Air compressor unit energy-saving control method and system and storage medium

Abstract

The present application relates to the technical field of industrial automation control, in particular to an air compressor unit energy-saving control method and system and a storage medium, comprising: obtaining real-time operation data of the air compressor unit; constructing an air compressor energy consumption digital twin model; obtaining a current working condition basic threshold value based on the air compressor energy consumption digital twin model; calculating a real-time air compressor unit state reference value to obtain a single machine operation time relative ratio; obtaining an operation time correction coefficient according to the single machine operation time relative ratio; obtaining a start-stop frequency correction coefficient according to the average start-stop frequency; correcting the current working condition basic threshold value based on the operation time correction coefficient and the start-stop frequency correction coefficient to obtain a current optimal control threshold value. The present application constructs a modular mechanism model, embeds physical laws as constraints; uses LSTM neural network to learn time sequence characteristics, and integrates mechanism constraints into training through a hybrid loss function, so that the model has both physical interpretability and precision.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, and specifically to an energy-saving control method, system and storage medium for air compressor units. Background Technology

[0002] In modern industrial production such as cigarette manufacturing, compressed air systems serve as a key power source, widely used in pneumatic actuation, valve driving, equipment cleaning, and process assurance. However, existing systems generally suffer from low operating efficiency and high energy consumption, especially under conditions of frequent fluctuations in air load, where energy waste is particularly pronounced. In cigarette production, the energy consumption of compressed air systems accounts for a significant proportion of a company's total energy consumption, directly impacting production costs and economic efficiency.

[0003] The air compression system commonly used in cigarette factories typically consists of three screw air compressors (model: ZR250VSD-FF, rated displacement: 40 m³ / min) and two centrifugal air compressors (model: ZH630+-8, rated displacement: 120 m³ / min). The actual air demand of the system usually fluctuates between 140 and 180 m³ / min. Due to uneven load distribution, it is common for one screw compressor to operate at a displacement below 15 m³ / min while another exceeds 40 m³ / min. This inefficient operating mode not only wastes electrical energy but also leads to frequent start-ups and shutdowns or prolonged low-load operation, accelerating unit wear and shortening equipment lifespan. Currently, the air compressors in this system communicate via a CAN bus and exchange data with the system PLC through an RS485 interface via a centrifugal compressor. However, due to limitations in the device's communication protocol, the pressure setpoint cannot be continuously adjusted via remote communication; it can only switch between a few preset pressure values. This severely restricts the system's ability to make precise adjustments based on real-time load.

[0004] Therefore, developing a control strategy that can dynamically optimize unit operation and improve the overall energy efficiency of the system to achieve energy saving and consumption reduction in the air compressor system has significant engineering application value and economic benefits. Summary of the Invention

[0005] The purpose of this invention is to provide an energy-saving control method, system, and storage medium for air compressor units, in order to solve the technical problems in the prior art, such as single control strategy, lack of equipment status perception, single dimension of anomaly detection, and lack of model evolution capability.

[0006] To achieve the above objectives, embodiments of the present invention provide an energy-saving control method for an air compressor unit, comprising: Obtain real-time operating data of the air compressor unit; Construct a digital twin model of air compressor energy consumption; The basic threshold for the current operating condition is obtained based on the digital twin model of the air compressor's energy consumption. Real-time calculation of the air compressor unit's status baseline value to obtain the relative comparison of single unit operating time; The running time correction coefficient is obtained based on the relative ratio of the single machine running time; The start-stop frequency correction coefficient is obtained based on the average start-stop frequency. The current operating condition base threshold is corrected based on the running time correction coefficient and the start / stop frequency correction coefficient to obtain the current optimal control threshold.

[0007] Optionally, constructing a digital twin model includes: Preprocess the historical operating data of the air compressor unit; A modular mechanism model is constructed based on the preprocessed data; Construct a temporal neural network model; The temporal neural network model is trained based on the aforementioned mechanism model to obtain a fusion model.

[0008] Optionally, the modular mechanism model includes: Thermodynamic sub-model, fluid dynamics sub-model, mechanical loss sub-model, and operating state sub-model; The thermodynamic sub-model is used to calculate the theoretical energy consumption and exhaust temperature of the compression process based on the ideal gas compression law. The fluid dynamics sub-model is used to calculate the pressure loss of the gas duct and pipeline system, and to correct the energy consumption calculation results of the thermodynamic sub-model. The mechanical loss sub-model is used to calculate mechanical wear and additional losses from start-stop based on cumulative operating time and start-stop frequency. The operating status sub-model is used to associate and bind the equipment's start / stop status, fault codes, and lubricating oil pressure parameters with the energy consumption model.

[0009] Optionally, training the temporal neural network model based on the mechanistic model to obtain the fusion model includes: Construct a hybrid loss function based on formula (1). (1) in, For loss function, The energy consumption value predicted by the model. This represents the actual standard energy consumption value. For mechanism constraint terms, This is the constraint coefficient.

[0010] Optionally, the air compressor unit's state baseline value is calculated in real time to obtain the relative ratio of single-unit operating time, including: Calculate the average operating time and average number of start-stop cycles of the air compressor unit; Calculate the relative ratio of single-machine running time according to formula (2). (2) in, Indicates the first The aging degree of the equipment Indicates the first The cumulative operating time of the equipment This represents the average operating time of the air compressor unit.

[0011] Optionally, obtaining the runtime correction coefficient based on the relative ratio of single-machine runtime includes: Based on the distribution of the relative ratios of single-machine running times, the value of the running time correction coefficient is determined: When the relative ratio of the individual running times of all air compressors is within the preset uniform range, it is determined that the equipment is aging uniformly, and the running time correction coefficient is set as the base value. When the relative ratio of the single running time of a single air compressor exceeds the uniform range but does not reach the significant difference threshold, and the other equipment is not lower than the preset lower limit threshold, it is determined that the aging of a single equipment is slightly higher than the average level, and the running time correction coefficient is set as the first correction value, which is greater than the base value. When the relative ratio of the single running time of a single air compressor exceeds the significant difference threshold, or when the relative ratio of the single running time of a single device is lower than the preset lower limit threshold, it is determined that the aging difference between devices is significant, and the running time correction coefficient is set as the second correction value, which is greater than the first correction value.

[0012] Optionally, the start-stop frequency correction factor obtained based on the average start-stop frequency includes: Based on the interval in which the average start-stop frequency falls, determine the value of the start-stop frequency correction coefficient: When the average start-stop frequency is within the first preset range, the start-stop is determined to be normal, and the start-stop frequency correction coefficient is set as the base value; When the average start-stop frequency is within the second preset range, it is determined that the start-stop frequency is too frequent, and the start-stop frequency correction coefficient is set to the first correction value, which is less than the base value. When the average start-stop frequency is within the third preset range, it is determined that the start-stop is too frequent, and the start-stop frequency correction coefficient is set to the second correction value, which is less than the first correction value.

[0013] Optionally, correcting the current operating condition base threshold based on the running time correction coefficient and the start / stop frequency correction coefficient to obtain the current optimal control threshold includes: The current optimal control threshold is obtained according to formulas (3) and (4). (3) (4) in, The final average threshold, This is the final single-machine priority threshold. The threshold for equal distribution, This is the priority threshold for a single machine. This is a runtime correction factor. This is a correction factor for the number of start-stop cycles.

[0014] On the other hand, the present invention also provides an energy-saving control system for an air compressor unit, the system including a processor configured to perform any of the methods described above.

[0015] In another aspect, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement any of the methods described above.

[0016] The beneficial effects of this invention are: The embodiments of the present invention introduce a runtime correction coefficient. With start-stop frequency correction factor The basic threshold of digital twins is double-corrected. Adjust the load distribution according to the differences in equipment aging to avoid aging units operating at full load for a long time; Narrow the mode switching range based on the frequency of start-stop operations to reduce unnecessary loading and unloading. Final threshold. and It has achieved a leap from single energy efficiency optimization to multi-objective synergistic optimization of energy efficiency, lifespan, and stability.

[0017] This invention constructs a modular mechanistic model encompassing thermodynamics, fluid mechanics, mechanical losses, and operational states, embedding physical laws as constraints. Simultaneously, it employs an LSTM neural network to learn the complex temporal characteristics of actual operational data, and integrates these mechanistic constraints into the training process through a hybrid loss function. The resulting digital twin model possesses both physical interpretability and data-driven accuracy, with an energy consumption simulation error ≤ ±2%.

[0018] The implementation of this invention supports incremental training based on newly acquired data after model deployment. When simulation errors exceed limits or significant changes in equipment status occur (such as major repairs or component replacements), the parameters of the front-layer network are frozen, and only the fully connected layers of the back layer are fine-tuned, enabling rapid adaptation to new operating conditions. This mechanism avoids frequent full retraining, ensuring that the digital twin maintains a high degree of consistency with the physical entity throughout its entire lifecycle.

[0019] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 A flowchart of an energy-saving control method for an air compressor unit according to an embodiment of the present invention; Figure 2 A flowchart illustrating a method for constructing a digital twin model of air compressor energy consumption according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the real-time display of the operating status and pressure setting of an air compressor according to an embodiment of the present invention. Figure 4 A scatter plot of the electrical ratio of air compressor No. 1 in 2023 and 2024 according to one embodiment of the present invention; Figure 5 A scatter plot of the electrical ratio of air compressor No. 2 in 2023 and 2024 according to an embodiment of the present invention; Figure 6 This is a scatter plot of the electrical ratio of air compressor No. 3 in 2023 and 2024 according to one embodiment of the present invention. Detailed Implementation

[0021] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0022] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0023] like Figure 1 The diagram shows a flowchart of an energy-saving control method for an air compressor unit according to an embodiment of the present invention. Figure 1 The control method may include the following steps: In step S10, real-time operating data of the air compressor unit is acquired; In step S11, a digital twin model of the air compressor's energy consumption is constructed; In step S12, the basic threshold of the current operating condition is obtained based on the digital twin model of the air compressor energy consumption; In step S13, the air compressor unit status reference value is calculated in real time to obtain the relative ratio of single unit running time; In step S14, the running time correction coefficient is obtained based on the relative ratio of single machine running time; In step S15, the start-stop frequency correction coefficient is obtained based on the average start-stop frequency; In step S16, the current operating condition base threshold is corrected based on the running time correction coefficient and the start-stop frequency correction coefficient to obtain the current optimal control threshold.

[0024] In such Figure 1 In the energy-saving control method for the air compressor unit shown, step S10 is used to obtain real-time operating data of the air compressor unit. In this embodiment, the method for obtaining real-time operating data of the air compressor unit can be to collect the following key operating parameters in real time through the iFix platform: the current pressure setpoint of the air compressor (such as setpoint 1 and setpoint 2), operating status (such as start / stop status, fault code), operating time (cumulative hours), outlet flow rate (unit: m³ / min), inlet temperature and humidity, air compressor power, etc.

[0025] Step S11 is used to construct a digital twin model of the air compressor's energy consumption. In this embodiment, the specific method for constructing the digital twin model of the air compressor's energy consumption in step S11 can be of various forms known to those skilled in the art. In one example of the present invention, step S11 may include, for example... Figure 2 The steps shown are described. Figure 2 In this context, step S11 may include: In step S20, data is collected and preprocessed; In step S21, a modular mechanism model is constructed based on the preprocessed data; In step S22, a temporal neural network model is constructed; In step S23, the temporal neural network model is trained based on the mechanistic model to obtain a fusion model.

[0026] In such Figure 2In the method shown, step S20 is used to collect and preprocess data. In this example, the collected data may include basic physical equipment data, historical operating data, and real-time operating condition data. The basic physical equipment data includes records for each air compressor's rated power, air duct structural parameters (friction coefficient, clearance), core component materials (rotor / bearing), and factory energy efficiency curves (flow-power relationship). This basic physical data provides a basis for subsequent mechanistic models. Historical operating data can be derived from cumulative operating data of ≥28,000 hours from 2023-2024, including: real-time flow rate, input power, exhaust temperature, lubricating oil pressure, ambient temperature, pipeline pressure, and cumulative operating time. This historical data provides learning samples for subsequent data-driven models. Real-time operating condition data can be collected via Kepserver and stored in an Access database for the past 8 hours, with a collection frequency of 1 second, covering compressed air demand flow rate, grid voltage, pipeline leakage, and start / stop frequency.

[0027] After collecting the necessary data, it needs to be preprocessed before it can be used for modeling. In this example, data preprocessing includes data cleaning, normalization, feature selection, and data partitioning. Specifically, data cleaning is used to remove invalid records such as those from downtime due to malfunctions (e.g., excessive exhaust temperature) and data transmission interruptions, retaining at least 95% of valid data. Normalization uses the Min-Max normalization formula: Parameters such as flow rate (0~40 m³ / min), power (0~rated value), and temperature (-10~40℃) were standardized to the [0,1] range. Pearson correlation analysis was used for feature selection, retaining core features with a correlation ≥0.8 with energy consumption (flow rate, pipeline pressure, ambient temperature, cumulative operating time) and removing redundant parameters (such as equipment number and data collection timestamp). Historical data was split into a training set (modeling) and a test set (validation) in a 7:3 ratio to ensure that both sets of data cover the same operating conditions.

[0028] Step S21 is used to construct a modular mechanism model based on the preprocessed data. Based on thermodynamics and fluid mechanics principles, the modular mechanism model reconstructs the equipment's operating mechanism. In this embodiment, the modular mechanism model includes: a thermodynamic sub-model, a fluid mechanics sub-model, a mechanical loss sub-model, and an operating state sub-model. The thermodynamic sub-model is used to calculate the theoretical energy consumption and exhaust temperature of the compression process based on the ideal gas compression law. The fluid mechanics sub-model is used to calculate the pressure loss of the gas duct and pipeline system and correct the energy consumption calculation results of the thermodynamic sub-model. The mechanical loss sub-model is used to calculate mechanical wear and additional losses from start-stop operations based on cumulative operating time and start-stop frequency. The operating state sub-model is used to associate and bind the equipment's start-stop status, fault codes, and lubricating oil pressure parameters with the energy consumption model.

[0029] Specifically, in this example, the thermodynamic sub-model is based on the ideal gas compression law, and the input parameters include inlet air temperature (real-time acquisition), inlet air pressure (pipeline pressure), and pressure setpoint (6.6 / 6.9 bar). The output results are exhaust temperature and energy consumption per unit volume of compressed air. A calibration operation is performed: 100 sets of historical data (different pressure / temperature combinations) are input, and the compression efficiency coefficient in the model is adjusted so that the deviation between the calculated energy consumption and the actual energy consumption is ≤ ±2%.

[0030] In this example, the fluid dynamics sub-model models the air compressor duct and piping system, calculating target values ​​using key parameters. These key parameters include the friction coefficient of the duct inner wall (0.01~0.03) and the piping leakage rate (dynamically updated based on real-time data acquisition). The target values ​​include pressure loss (ΔP) at different flow rates, and the energy consumption calculation results from the corrected thermodynamic model are adjusted (energy consumption adjustment +1.2% for every 0.01 MPa increase in pressure loss).

[0031] In this example, the correlation factors for the mechanical loss sub-model are cumulative running time (t) and number of start-stop cycles (s). Modeling is based on the wear coefficient and start-stop losses, where the wear coefficient... ( (Unit: hour) ; , Start-up and shutdown losses ( (Unit: times / hour), added to total energy consumption. Adjustments are made by using air compressor operating data for different operating durations. A coefficient is used to ensure that the loss calculation error is ≤ ±1%.

[0032] In this example, the mapping parameters of the operating state sub-model include start / stop status (0 = stopped / 1 = running), fault codes (such as exhaust temperature exceeding the standard = 001), and lubricating oil pressure (normal range 0.2~0.4MPa). These state parameters are then bound to the energy consumption model.

[0033] Step S22 is used to construct the temporal neural network model. In this example, a lightweight LSTM architecture is adopted (adapting to the small sample size and high generalization requirements of industrial scenarios, avoiding excessive complexity). Combined with the temporal characteristics of air compressor unit energy consumption, a four-layer network structure is built: input layer - LSTM hidden layer - fully connected layer - output layer. The specific architecture parameters are as follows: The input layer has dimensions of [None (number of samples) × 10 (time steps) × 8 (input features)], receives preprocessed temporal feature samples, has no activation function, and only performs data transfer.

[0034] The LSTM hidden layers consist of two cascaded LSTM layers, balancing temporal feature extraction with model lightweighting: The first LSTM layer has 64 neurons, with Dropout=0.2 enabled (to prevent overfitting) to preserve long-term dependencies in temporal features; the second LSTM layer has 32 neurons, with Dropout=0.2 enabled, to perform dimensionality reduction and refinement of the features extracted by the first layer. The activation function is tanh (adapted to feature extraction from temporal data), and the forget gate uses the sigmoid function.

[0035] One fully connected layer with 16 neurons is set up, and the activation function is ReLU. The temporal features extracted by the LSTM layer are nonlinearly mapped to adapt to the complex relationship between energy consumption and features.

[0036] The output layer is set to a fully connected layer with 2 neurons and no activation function (adapted to the regression prediction of continuous energy consumption values), directly outputting the normalized energy consumption values ​​of mode A and mode B.

[0037] Step S23 is used to train the temporal neural network model based on the mechanistic model to obtain a fusion model. In this example, based on the TensorFlow / PyTorch framework, supervised learning mini-batch gradient descent is used for training, and a loss function constrained by the mechanistic model is fused to achieve iterative updates of the model parameters. Further, the specific method for training the temporal neural network model based on the mechanistic model in step S23 may include the following steps: In step S30, the training parameters are set; In step S31, the loss function is defined; In step S32, the training process is performed; In step S33, the training process is monitored.

[0038] Step S30 is used for training parameter settings, adapting to the characteristics of industrial data, setting lightweight training parameters to avoid overfitting and excessive training time. Batch size: 32 (small batch training, improving model generalization ability); Number of training epochs: initially set to 100 epochs, enabling early stopping mechanism; Optimizer: Adam is used (fast convergence speed, adaptive learning rate), with an initial learning rate set to 0.001; Learning rate decay: step decay is used, multiplying the learning rate by 0.5 every 20 epochs to avoid oscillations in later training; Early stopping mechanism: monitor the MAPE of the validation set, if the MAPE of the validation set does not decrease for 10 consecutive epochs, immediately stop training and save the current optimal model parameters (to prevent overfitting).

[0039] Step S31 is used to define the loss function. Specifically, a hybrid loss function of mean squared error (MSE) and mechanistic constraint term is adopted to ensure that the model training results are consistent with the actual data and do not deviate from the operating mechanism of the air compressor. The hybrid loss function is constructed using formula (1): (1) in, For loss function, The energy consumption value predicted by the model. This represents the actual standard energy consumption value. These are mechanistic constraints, including constraints such as non-negative energy consumption and monotonically increasing energy consumption with flow rate. If the mechanism is violated, The penalty value is 0 otherwise; This represents the constraint coefficient. In this example, The value is 0.01, which is used to balance data fitting and mechanistic constraints and avoid excessive penalty.

[0040] Step S32 executes the training process. First, the training set data is input into the model in batches, and the energy consumption prediction values ​​for modes A and B are obtained through forward propagation. Next, the mixture loss function value is calculated, and the loss value is passed to each layer of the network through the backpropagation algorithm to update the model parameters. After each training epoch, the model accuracy is evaluated using validation set data, and MAPE, MAE, and RMSE are recorded. If the early stopping mechanism is triggered or the preset number of training epochs is reached, training is stopped, and the optimal model file and normalized parameter file are saved for subsequent deployment and prediction.

[0041] Step S33 is used to monitor the training process, and monitor the loss value and accuracy index of the training set and validation set in real time. If the MAPE of the training set continues to decrease and the MAPE of the validation set increases, it is determined that the model is overfitting, and the Dropout coefficient or learning rate is adjusted immediately.

[0042] After completing the iterative training of the model, a virtual-physical real-time mapping link needs to be established. In this example, the overall architecture of the communication link deployment is as follows: Data Acquisition Server (Kepserver) → Local KEPServerEX (data acquisition) → Local Access Database (data storage) → Digital Twin Model (data reading). The core guarantee is a 1-second acquisition / reading frequency to match the real-time requirements of the digital twin model. To ensure state synchronization, the data transmission rhythm is that the physical device uploads core parameters (flow, pressure, power, status code) every 1 second, and the server sends simulation results / threshold data every 1 second. Synchronization verification is performed: after receiving data, the virtual model verifies its integrity using CRC32; if the verification fails, a retransmission is triggered. Anomaly handling is implemented: when the physical device parameters exceed the normal range (e.g., flow > 45 m³ / min), the virtual model immediately marks the abnormal condition and locks the current simulation parameters for easy traceability.

[0043] After training and validation, the model is deployed to an industrial-grade server to provide real-time energy consumption simulation for the digital twin system. Simultaneously, considering equipment aging and changes in operating conditions of the air compressor unit, incremental training based on newly acquired data is supported to ensure long-term model accuracy stability. Model deployment includes deploying the optimal model file and normalized parameter file to the local digital twin server and interfacing with a real-time data reading program from an Access database. Data preprocessing and prediction encapsulation functions are written to achieve a fully automated workflow from real-time data reading → normalization → model prediction → denormalization → output of energy consumption simulation values, adapting to real-time simulation requirements with a 1-second sampling period.

[0044] Incremental training is used to ensure that the digital twin maintains a high degree of consistency with the physical entity throughout its entire lifecycle. The specific methods for incremental training include: when operators find that the model simulation error is >±3% (e.g., due to equipment overhaul, core component replacement, or shift in operating conditions), incremental training is initiated. It does not require retraining the entire dataset; training is performed using only newly added 1-second sampled valid data (≥500 hours). The newly added valid data is extracted, processed according to the preprocessing rules of this method, and mixed with a small amount of typical data from the original training set. The optimal parameters of the already trained model are loaded, the parameters of the front-layer LSTM network are frozen, and only the fully connected layers and the output layer are trained. Mini-batch training (Epoch=20) is performed using a smaller learning rate (0.0001) to quickly adapt to new operating conditions / new equipment states. After incremental training is completed, the model is verified through on-site testing. Once the accuracy meets the standards, the original deployed model is replaced, completing the model iteration update.

[0045] In this example, a digital twin is created from the historical operating data of the screw air compressor from 2022 to 2023. The model is deployed on a server, and every 3 minutes, the real-time data of the current air compressor is processed by the model to derive the initial baseline threshold and the average threshold. and single-machine priority threshold Then, the final threshold setting value is calculated based on the running time and start / stop frequency. Step S12 is used to obtain the basic threshold for the current operating condition based on the digital twin model of air compressor energy consumption. Specifically, the digital twin model calculates the basic threshold based on real-time operating parameters through the following process: Simulation boundary settings: Fixed pressure setpoints (Mode A: 6.9 + 6.6; Mode B: 6.9 + 6.9), with the total flow rate as the variable. (Range: 25~60m³ / min, covering traditional thresholds and operating condition fluctuation ranges); Energy consumption curve plotting: Simulation of different operating conditions for the current operating conditions. Draw EA and EB below , Two energy consumption curves; find the intersection of the two energy consumption curves. (EA=EB), then ; ( (To avoid frequent switching of critical points) Step S13 is used to calculate the real-time baseline value of the air compressor unit to obtain the relative ratio of single-unit operating time. In this example, the three screw compressors are considered as a whole, and the equipment baseline is calculated in real time to provide a basis for the correction factor. Calculate the average operating time: (Overall aging baseline of the unit). Calculate the relative ratio of single-unit operating time according to formula (2). (2) in, Indicates the first The aging degree of the equipment Indicates the first The cumulative operating time of the equipment This represents the average operating time of the air compressor unit. Reflecting the difference between the aging of a single unit and the average level, The older it is, the more severe the aging.

[0046] Calculate the average start-stop frequency: This serves as a benchmark for the overall start-up and shutdown losses of the unit.

[0047] Step S14 is used to obtain the running time correction factor based on the relative ratio of single-machine running time. Severely aging equipment ( (A value that is too high) results in a faster increase in energy consumption under full load, requiring appropriate adjustment of the threshold to avoid prolonged full-load operation while ensuring optimal energy consumption. Specifically, in this example, the value of the operating time correction coefficient is determined based on the distribution of the relative ratios of the single-machine operating times: When all That is, when aging is uniform, At this point, the base threshold does not need to be adjusted, and the energy consumption comparison logic remains unchanged; when and That is, when the aging of a single unit slightly exceeds the average, The value is At this time, it is appropriate to increase ,reduce Expand the scope of application of Mode B and distribute the load of aging equipment; when or That is, the aging of individual units varies significantly. The value is At this point, further improvement ,reduce Prioritize using Mode B to avoid prolonged full-load operation of aging equipment, while utilizing new equipment ( High efficiency.

[0048] Step S15 is used to obtain a start-stop frequency correction coefficient based on the average start-stop frequency. More frequent start-stops increase energy consumption during mode switching, necessitating a narrower threshold range to reduce the number of switching operations while ensuring optimal energy efficiency. In this example, the value of the start-stop frequency correction coefficient is determined based on the range in which the average start-stop frequency falls: when That is, normal start-stop is At this point, the base threshold remains unchanged; when That is, start / stop frequency offset. The value is At this time, it is appropriate to reduce ,improve e Narrowing the mode switching range; when That is, when the start and stop are too frequent. The value is At this point, the price drops significantly. ,improve Prioritize maintaining the current mode to avoid exacerbating start-stop issues during switching.

[0049] Step S16 is used to correct the current operating condition base threshold based on the running time correction coefficient and the start-stop frequency correction coefficient to obtain the current optimal control threshold. In this example, formulas (3) and (4) are used to obtain the current optimal control threshold, ensuring that both energy consumption comparison requirements and equipment status are met: (3) (4) in, The final average threshold, This is the final single-machine priority threshold. The threshold for equal distribution, This is the priority threshold for a single machine. This is a runtime correction factor. This is a correction factor for the number of start-stop cycles.

[0050] The threshold results are stored in the ifix database, and the model is iterated and optimized every quarter: offline training and validation are performed based on historical data from the most recent year, the optimized version is run in parallel with the current version, and the running effect is compared in real time (for 24 hours). After a gray-scale release without any anomalies, the online version is replaced.

[0051] like Figure 3As shown, a new air compressor pressure setting screen is created in iTunes, displaying key operating parameters of the air compressor to facilitate operators' observation of its operating status. A pressure setting button is added, allowing operators to change the selected pressure setting value remotely.

[0052] Add time-based scheduling to the ifix scheduling mechanism, performing a conditional check every 3 minutes and adjusting the pressure setpoint based on the result. The control logic is as follows: When the sum of the outlet flow rates of the screw air compressor is greater than When the pressure setting value of the air compressor is changed from 6.6 in setting value 1 to pressure setting value 2 (the pressure setting value 2 of all air compressors is 6.9), the current pressure setting value of the air compressors is: 6.9 for two air compressors and 6.5 for one air compressor. Under the current setting state, the air compressor with a pressure setting value of 6.9 will bear the average flow of compressed air required.

[0053] When the sum of the outlet flow rates of the screw air compressor is at and When the pressure is between t and t, the current pressure setting of the air compressor remains unchanged.

[0054] When the sum of the outlet flow rates of the screw air compressor is less than When the pressure setting value is 6.6, the pressure setting value of the air compressor with pressure setting value 1 is changed to pressure setting value 1 to reduce its air output until it is shut down. That is, the current pressure setting values ​​of the air compressors are: one air compressor is 6.9, one air compressor is 6.6, and one air compressor is 6.5.

[0055] like Figure 4 , 5 Figure 6 shows a scatter plot of the electrical efficiency of the three screw air compressors in 2023 and 2024. The plot clearly shows that the electrical efficiency in 2024 was higher than that in 2023. According to statistical analysis, the electrical efficiency of the three screw air compressors at Ningbo Cigarette Factory decreased by 2.67% in 2024 compared to 2023. Specifically, the total power consumption of the three screw air compressors in 2024 was 1,540,385 kWh. Therefore, it can be calculated that by implementing this control strategy, the company achieved a power saving of 41,128 kWh.

[0056] On the other hand, the present invention also provides an energy-saving control system for an air compressor unit, the system including a processor configured to perform any of the methods described in the energy-saving control method for an air compressor unit.

[0057] In another aspect, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement any of the methods described in the air compressor unit energy-saving control method.

[0058] The beneficial effects of this invention are: The embodiments of the present invention introduce a runtime correction coefficient. With start-stop frequency correction factor The basic threshold of digital twins is double-corrected. Adjust the load distribution according to the differences in equipment aging to avoid aging units operating at full load for a long time; Narrow the mode switching range based on the frequency of start-stop operations to reduce unnecessary loading and unloading. Final threshold. and It has achieved a leap from single energy efficiency optimization to multi-objective synergistic optimization of energy efficiency, lifespan, and stability.

[0059] This invention constructs a modular mechanistic model encompassing thermodynamics, fluid mechanics, mechanical losses, and operational states, embedding physical laws as constraints. Simultaneously, it employs an LSTM neural network to learn the complex temporal characteristics of actual operational data, and integrates these mechanistic constraints into the training process through a hybrid loss function. The resulting digital twin model possesses both physical interpretability and data-driven accuracy, with an energy consumption simulation error ≤ ±2%.

[0060] The implementation of this invention supports incremental training based on newly acquired data after model deployment. When simulation errors exceed limits or significant changes in equipment status occur (such as major repairs or component replacements), the parameters of the front-layer network are frozen, and only the fully connected layers of the back layer are fine-tuned, enabling rapid adaptation to new operating conditions. This mechanism avoids frequent full retraining, ensuring that the digital twin maintains a high degree of consistency with the physical entity throughout its entire lifecycle.

[0061] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section.

[0062] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0063] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0064] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0065] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0066] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0067] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0068] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0069] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0070] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. An energy-saving control method for an air compressor unit, characterized in that, The control method includes: Obtain real-time operating data of the air compressor unit; Construct a digital twin model of air compressor energy consumption; The basic threshold for the current operating condition is obtained based on the digital twin model of the air compressor's energy consumption. Real-time calculation of the air compressor unit's status baseline value to obtain the relative comparison of single unit operating time; The running time correction coefficient is obtained based on the relative ratio of the single machine running time; The start-stop frequency correction coefficient is obtained based on the average start-stop frequency. The current operating condition base threshold is corrected based on the running time correction coefficient and the start / stop frequency correction coefficient to obtain the current optimal control threshold.

2. The control method according to claim 1, characterized in that, Building a digital twin model includes: Collect data and preprocess it; A modular mechanism model is constructed based on the preprocessed data; Construct a temporal neural network model; The temporal neural network model is trained based on the aforementioned mechanism model to obtain a fusion model.

3. The control method according to claim 2, characterized in that, The modular mechanism model includes: Thermodynamic sub-model, fluid dynamics sub-model, mechanical loss sub-model, and operating state sub-model; The thermodynamic sub-model is used to calculate the theoretical energy consumption and exhaust temperature of the compression process based on the ideal gas compression law. The fluid dynamics sub-model is used to calculate the pressure loss of the gas duct and pipeline system, and to correct the energy consumption calculation results of the thermodynamic sub-model. The mechanical loss sub-model is used to calculate mechanical wear and additional losses from start-stop based on cumulative operating time and start-stop frequency. The operating status sub-model is used to associate and bind the equipment's start / stop status, fault codes, and lubricating oil pressure parameters with the energy consumption model.

4. The control method according to claim 2, characterized in that, Training the temporal neural network model based on the aforementioned mechanism model to obtain the fusion model includes: Construct a hybrid loss function based on formula (1). ，（1） in, For loss function, The energy consumption value predicted by the model. This represents the actual standard energy consumption value. For mechanism constraint terms, This is the constraint coefficient.

5. The control method according to claim 1, characterized in that, Real-time calculation of the air compressor unit's status baseline value to obtain the relative ratio of single-unit operating time includes: Calculate the average operating time and average number of start-stop cycles of the air compressor unit; Calculate the relative ratio of single-machine running time according to formula (2). ，（2） in, Indicates the first The aging degree of the equipment Indicates the first The cumulative operating time of the equipment This represents the average operating time of the air compressor unit.

6. The control method according to claim 1, characterized in that, The runtime correction factor is obtained based on the relative ratio of the single-machine runtime, including: Based on the distribution of the relative ratios of single-machine running times, the value of the running time correction coefficient is determined: When the relative ratio of the individual running times of all air compressors is within the preset uniform range, it is determined that the equipment is aging uniformly, and the running time correction coefficient is set as the base value. When the relative ratio of the single running time of a single air compressor exceeds the uniform range but does not reach the significant difference threshold, and the other equipment is not lower than the preset lower limit threshold, it is determined that the aging of a single equipment is slightly higher than the average level, and the running time correction coefficient is set as the first correction value, which is greater than the base value. When the relative ratio of the single running time of a single air compressor exceeds the significant difference threshold, or when the relative ratio of the single running time of a single device is lower than the preset lower limit threshold, it is determined that the aging difference between devices is significant, and the running time correction coefficient is set as the second correction value, which is greater than the first correction value.

7. The control method according to claim 1, characterized in that, The start-stop frequency correction factor is obtained based on the average start-stop frequency, including: Based on the interval in which the average start-stop frequency falls, determine the value of the start-stop frequency correction coefficient: When the average start-stop frequency is within the first preset range, the start-stop is determined to be normal, and the start-stop frequency correction coefficient is set as the base value; When the average start-stop frequency is within the second preset range, it is determined that the start-stop frequency is too frequent, and the start-stop frequency correction coefficient is set to the first correction value, which is less than the base value. When the average start-stop frequency is within the third preset range, it is determined that the start-stop is too frequent, and the start-stop frequency correction coefficient is set to the second correction value, which is less than the first correction value.

8. The control method according to claim 1, characterized in that, The current operating condition base threshold is corrected based on the running time correction coefficient and the start-stop frequency correction coefficient to obtain the current optimal control threshold, including: The current optimal control threshold is obtained according to formulas (3) and (4). ，（3） ，（4） in, The final average threshold, This is the final single-machine priority threshold. The threshold for equal distribution, This is the priority threshold for a single machine. This is a runtime correction factor. This is a correction factor for the number of start-stop cycles.

9. An energy-saving control system for an air compressor unit, characterized in that, The system includes a processor configured to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 8.

Images