Gas compressor frequency conversion control method, device and equipment

Abstract

The invention relates to a gas compressor frequency conversion control method, device and equipment, and belongs to the technical field of industrial automation control. According to the method, device and equipment, multi-dimensional sensing data of the running state, the process requirement and the environment condition of a gas compressor are fused, feature extraction and fusion are carried out through a pre-trained state evaluation model, and the gas compressor frequency conversion control precision is improved. Therefore, continuous health degree indexes and accurate state levels are output, quantitative diagnosis and foresight early warning of sub-health and early faults of the compressor are achieved, and the problems that a traditional system cannot accurately recognize the sub-health state and the early faults of equipment, safety early warning of the equipment is difficult to consider, and the working efficiency is low are fundamentally solved. The operation reliability is low; the energy efficiency is poor; and the maintenance cost is high.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, specifically to a variable frequency control method, device, and equipment for a gas compressor. Background Technology

[0002] As a core power equipment in industrial production, the operational stability, energy efficiency, and safety reliability of gas compressors directly affect production continuity and operating costs. Currently, the widely adopted variable frequency control scheme in the industrial sector primarily uses the pipeline outlet pressure as the core feedback variable, adjusting the motor frequency through a PID controller to achieve stable control of the gas supply pressure.

[0003] However, existing variable frequency control systems rely on only a few process parameters such as pressure and temperature for control decisions. They lack a comprehensive understanding of the compressor's mechanical state (such as bearing wear and housing vibration), electrical performance (such as current and voltage fluctuations), and environmental conditions (such as inlet temperature and humidity). This makes it impossible to accurately identify the sub-healthy state and early faults of the equipment. Often, the protection mechanism is only passively triggered when the fault has seriously affected the operation, which increases maintenance costs and affects production continuity.

[0004] Therefore, the existing gas compressor frequency conversion control system lacks full-dimensional state perception, making it impossible to accurately identify the sub-health state and early faults of the equipment, and it is difficult to take into account the equipment safety warning, resulting in low operational reliability, poor energy efficiency and high maintenance costs. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a gas compressor frequency conversion control method, device and equipment to overcome the problems of low operational reliability, poor energy efficiency and high maintenance cost caused by the lack of full-dimensional state perception in the current gas compressor frequency conversion control system, which makes it impossible to accurately identify the sub-health state and early faults of the equipment, and difficult to take into account the equipment safety warning.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] On the one hand, a variable frequency control method for a gas compressor, applied to variable frequency control equipment, the method comprising: In response to the variable frequency monitoring request of the gas compressor, the gas compressor's full-dimensional sensing data is acquired through a sensor network. The sensor network is deployed in the gas compressor body, motor, inlet and outlet pipelines, cooling system, and environment. The full-dimensional sensing data includes the gas compressor's operating status, process requirements, and environmental conditions. The full-dimensional sensor data is processed and features are extracted to construct a standardized feature matrix; The standardized feature matrix is ​​input into a pre-trained state evaluation model for analysis, outputting the discrete state level and continuous health index of the gas compressor, and generating a comprehensive performance score based on the health index and real-time energy efficiency ratio; wherein, the discrete state level includes healthy state, sub-healthy state, early warning state, and alarm state; Based on the comprehensive performance score, the discrete state level, and the real-time pressure deviation, a target value for the motor drive frequency is generated through a multi-layer logic strategy. The target value of the motor drive frequency is sent to the frequency converter to control the motor frequency of the gas compressor.

[0008] Optionally, the process of processing and extracting features from the full-dimensional sensing data to construct a standardized feature matrix includes: Anomaly detection and missing value imputation are performed on the full-dimensional sensor data to obtain processed data; Extract time-domain features, frequency-domain features, and sequence trend features from the processed data; The full-dimensional sensing data, along with the extracted time-domain features, frequency-domain features, and sequence trend features, are aligned and recombined in time sequence to form the standardized feature matrix.

[0009] Optionally, the step of performing anomaly detection and missing value imputation on the full-dimensional sensing data to obtain processed data includes: Based on the full-dimensional sensing data, the 3σ criterion based on sliding window and the isolated forest algorithm are used to identify and remove transient interference and sensor failure data. Missing values ​​for non-critical parameters are filled using time series linear interpolation; for abnormal critical parameters, a data quality alarm is triggered to update the critical parameters; the updated critical parameters, the filled non-critical parameters, and other normal data are used as the processed data.

[0010] Optionally, extracting time-domain features, frequency-domain features, and sequence trend features from the processed data includes: Extract the root mean square value, peak factor, and impulse index from the time-domain features of the processed data; The vibration signal in the frequency domain features is extracted from the processed data and rapidly transformed to extract the amplitude energy of specific harmonics as fault symptom features. Pressure and flow sequence features are extracted from the processed data, and their short-term fluctuation variance and trend slope are calculated.

[0011] Optionally, the state evaluation model adopts a multi-branch deep neural network architecture; The input layer receives the standardized feature matrix, and different branches process different types of feature data in parallel. Each branch extracts high-dimensional abstract features through convolutional layers, and the features are cross-complemented through fusion layers to generate a comprehensive state depth representation vector; The top classifier outputs discrete status levels of healthy, sub-healthy, early warning, and alarm through the Softmax function, while the regressor outputs a continuous health index through nonlinear normalization.

[0012] Optionally, generating a comprehensive performance score based on the health index and real-time energy efficiency ratio includes: The real-time input power of the gas compressor is determined based on the sensor network, and the real-time energy efficiency ratio is calculated based on the theoretical power value. Based on the real-time energy efficiency ratio and the health index, a comprehensive performance score is calculated using the comprehensive performance scoring rules; wherein, the comprehensive performance scoring rules are as follows:

[0013] CPS is the overall performance score; w H The health weighting coefficient is dynamically set based on discrete status levels; HI is the continuous health index; HImin is the effective threshold for the health index; w E η is the energy efficiency weighting coefficient, and satisfies wH+wE=1; ηnorm is the normalized real-time energy efficiency ratio; λ is the load range correction factor.

[0014] Optionally, the multi-layer logic strategy includes: The safety protection sub-logic is used to prioritize generating frequency reduction or shutdown commands when the status is alarm. The energy efficiency optimization sub-logic is used to initiate frequency optimization based on the comprehensive performance score under safe conditions. The pressure control sub-logic is used to calculate the base frequency adjustment based on the real-time pressure deviation. The target value of the motor drive frequency is the result of weighted fusion of the output of the pressure control sub-logic after being corrected by the safety protection sub-logic and the energy efficiency optimization sub-logic; the priority of the safety protection sub-logic, the energy efficiency optimization sub-logic, and the pressure control sub-logic decreases sequentially.

[0015] Optionally, after sending the target value of the motor drive frequency to the frequency converter to control the motor frequency of the gas compressor to be adjusted to the target value of the motor drive frequency, the method further includes: Acquire real-time sensing data collected by the sensor network to evaluate the control effect; The entire process data related to this frequency conversion control, as well as the control effect, are stored as case samples for future reference.

[0016] In another aspect, a variable frequency control device for a gas compressor includes: The acquisition module is used to respond to the frequency conversion monitoring request of the gas compressor and acquire full-dimensional sensing data of the gas compressor through a sensor network. The sensor network is deployed in the gas compressor body, motor, inlet and outlet pipelines, cooling system and environment. The full-dimensional sensing data covers the operating status, process requirements and environmental conditions of the gas compressor. The construction module is used to process and extract features from the full-dimensional sensor data to construct a standardized feature matrix; The analysis module is used to input the standardized feature matrix into a pre-trained state evaluation model for analysis, output the discrete state level and continuous health index of the gas compressor, and generate a comprehensive performance score based on the health index and real-time energy efficiency ratio; wherein, the discrete state level includes healthy state, sub-healthy state, early warning state, and alarm state; The control module is used to generate a target value for the motor drive frequency based on the comprehensive performance score, the discrete state level, and the real-time pressure deviation through a multi-layer logic strategy; and to send the target value for the motor drive frequency to the frequency converter to control the motor frequency of the gas compressor.

[0017] In another aspect, an electronic device includes a processor and a memory, wherein the processor is connected to the memory: The processor is used to call and execute the program stored in the memory; The memory is used to store the program, which is used to execute at least the gas compressor frequency conversion control method described above.

[0018] The technical solution provided in this application has at least the following beneficial effects: In the technical solution of this invention, by integrating the operating status, process requirements and environmental conditions of the gas compressor, namely multi-dimensional sensing data such as mechanical vibration, electrical, and thermal, and using a pre-trained state assessment model for feature extraction and fusion, a continuous health index and accurate state level are output. This achieves quantitative diagnosis and proactive early warning of compressor sub-health and early failures, fundamentally solving the technical problems of traditional systems being unable to accurately identify equipment sub-health and early failures, making it difficult to take into account equipment safety warnings, resulting in low operational reliability, poor energy efficiency and high maintenance costs. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flow chart of a variable frequency control method for a gas compressor provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a gas compressor frequency converter control device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a gas compressor frequency converter control device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0022] As described in the background section, existing variable frequency control systems rely solely on a few process parameters such as pressure and temperature for control decisions. They lack comprehensive perception of the compressor's mechanical state (such as bearing wear and casing vibration), electrical performance (such as current and voltage fluctuations), and environmental conditions (such as inlet temperature and humidity). This results in the inability to accurately identify the sub-health state and early faults of the equipment. Often, the protection mechanism is only passively triggered when the fault has already seriously affected the operation, which increases maintenance costs and affects production continuity.

[0023] In other words, existing control methods often lead to passive control decisions, typically triggering protection only when a fault has already significantly impacted production operations. This makes predictive maintenance difficult, affects production continuity, and may increase maintenance costs.

[0024] Therefore, the existing gas compressor frequency conversion control system lacks full-dimensional state perception, making it impossible to accurately identify the sub-health state and early faults of the equipment, and it is difficult to take into account the equipment safety warning, resulting in low operational reliability, poor energy efficiency and high maintenance costs.

[0025] Furthermore, traditional control objectives focus on maintaining pressure stability, while energy-saving optimization and ensuring long-term equipment safety are often independent or even conflicting goals. Simple threshold protection struggles to achieve a dynamic balance between ensuring equipment safety, meeting process requirements, and optimizing energy efficiency. Finally, once the parameters of a control system are set, they typically remain unchanged for a long period, unable to self-optimize and adjust based on equipment performance degradation due to wear, changes in the external environment (such as seasonal temperature differences), or historical operating data. This can lead to a gradual decrease in the overall energy efficiency of long-term operation.

[0026] Therefore, existing technologies have significant shortcomings in terms of comprehensively sensing equipment status, dynamically balancing multiple control objectives, and achieving adaptive optimization, and a more intelligent and comprehensive control method is urgently needed.

[0027] Based on this, in order to solve at least one of the above-mentioned technical problems, embodiments of the present invention provide a variable frequency control method, apparatus and device for a gas compressor.

[0028] Figure 1 This is a flowchart illustrating a variable frequency control method for a gas compressor provided in an embodiment of the present invention, applicable to variable frequency control equipment. Please refer to [link / reference]. Figure 1 This embodiment may include the following steps: Step S101: In response to the variable frequency monitoring request of the gas compressor, acquire full-dimensional sensing data of the gas compressor through a sensor network; the sensor network is deployed on the gas compressor body, motor, inlet and outlet pipelines, cooling system and environment, and the full-dimensional sensing data includes the operating status of the gas compressor, process requirements and environmental conditions.

[0029] In this context, "full-dimensional sensor data" refers to a dataset encompassing multiple physical and logical dimensions of gas compressor operation. In this embodiment, full-dimensional sensor data can specifically include vibration data characterizing the mechanical health of the equipment, electrical parameter data characterizing the operating conditions of the electrical system, process data characterizing production task requirements, and environmental condition data affecting the equipment's operating baseline. This "full-dimensional" characteristic of the data is fundamental to achieving a comprehensive and in-depth understanding of the compressor's condition.

[0030] In a specific gas compressor frequency conversion control process, the gas compressor frequency conversion control method of this embodiment can be integrated into the frequency conversion control equipment, and the frequency conversion control equipment (which can be a frequency converter) can perform frequency conversion control.

[0031] The variable frequency monitoring request for the gas compressor can be a start-up signal or a pre-defined signal to initiate variable frequency control; this application does not specify a particular signal. The request can be triggered by a pre-set periodic task, such as executing once per minute; it can also be triggered by a specific event, such as abnormal fluctuations in pipeline pressure; or it can be manually initiated by maintenance personnel via a host computer. Once the request is responded to, the method proceeds... Figure 1 The first step shown is to obtain the full-dimensional data S101.

[0032] Sensor networks can be widely deployed on the gas compressor body (such as housing and bearing housing), drive motor, inlet and outlet pipelines, cooling system (such as cooling water pipelines and fans), and the surrounding environment of the equipment. The raw data stream undergoes protocol conversion and timestamp alignment through an industrial gateway to form data packets with a unified timing sequence, which are then sent to the frequency converter control equipment.

[0033] The operating status data may include: the triaxial vibration acceleration spectrum of the bearing and housing collected by the vibration sensor, the real-time three-phase current, voltage and power factor collected by the motor smart meter, the exhaust temperature of each cylinder, the main shaft bearing temperature and the motor winding temperature collected by the temperature sensor. The data required for the process can include the target pressure value, actual pressure value and instantaneous flow value of the pipeline outlet, which are collected in real time by flow meters and pressure transmitters; Environmental data may include intake air temperature, intake air humidity, and cooling water intake temperature collected by environmental sensors.

[0034] Step S102: Process and extract features from the full-dimensional sensing data to construct a standardized feature matrix.

[0035] The standardized feature matrix refers to a structured and standardized data format formed by organizing raw, heterogeneous, full-dimensional sensor data after a series of processing steps (such as cleaning, synchronization, transformation, and extraction).

[0036] Understandably, the raw full-dimensional sensor data may contain noise, missing data, or inconsistent formats, which could affect the accuracy of the analysis if used directly. Therefore, in this embodiment, this step first performs necessary processing on the raw full-dimensional sensor data, such as data cleaning, outlier removal, and data normalization. Subsequently, deeper features that more effectively characterize the device status are extracted from the processed data; that is, feature extraction is performed. After processing and feature extraction, all data and features are reorganized and aligned according to a unified timestamp, constructing a standardized feature matrix. This matrix is ​​a highly structured data volume, where each row represents a moment in time, and each column represents a specific feature dimension.

[0037] In some embodiments, the processing and feature extraction of the full-dimensional sensing data to construct a standardized feature matrix includes: Anomaly detection and missing value imputation are performed on the full-dimensional sensor data to obtain processed data; Extract time-domain features, frequency-domain features, and sequence trend features from the processed data; The full-dimensional sensing data, along with the extracted time-domain features, frequency-domain features, and sequence trend features, are aligned and recombined in time sequence to form the standardized feature matrix.

[0038] Understandably, by performing anomaly detection and missing value imputation on full-dimensional sensor data, the accuracy of the data is improved, and a standardized feature matrix is ​​constructed by extracting different features.

[0039] In some embodiments, the step of performing anomaly detection and missing value imputation on the full-dimensional sensing data to obtain processed data includes: Based on the full-dimensional sensing data, the 3σ criterion based on sliding window and the isolated forest algorithm are used to identify and remove transient interference and sensor failure data. Missing values ​​for non-critical parameters are filled using time series linear interpolation; for abnormal critical parameters, a data quality alarm is triggered to update the critical parameters; the updated critical parameters, the filled non-critical parameters, and other normal data are used as the processed data.

[0040] For example, anomaly detection can combine the sliding window-based 3σ criterion with the Isolation Forest algorithm. The 3σ criterion excels at quickly identifying transient disturbance data that deviates from the normal distribution, while the Isolation Forest algorithm is more effective at identifying isolated data points caused by persistent sensor failures. Combining the two can efficiently identify and eliminate these two main types of anomalous data.

[0041] For handling missing data, different strategies can be adopted based on the importance of the parameters. For temporary missing non-critical parameters (such as ambient humidity), time series linear interpolation can be used to fill in the gaps to ensure the integrity of the data sequence. However, for abnormal or missing critical parameters (such as main bearing temperature and exhaust pressure), they should not be filled in easily. Instead, a data quality alarm should be triggered immediately to prompt maintenance personnel to check the sensors or related circuits and wait for updated reliable data. Finally, the updated critical parameters, filled non-critical parameters, and other normal data are integrated as processed data and proceed to the next stage.

[0042] In some embodiments, extracting time-domain features, frequency-domain features, and sequence trend features from the processed data includes: Extract the root mean square value, peak factor, and impulse index from the time-domain features of the processed data; The vibration signal in the frequency domain features is extracted from the processed data and rapidly transformed to extract the amplitude energy of specific harmonics as fault symptom features. Pressure and flow sequence features are extracted from the processed data, and their short-term fluctuation variance and trend slope are calculated.

[0043] For example, in this application, the collected full-dimensional sensor data can be cleaned based on the sliding window 3σ criterion and the isolated forest algorithm to remove instantaneous data spikes caused by electromagnetic interference; a 5-second data gap was identified in the cooling water inlet temperature sensor, and since this parameter is not critical, the system automatically fills it in using time series linear interpolation. The processed data is used for feature extraction: the vibration signal is subjected to FFT transformation to extract the amplitude energy of the 1X and 2X harmonics; the root mean square value and peak factor of the current signal are calculated; and the fluctuation variance and trend slope of the outlet pressure sequence over the past minute are calculated. Finally, all the full-dimensional sensor data and extracted features are aligned and recombined to form a standardized feature matrix containing hundreds of dimensions of features.

[0044] Step S103: Input the standardized feature matrix into the pre-trained state evaluation model for analysis, output the discrete state level and continuous health index of the gas compressor, and generate a comprehensive performance score based on the health index and real-time energy efficiency ratio; wherein, the discrete state level includes healthy state, sub-healthy state, early warning state, and alarm state.

[0045] The condition assessment model is an algorithmic model pre-trained with a large amount of historical data, capable of analyzing and judging the current operating status of the gas compressor. This model takes a standardized feature matrix as input, and its core function is to parse complex data patterns and output quantitative and qualitative assessment results of the equipment status. It is the core intelligent engine for realizing the transformation from data to insight.

[0046] For example, the real-time energy efficiency ratio (EER) of a gas compressor can be calculated based on real-time power and flow data collected by a sensor network. Then, the EER is combined with the continuous health index output by the condition assessment model, and a comprehensive performance score is generated using a pre-defined weighted fusion algorithm. This comprehensive performance score considers both the equipment's reliability (health) and economy (energy efficiency), providing a single, comprehensive evaluation benchmark for subsequent decision-making.

[0047] In this application, the condition assessment model, after receiving the standardized feature matrix, performs in-depth analysis and outputs two results. The first is the discrete condition level of the gas compressor. This level categorizes the complex state of the equipment into several classes, such as "healthy state," "sub-healthy state," "early warning state," and "alarm state," providing clear qualitative basis for subsequent control decisions. The second is the continuous health index, a numerical value, for example, between 0 and 1, which quantifies the health level of the equipment; a higher value indicates a better condition. Compared to the discrete level, the health index can more sensitively reflect subtle changes in the equipment's condition and long-term degradation trends.

[0048] In some embodiments, the state evaluation model employs a multi-branch deep neural network architecture; The input layer receives the standardized feature matrix, and different branches process different types of feature data in parallel. Each branch extracts high-dimensional abstract features through convolutional layers, and the features are cross-complemented through fusion layers to generate a comprehensive state depth representation vector; The top classifier outputs discrete status levels of healthy, sub-healthy, early warning, and alarm through the Softmax function, while the regressor outputs a continuous health index through nonlinear normalization.

[0049] For example, the standardized feature matrix is ​​input into a pre-trained state assessment model. Different feature processing branches of the state assessment model work in parallel and extract their respective deep features through convolutional layers. The deep features interact through a fusion layer to generate a comprehensive state deep representation vector. Based on the comprehensive state deep representation vector, the highest probability output by the top Softmax classifier points to "healthy state," while the regressor outputs a continuous health index of 0.98. Simultaneously, the system calculates the real-time energy efficiency ratio based on real-time power and flow, and substitutes it into the CPS calculation formula. Because it is in a "healthy state," the health weight wH is set to a low 0.2, resulting in a final comprehensive performance score of 95, indicating that the device is in a very ideal state.

[0050] Specifically, different feature processing branches of the state assessment model are used to process vibration spectrum features, thermodynamic time series features, and electrical parameter features in parallel.

[0051] Step S104: Based on the comprehensive performance score, the discrete state level, and the real-time pressure deviation, generate a target value for the motor drive frequency through a multi-layer logic strategy.

[0052] In some embodiments, generating a comprehensive performance score based on the health index and the real-time energy efficiency ratio includes: The real-time input power of the gas compressor is determined based on the sensor network, and the real-time energy efficiency ratio is calculated based on the theoretical power value. Based on the real-time energy efficiency ratio and the health index, a comprehensive performance score is calculated using the comprehensive performance scoring rules; wherein, the comprehensive performance scoring rules are as follows:

[0053] CPS is the overall performance score, a value between 0 and 100; w H This is the health weighting coefficient, dynamically set according to the discrete state levels. For example, under the "healthy state", w H It might be set to 0.3, but in a "sub-healthy state," it would increase to 0.7 to indicate a higher level of concern for the equipment's health; HI stands for Continuous Health Index; HI min The HI (Health Index) threshold is set to 0.5. When the HI falls below this value, it indicates a severe deterioration in the device's condition, and the score will drop rapidly. E η is the energy efficiency weighting coefficient, which satisfies wH+wE=1; ηnorm is the normalized real-time energy efficiency ratio, which is obtained by comparing the current energy efficiency ratio with the optimal energy efficiency ratio under this operating condition, so as to eliminate the influence of the operating condition itself on the energy efficiency evaluation; λ is the load range correction factor, which is used to compensate for the difference in evaluation scale under high load and low load, and ensure the fairness of the scoring.

[0054] Wherein, real-time energy efficiency ratio = (theoretical power value or equivalent value corresponding to output air volume) / real-time input electrical power.

[0055] Understandably, the formula provided in this example, by adjusting the weighting coefficients wH and wE, allows the system to intelligently balance "ensuring safety" and "seeking energy conservation." When the equipment is in good health, the system focuses more on energy efficiency; however, once a potential health hazard is detected, the decision-making balance immediately shifts towards ensuring equipment reliability. This formulaic scoring mechanism transforms a fuzzy, multi-objective optimization problem into a clear, single-valued optimization objective, simplifying subsequent decision-making logic.

[0056] In some embodiments, the multi-layer logic strategy includes: The safety protection sub-logic is used to prioritize generating frequency reduction or shutdown commands when the status is alarm. The energy efficiency optimization sub-logic is used to initiate frequency optimization based on the comprehensive performance score under safe conditions. The pressure control sub-logic is used to calculate the base frequency adjustment based on the real-time pressure deviation. The target value of the motor drive frequency is the result of weighted fusion of the output of the pressure control sub-logic after being corrected by the safety protection sub-logic and the energy efficiency optimization sub-logic.

[0057] Specifically, in a more concrete implementation, the multi-layer decision-making strategy can consist of three distinct sub-logic that work collaboratively with different priorities. The first layer is the safety protection sub-logic, which has the highest decision priority. This logic continuously monitors discrete state levels, and once the state switches to an "alarm state," it immediately and unconditionally generates instructions to reduce the frequency to a preset safe speed or to shut down the machine directly to prevent catastrophic failures.

[0058] The second layer is the energy efficiency optimization sub-logic. This logic is activated in a safe state (i.e., non-alarm state). It determines whether the current overall performance score is below a certain set threshold (e.g., 70 points) and whether the real-time pressure deviation is within the allowable range. If the conditions are met, it indicates that there is room for optimization at the current operating point and the process conditions allow for adjustment. This logic then initiates an online optimization algorithm. For example, it fine-tunes the motor frequency in a preset step size (e.g., 0.1Hz) and closely monitors changes in the overall performance score or real-time energy efficiency ratio. Through strategies similar to gradient descent or hill climbing, it gradually searches for the locally optimal operating frequency under the current operating conditions.

[0059] The third layer is the pressure control sub-logic. This is a basic control logic whose main goal is to maintain stable pipeline pressure. It can be a traditional PID controller that calculates a basic frequency adjustment based on real-time pressure deviation. The final output target value for the motor drive frequency is the result of a weighted fusion of the output from the pressure control sub-logic after corrections from the first two layers of logic. For example, the shutdown command from the safety logic will directly override all other outputs; the optimal frequency point found by the energy efficiency optimization logic will be used as the target setpoint for the pressure control logic; during normal operation, the final frequency may be a weighted sum of the pressure PID output and the energy efficiency fine-tuning.

[0060] For example, when the calculated comprehensive performance score is 95, the discrete state level is healthy, and the real-time pressure deviation is -0.01MPa, the safety protection sub-logic detects the "healthy" state and therefore remains silent; the energy efficiency optimization sub-logic determines that the score of 95 is higher than its optimization start threshold (e.g., 85), therefore it judges that the current operating point is close to the optimal, and does not start the computationally intensive online optimization; the decision-making power is handed over to the pressure control sub-logic, whose PID controller calculates a frequency fine-tuning amount of +0.05Hz based on the small negative deviation of -0.01MPa. This frequency fine-tuning amount is the target value of the motor drive frequency, and the target value of the motor drive frequency is sent to the gas compressor for variable frequency control.

[0061] Step S105: Send the target value of the motor drive frequency to the frequency converter to control the motor frequency of the gas compressor.

[0062] For example, after obtaining the target value of the motor drive frequency +0.05Hz, the inverter smoothly increases the motor frequency from 45.00Hz to 45.05Hz.

[0063] In some embodiments, after sending the target value of the motor drive frequency to the frequency converter to control the motor frequency of the gas compressor to be adjusted to the target value of the motor drive frequency, the method further includes: Acquire real-time sensing data collected by the sensor network to evaluate the control effect; The entire process data related to this frequency conversion control, as well as the control effect, are stored as case samples for future reference.

[0064] For example, after controlling the motor frequency conversion, the system immediately begins to monitor the adjustment effect and finds that the pipeline pressure rises and stabilizes near the target value in a short period of time. The system automatically packages the data from the entire process of data acquisition to control execution into a case sample marked "successful pressure fine-tuning in a healthy state", encrypts it, and uploads it to the case library on the cloud server. This provides training samples for the regular optimization of future models and can also be viewed by different users.

[0065] Based on this, the present invention also provides a specific embodiment, for example, when a slight crack appears in an impeller of a gas compressor. In steps S101 and S102, the vibration sensor captures the asynchronous vibration signal and forms a new energy peak in a high-frequency band of the FFT spectrum, extracts features to construct a standardized feature matrix; and inputs the standardized feature matrix into the state assessment model. In step S103, the vibration processing branch is sensitive to this new high-frequency energy peak, causing a significant change in its output feature vector. After the fusion layer, the final deep representation vector deviates from the ideal healthy state. Therefore, the output of the Softmax classifier becomes "sub-healthy state," and the health index HI output by the regressor also drops to 0.70. At this time, when calculating the comprehensive performance score CPS, because the state becomes "sub-healthy," the health weight wH is dynamically increased to 0.6. Even if the energy efficiency ratio may still be high at this time, due to the significant decrease in HI and the increase in wH, the final CPS score may plummet to 65 points.

[0066] In the S104 decision-making step, for the "sub-healthy state" and the low score of 65, the safety protection logic remains silent, but the energy efficiency optimization sub-logic is activated. It determines that there is significant optimization potential (or risk) at the current operating point, and thus begins small-step frequency exploration near the current frequency to find areas that can reduce vibration and improve the overall score. Simultaneously, the system sends a warning message to the central control room: "Equipment has entered a sub-healthy state; inspection recommended." The final target frequency value will be a trade-off between the energy efficiency optimization result and the pressure PID control requirements; the system will choose a slightly lower frequency to reduce stress on the cracked impeller.

[0067] In step S105, the complete event chain, including "sub-health diagnosis - triggering early warning - initiating energy efficiency optimization," is stored as an early failure case sample. When other compressors of the same type exhibit similar data patterns, the system, based on its learned experience, can make diagnoses and responses more quickly and accurately.

[0068] It is understood that the technical solution provided by the embodiments of the present invention has the following beneficial effects: In the technical solution of this invention, by integrating multi-dimensional sensing data such as mechanical vibration, electrical, and thermal, and using a pre-trained state assessment model for feature extraction and fusion, a continuous health index and accurate state level are output, thereby realizing quantitative diagnosis and forward-looking early warning of compressor sub-health and early failure. This fundamentally solves the technical problems of traditional systems being unable to accurately identify equipment sub-health and early failure, making it difficult to take into account equipment safety early warning, resulting in low operational reliability, poor energy efficiency, and high maintenance costs.

[0069] Furthermore, this invention pioneers a multi-layered decision-making mechanism that prioritizes "safety status" as the highest-priority trigger condition and uses "comprehensive performance scoring" as the optimization guide. Through a dynamic weight fusion algorithm, the system can automatically adjust the weights between energy efficiency optimization and pressure closed-loop control based on real-time health status. Under the premise of ensuring equipment operation safety, it dynamically explores energy efficiency potential and minimizes process disturbances, effectively solving the problem of "rigid" conflicts between multiple objectives.

[0070] By automatically constructing "decision-outcome" sample pairs and periodically retraining the model and strategy, this invention enables the control system to continuously learn from historical operating data, adapt to the aging and degradation of equipment performance and changes in operating conditions, and achieve a leap from "fixed parameter control" to "continuous evolution and optimization", ensuring optimal control performance throughout the entire life cycle.

[0071] The methods and apparatus disclosed in this application have a wide range of applications. In energy-intensive industries such as petrochemicals, steel metallurgy, and textile manufacturing, gas compressors are among the main energy-consuming devices. Applying this application, its powerful energy efficiency optimization sub-logic allows for refined energy efficiency management of compressors, continuously unlocking their energy-saving potential. Calculations show that even a 1% improvement in energy efficiency for a large compressor can save on annual electricity costs, providing strong technical support for enterprises to achieve their cost reduction and efficiency improvement goals.

[0072] In fields with extremely high requirements for production continuity, such as semiconductor manufacturing, food, and pharmaceuticals, any unplanned downtime can lead to huge economic losses. One technical feature of this application is its predictive maintenance capability. By identifying and warning of sub-optimal production conditions early, companies can transform unexpected failures into planned preventative maintenance, scheduling maintenance windows during planned downtime periods, thereby maximizing the stable operation of the production line.

[0073] For compressor stations deployed in remote areas or unattended environments, such as booster stations for natural gas pipelines and offshore oil platforms, manual inspections are costly and slow to respond. The remote monitoring and intelligent decision-making capabilities provided in this application are therefore particularly important. The system can perform expert-level health assessments of equipment 24 / 7 and transmit diagnostic results and maintenance recommendations to a central control center via the network, enabling lean "remote control" management of remote assets and reducing the manpower requirements and safety risks of on-site maintenance.

[0074] Furthermore, the adaptive learning capability of this application enables it to easily cope with complex and ever-changing operating conditions. For example, in some regions, seasonal temperature and humidity variations have a significant impact on the compressor's intake conditions, making it difficult for traditional fixed-parameter controllers to maintain optimal efficiency throughout the year. The system of this application can continuously learn the relationship between environmental changes and equipment performance, dynamically adjusting the control strategy, and automatically finding the most economical operating mode regardless of whether it is a hot summer or a cold winter.

[0075] The architecture of this application is also scalable. For large enterprises with a large fleet of compressors, case samples collected from each individual inverter control device 20 can be aggregated to a cloud platform. In the cloud, more powerful computing resources can be used to perform joint modeling and in-depth analysis of data from the entire fleet, discovering more universal fault modes and optimization strategies. The optimized model and knowledge base can then be distributed to control devices at various edge devices to achieve collaborative evolution, thereby building an intelligent equipment operation and maintenance system covering the entire enterprise.

[0076] Based on a general inventive concept, the present invention also provides a gas compressor frequency converter control device for implementing the above-described method embodiments. Figure 2This is a schematic diagram of the structure of a gas compressor frequency converter control device provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the device may include: The acquisition module 21 is used to respond to the frequency conversion monitoring request of the gas compressor and acquire full-dimensional sensing data of the gas compressor through a sensor network. The sensor network is deployed in the gas compressor body, motor, inlet and outlet pipelines, cooling system and environment. The full-dimensional sensing data covers the operating status, process requirements and environmental conditions of the gas compressor. Module 22 is used to process and extract features from the full-dimensional sensing data to construct a standardized feature matrix; Analysis module 23 is used to input the standardized feature matrix into a pre-trained state evaluation model for analysis, output the discrete state level and continuous health index of the gas compressor, and generate a comprehensive performance score based on the health index and real-time energy efficiency ratio; wherein, the discrete state level includes healthy state, sub-healthy state, early warning state, and alarm state; The control module 24 is used to generate a target value for the motor drive frequency based on the comprehensive performance score, the discrete state level, and the real-time pressure deviation through a multi-layer logic strategy; and to send the target value for the motor drive frequency to the frequency converter to control the motor frequency of the gas compressor to be adjusted to the target value for the motor drive frequency.

[0077] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0078] Based on a general inventive concept, the present invention also provides a variable frequency control device for a gas compressor, used to implement the above-described method embodiments. Figure 3 This is a schematic diagram of the structure of a gas compressor frequency converter control device provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the device may include a processor 31 and a memory 32, with the processor 31 connected to the memory 32. The processor 31 is used to call and execute a program stored in the memory 32; the memory 32 is used to store the program, which is at least used to execute the gas compressor frequency conversion control method in the above embodiments.

[0079] The specific implementation scheme of the gas compressor frequency conversion control device provided in this application can be referred to the implementation scheme of the gas compressor frequency conversion control method in any of the above embodiments, and will not be repeated here.

[0080] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0081] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0082] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0083] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0084] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0085] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0086] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0087] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0088] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A variable frequency control method for a gas compressor, characterized in that, Applied to frequency converter control equipment, the method includes: In response to the variable frequency monitoring request of the gas compressor, the gas compressor's full-dimensional sensing data is acquired through a sensor network. The sensor network is deployed in the gas compressor body, motor, inlet and outlet pipelines, cooling system, and environment. The full-dimensional sensing data includes the gas compressor's operating status, process requirements, and environmental conditions. The full-dimensional sensor data is processed and features are extracted to construct a standardized feature matrix; The standardized feature matrix is ​​input into a pre-trained state evaluation model for analysis, outputting the discrete state level and continuous health index of the gas compressor, and generating a comprehensive performance score based on the health index and real-time energy efficiency ratio; wherein, the discrete state level includes healthy state, sub-healthy state, early warning state, and alarm state; Based on the comprehensive performance score, the discrete state level, and the real-time pressure deviation, a target value for the motor drive frequency is generated through a multi-layer logic strategy. The target value of the motor drive frequency is sent to the frequency converter to control the motor frequency of the gas compressor.

2. The method according to claim 1, characterized in that, The process of processing and extracting features from the full-dimensional sensing data to construct a standardized feature matrix includes: Anomaly detection and missing value imputation are performed on the full-dimensional sensor data to obtain processed data; Extract time-domain features, frequency-domain features, and sequence trend features from the processed data; The full-dimensional sensing data, along with the extracted time-domain features, frequency-domain features, and sequence trend features, are aligned and recombined in time sequence to form the standardized feature matrix.

3. The method according to claim 2, characterized in that, The process of performing anomaly detection and missing value imputation on the full-dimensional sensing data to obtain processed data includes: Based on the full-dimensional sensing data, the 3σ criterion based on sliding window and the isolated forest algorithm are used to identify and remove transient interference and sensor failure data. Missing values ​​for non-critical parameters are filled using time series linear interpolation; for abnormal critical parameters, a data quality alarm is triggered to update the critical parameters; the updated critical parameters, the filled non-critical parameters, and other normal data are used as the processed data.

4. The method according to claim 2, characterized in that, The extraction of time-domain features, frequency-domain features, and sequence trend features from the processed data includes: Extract the root mean square value, peak factor, and impulse index from the time-domain features of the processed data; The vibration signal in the frequency domain features is extracted from the processed data and rapidly transformed to extract the amplitude energy of specific harmonics as fault symptom features. Pressure and flow sequence features are extracted from the processed data, and their short-term fluctuation variance and trend slope are calculated.

5. The method according to claim 1, characterized in that, The state assessment model adopts a multi-branch deep neural network architecture; The input layer receives the standardized feature matrix, and different branches process different types of feature data in parallel. Each branch extracts high-dimensional abstract features through convolutional layers, and the features are cross-complemented through fusion layers to generate a comprehensive state depth representation vector; The top classifier outputs discrete status levels of healthy, sub-healthy, early warning, and alarm through the Softmax function, while the regressor outputs a continuous health index through nonlinear normalization.

6. The method according to claim 1, characterized in that, The process of generating a comprehensive performance score based on the health index and real-time energy efficiency ratio includes: The real-time input power of the gas compressor is determined based on the sensor network, and the real-time energy efficiency ratio is calculated based on the theoretical power value. Based on the real-time energy efficiency ratio and the health index, a comprehensive performance score is calculated using the comprehensive performance scoring rules; wherein, the comprehensive performance scoring rules are as follows: CPS is the overall performance score; w H The health weighting coefficient is dynamically set based on discrete status levels; HI is the continuous health index; HImin is the effective threshold for the health index; w E η is the energy efficiency weighting coefficient, and it satisfies wH+wE=1; ηnorm is the normalized real-time energy efficiency ratio; λ is the load range correction factor.

7. The method according to claim 1, characterized in that, The multi-layer logic strategy includes: The safety protection sub-logic is used to prioritize generating frequency reduction or shutdown commands when the status is alarm. The energy efficiency optimization sub-logic is used to initiate frequency optimization based on the comprehensive performance score under safe conditions. The pressure control sub-logic is used to calculate the base frequency adjustment based on the real-time pressure deviation. The target value of the motor drive frequency is the result of weighted fusion of the output of the pressure control sub-logic after being corrected by the safety protection sub-logic and the energy efficiency optimization sub-logic; the priority of the safety protection sub-logic, the energy efficiency optimization sub-logic, and the pressure control sub-logic decreases sequentially.

8. The method according to claim 1, characterized in that, After sending the target value of the motor drive frequency to the frequency converter to control the motor frequency of the gas compressor to be adjusted to the target value of the motor drive frequency, the method further includes: Acquire real-time sensing data collected by the sensor network to evaluate the control effect; The entire process data related to this frequency converter control, as well as the control effect, are stored as case samples for future reference.

9. A variable frequency control device for a gas compressor, characterized in that, include: The acquisition module is used to respond to the frequency conversion monitoring request of the gas compressor and acquire full-dimensional sensing data of the gas compressor through a sensor network. The sensor network is deployed in the gas compressor body, motor, inlet and outlet pipelines, cooling system and environment. The full-dimensional sensing data covers the operating status, process requirements and environmental conditions of the gas compressor. The construction module is used to process and extract features from the full-dimensional sensor data to construct a standardized feature matrix; The analysis module is used to input the standardized feature matrix into a pre-trained state evaluation model for analysis, output the discrete state level and continuous health index of the gas compressor, and generate a comprehensive performance score based on the health index and real-time energy efficiency ratio; wherein, the discrete state level includes healthy state, sub-healthy state, early warning state, and alarm state; The control module is used to generate a target value for the motor drive frequency based on the comprehensive performance score, the discrete state level, and the real-time pressure deviation through a multi-layer logic strategy; and to send the target value for the motor drive frequency to the frequency converter to control the motor frequency of the gas compressor.

10. An electronic device, characterized in that, It includes a processor and a memory, wherein the processor is connected to the memory: The processor is used to call and execute the program stored in the memory; The memory is used to store the program, which is at least used to execute the gas compressor frequency conversion control method according to any one of claims 1-8.

Images