time required for core energy efficiency indicators to recover to the baseline steady-state value from the moment of switching;

[0020] The time length is marked as the energy efficiency recovery lag time;

[0021] Core energy efficiency indicators include spindle current stability indicators or thermal balance state indicators.

[0022] Preferred methods for predicting time-varying nonlinear drift trends in energy efficiency include:

[0023] Obtain the initial energy efficiency status of the current device;

[0024] Based on the energy efficiency attenuation coefficient, the energy efficiency decline trajectory of the equipment under continuous load is simulated;

[0025] If a state switching node is detected in the work sequence, the energy efficiency decline trajectory is compensated and corrected based on the energy efficiency recovery lag time, and a dynamic energy efficiency curve containing the recovery cycle is generated.

[0026] The dynamic energy efficiency curve is used as a time-varying nonlinear drift trend of energy efficiency.

[0027] Preferably, the method for calculating the energy entropy of a job sequence includes:

[0028] Obtain the estimated energy consumption value for each task in the candidate job sequence;

[0029] The projected energy consumption is converted back to power per unit time and then expanded along the time axis to construct a time series distribution of the projected energy consumption.

[0030] Calculate the standard deviation or variance of the time series distribution as a measure of the degree of volatility;

[0031] The intensity of fluctuations is normalized and mapped to a preset numerical range to obtain the energy consumption entropy of the work sequence;

[0032] Among them, the lower the value of the energy consumption entropy of the task sequence, the smoother the energy consumption fluctuation of the task sequence in the time dimension, and the higher the degree of matching with the energy efficiency breathing rhythm of the equipment.

[0033] Preferably, the method for calculating the global energy efficiency-flux coupling ratio includes:

[0034] Obtain the effective output throughput of the candidate job sequence;

[0035] Calculate the total cumulative energy consumption of the candidate job sequence over the entire cycle;

[0036] Obtain the preset equipment health loss weighting factor;

[0037] The product of the total lifecycle cumulative energy consumption and the weighting factor of equipment health loss is calculated as the comprehensive energy cost.

[0038] Calculate the ratio of effective output flux to comprehensive energy consumption cost, and use this ratio as the global energy efficiency-flux coupling ratio.

[0039] Preferably, the method for generating a resource allocation scheme that can avoid the critical low point of energy efficiency includes:

[0040] Preset energy efficiency threshold;

[0041] When the drift prediction module predicts that the energy efficiency level of the device is lower than the energy efficiency threshold, it is marked as a potential energy efficiency critical low point.

[0042] In the sequence generation module, low-load tasks or short-time standby instructions are inserted into the preceding time window for potential energy efficiency critical low points to trigger the energy efficiency recovery mechanism.

[0043] Recalculate the global energy efficiency-flux coupling ratio after inserting low-load tasks or short-term standby instructions until the job sequence corresponding to the maximized global energy efficiency-flux coupling ratio is obtained.

[0044] Preferably, it also includes a dynamic feedback adjustment module, used for:

[0045] During the execution of the optimal job sequence, the actual energy efficiency value of the equipment is monitored in real time;

[0046] Calculate the deviation between the actual energy efficiency value and the predicted time-varying nonlinear drift trend of energy efficiency;

[0047] If the deviation exceeds the preset safety tolerance, a rescheduling command is triggered, and the parameters in the energy efficiency degradation dynamics model are updated using the current actual energy efficiency value.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. This invention constructs a dynamic model of energy efficiency decay that conforms to physical characteristics and is dimensionally consistent. It explicitly defines the evolution relationship of energy efficiency state with time and load using a system of mixed-state differential equations and introduces a benchmark unit energy consumption as a normalized denominator to ensure the consistency of the physical meaning of the model calculation. By establishing a nonlinear mapping relationship between load intensity and energy efficiency decay coefficient and combining it with the dynamic calibration of accelerated aging factor by ambient temperature, it achieves accurate prediction of the time-varying nonlinear drift trend of energy efficiency of equipment under continuous variable load operation. This effectively breaks through the limitation of treating energy efficiency as a static constant or simple average value in traditional scheduling and avoids production scheduling failures caused by estimation errors.

[0050] 2. This invention quantifies the energy efficiency recovery lag time, identifies the physical recovery cycle of equipment after switching from high load to low load, and introduces spindle current stability or thermal balance state as the core judgment index to establish a recovery mechanism based on physical inertia. By actively inserting low-load tasks or short-term standby instructions within the time window preceding the predicted energy efficiency critical low point, the energy efficiency recovery mechanism is triggered. By adopting a proactive intervention strategy of retreating to advance, the invention effectively avoids equipment thermal saturation or accuracy drift caused by continuous high-load operation, ensuring processing yield while preventing the risk of forced shutdown due to deep fatigue.

[0051] 3. This invention introduces the energy consumption entropy index of the task sequence to quantitatively evaluate the energy consumption fluctuation characteristics of candidate task sequences. It also employs time-domain resampling projection and dynamic normalization strategies to ensure that the evaluation results truly reflect the matching degree between the task sequence and the energy efficiency rhythm of the equipment. By constructing a global energy efficiency-flux coupling ratio that includes a weighted factor for equipment health loss, it incorporates output efficiency, energy consumption, and equipment aging costs into a unified optimization framework, thereby selecting the optimal task sequence with smooth energy consumption transition. This effectively avoids the mechanical shock caused by drastic power jumps and maximizes the comprehensive benefits of output throughput and equipment health lifespan.

[0052] 4. This invention uses a dynamic feedback adjustment module and an online parameter update mechanism to monitor the deviation between the actual energy efficiency value and the predicted trend in real time. It also uses attenuation phase gating logic to filter effective data and combines recursive least squares method to adaptively calibrate the dynamic model parameters. This mechanism endows the system with the ability to self-evolve in response to physical aging of equipment and changes in the environment, solves the technical problem that static models gradually fail due to equipment characteristic drift after long-term operation, and ensures the prediction accuracy and decision reliability of the resource scheduling system throughout its entire life cycle. Attached Figure Description

[0053] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0054] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0056] Example 1:

[0057] Please see Figure 1 The production line resource scheduling and management system based on energy efficiency data includes: a data acquisition module, configured to collect real-time operating status data of production line equipment and obtain raw operating data including load intensity, instantaneous power and environmental parameters;

[0058] The state quantification module is configured to analyze the energy efficiency change characteristics of the equipment under continuous operation based on the original operating data, extract the energy efficiency decay coefficient and energy efficiency recovery lag time, and construct an energy efficiency decay dynamic model based on the energy efficiency decay coefficient and energy efficiency recovery lag time.

[0059] The drift prediction module is configured to predict the time-varying nonlinear drift trend of equipment energy efficiency under different operating durations based on the energy efficiency decay dynamic model and combined with the load requirements of the production tasks to be scheduled, and identify the critical low point of energy efficiency index below the preset threshold.

[0060] The sequence generation module is configured to generate multiple candidate job sequences based on production delivery constraints, and calculate the job sequence energy consumption entropy corresponding to each candidate job sequence, which characterizes the energy consumption fluctuation characteristics, in order to evaluate the matching degree between the task sequence and the equipment energy efficiency rhythm.

[0061] The global optimization module is configured to combine the time-varying nonlinear drift trend of energy efficiency with the energy consumption entropy of the job sequence, calculate the global energy efficiency-flux coupling ratio that characterizes the comprehensive benefits of output and energy consumption, and select the optimal job sequence from the candidate job sequences based on the global energy efficiency-flux coupling ratio to generate a resource allocation scheme that can avoid the critical low point of energy efficiency.

[0062] This embodiment adopts an analogy between metabolic homeostasis in biology and exercise physiology, treating the device as a dynamic living organism with fatigue and recovery mechanisms. The system's data acquisition module, as the sensing end, is deployed at key nodes such as the spindle motor, servo driver, and cooling system, with a set acquisition frequency of [missing information]. The above method captures transient changes to obtain raw operational data. ;

[0063] The state quantization module utilizes this data to analyze the nonlinear energy efficiency degradation law of the equipment under continuous operation, and extracts the energy efficiency degradation coefficient that characterizes the rate of energy efficiency decline. And the energy efficiency recovery hysteresis time, which characterizes the physical time required to recover to steady state after switching from high load to low load. Construct a description of energy efficiency status Over time and load The dynamic model of energy efficiency degradation evolution; where the energy efficiency degradation coefficient is... With load The specific functional relationship and its acquisition method will be described in detail in Example 2. In this example, the relationship is directly called to construct the model.

[0064] Specifically, in order to overcome the functional limitations and correct the dimensional mismatch problem in the original model, the hybrid state refers to the hybrid dynamic characteristics of the device's energy efficiency state evolution in the continuous time dimension, coupled with the discrete logical states caused by the switching of different task loads, namely processing state, standby state, and recovery state.

[0065] In this embodiment, the dynamic model is defined as the following system of mixed-state differential equations, and its calculation formula is as follows:

[0066]

[0067] in, A load threshold for determining whether equipment has entered the recovery period, such as 5% of the rated load; The specific value is determined based on the reference power fluctuation range of the equipment spindle during no-load operation, aiming to distinguish between effective cutting load and no-load loss, and avoid misjudging normal no-load loss as energy efficiency degradation.

[0068] This formula introduces a benchmark unit energy consumption. ,unit: As a normalized denominator, it is used to offset the energy efficiency degradation coefficient. physical units This ensures the dimensionless rate of change of state on the left side of the equation. The terms on the right-hand side are strictly dimensionally consistent; the mathematical expression explicitly defines the input parameters. With state variables The deterministic evolutionary relationship between them ensures the reproducibility of the model construction process and the correctness of the physical principles; the drift prediction module, as the prediction center, uses this model to predict the time-varying nonlinear drift trend of the equipment's energy efficiency under different operating durations. And identify curves that are below a preset threshold. The critical low point of energy efficiency;

[0069] Based on this, the sequence generation module generates a set of candidate job sequences according to production delivery constraints. Specifically, to avoid the singularity of the solution space and ensure the executability of the scheduling algorithm, the system adopts a hybrid generation strategy of heuristic rules + random perturbation: an initial seed sequence is generated based on the earliest delivery date (EDD) and shortest processing time (SPT) rules; a random exchange mutation operation is performed on the seed sequence, that is, the execution order of two tasks in the sequence is randomly exchanged, thereby expanding the generation to include... For example, a set of 50 differentiated candidate job sequences. ;

[0070] The system calculates the job sequence energy entropy for each sequence. This metric aims to assess the matching degree between the task sequence and the energy efficiency rhythm of the equipment; the global optimization module calculates the global energy efficiency-flux coupling ratio, which characterizes the overall benefits of output and energy consumption. Filter out Maximize the optimal job sequence ;

[0071] In this embodiment, under the scenario of high-precision machining with continuous variable load, an energy efficiency decay and recovery model that conforms to physical characteristics and is dimensionally consistent is constructed, which breaks through the limitation of treating energy efficiency as a static constant in traditional scheduling. The system can accurately predict and actively avoid equipment thermal saturation or precision drift caused by continuous high-load operation, and significantly reduce hidden energy consumption and equipment wear while ensuring delivery time.

[0072] Example 2:

[0073] The method for extracting the energy efficiency attenuation coefficient includes: selecting a historical operating data segment of the equipment under a specific continuous load intensity; calculating the rate of increase of unit output energy consumption per unit time within the segment; using the rate of increase as the energy efficiency attenuation coefficient under the corresponding load intensity; constructing a mapping relationship library between load intensity and energy efficiency attenuation coefficient; and matching the corresponding energy efficiency attenuation coefficient from the mapping relationship library based on the preset load intensity of the current production task to be scheduled, which is used to characterize the rate at which the equipment enters the low energy efficiency zone.

[0074] The system selects equipment under a specific continuous load intensity The system uses historical operational data segments covering the time interval from cold start-up or steady-state operation until a significant decline in energy efficiency. To quantify the degree of energy efficiency degradation, the system calculates the rate of increase in energy consumption per unit output per unit time and defines it as the energy efficiency degradation coefficient. ;

[0075] In this process, the energy consumption per unit output Specifically defined as the real-time instantaneous power of the device. The ratio of the production rate to the effective output rate; in machining scenarios, this output rate is expressed as the material removal rate (MRR), i.e. This eliminates the impact of machining process differences on energy consumption measurement; to prevent energy consumption during non-cutting strokes, such as rapid positioning or tool changes. Numerical singularities arise, and the system introduces a state-gated filtering mechanism: only when real-time... Greater than the preset effective cutting threshold, such as Only when, The calculation and sampling ensure that the extracted attenuation coefficient purely reflects the degradation of processing energy efficiency; furthermore, to avoid... Slightly above the threshold but still extremely low, causing When numerical values ​​explode, the system uses a soft threshold function for smoothing: setting an upper limit for the smooth transition range. ,when When the correction formula is used, it is:

[0076]

[0077] Suppressing noise amplification at low speeds; the calculation formula is as follows:

[0078]

[0079] For discrete historical operating data, the specific calculation steps are as follows: Select the steady-state time window under the load intensity and extract the corresponding unit output energy consumption sequence points. Linear regression fitting was performed using the least squares method, and the slope of the fitted line was determined as follows: To eliminate data noise interference;

[0080] in, For load strength The energy efficiency degradation coefficient at that time is derived from historical data fitting and its physical meaning is the rate of energy efficiency decline, with units of 1000 kJ / m². ; Energy consumption per unit of output, derived from the ratio of real-time power to output rate; The continuous operation time is derived from the system clock; A specific load intensity level, derived from equipment control commands;

[0081] System build load strength With energy efficiency attenuation coefficient mapping relationship library In the specific construction, the least squares method is used for multiple discrete sets of data. The experimental data points were fitted using a quadratic polynomial regression to obtain a continuous function, the calculation formula of which is:

[0082]

[0083] in, The fitting parameters are the load strength during this fitting process. Normalized to The range, relative to the spindle's rated power, is used to ensure the numerical stability of the least squares fitting algorithm and to unify the input domain, in order to establish that the higher the load, The larger the value, the nonlinear relationship;

[0084] Specifically, the fitting parameters are determined based on the following: at least 50 sets of steady-state energy consumption data of the equipment under different load rates are collected as the sample set; Characterizing the nonlinear loss gain under high load, the typical value range is: ; Characterizing linear load loss, the typical value range is: ;

[0085] Characterizing the drift of foundation no-load loss, the typical value range is: The required fitting accuracy determines the coefficient. If the value is lower than this, recalibration is required. Simultaneously, the system will operate the equipment under standard rated conditions, specifically defined as: a constant load of 80% of the rated torque, tool flank wear VB < 0.05mm (indicating a new tool), and ambient temperature controlled within [specific range]. The measured unit output energy consumption value is recorded as the baseline unit energy consumption. , among which, unit In this context, "unit" refers to a standardized output unit, specifically defined in a cutting scenario as the amount of material removed per unit volume. ; The physical meaning of represents the net electrical energy consumed by the equipment to remove a unit volume of material under ideal conditions without thermal fatigue and mechanical wear; it serves as a physical reference benchmark for dimensionless processing in the subsequent drift prediction module.

[0086] Preset load intensity in response to current pending production tasks The system retrieves data from the mapping database. Matching the corresponding ;like If there is no direct corresponding record in the database, then a continuous function is used. The calculated value is used for subsequent drift prediction.

[0087] In this embodiment, under the scenario of mixed production of multiple varieties and small batches, the contribution of different processing tasks to loss is accurately quantified by establishing a nonlinear mapping relationship between load and attenuation coefficient. This method provides a quantitative basis for predicting the inflection point of energy efficiency and effectively avoids the cumulative error caused by the traditional one-size-fits-all energy consumption estimation.

[0088] Example 3:

[0089] The method for extracting the energy efficiency recovery lag time by the state quantification module includes: identifying the switching moment when the equipment switches from a high-load state to a low-load or standby state; monitoring the time length required for the core energy efficiency indicators to recover to the reference steady-state value from the switching moment; marking the time length as the energy efficiency recovery lag time; the core energy efficiency indicators include the spindle current stability indicator or the thermal balance state indicator.

[0090] The system identifies the moment when equipment switches from a high-load state, such as full-load cutting, to a low-load or standby state. The system monitors core energy efficiency indicators from From the point of recovery to the reference steady-state value Required time length; here, the reference steady-state value It is not a static constant, but rather the duration of the device's most recent long-term standby. The arithmetic mean of the core indicators during the period, and based on the current ambient temperature. Dynamic drift compensation is performed, and the compensation formula is as follows: An adaptive threshold is then determined to eliminate the interference of environmental temperature changes on the decision logic;

[0091] It is important to note the dynamic threshold at this physical layer. Numerically, this corresponds to the normalized energy efficiency state in the subsequent drift prediction model. This establishes a mapping anchor between physical monitoring data and a dimensionless mathematical model, resolving the logical connection issue between physical benchmark drift and model setpoints. In this process, the core energy efficiency indicators specifically selected are either the spindle current stability index reflecting the frictional resistance of mechanical components or the thermal equilibrium index reflecting the temperature state of key components. Specifically, at the algorithm implementation level, the spindle current stability index... Defined as the non-cutting retraction phase of the equipment, i.e. The formula for calculating the sliding average reference value of the spindle load current during command execution is as follows:

[0092]

[0093] in, To smooth the window size, take Each sampling point is used to filter out electromagnetic noise. Feedback for servo drivers Shaft torque current component; this indicator can sensitively reflect changes in mechanical frictional resistance caused by thermal deformation, and frictional torque and It is directly proportional, thus serving as a quantitative proxy variable for determining physical thermal equilibrium; the system marks this time length as the energy efficiency recovery lag time. The calculation formula is as follows:

[0094]

[0095] in, Energy efficiency recovery lag time, derived from monitoring calculations, physically refers to the lag period of equipment thermal inertia or stress release, and is measured in units of... ; For core energy efficiency indicators to enter The error band and the moment when it remains stable; where the error band threshold is... Set as the reference steady-state value of Furthermore, the criterion for maintaining stability is that the indicator remains within the error band for a duration exceeding a preset anti-noise time window, such as... ; This refers to the moment when the load decreases by a step.

[0096] In this embodiment, the recovery lag effect caused by the physical inertia of the equipment is quantified in the process switching scenario of precision manufacturing. By introducing this lag time parameter, the system can avoid forcibly loading high-precision tasks before the equipment has completed its physiological recovery, thereby preventing the yield reduction and rework energy consumption caused by residual thermal deformation.

[0097] Example 4:

[0098] The method for predicting the time-varying nonlinear drift trend of energy efficiency includes: obtaining the initial energy efficiency state of the current equipment; simulating the energy efficiency decline trajectory of the equipment under continuous load based on the energy efficiency decay coefficient; if a state switching node is detected in the work sequence, compensating and correcting the energy efficiency decline trajectory based on the energy efficiency recovery lag time to generate a dynamic energy efficiency curve containing the recovery period; and using the dynamic energy efficiency curve as the time-varying nonlinear drift trend of energy efficiency.

[0099] The system obtains the initial energy efficiency status of the current equipment by monitoring data in real time. ,scope To ensure the executability of the prediction model in a computer and to accurately handle state transitions, the system uses a discrete-time stepping method instead of abstract continuous integration, and sets a time step. ,For example The specific iterative logic for drift prediction is as follows: For each time step in the prediction time domain... : Get the current job sequence Preset load intensity at any time ;like Under high-load processing conditions, the energy efficiency state evolves according to the decay dynamics. To ensure the consistency of physical dimensions and to correspond with the modified model in the implementation of Example 1, the system introduces the benchmark unit energy consumption determined in Example 2. Energy consumption per unit output under standard operating conditions when the equipment is in brand new condition. Unit: As the normalized denominator, the calculation formula is as follows:

[0100]

[0101] in, The energy efficiency degradation coefficient extracted in the aforementioned steps, physically representing the rate of increase in energy consumption per unit output, is expressed in units of: ; This is a dimensionless accelerated aging factor, with a default value of 1.0; the specific determination method is as follows: the system collects ambient temperature data. And according to the formula:

[0102]

[0103] in, Standard room temperature, For the preset temperature sensitivity coefficient, such as The coefficient The determination is based on the thermal aging derating curve provided by the equipment manufacturer, or on the Arrhenius equation through accelerated life experiment fitting, which characterizes the gain ratio of the aging rate for every 1°C increase in temperature, thereby quantifying the accelerating effect of high temperature environment on component aging.

[0104] Specifically, The range of values ​​is usually 100. Its calibration method is based on the Arrhenius model: By using two different high temperature points, for example and Accelerated aging tests were conducted to measure the rate of change in energy efficiency degradation and to determine the activation energy. Thus, the linearization coefficients are determined. For precision CNC machine tool spindles, empirical values ​​are usually used. That is, for every increase in temperature The aging rate increases by approximately ;

[0105] This formula accurately describes the energy efficiency status. The dimensionless decrease resulting from the accumulation of physical losses; it is worth noting that this recursive model is a first-order linearized approximation of the energy efficiency degradation process, and its effectiveness is based on the relatively small energy efficiency drift within a single production cycle, for example... The engineering assumptions ensure both computational efficiency and the accuracy requirements of the scheduling layer.

[0106] If a state transition node is detected in the job sequence, i.e. It is in a low-load or standby state, and the duration meets the energy efficiency recovery lag time. If the triggering condition is met, compensation correction will be initiated; in this step, the system will restore the physical-level energy efficiency lag time. Defined as the time required to recover to the steady-state error band, it is converted into the relaxation time constant in the mathematical model. Based on the characteristics of a first-order system, take The coefficient 3 is chosen because the first-order exponential response can reach more than 95% of its steady-state value at a time constant of 3, thus making the relaxation process in the mathematical model consistent with the time observed by physics to recover to the steady-state error band. Logical alignment;

[0107] The derivation of this relationship is based on the step response characteristic formula of a first-order linear time-invariant system. When time When the system response value reaches the steady-state value This is related to the time required to recover to the steady-state error band, i.e., to reach 95% of the steady-state value, as defined at the physical level. Mathematically, they correspond exactly; the energy efficiency state recovers according to the exponential relaxation mechanism, and its calculation formula is:

[0108]

[0109] in, This is the reference steady-state energy efficiency value of the equipment; in this embodiment, although this value is set as a normalized constant in the dimensionless model. However, its physical meaning strictly corresponds to the adaptive threshold determined by dynamic drift compensation in the implementation of Example 3. ;

[0110] This mapping mechanism ensures that the complete recovery in the model does not point to a fixed theoretical value, but rather to the actual optimal thermal equilibrium state that the device can achieve under the current ambient temperature, thus solving the problem of consistency between the logical branches of the model definition and the physical reality; during mathematical calculations, the system... The mapping relationship handles the drift of the physical reference, that is, when the physical reference... When the temperature rises and the denominator shifts upward, the energy efficiency state within the model is ensured by adaptively adjusting the denominator. Always adapting In-space operations, where, for The core energy efficiency indicators that are monitored in real time;

[0111] All Moment By connecting them sequentially, a dynamic energy efficiency curve containing both fatigue decay and intermittent recovery characteristics is generated. As a time-varying nonlinear drift trend in energy efficiency;

[0112] This embodiment constructs a piecewise iterative dynamic equation and strictly constrains the dimensional mapping relationship of physical parameters, thereby achieving accurate mathematical reproduction of continuous load simulation and recovery cycle compensation, effectively avoiding the physical meaning conflict caused by simple linear superposition.

[0113] Example 5:

[0114] The method for calculating the energy consumption entropy of the job sequence in the sequence generation module includes: obtaining the expected energy consumption value corresponding to each task in the candidate job sequence; restoring the expected energy consumption value to power per unit time and expanding it according to the time axis to construct the time series distribution of the expected energy consumption value; calculating the standard deviation or variance of the time series distribution as the degree of fluctuation; normalizing the degree of fluctuation and mapping it to a preset numerical range to obtain the energy consumption entropy of the job sequence; wherein, the lower the value of the energy consumption entropy of the job sequence, the smoother the energy consumption fluctuation of the task sequence in the time dimension, and the higher the degree of matching with the energy efficiency breathing rhythm of the equipment.

[0115] The system retrieves each task from the candidate job sequence. Corresponding estimated energy consumption value To ensure the physical meaning of the data, here... The calculation formula is set as follows:

[0116]

[0117] in, For the task Under load Rated power below, The standard working hours for each task; the system constructs a time series distribution according to the task execution order. In this step, to ensure that the energy consumption entropy accurately reflects the device's breathing rhythm over time rather than simply differences in task scale, and to avoid misjudging long-duration low-energy-consumption tasks and short-duration high-energy-consumption tasks as having no fluctuations due to similar total energy consumption values, the system performs time-domain resampling projection: the estimated energy consumption value for each task is... Reduced to energy consumption per unit time, i.e., power. And according to the timeline at a fixed frequency, such as Expanding, this forms a continuous time series reflecting instantaneous load fluctuations. ;

[0118] Calculate the standard deviation of the sequence. This quantifies the severity of the fluctuations; to eliminate the influence of dimensions and adapt to the optimization model, the system will... Normalization mapping to a preset numerical range Thus, the energy consumption entropy of the job sequence can be obtained. The calculation formula is as follows:

[0119]

[0120] in, The energy consumption entropy of the job sequence originates from statistical calculations and specifically refers to a statistical index that characterizes the degree of orderliness of energy consumption fluctuations. Its mathematical essence is the power distribution dispersion after time-domain resampling and normalization. It borrows the concept of thermodynamic entropy to describe the degree of disorder in the energy state of the system. Predict the standard deviation of the energy consumption distribution for candidate sequences; These represent the extreme values ​​of energy consumption fluctuations in historical statistical data; the source and update mechanism of historical statistical data are as follows: the system maintains a database of length [length missing]. ,like A sliding window for the standard deviation of energy consumption fluctuations in historical operation sequences;

[0121] When the system starts up for the first time or when data is missing, the default settings are as follows: The theoretical maximum fluctuation value is estimated based on the rated power of the equipment; the specific estimation process is as follows:

[0122]

[0123] in, Rated power of the equipment For no-load power, assume the standard deviation of energy consumption under the worst-case scenario of drastic switching across the entire range;

[0124] As the job progresses, the newly calculated data will be updated in real time. Save to a sliding window and update dynamically. The sliding window mechanism effectively avoids the permanent distortion of extreme values ​​and outlier locking caused by sudden environmental changes.

[0125] For the preset regularization constant, such as This is used to prevent the overlap of historical extreme values. This leads to a division by zero error; value... It is based on the machine precision limit setting of the system's double-precision floating-point arithmetic, ensuring that... and When the values ​​are extremely close, the regularization term can ensure computational stability without interfering with the relative ordering of entropy values.

[0126] Specifically, considering the non-stationarity of the production environment, to prevent clipping distortion (i.e., multiple drastically fluctuating sequences exceeding historical extremes becoming indistinguishable due to forced truncation to 1.0), the system implements a batch dynamic normalization strategy: before calculating the entropy value, it iterates through the standard deviations of fluctuations of all candidate sequences in the current batch and extracts the maximum value of the batch. ;

[0127] Set a normalization upper bound At this point, when the system calculates the energy consumption entropy H of the current batch of jobs, it will change the original formula. Replace with That is, using the correction formula Calculations are performed based on this dynamic upper bound. The value ensures that even the worst sequence can remain intact. Within the interval and retaining relative gradient information, it effectively guides subsequent optimization algorithms to distinguish high-risk sequences; after calculation, the system asynchronously updates historical extreme values. ;

[0128] In energy-sensitive production scheduling scenarios, this embodiment minimizes energy entropy and tends to generate work sequences with smooth energy transitions. This strategy effectively avoids mechanical shocks and transient high energy consumption caused by drastic power jumps, making the equipment operation rhythm closer to deep breathing rather than rapid panting, thereby extending the equipment's service life.

[0129] Example 6:

[0130] The method for calculating the global energy efficiency-flux coupling ratio includes: obtaining the effective output flux of the candidate job sequence; calculating the cumulative energy consumption of the candidate job sequence over the entire cycle; obtaining the preset equipment health loss weighting factor; calculating the product of the cumulative energy consumption over the entire cycle and the equipment health loss weighting factor as the comprehensive energy cost; and calculating the ratio of the effective output flux to the comprehensive energy cost, and using the ratio as the global energy efficiency-flux coupling ratio.

[0131] The system obtains the effective output throughput of candidate job sequences per unit time. To ensure To ensure physical consistency in multi-variety mixed-flow production scenarios and avoid measurement distortion caused by differences in product specifications, here... Strictly defined as standard equivalent output ;

[0132] The system uses a pre-defined product-complexity coefficient table to convert the quantity of different types of products into equivalent quantities of standard products. This table is constructed using the standard working hour ratio method: the product with the highest historical cumulative production quantity and total cumulative output in the production line is selected as the benchmark product, with a coefficient set to 1.0. For any other product... The complexity coefficient is calculated using the following formula:

[0133]

[0134] in, This is the standard operating cycle for this product. The standard operating cycle for a benchmark product is determined by extracting theoretical processing time from simulation results of computer-aided manufacturing software and adding a preset standard auxiliary operation time, or by selecting the average time of the 10% most efficient samples in the product's historical production records as the standard value. It can accurately reflect the output rate of the value stream;

[0135] Defined as the total time span from the start of the first task to the completion of the last task in the job sequence, in units of: This includes machining time, tool change time, and all intermediate waiting times; simultaneously, based on the aforementioned predicted dynamic energy efficiency curve... The cumulative energy consumption over the entire cycle is calculated by discrete summation:

[0136]

[0137] It should be noted that here... This refers to the production tasks to be scheduled at any given time. The rated power corresponding to the load command is retrieved from the equipment specification database, rather than the actual power collected by the sensor in real time, in order to predict future energy consumption.

[0138] The key point is that, in order to respond to the technical requirement of global optimization based on the energy consumption entropy of the work sequence in Example 1, this example defines a composite weighting factor for equipment health loss; the system retrieves a preset benchmark health loss index from the equipment database. and fluctuation sensitivity coefficient And combined with the energy consumption entropy calculated from the current sequence Constructing the dynamic weighting factor that actually takes effect :

[0139]

[0140] This step concretizes the preset factor of the embodiment into a comprehensive coefficient coupled with preset parameters and dynamic variables, thereby achieving the penalty for non-stationary sequences;

[0141] Here, The range of values ​​is normalized to , The characterization equipment is in brand new condition. The characterization equipment is nearing the end of its service life; The specific value is determined by the device's current cumulative effective operating time. With respect to the manufacturer's specified design life The decision is made using the following formula:

[0142]

[0143] in, The failure rate curve shape parameter is usually taken as... To simulate the accelerated aging effect in the later stages of wear; The value is set based on the shape parameter characterizing the wear and tear failure period in the Weibull distribution. This indicates that the equipment has entered a stage of accelerated wear and tear.

[0144] Alternatively, it can be obtained through periodic precision inspection results, such as spindle rotation error, by looking up a table. For example: rotation error hour gyration error hour ;

[0145] coefficient The settings were obtained through experimental calibration, specifically by measuring the vibration spectrum changes of the spindle bearing under different energy consumption entropy levels and calculating the proportional relationship between the spectral centroid drift and the entropy value; for example... The physical meaning is: the more drastic the energy consumption fluctuation, that is... The larger the value, the higher the weight of aging and wear-out costs for the equipment, and the older the equipment, the more significant the cost. The larger the value, the heavier the penalty for additional losses caused by fluctuations;

[0146] The system calculates the product of the total energy consumption over the entire lifecycle and the dynamic weighting factor, which is taken as the comprehensive energy cost. The factor is defined as the incremental part, and the calculation formula is set as follows:

[0147]

[0148] This embodiment employs the latter logic to align with physical intuition; it calculates the ratio of effective output flux to comprehensive energy cost, and uses a formula to restore the output flux to the standard equivalent total output for calculation, thereby eliminating the nonlinear interference of time span on energy efficiency evaluation. The physical meaning is clearly defined as the standard output per unit of comprehensive cost, and the specific formula is:

[0149]

[0150] A time span factor is introduced to correct the dimensions, resulting in the global energy efficiency-flux coupling ratio; specifically, for the original ratio... Dimensions of existence This embodiment addresses the problem of lacking intuitive physical meaning by effectively generating flux. Integral reduction over time to standard equivalent output ,Right now Alternatively, it can be understood as amortizing the energy cost in the denominator into a power cost; the corrected calculation formula is:

[0151]

[0152] at this time, The physical unit is , which is output per unit of energy consumption, accurately represents the effective output per unit of comprehensive energy consumption cost in a physical sense, thereby achieving a three-dimensional balance between output efficiency, energy consumption and equipment health.

[0153] This embodiment ensures that the scheduling scheme not only pursues high output but also actively adapts to the smooth operation rhythm of the equipment by explicitly coupling energy consumption entropy in the cost function, thereby maximizing the true comprehensive benefits.

[0154] Example 7:

[0155] The method for the global optimization module to generate a resource allocation scheme that can avoid the energy efficiency critical low point includes: preset an energy efficiency threshold; when the drift prediction module predicts that the energy efficiency level of the device is lower than the energy efficiency threshold, it is marked as a potential energy efficiency critical low point; in the sequence generation module, for the preceding time window of the potential energy efficiency critical low point, a low-load task or short-time standby instruction is inserted to trigger the energy efficiency recovery mechanism; the global energy efficiency-flux coupling ratio after inserting the low-load task or short-time standby instruction is recalculated until the job sequence corresponding to the maximized global energy efficiency-flux coupling ratio is obtained.

[0156] System preset energy efficiency threshold The drift prediction module predicts the equipment energy efficiency level. At that time, the moment Marked as a potential low point for energy efficiency; targeting Preceding time window ,in, For example, a preset search backtracking window minutes; distinct from differential step size The system attempts to insert low-load tasks, such as detection processes or short-term standby instructions, into the sequence generation module, aiming to utilize energy efficiency to recover lag time. Its characteristics trigger the energy efficiency recovery mechanism, restoring the energy efficiency status to the safe zone;

[0157] The system recalculates the global energy efficiency-flux coupling ratio after the intervention command is inserted. In this step, the specific logic for recalculation includes timeline reconstruction and full sequence replay: the system will recalculate the estimated start time of all tasks to be scheduled after the insertion point. Shift backward by an amount equal to the duration of the insertion instruction. This updates the total span time of the sequence, calculated using the following formula:

[0158]

[0159] Based on the updated timeline, the drift prediction module is rerun to generate a new dynamic energy efficiency curve. ;Although An increase in output flux will lead to The decline, but the insertion of a rest period brought about The recovery will significantly reduce the instantaneous energy consumption of subsequent high-load tasks. This reduces cumulative energy consumption over the entire lifecycle. The system finds the answer by balancing the magnitude of change in these two factors. The global maximum value;

[0160] A heuristic search algorithm is used to adjust the insertion position and duration: A set step size, such as 1 minute, is used to traverse all feasible insertion points within the preceding time window. At each insertion point, the standby duration is incremented by a set increment, such as 30 seconds. The results of each adjustment are calculated. Compare all generated The value is used to select the insertion strategy corresponding to the maximum value as the optimal solution, and this process is repeated until the maximum value is obtained. The corresponding task sequence;

[0161] In a continuous production scenario facing the risk of equipment thermal saturation, this embodiment adopts a proactive intervention strategy of retreating to advance. By intelligently inserting short periods of inefficiency or downtime, the system effectively avoids prolonged forced shutdowns or large-scale scrap generation caused by equipment entering a deep fatigue zone, thereby achieving a higher energy efficiency output ratio on a global time scale.

[0162] Example 8:

[0163] The production line resource scheduling and management system based on energy efficiency data also includes a dynamic feedback adjustment module, which is used to: monitor the actual energy efficiency value of the equipment in real time during the execution of the optimal work sequence; calculate the deviation between the actual energy efficiency value and the predicted time-varying nonlinear drift trend of energy efficiency; if the deviation exceeds the preset safety tolerance, trigger a rescheduling instruction and update the parameters in the energy efficiency decay dynamic model using the current actual energy efficiency value.

[0164] During execution, the dynamic feedback adjustment module monitors the actual energy efficiency value in real time. ; Calculate its relationship with the predicted value Deviation:

[0165]

[0166] like Exceeding the preset safety tolerance ,For example The system triggers a rescheduling instruction;

[0167] Meanwhile, in order to achieve closed-loop adaptation, the system executes a parameter update mechanism; specifically, it addresses the core load-attenuation coefficient mapping relationship in the energy efficiency degradation dynamics model.

[0168]

[0169] The system collects the latest Actual operating data within each time window ;in, The inverse calculation must follow the inverse dynamic equations consistent with the drift prediction module to ensure closed-loop physical dimensions. The calculation formula is as follows:

[0170]

[0171] in, This is the baseline unit energy consumption for the equipment. To accelerate aging, all parameters were kept consistent with those defined in Example 4; Where is the sampling time step; to suppress the interference of high-frequency noise from the field sensors on the differential calculation, the above formula... Using a sliding window, window length The original monitoring values ​​were obtained by smoothing and filtering them. The specific filtering algorithm used was the simple moving average method. That is, its calculation formula is:

[0172]

[0173] in, for The original energy efficiency state observation at time t is defined by the following formula:

[0174]

[0175] in, As a benchmark unit energy consumption, The energy consumption per unit output is calculated in real time based on the method in Example 2; when At that time, the calculated This characterizes the energy efficiency decline; based on this observation, high-frequency random noise from the sensor is eliminated; crucially, to prevent negative attenuation data generated during the recovery or standby period from contaminating the attenuation model, the system introduces attenuation phase gating before performing parameter updates. Logic: Only if the condition is met At that time, among them, Defined as the power threshold when the equipment is running under no-load conditions, it is usually set as the rated power of the spindle. This is used to distinguish between cutting and idling states. Only when the equipment is operating at effective load and the measured energy efficiency is indeed decreasing is the data point allowed to enter the regression queue; otherwise, the data point will be marked as recovery data for hysteresis calibration. And not involved in the attenuation coefficient The system updates the data; using the valid data filtered by DPG, the system employs recursive least squares method. Online update of polynomial coefficient vector To ensure the white-box reproducibility of the algorithm, the specific iterative update steps are as follows: Define the vector of parameters to be estimated, and its calculation formula is:

[0176]

[0177] The observed regression vector is calculated using the following formula:

[0178]

[0179] The Kalman gain is calculated using the following formula:

[0180]

[0181] in, Let covariance be the initial value. Set as Characterizes initial uncertainty; selection Such a large initial value is to assign higher weight to new data in the early stages of the algorithm, so that the parameter estimates can quickly converge from the initial guesses to near the true values, avoiding the long-term impact of initial bias on subsequent estimates; the updated parameter estimates are calculated using the following formula:

[0182]

[0183] The updated covariance matrix is ​​calculated using the following formula:

[0184]

[0185] To ensure algorithm stability, the system incorporates a convergence and freezing mechanism for parameter updates: calculating the trace of the prediction error covariance matrix. ,when ,like When the parameter estimation has converged, parameter updates should be stopped to prevent overfitting noise; or when a step change in equipment operating conditions is detected, such as a change in load rate. Reset This is to reactivate the algorithm's ability to track new features;

[0186] Through this explicit matrix operation, the system can automatically calibrate the prediction model as the physical characteristics of the equipment age and drift, which manifests as changes in coefficients, thus maintaining long-term scheduling prediction accuracy.

[0187] This embodiment endows the system with self-evolution capability through a clear parameter update algorithm and data validity filtering mechanism, solving the technical problem that static models gradually become ineffective after long-term operation.

[0188] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A production line resource scheduling and management system based on energy efficiency data, characterized in that, include: The data acquisition module is configured to collect real-time operating status data of production line equipment, and obtain raw operating data including load intensity, instantaneous power and environmental parameters. The state quantification module is configured to analyze the energy efficiency change characteristics of the equipment under continuous operation based on the original operating data, extract the energy efficiency decay coefficient and energy efficiency recovery lag time, and construct an energy efficiency decay dynamic model based on the energy efficiency decay coefficient and the energy efficiency recovery lag time. The drift prediction module is configured to predict the time-varying nonlinear drift trend of the equipment's energy efficiency under different operating durations based on the energy efficiency decay dynamic model and the load requirements of the production tasks to be scheduled, and to identify the critical low point of energy efficiency index below the preset threshold. The sequence generation module is configured to generate multiple candidate job sequences based on production delivery constraints, and calculate the job sequence energy consumption entropy corresponding to each candidate job sequence, which characterizes the energy consumption fluctuation characteristics, in order to evaluate the matching degree between the task sequence and the equipment energy efficiency rhythm. The global optimization module is configured to combine the time-varying nonlinear drift trend of energy efficiency with the energy consumption entropy of the job sequence, calculate the global energy efficiency-flux coupling ratio that characterizes the comprehensive benefits of output and energy consumption, and select the optimal job sequence from the candidate job sequences based on the global energy efficiency-flux coupling ratio to generate a resource allocation scheme that can avoid the critical low point of energy efficiency. Methods for extracting the energy efficiency degradation coefficient include: Select a segment of historical operating data of the equipment under a specific continuous load intensity. This segment covers the time interval from cold start or steady-state operation to a significant decrease in energy efficiency. Calculate the rate of increase of energy consumption per unit output per unit time within this segment; The rate of rise is used as the energy efficiency attenuation coefficient under the corresponding load intensity; Construct a mapping library between load intensity and energy efficiency degradation coefficient; Based on the preset load intensity of the current production tasks to be scheduled, the corresponding energy efficiency attenuation coefficient is matched from the mapping relationship library to characterize the rate at which the equipment enters the low energy efficiency zone. Methods for extracting energy efficiency recovery lag time include: Identify the moment when a device switches from a high-load state to a low-load or standby state; The time required for core energy efficiency indicators to recover to the baseline steady-state value from the moment of switching; The time length is marked as the energy efficiency recovery lag time; The core energy efficiency indicators include spindle current stability indicators or thermal balance state indicators. Methods for calculating the global energy efficiency-flux coupling ratio include: Obtain the effective output throughput of the candidate job sequence; Calculate the total cumulative energy consumption of the candidate job sequence over the entire cycle; Obtain the preset equipment health loss weighting factor; The product of the total lifecycle cumulative energy consumption and the weighting factor of the equipment health loss is calculated as the comprehensive energy cost. Calculate the ratio of the effective output flux to the comprehensive energy consumption cost, and use the ratio as the global energy efficiency-flux coupling ratio; The system also includes a dynamic feedback adjustment module for: During the execution of the optimal job sequence, the actual energy efficiency value of the equipment is monitored in real time; Calculate the deviation between the actual energy efficiency value and the predicted time-varying nonlinear drift trend of energy efficiency; If the deviation exceeds the preset safety tolerance, a rescheduling command is triggered, and the parameters in the energy efficiency degradation dynamics model are updated using the current actual energy efficiency value.

2. The production line resource scheduling and management system based on energy efficiency data according to claim 1, characterized in that, Methods for predicting the time-varying nonlinear drift trend of energy efficiency include: Obtain the initial energy efficiency status of the current device; Based on the energy efficiency attenuation coefficient, the energy efficiency decline trajectory of the equipment under continuous load is simulated; If a state switching node is detected in the work sequence, the energy efficiency decline trajectory is compensated and corrected based on the energy efficiency recovery lag time, and a dynamic energy efficiency curve containing the recovery cycle is generated. The dynamic energy efficiency curve is used as a time-varying nonlinear drift trend of energy efficiency.

3. The production line resource scheduling and management system based on energy efficiency data according to claim 1, characterized in that, Methods for calculating the energy entropy of a job sequence include: Obtain the estimated energy consumption value for each task in the candidate job sequence; The predicted energy consumption value is converted back to power per unit time and expanded along the time axis to construct a time series distribution of the predicted energy consumption value; Calculate the standard deviation or variance of the time series distribution as the degree of fluctuation. The intensity of the fluctuation is normalized and mapped to a preset numerical range to obtain the energy consumption entropy of the work sequence; The lower the value of the energy consumption entropy of the task sequence, the smoother the energy consumption fluctuation of the task sequence in the time dimension, and the higher the degree of matching with the energy efficiency breathing rhythm of the equipment.

4. The production line resource scheduling and management system based on energy efficiency data according to claim 1, characterized in that, Methods for generating resource allocation schemes that can avoid the critical low point of energy efficiency include: Preset energy efficiency threshold; When the drift prediction module predicts that the energy efficiency level of the device is lower than the energy efficiency threshold, it is marked as a potential energy efficiency critical low point. In the sequence generation module, a low-load task or short-time standby instruction is inserted into the preceding time window of the potential energy efficiency critical low point to trigger the energy efficiency recovery mechanism. Recalculate the global energy efficiency-flux coupling ratio after inserting the low-load task or the short-time standby instruction, until the job sequence corresponding to the maximized global energy efficiency-flux coupling ratio is obtained.

Images