An environment compensation-based compressed air system energy efficiency intelligent evaluation method

Abstract

The application discloses an environmental compensation-based compressed air system energy efficiency intelligent evaluation method. The method first collects operation performance parameters, air quality parameters and environmental parameters, establishes a thermophysical mapping by using an enthalpy humidity chart and forms a quality coupling characteristic data set; meanwhile, a theoretical energy efficiency benchmark surface and a thermodynamic constraint set are constructed based on an isentropic compression principle and historical optimal samples. Subsequently, a nonlinear machine learning model is trained to learn the nonlinear mapping relationship between environmental parameter changes and performance and quality offset. The model converts the operation parameters under real-time non-standard environments into equivalent operation performance parameters under standard environmental conditions. During evaluation, the equivalent operation parameters are mapped to the theoretical energy efficiency benchmark surface, and the evaluation results reflecting the system energy efficiency deviation and performance degradation degree are obtained by calculating the distance vector. The method effectively eliminates environmental interference and significantly improves the accuracy and comparability of the compressed air system energy efficiency evaluation.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and in particular to an intelligent energy efficiency assessment method for compressed air systems based on environmental compensation. Background Technology

[0002] Compressed air systems are widely used in industrial settings, and their energy efficiency directly impacts energy consumption and equipment operating costs. Current technologies generally monitor system operation based on data collected by sensors, such as power consumption, output air volume, pressure stability, and air quality parameters, and analyze energy efficiency performance using thermodynamic models or empirical formulas. Simultaneously, some technologies utilize environmental information such as temperature, humidity, and atmospheric pressure to correct for operational data, mitigating the impact of environmental differences. Furthermore, with improved data acquisition capabilities, related fields are beginning to explore the introduction of machine learning methods to model the operational patterns of compressed air systems.

[0003] However, existing methods are mostly based on linear or quasi-linear correction strategies, which are difficult to effectively express the highly nonlinear relationship between environmental factors and operational performance. As a result, the interference of environmental changes on energy efficiency assessment results is still significant. At the same time, the multidimensional coupling relationship between air quality parameters and environmental conditions often lacks systematic modeling, making it impossible to accurately convert data collected under non-standard environmental conditions to comparable standard environmental conditions. Traditional data-driven methods are also prone to bias due to the lack of physical constraints, making it difficult to obtain robust and physically consistent energy efficiency evaluation results.

[0004] Therefore, there is an urgent need to introduce machine learning technology that combines physical constraints and nonlinear modeling capabilities to build a technical solution that can achieve environmental compensation and improve the consistency of energy efficiency assessment. Summary of the Invention

[0005] This application provides an intelligent energy efficiency assessment method for compressed air systems based on environmental compensation, thereby improving the accuracy of compressed air system energy efficiency assessment.

[0006] This application provides a method for intelligent energy efficiency assessment of compressed air systems based on environmental compensation, including: The system collects the operating performance parameters and air quality parameters of the compressed air system, and simultaneously collects environmental parameters including temperature, humidity and atmospheric pressure. It uses enthalpy-humidity diagrams to establish the thermal property mapping relationship between environmental parameters and air quality parameters, and aligns them in time order to form a quality coupling feature dataset. Based on the principle of isentropic compression and combined with historical best energy efficiency samples, a theoretical energy efficiency benchmark surface corresponding to standard environmental conditions is constructed, and a set of thermodynamic constraints is formed to limit the ideal boundary of operating performance based on the theoretical energy efficiency benchmark surface corresponding to standard environmental conditions. Based on the quality coupling feature dataset and the thermodynamic constraint set, a nonlinear machine learning model is trained to learn the nonlinear mapping relationship between the performance offset caused by changes in environmental parameters and the air quality offset. The real-time collected environmental parameters, actual operating performance parameters, and air quality parameters are input into the trained nonlinear machine learning model, which converts the operating performance parameters and air quality parameters under non-standard environments into equivalent operating performance parameters under corresponding standard environmental conditions. By mapping equivalent operating parameters to a theoretical energy efficiency reference surface and calculating the distance vector between parameter points and the reference surface, an energy efficiency assessment result reflecting the deviation of system energy efficiency and the degree of performance degradation can be obtained.

[0007] The beneficial effects of the technical solution provided in this application include: (1) A coupled modeling approach combining environmental parameters, operational performance parameters, and air quality parameters is adopted. Environmental disturbance compensation is achieved through a nonlinear machine learning model, enabling operational data under different temperature, humidity, atmospheric pressure, and air quality conditions to be uniformly converted to standard environmental conditions. This eliminates the evaluation bias caused by environmental differences and achieves energy efficiency comparability across equipment and time periods. (2) A theoretical energy efficiency benchmark is constructed using the isentropic compression principle and used as a thermodynamic constraint to constrain the model training process. This ensures that the model output has physical consistency, reflecting not only the inherent energy efficiency level of the system but also avoiding the unconstrained fitting error that easily occurs in pure data-driven methods, thus obtaining more reliable energy efficiency evaluation results. (3) Multidimensional coupled feature data is introduced during the training process, enabling the model to learn the combined influence of environmental changes on performance and air quality deviations. This overcomes the complex nonlinear relationships that traditional linear compensation methods struggle to handle and improves the energy efficiency inference capability under varying operating conditions. (4) By converting real-time operating parameters into equivalent standard environmental parameters and mapping them to the theoretical energy efficiency benchmark, the energy efficiency deviation and performance degradation of the system can be directly quantified, making energy efficiency diagnosis more intuitive and timely, which is conducive to the rapid identification of potential energy consumption anomalies, equipment aging and operation optimization space. Attached Figure Description

[0008] Figure 1 This is a flowchart of an intelligent energy efficiency assessment method for compressed air systems based on environmental compensation, provided in the first embodiment of this application. Detailed Implementation

[0009] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0010] The first embodiment of this application provides a method for intelligent energy efficiency assessment of compressed air systems based on environmental compensation. Please refer to... Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1 The first embodiment of this application provides a detailed description of an intelligent energy efficiency assessment method for compressed air systems based on environmental compensation.

[0011] Step S101: Collect the operating performance parameters and air quality parameters of the compressed air system, and simultaneously collect environmental parameters including temperature, humidity and atmospheric pressure. Use the enthalpy-humidity diagram to establish the thermal property mapping relationship between environmental parameters and air quality parameters, and align them in time order to form a quality coupling feature dataset.

[0012] In step S101, it is necessary to comprehensively collect the operating performance parameters, air quality parameters, and environmental parameters of the compressed air system. Based on this, a thermal property mapping relationship between environmental parameters and air quality parameters is established based on the enthalpy-humidity diagram. Then, a quality coupling feature dataset for subsequent model training is constructed through time alignment.

[0013] First, it is necessary to collect the operating performance parameters of the compressed air system. Operating performance parameters refer to data that reflect the energy efficiency and operating conditions of the compressed air system over a certain period. In this technical solution, the operating performance parameters include at least the system's instantaneous power consumption, output air volume (i.e., compressed air flow rate), and pressure stability. Power consumption can be obtained through the system's power metering module, generally in kWh or W; output air volume can be measured by a mass flow meter or volumetric flow meter, and the unit can be m³ / s. 3 / min; Pressure stability reflects the fluctuation range of pipeline pressure, which can be acquired in real time by pressure sensors and expressed in the form of standard deviation or maximum-minimum difference. For example, pressure values ​​are recorded within a 10-second time window and the standard deviation of the pressure sequence within that time window is calculated. All the above parameters need to be recorded with the same sampling period, such as once every 1 second or every 10 seconds, to ensure the temporal correlation between parameters.

[0014] Secondly, air quality parameters need to be collected. Air quality parameters reflect the quality characteristics of compressed air at the output end, and at least include dew point, oil content, and particulate matter concentration. Dew point can be measured using an online dew point meter; its value reflects the moisture content in the air and is usually expressed in degrees Celsius. Oil content can be obtained using an oil vapor detector or particulate monitor, and the unit is usually mg / m³. 3Particulate matter concentration is used to represent the content of solid particles in the air. Its value can be obtained using a laser particle counter. Typically, the number of particles is measured separately for different particle sizes (e.g., 0.1 μm, 0.3 μm, 0.5 μm, 1 μm, etc.) and recorded as particles per cubic meter or in grade terms. The sampling period for air quality parameters is consistent with that for operational performance parameters to facilitate subsequent data coupling.

[0015] Next, environmental parameters such as temperature, humidity, and atmospheric pressure need to be collected. These parameters directly affect air density, air moisture content, and compression work requirements, and therefore need to be recorded in real time using a dedicated environmental monitoring module. Ambient temperature can be obtained using a thermistor or thermocouple with an accuracy of ±0.1 ℃; relative humidity can be measured using a capacitive humidity sensor, in %RH; atmospheric pressure can be obtained using a MEMS barometer, in hPa or kPa. These environmental parameters also need to be collected at the same sampling period as the aforementioned performance parameters.

[0016] After collecting operational performance parameters, air quality parameters, and environmental parameters, it is necessary to establish a thermophysical mapping relationship between environmental parameters and air quality parameters using an enthalpy-humidity chart. An enthalpy-humidity chart is an engineering tool characterizing the relationship between air temperature, humidity, enthalpy, and moisture content. In this technical solution, it is used to convert ambient temperature and humidity into absolute air moisture content (unit: g / kg dry air). For example, under conditions of 30 ℃ and 70% relative humidity, the absolute air moisture content can be found to be approximately 18.5 g / kg on the enthalpy-humidity chart. When the conditions change to a temperature of 20 ℃ and humidity of 40%, the mapped moisture content is approximately 5.0 g / kg. Through this mapping, the influence of environmental factors can be transformed into the expected changing trend of air quality parameters, providing an interpretable physical basis for the subsequent identification of air quality shifts in the model.

[0017] Finally, all collected data needs to be time-series aligned. Time-series alignment refers to interpolating, synchronizing, or resampling data from different sensors on a unified time axis, ensuring that each time point simultaneously includes operational performance parameters, air quality parameters, and environmental parameters. For example, if air quality parameters are collected every 2 seconds and operational performance parameters are collected every 1 second, linear interpolation can be used to interpolate the air quality parameters to once per second, presenting all parameters at the same sampling frequency. After time-series alignment, a quality coupling feature dataset is formed, which includes environmental factors, air quality characteristics, and system operational performance. This dataset is stored in matrix form, with each row representing a complete operating condition sample at a specific time point, for example, organized as: [temperature, humidity, air pressure, absolute moisture content, power consumption, output air volume, pressure stability, dew point, oil content, particulate concentration].

[0018] Through the above data acquisition, thermal property mapping and time series alignment process, the final quality coupling feature dataset fully expresses the combined impact of environmental changes on operational performance and air quality, providing basic data support for training nonlinear machine learning models in subsequent steps.

[0019] Furthermore, the system collects operating performance parameters and air quality parameters of the compressed air system, and simultaneously collects environmental parameters including temperature, humidity, and atmospheric pressure. It then uses an enthalpy-humidity diagram to establish a thermophysical property mapping relationship between these environmental parameters and air quality parameters, aligning them chronologically to form a quality-coupled feature dataset, including: At each operating node, multiple sensors for power consumption, output gas volume, pressure, dew point, oil content, and particulate matter are activated, and the internal clocks of the sensors are synchronized with a unified time base to obtain raw data with consistent timestamps. The synchronized ambient temperature, relative humidity, and atmospheric pressure are sequentially converted into absolute humidity content, specific enthalpy, and saturated humidity. The thermal property parameters are then combined with the corresponding air quality parameters to form a first characteristic sequence containing the coupling relationship between thermal property quality. The particulate matter concentration and oil content in the first feature sequence are corrected for deviation. By comparing the average value of the stable segment of adjacent sampling segments, the data difference caused by sensor zero drift is eliminated, and the corrected second feature sequence is obtained. Based on the actual pipeline layout, the difference between ambient temperature and humidity and compressed air outlet temperature is used to calculate the local water vapor condensation amount, and the dew point and absolute moisture content are compensated according to the local water vapor condensation amount to form a third characteristic sequence that reflects the true thermal and humid state. The third feature sequence is interpolated and resampled according to the selected uniform sampling period to align all parameters on the same time axis and organize them in chronological order into a quality coupled feature dataset containing environmental thermal properties, air quality and operational performance.

[0020] When synchronously acquiring compressed air system operating performance parameters, air quality parameters, and environmental parameters such as temperature, humidity, and atmospheric pressure, it is necessary to ensure that all sensors operate under the same time reference to avoid time misalignment between different data sources. In this embodiment, "multi-point sensors" refer to independent measurement units installed at different locations or on different pipe sections, including power metering modules for measuring power consumption, flow sensors for measuring output air volume, pressure sensors for measuring pressure, dew point meters for measuring dew point, oil vapor detectors for measuring oil content, and laser particle counters for measuring particulate matter. Since the sampling clocks within these sensors are independent of each other, their readings at the same physical moment may have millisecond-level or even second-level errors under default conditions. Therefore, clock synchronization using a unified time reference is required. Clock synchronization is typically achieved through a unified network time protocol (such as NTP) or hardware pulse signals, ensuring that all sensors output data with the same timestamp.

[0021] To convert ambient temperature, relative humidity, and atmospheric pressure into absolute moisture content, specific enthalpy, and saturated humidity of air, it is necessary to rely on the air thermophysical properties described by the enthalpy-humidity chart. The enthalpy-humidity chart is an engineering tool reflecting the relationship between air temperature, moisture content, and enthalpy, and is widely used in HVAC engineering. This patent obtains three types of thermophysical parameters through enthalpy-humidity chart lookup or calculation. Absolute moisture content represents the mass of water vapor contained in one kilogram of dry air; specific enthalpy represents the energy content of the air; and saturated humidity represents the maximum moisture-holding capacity of the air at the current temperature. For example, when the ambient temperature is 30°C, the relative humidity is 70%, and the atmospheric pressure is 100 kPa, the absolute moisture content can be found to be approximately 18 g / kg, the saturated humidity to be approximately 26 g / kg, and the specific enthalpy to be approximately 72 kJ / kg on the enthalpy-humidity chart. This step combines these thermophysical parameters with the dew point, oil content, and particulate matter concentration at the same time to form a first characteristic sequence that includes the coupling relationship between air properties and environmental conditions.

[0022] Since particulate matter sensors and oil content sensors are often affected by zero-point drift, long-term operation may cause their output to shift. To prevent sensor errors from propagating to subsequent steps, deviation correction is needed for the first characteristic sequence. Zero-point drift typically manifests as a slow shift or abnormal increase / decrease in sensor output during a stable operating period without any actual change. This implementation identifies the drift trend by comparing the average value of stable segments within adjacent time periods, and then subtracts this drift amount from the current sequence to achieve effective correction. For example, if a particulate matter sensor outputs 1000 particles / m³ in two stable operating periods... 3 With 1100 / m 3 If the actual working conditions remain unchanged, then it can be assumed that the latter has changed by 100 units / m. 3The drift can be removed from the second output segment by eliminating the deviation. The output after correction in this step constitutes the second feature sequence.

[0023] After obtaining the second characteristic sequence, the impact of pipeline layout on the thermal and humidity state of the air needs to be further considered. Temperature differences between different pipe sections can lead to water vapor condensation, especially at bends, diameter changes, or areas near the outlet of the refrigerated dryer where temperatures decrease. To calculate the local water vapor condensation amount, the difference between the ambient temperature and humidity and the outlet air temperature is required. The so-called "local water vapor condensation amount" refers to the mass of water vapor precipitated from the air due to air cooling. The condensation amount can be estimated by comparing the saturated humidity of the air at different temperatures. For example, when the saturated humidity of air is 26 g / kg at 30°C, but only 15 g / kg at 20°C, if the air is not dehumidified, approximately 11 g / kg of condensate water will be generated during the cooling process. This step adjusts the dew point and absolute moisture content based on this condensation amount to make the calculation results closer to the actual cooling and water precipitation process experienced by the air in the pipeline, ultimately obtaining the third characteristic sequence that reflects the true thermal and humidity state.

[0024] To ensure that all parameters from different sources and with different sampling frequencies are processed uniformly on the same time axis, interpolation resampling of the third feature sequence is required. The purpose of interpolation resampling is to present all parameters with the same time step, such as one sampling point every 1 second or every 500 milliseconds. If the dew point meter samples every 2 seconds and the pressure sensor samples every 0.5 seconds, linear interpolation can be used to pad the dew point data to once per second, making it consistent with the sampling frequency of other parameters. After this step, the data points of each parameter are uniformly aligned on the same time axis, forming a complete quality-coupled feature dataset in chronological order.

[0025] This quality coupling feature dataset contains full correlation information on environmental thermal properties, air quality, and operational performance, which can be directly used for any subsequent analysis and calculation.

[0026] Step S102: Based on the principle of isentropic compression and combined with historical best energy efficiency samples, construct a theoretical energy efficiency benchmark surface corresponding to standard environmental conditions, and form a set of thermodynamic constraints to limit the ideal boundary of operating performance based on the constructed theoretical energy efficiency benchmark surface corresponding to standard environmental conditions.

[0027] In step S102, a theoretical energy efficiency benchmark surface under corresponding standard environmental conditions needs to be constructed based on the isentropic compression principle and combined with historical best energy efficiency samples. This surface forms a set of thermodynamic constraints to define the ideal boundary of operational performance. The purpose of this step is to establish an "ideal performance boundary" that can serve as a reference for energy efficiency assessment, providing a stable and interpretable physical basis for subsequent model training and evaluation processes.

[0028] Isentropic compression is an engineering method used to describe the minimum energy required to compress air under completely adiabatic conditions with no energy loss. Under isentropic conditions, the theoretical minimum compression work required to compress air from the inlet pressure to the outlet pressure is unaffected by temperature loss and does not include additional energy consumption such as mechanical friction; it can be considered an "optimal possible performance." In practical engineering, the theoretical energy value required for isentropic compression can usually be obtained by consulting air thermodynamic property tables or using widely used engineering software (such as REFPROP, CoolProp, etc.). For example, given the inlet pressure, outlet pressure, and air properties, the software will directly provide the theoretical minimum power requirement for the compression process.

[0029] To establish a theoretical energy efficiency benchmark, the first step is to determine the standard environmental conditions. Standard environmental conditions typically refer to operating conditions with a temperature of 20°C, relative humidity of 50%, and atmospheric pressure of approximately 101.325 kPa. These are commonly used reference conditions in compressed air energy efficiency assessment. Under these environmental conditions, the theoretical minimum energy consumption values ​​for different operating points need to be calculated. For example, an air compressor may operate in different load ranges, such as 50% load, 75% load, or 100% full load. The corresponding outlet pressure for each operating range may also be different, such as 0.6 MPa, 0.7 MPa, or 0.8 MPa. This step requires calculating the theoretical minimum energy consumption values ​​for each of these operating points, forming a set of theoretical benchmarks representing the ideal state.

[0030] To more closely reflect actual equipment performance, this step, in addition to the isentropic theoretical calculations, also requires correction based on historical best-efficiency samples. These historical best-efficiency samples are obtained by long-term monitoring of compressed air system operation data, selecting the operating segments with the lowest unit energy consumption and highest operating efficiency under environmental conditions as close to standard as possible. These segments reflect the optimal performance the equipment can achieve in real-world operating environments and therefore have engineering significance. For example, if the unit energy consumption of a certain segment of actual operating data is slightly higher than the isentropic theoretical value, this data point can be included in the benchmark construction as "actually achievable ideal performance" to avoid the problem of unrealistic benchmarks caused by overly ideal theoretical values.

[0031] After integrating the isentropic theoretical benchmark with historical best energy efficiency sample points, a continuous multidimensional benchmark surface needs to be constructed using interpolation or surface fitting. Common engineering methods, such as cubic spline interpolation, radial basis function fitting, or polynomial surface fitting, can be employed to ensure that it provides the theoretically optimal energy efficiency value for the corresponding standard environment at any operating point. This benchmark surface is referred to as the "theoretical energy efficiency benchmark surface" in this technical solution, and it can provide a reference value for ideal energy efficiency throughout the entire operating range.

[0032] Based on this theoretical energy efficiency benchmark, a set of rules for constraining model predictions can be further derived, namely, the thermodynamic constraint set. This set defines a physically acceptable performance range. For example, it requires that the model's predicted equivalent energy consumption not be lower than a certain percentage of the theoretical minimum energy consumption, or that the predicted pressure stability or air quality deviation must be within the range achievable by historical best samples. The role of the thermodynamic constraint set is to ensure that the machine learning model does not produce results that violate physical laws during training and inference, thus ensuring the consistency and credibility of the final energy efficiency assessment results.

[0033] Through the above steps, a fully operable standard environmental theoretical energy efficiency benchmark surface can be constructed, and a set of thermodynamic constraints with clear engineering boundaries can be established, providing a reliable reference framework for the training and evaluation of machine learning models in subsequent steps.

[0034] Furthermore, based on the principle of isentropic compression and combined with historical best energy efficiency samples, a theoretical energy efficiency benchmark surface corresponding to standard environmental conditions is constructed. Based on this benchmark surface, a set of thermodynamic constraints is formed to define the ideal boundary of operational performance, including: Under standard environmental conditions, the theoretical minimum energy consumption of multiple typical operating points is calculated by using the inlet pressure, outlet pressure and thermal properties of the compressed air system as inputs, through isentropic compression relationship. The theoretical minimum energy consumption of each operating point is then combined with the corresponding flow range to form an initial theoretical energy consumption dataset. The initial theoretical energy consumption dataset is compared with the optimal energy efficiency samples selected from historical operation records, and the theoretical minimum energy consumption of each operation point is corrected according to the deviation range that the actual equipment can achieve under optimal operating conditions, to obtain the second energy consumption dataset after equipment characteristic calibration. The continuity of energy consumption values ​​of adjacent operating points in the second energy consumption dataset is checked. By calculating the gas volume gradient and energy consumption gradient between operating points, discrete energy consumption points caused by short-term abnormal operating conditions are removed, and intermediate operating condition values ​​with smooth transition are re-inserted to obtain a continuous and fitable third energy consumption dataset. The third energy consumption dataset is interpolated in two dimensions according to pressure level and flow level to obtain a continuous theoretical energy efficiency reference surface covering the entire operating range. The minimum energy consumption curve corresponding to the same pressure level in the continuous theoretical energy efficiency reference surface is extracted again as the performance limit curve to form a complete ideal performance boundary. Based on the ideal performance boundary, the acceptable upper limit of energy consumption, the upper limit of pressure fluctuation, and the upper limit of air humidity are calculated respectively. The upper limit values ​​are combined with the corresponding optimal values ​​on the continuous theoretical energy efficiency reference surface to form a set of thermodynamic constraint intervals, so that the predicted operating performance shall not be lower than the theoretical limit or higher than the physically achievable range, thus forming a set of thermodynamic constraints that limit the ideal boundary of operating performance.

[0035] In constructing a theoretical energy efficiency benchmark based on the isentropic compression principle and historical best energy efficiency samples, it is necessary to first clarify the meaning of "standard environmental conditions." Standard environmental conditions refer to industry-recognized comparison benchmarks, typically using a temperature of 20°C, relative humidity of 50%, and atmospheric pressure of 101.325 kPa, to eliminate the incomparability of compressed air energy efficiency data under different environments. Under these conditions, the inlet pressure, outlet pressure, and thermal properties of the compressed air system are standardized and used as the basic input for calculating isentropic compression energy consumption. The isentropic compression relationship is a commonly used physical model in the theoretical efficiency analysis of air compressors. It describes the minimum work required to increase air pressure from inlet pressure to outlet pressure under ideal conditions with no heat exchange and no losses. Its calculation is usually based on physical properties such as inlet specific enthalpy, outlet specific enthalpy, and specific heat ratio. This manual does not introduce explicit mathematical formulas but uses an implementable method to explain its calculation logic. For example, with an inlet pressure of 100 kPa and an outlet pressure of 700 kPa, the theoretical outlet state of air after isentropic pressurization can be obtained by consulting the standard air enthalpy-humidity chart, including the increase in specific enthalpy. Multiplying this specific enthalpy difference by the flow rate yields the theoretical minimum energy consumption at a given operating point. By repeating the above calculations in different typical flow ranges, an initial theoretical energy consumption dataset covering multiple operating points can be generated.

[0036] After obtaining the initial theoretical energy consumption dataset, it needs to be compared with the optimal energy efficiency samples selected from historical operating records. The optimal energy efficiency samples refer to a set of records exhibiting the lowest unit energy consumption during long-term operation, typically corresponding to the operating point when the equipment is in good condition, adequately cooled, has no pipeline leaks, and is adequately lubricated. Since the theoretical minimum energy consumption assumes no losses, while actual equipment inevitably experiences factors such as mechanical friction, valve resistance, and flow losses, the theoretical value must be corrected using the optimal samples to reflect the actual achievable performance of the equipment. For example, when the theoretical minimum energy consumption is calculated to be 1.00 kWh / m³... 3 The historical best sample size was 1.15 kWh / m³. 3 Then, based on the equipment's optimal performance, the theoretical energy consumption at that operating point can be adjusted to a practically achievable value between the two, such as 1.12 kWh / m³. 3 And all operating points are corrected in the same way to form a second energy consumption dataset.

[0037] To further ensure the continuity of the energy efficiency benchmark, it is necessary to check the energy consumption change trends of adjacent operating points in the second energy consumption data set. Continuity checking involves comparing whether the rate of change in gas volume and the rate of change in energy consumption are proportional between two adjacent flow intervals. If a point experiences a sudden increase or decrease in energy consumption, while the operating points before and after it remain stable, this point is likely unrepresentative due to insufficient cooling, instantaneous pressure disturbances, or sensor errors, and should be removed from the list. Those skilled in the art can understand this process through simple examples, such as a flow rate from 1.0 m³ / s... 3 / min to 1.2 m 3 Energy consumption should increase by about 5% per minute according to normal patterns. If a record point increases by 20%, it should be considered an outlier. After removing outliers, to avoid gaps in the data, it is necessary to re-insert the interpolated intermediate operating conditions to make the energy consumption change present a smooth curve, thus forming the third energy consumption dataset.

[0038] Based on the third energy consumption dataset, two-dimensional interpolation is performed according to pressure and flow levels to form a continuous and fitable energy efficiency benchmark surface across the entire operating range. Two-dimensional interpolation means that energy consumption varies not only with flow rate but also with pressure; therefore, interpolation needs to be performed on all discrete operating points in a two-dimensional space to ensure that a corresponding benchmark energy consumption value can be obtained under any pressure and flow rate combination. Within this continuous energy efficiency benchmark surface, it is further necessary to extract the curve with the lowest energy consumption value corresponding to the same pressure level, which is then used as the performance limit curve. The performance limit curve reflects the lowest energy consumption value that the equipment can achieve under a certain pressure under ideal conditions, and is therefore a key basis for constructing the ideal boundary of operating performance.

[0039] After obtaining the performance limit curve, it is necessary to calculate the acceptable upper limits for energy consumption, pressure fluctuation, and air moisture content based on this curve. The upper limit for energy consumption can be obtained by increasing the performance limit curve by an acceptable deviation, for example, a deviation of no more than 10% determined according to the manufacturer's specifications or industry standards. The upper limit for pressure fluctuation can be determined based on the allowable stability range of the pressure regulation system, for example, within 0.2 bar above and below the rated pressure. The upper limit for air moisture content can be determined based on dew point limits and the performance of the outlet refrigerated dryer; the maximum permissible absolute moisture content can be determined by consulting the enthalpy-humidity chart. These upper limits are combined with the optimal values ​​in the energy efficiency reference surface, and all operating points are restricted in the form of intervals, ensuring that they do not fall below the theoretical limit or exceed the physical range achievable by the equipment, thus forming a complete set of thermodynamic constraints.

[0040] Furthermore, the continuity check of energy consumption values ​​at adjacent operating points in the second energy consumption dataset is performed. By calculating the gas volume gradient and energy consumption gradient between operating points, discrete energy consumption points caused by short-term abnormal operating conditions are eliminated, and intermediate operating condition values ​​with smooth transitions are re-inserted to obtain a continuously fit third energy consumption dataset, including: The running points in the second energy consumption dataset are sorted in ascending order of gas volume, and the gas volume gradient representing the rate of change of gas volume is calculated for adjacent running points after sorting, so as to obtain the first gradient sequence used to judge the stability of the change of running points. The first gradient sequence is compared with the energy consumption changes in the second energy consumption dataset. The energy consumption gradient representing the rate of energy consumption change is calculated for adjacent operating points. The gas volume gradient and the energy consumption gradient are combined to obtain a joint gradient judgment sequence for identifying short-term operating condition anomalies. Based on the joint gradient determination sequence, the operating points that simultaneously satisfy the sudden change in gas volume gradient and the deviation of energy consumption gradient from the continuous trend are marked as discrete energy consumption points. After removing discrete energy consumption points, a local transition interval is constructed based on the relationship between gas volume and energy consumption changes of the operating points before and after removal, and the parameters of the transition interval that need to be supplemented are obtained. Based on the transition interval parameters, multiple intermediate operating points with increasing gas volume and smooth energy consumption change characteristics are generated between the eliminated operating points, and the intermediate operating points are inserted into the corresponding positions so that the gas volume changes monotonically with the sequence and the energy consumption change remains continuous. After inserting intermediate operating points, the operating points are reorganized into a continuous energy consumption sequence according to gas volume order to form a third energy consumption dataset that can reflect the real changing trend of operating conditions and can be used for subsequent interpolation fitting.

[0041] During the continuity check of the second energy consumption dataset, it is necessary to first clarify the meanings of the so-called gas volume gradient and energy consumption gradient. The gas volume gradient represents the rate of change of the gas volume parameter between adjacent operating points. This is obtained by subtracting the gas volume value of the previous operating point from the gas volume value of the subsequent operating point. This change is then considered as the rate of change based on the sequential interval between the two points (usually one sampling period). The energy consumption gradient is calculated similarly to the gas volume gradient, also reflecting the trend of energy consumption change with gas volume through the difference in energy consumption values. Since this step aims to identify abnormal operating conditions, it is necessary to simultaneously observe abrupt changes in gas volume and whether energy consumption changes deviate from the normal trend. The combination of these two factors constitutes the basis for determining whether continuity has been disrupted.

[0042] To ensure a consistent order of change in the data, all operating points in the second energy consumption dataset should first be sorted in ascending order of gas volume, ensuring that the gas volume value of each operating point increases sequentially. Those skilled in the art typically use a simple ascending order when processing such sequences. For example, if the gas volumes of three operating points are 0.5, 0.2, and 0.8 cubic meters per minute, the sorted values ​​would be 0.2, 0.5, and 0.8. After sorting, the gas volume gradient is calculated for adjacent operating points, which is the latter's gas volume minus the former's. For example, if one operating point has a gas volume of 0.2 and the next has a gas volume of 0.5, the gas volume gradient is 0.3, indicating that the system's gas volume increase within this range is normal. By performing this operation on all operating points, a continuous gradient sequence can be formed to observe whether the gas volume changes smoothly.

[0043] After obtaining the gas volume gradient, it is necessary to further calculate the energy consumption gradient for each pair of adjacent operating points, i.e., the difference in energy consumption values. For example, if the energy consumption of one operating point is 12 kWh and the next point is 20 kWh, then the energy consumption gradient is 8 kWh. If the energy consumption gradients between most operating points are within a small range, but a pair of operating points suddenly shows an excessively large energy consumption gradient, it indicates that the operating point may be under unstable conditions such as short-term operating fluctuations, sensor malfunctions, or sudden load changes. Subsequently, the gas volume gradient and energy consumption gradient are combined, and a joint gradient determination sequence is formed by comparing the reasonable range of the gas volume gradient with the continuity of the energy consumption gradient. For example, when an adjacent point with a gas volume gradient of 0.1 suddenly jumps to an energy consumption gradient of 15 kWh, while the energy consumption gradients of surrounding operating points are usually no more than 3 kWh, then the operating data corresponding to this point can be identified as a discrete energy consumption point.

[0044] After discrete energy consumption points are identified and removed, a gap in operating points will appear in the dataset. At this point, a local transition interval needs to be constructed based on the relationship between the gas volume changes and energy consumption changes of the two normal operating points before and after removal. This transition interval is used to generate intermediate operating points for a smooth transition. The transition interval can be understood as the range of operating condition changes between two adjacent normal operating points. If the gas volume of the previous operating point is 0.4 and the energy consumption is 10 kWh, and the gas volume of the next operating point is 0.6 and the energy consumption is 14 kWh, then it can be inferred that within the gas volume range of 0.4 to 0.6, the system's energy consumption should increase steadily with the increase in gas volume. If a discrete point in the middle represents a significant jump, such as an energy consumption of 25 kWh, then after removing that point, several intermediate operating points need to be added. For example, representative operating points can be generated at gas volumes of 0.46, 0.52, and 0.58 kWh. Based on the energy consumption trends of these points, reasonable energy consumption values ​​can be calculated for these points, such as 11.5, 12.5, and 13.5 kWh respectively, to ensure that energy consumption exhibits a linear or slightly curved smooth change. Either linear interpolation or local spline interpolation can be used, as long as it ensures that no new abnormal abrupt changes in energy consumption occur.

[0045] After insertion, the operating points containing the newly added intermediate operating points need to be re-sorted by air volume to ensure the entire sequence exhibits a monotonic change relationship from smallest to largest. The operating point sequence after this processing will simultaneously possess air volume continuity, energy consumption continuity, and energy consumption distribution characteristics that conform to physical change trends. This will accurately reflect the energy efficiency changes of the compressed air system under different operating conditions, allowing the sequence to be subsequently used to fit two-dimensional or three-dimensional theoretical energy efficiency benchmarks, ultimately serving as a third energy consumption dataset for subsequent analysis.

[0046] Step S103: Based on the quality coupling feature dataset and the thermodynamic constraint set, train a nonlinear machine learning model to learn the nonlinear mapping relationship between the performance offset caused by changes in environmental parameters and the air quality offset.

[0047] In step S103, the quality coupling feature dataset constructed in step S101, combined with the thermodynamic constraint set established in step S102, is used to train the nonlinear machine learning model. This enables the model to learn the combined influence of environmental condition changes on operating performance parameters and air quality parameters, and to accurately identify the performance and quality offsets caused by these influences. The purpose of this step is to construct an environmental compensation model that accurately expresses the multi-parameter coupling relationship and conforms to the physical operating boundary, allowing it to convert operating data collected under non-standard environments into equivalent parameters under standard environments in subsequent steps.

[0048] To ensure the feasibility of model training, the organization of the training data needs to be clearly defined. The quality coupling feature dataset obtained in step S101 is essentially a multi-dimensional time series matrix, where each row represents the complete operating state at a certain time point, including parameters such as ambient temperature, relative humidity, atmospheric pressure, absolute humidity content of air, system power consumption, output gas volume, pressure stability, dew point, oil content, and concentration of particulate matter of different sizes. This step uses this dataset as the input sample data for the model. To ensure that the model can identify the performance deviation caused by changes in environmental parameters, these samples need to be labeled, that is, to define the "theoretical corresponding value" of each operating point under standard environmental conditions. This theoretical value comes from the theoretical energy efficiency benchmark surface constructed in step S102. For example, when the output pressure of a certain actual operating point is 0.7 MPa and the output gas volume is 6 m³ / s... 3 When the energy consumption is / min, the corresponding ideal energy consumption or ideal specific power value under standard environment can be found through the theoretical energy efficiency reference surface. This ideal value can be regarded as the "standard environment label" for this operating point.

[0049] To enable machine learning models to learn the offset relationships caused by environmental changes, these labels need to be combined with actual collected operational performance parameters to form an offset. For example, when the actual ambient temperature is high, causing a decrease in air density, the specific work per unit volume of air may increase. In this case, the offset can be represented by the "difference between the actual specific work and the standard specific work." This difference can be calculated in data processing by subtracting the standard value corresponding to the reference surface from the actual value, and can be reproduced as long as the input and target data are clearly defined.

[0050] After constructing the training data, it is necessary to determine the specific type of nonlinear machine learning model. This invention does not limit the specific algorithm, but the model must be able to handle multidimensional nonlinear relationships. Commonly used models include deep neural networks, gradient boosting trees, random forests, or support vector regression based on kernel functions. For example, deep neural networks can automatically extract the complex correlation between environmental parameters and operational performance parameters through multi-layered structures; gradient boosting trees can effectively learn the nonlinear effects of environmental changes on air quality shifts such as dew point, oil content, and particulate concentration. The choice of model depends on the data scale and actual engineering requirements; this invention requires that the model at least be able to express nonlinear mapping relationships.

[0051] During training, thermodynamic constraints need to be used as either "soft constraints" or "hard constraints" for the model. Soft constraints involve adding a penalty term to the model's error function; for example, if the model's output equivalent energy consumption is lower than the physical lower limit allowed by the theoretical energy efficiency benchmark, it is considered a prediction error and subject to additional penalties. Hard constraints, on the other hand, can directly prohibit the output of any predicted values ​​exceeding physical boundaries during backpropagation. This approach effectively prevents the model from learning inexplicable or physically inconsistent anomalous mappings, ensuring that the final model not only fits the data but also conforms to the physical operating characteristics of the compressed air system.

[0052] After training, the model will have the ability to map operational data collected under arbitrary environmental conditions to equivalent operational parameters under standard conditions. In other words, it learns a mapping rule that can separate the energy efficiency and air quality offsets caused by changes in temperature, humidity, and atmospheric pressure from the operational data.

[0053] To further enable those skilled in the art to understand how to train a nonlinear machine learning model in this step, a specific model example is given below. In this embodiment, a three-layer feedforward neural network with a simple structure but capable of expressing nonlinear relationships is used as the environment compensation model. This neural network consists of three parts: an input layer, a hidden layer, and an output layer.

[0054] In the input layer, all parameters related to the operating state in the quality coupling feature dataset constructed in step S101 are used as input, including ambient temperature, relative humidity, atmospheric pressure, absolute humidity content of air, current power consumption, output air volume, pressure fluctuation amplitude, dew point, oil content, and concentration of particulate matter of various sizes. For example, when there are 12 input parameters, the input layer contains 12 nodes. Those skilled in the art can implement this layer in any deep learning framework (such as TensorFlow, PyTorch, or Keras) by defining a tensor of size 12 as input, without additional complex operations.

[0055] The hidden layer is used to learn the nonlinear mapping relationship between changes in environmental parameters and performance bias. This embodiment uses a 32-node hidden layer with the ReLU activation function to enhance the model's ability to fit nonlinear relationships. The input to the hidden layer is the vector output from the input layer, and its output is a set of 32-dimensional intermediate features.

[0056] The output layer generates predicted performance offset values ​​for this operating state. The number of nodes in the output layer is determined by the training objective. In this embodiment, the model needs to simultaneously predict four types of indicators: equivalent unit energy consumption offset, equivalent gas volume offset, pressure fluctuation offset, and air quality offset. Therefore, the output layer is set to 4 nodes. Each node corresponds to one type of offset, and the output is a continuous value. Those skilled in the art can implement this function using a linear activation function, because the offset is essentially a numerical prediction problem and does not require probabilistic representation.

[0057] During training, the thermodynamic constraint set constructed in step S102 needs to be used as a constraint term in the loss function to ensure that the model's predictions do not exceed the physically permissible range. Specifically, when a predicted value output by the model is lower than the minimum allowable value of the theoretical energy efficiency benchmark, an additional penalty term is added, allowing the model to automatically correct the bias during training. For example, if the predicted equivalent unit energy consumption is lower than the theoretical minimum energy consumption, it is considered a constraint violation, and the corresponding error value is increased to correct the prediction back to a reasonable range in the next training round. Those skilled in the art can implement this by defining a custom loss function within a deep learning framework.

[0058] Through the above structure, the nonlinear machine learning model in this embodiment can effectively learn the performance offset caused by changes in environmental parameters after training, and realize the ability to map multidimensional operating states to equivalent operating parameters under standard environmental conditions.

[0059] Furthermore, the nonlinear machine learning model includes: The environmental disturbance coding unit is used to receive environmental thermophysical parameters such as temperature, relative humidity, atmospheric pressure and their specific enthalpy value and absolute humidity obtained by converting them through the enthalpy-humidity diagram, and to obtain an environmental offset vector to characterize the magnitude of environmental disturbance by performing a nonlinear scaling transformation on the rate of change of temperature and humidity through a multi-segment piecewise hyperbolic function. The quality offset deconstruction unit is used to receive air quality parameters such as dew point, oil content, and particulate matter concentration, as well as operating performance parameters such as power consumption, output air volume, and pressure. After aligning the received parameters by time, it inputs them into a residual deconstructor based on local trend analysis. By calculating the difference between the local linear trend and the global average trend of adjacent sampling points, a quality offset vector representing the contribution of air quality offset is obtained. The coupled response inference unit takes the environmental offset vector and the air quality offset vector as joint inputs and performs three types of operations in sequence through a three-stage nonlinear combination mechanism: environmental priority amplification, cross-sensitivity suppression, and energy efficiency correlation enhancement. Among them, environmental priority amplification enhances the sensitivity under high temperature and high humidity conditions through an exponential amplification factor; cross-sensitivity suppression reduces the collinearity effect by performing inverse scaling on the product term of the air quality offset vector and the environmental offset vector; and energy efficiency correlation enhancement forms the coupled offset response vector by introducing the instantaneous change rate of the operating performance parameters into the offset weight. The standard state equivalent restoration unit is used to take the coupled offset response vector and the operating performance parameters as inputs. Through continuous interval piecewise regression, an environmental offset compensation function is established for each operating performance parameter. The environmental offset compensation function adopts different fitting slopes in different flow, pressure and dew point intervals, and the compensated parameters are output as equivalent operating performance parameters under the corresponding standard environmental conditions. The offset consistency correction unit is used to perform physical consistency constraint correction on the equivalent operating performance parameters. By comparing the physical boundary relationship between the inlet specific enthalpy, the outlet specific enthalpy and the compression ratio, it identifies points that violate energy conservation or do not meet the physical constraints of the compressor, and corrects the abnormal points back to the allowable offset range to obtain the final standard state equivalent operating performance parameters.

[0060] When implementing the aforementioned nonlinear machine learning model, it is first necessary to clarify the working method of the environmental perturbation encoding unit in the model. This unit receives environmental parameters including temperature, relative humidity, atmospheric pressure, and specific enthalpy and absolute humidity calculated from the enthalpy-humidity diagram. Specific enthalpy is the total heat contained in a unit mass of air, an important physical quantity used to describe the thermal state of air; absolute humidity is the actual mass of water vapor contained in a unit volume of air, used to reflect the true humidity content. After obtaining these parameters, the unit needs to calculate the rate of change of temperature and humidity, i.e., the ratio of the difference between the current time and the previous sampling time to the sampling time interval. The higher the rate of change, the more significant the environmental abrupt change. In this embodiment, a piecewise hyperbolic function is used to nonlinearly scale the rate of change. The hyperbolic function can be constructed in a manner commonly used in the art, such as using the hyperbolic tangent function, which outputs an approximately linear value when the rate of change is small and a gradually saturating value when the rate of change is large, thereby enhancing the model's sensitivity to environmental abrupt changes in scenarios with rapid changes in high temperature or high humidity. By setting different scaling coefficients for different rate-of-change intervals—for example, using weak scaling for intervals with a rate of change less than 1°C / minute and strong scaling for intervals with a rate of change exceeding 5°C / minute—an environmental offset vector characterizing the intensity of environmental disturbances can be generated. Those skilled in the art can adjust the scaling interval and scale according to the numerical range to achieve a highly controllable nonlinear mapping.

[0061] The analysis then proceeds to the Quality Offset Decomposition Unit, which processes air quality parameters such as dew point, oil content, and particulate matter concentration, and combines them with operational performance parameters such as power consumption, output air volume, and pressure for joint analysis. Before entering the analysis, all parameters must be aligned to a unified timestamp to ensure they represent the same operational state at the same moment. This unit employs a residual decomposition method based on local trend analysis. The local trend refers to the slope of a linear regression calculated within a short time window (e.g., 10 or 30 seconds); while the global average trend is the average direction of change of the corresponding parameter calculated over a longer period (e.g., the past 30 minutes). The residual is the difference between the local trend and the global trend. Those skilled in the art can obtain the trend value using simple linear regression, without the need for complex models. For example, when calculating particulate matter concentration, if the local trend is +20 μg / m³ per minute... 3 The global trend is +2 μg / m² per minute. 3 Therefore, the residual is 18 μg / m 3 This represents a short-term abnormal offset. A set of values ​​can be obtained by performing the same calculations on parameters such as dew point and oil content. Combining these residuals forms the quality offset vector, which represents the instantaneous impact of air quality changes on system performance.

[0062] The coupled response inference unit combines the environmental offset vector and the air quality offset vector to generate a coupled offset response that simultaneously reflects environmental and air quality factors. This unit sequentially executes three stages: environmental priority amplification, cross-sensitivity suppression, and energy efficiency correlation enhancement. Environmental priority amplification is suitable for high-temperature and high-humidity conditions. When the specific enthalpy exceeds a certain threshold (e.g., 75 kJ / kg), this unit sets an exponential amplification factor to give higher weight to the highly sensitive components in the environmental offset vector, thereby enhancing the system's response to harsh environments. In the cross-sensitivity suppression stage, the air quality offset vector and the environmental offset vector are multiplied term by term to identify simultaneously changing collinear factors. These factors are then compressed using an inverse scaling factor (e.g., a scaling factor less than 1) to prevent air quality changes from being incorrectly interpreted as environmental changes. Energy efficiency correlation enhancement introduces the relative change value of operating performance parameters (e.g., the rate of change in power consumption) as an additional weight, multiplying it by the output of the first two stages to strengthen dimensions with high energy efficiency correlation. The final coupled offset response vector clearly reflects the combined effect of environment and air quality on system performance offset.

[0063] The standard-state equivalent reduction unit is used to combine the coupled offset response with the operating performance parameters to obtain the equivalent operating performance parameters under standard environmental conditions. This implementation employs a continuous interval piecewise regression method, dividing the operating performance parameters into different intervals based on pressure, flow rate, and dew point values, such as low flow rate, medium flow rate, and high flow rate. Each interval uses a set of independent linear or nonlinear fitting slopes to map the coupled offset response to the compensation amount of the operating parameters. For example, when the output gas volume is in the medium flow rate stage and the environmental offset is strong, a steeper slope can be selected for compensation, making the conversion result closer to the standard-state load level. Those skilled in the art can complete the interval regression using conventional least-squares fitting. The compensated parameters are the equivalent operating performance parameters achievable under standard environmental conditions.

[0064] Finally, the offset consistency correction unit checks whether the equivalent operating performance parameters satisfy the basic physical laws of the compressed air system, such as energy conservation and the correspondence between compression ratio and specific enthalpy changes. If the equivalent parameters lead to an unreasonable combination between the inlet specific enthalpy, outlet specific enthalpy, and compression ratio—for example, the outlet specific enthalpy does not increase with the increase of the compression ratio—then it is determined that this point does not conform to physical consistency. For such anomalies, this unit will adjust them back to a reasonable range by reducing the compensation amount by a certain percentage (e.g., 10% or 20%), so that the final output standard-state equivalent operating performance parameters satisfy the true physical boundaries of compressor operation in terms of both numerical value and self-consistency, and can thus be used for subsequent energy efficiency assessments.

[0065] Step S104: Input the real-time collected actual environmental parameters, actual operating performance parameters and air quality parameters into the trained nonlinear machine learning model, and convert the operating performance parameters and air quality parameters under non-standard environment into equivalent operating performance parameters under the corresponding standard environmental conditions.

[0066] In step S104, the real-time collected environmental parameters, operational performance parameters, and air quality parameters need to be input into the nonlinear machine learning model trained in step S103. This allows the model to determine the degree of difference between the current operating conditions and standard environmental conditions based on the real-time data, and automatically generate equivalent operational performance parameters corresponding to the standard environment. The purpose of this step is to perform environmental compensation on the real-time data, ensuring that the operating status can be compared using the same standard under any environmental conditions. Therefore, this step must clearly define the organization of the input data, the execution method of the model inference, and the specific presentation method of the compensation results.

[0067] In practical implementation, real-time environmental parameters include the current ambient temperature, relative humidity, and atmospheric pressure; real-time operational performance parameters include the current instantaneous power consumption, output gas volume, pressure fluctuation amplitude, and their trend compared to the previous time point; real-time air quality parameters include dew point, oil content, and particulate matter concentration. All these parameters must be recorded at a preset sampling period, such as every 1 second or every 5 seconds, to ensure sufficient timeliness of the input data and maintain the real-time nature of the compensation results. To ensure the model can correctly parse this input data, the input vectors need to be organized according to the same feature order as during the training phase. For example, a fixed input order can be used, namely ambient temperature, relative humidity, atmospheric pressure, absolute humidity (which can be calculated using the enthalpy-humidity diagram method in step S101), power consumption, output gas volume, pressure fluctuation amplitude, dew point, oil content, and particulate matter concentration. As long as the feature dimensions remain consistent, the model can correctly perform inference.

[0068] After inputting these real-time features into the model, the model automatically determines the degree of deviation of these real-time parameters relative to standard environmental conditions under the current environmental conditions, based on the nonlinear mapping relationship learned in step S103. For example, when the ambient temperature is high and the air humidity is high, the model can identify the deviation pattern of increased compression work due to decreased air density; when the ambient humidity increases and the dew point rises, the model can also identify the situation where changes in air quality cause deviations in gas moisture content. By identifying these deviation patterns, the model automatically converts the original operating data into equivalent values ​​corresponding to the standard environment, such as equivalent unit energy consumption, equivalent output gas volume, equivalent pressure stability, and equivalent air quality parameters.

[0069] To better understand this conversion process, an example can be used. Assume that at a certain moment, the actual ambient temperature is 35℃, the relative humidity is 70%, the atmospheric pressure is 98 kPa, the power consumption is 15 kW, and the output air volume is 4.5 m³. 3 / min. Due to the higher ambient temperature, the air density is lower than in a standard environment (20℃), resulting in a relatively lower output air volume. During inference, the model calculates the equivalent output air volume under standard environmental conditions for this running point based on the patterns learned during the training phase, for example, it might convert it to 5.1 m³ / min. 3 / min. Similarly, if the actual unit energy consumption is amplified due to high humidity conditions, the model can be equivalent to a lower specific power under standard conditions. For example, the actual unit energy consumption is 3.5 kWh / m 3 However, the model may convert it to 3.1 kWh / m³ under standard conditions. 3 This reflects the equipment's true capabilities, unaffected by environmental differences.

[0070] The equivalent operating parameters output by the model are typically a set of values, including equivalent power consumption, equivalent output air volume, equivalent pressure quality, and equivalent air quality parameters. These parameters will serve as the direct basis for calculating energy efficiency deviation in subsequent steps. Therefore, in practice, the model output data should be stored locally or transmitted in real time to the energy efficiency calculation module. This approach ensures that operating data under different environmental conditions are standardized to the same standard, thereby guaranteeing the comparability and accuracy of subsequent energy efficiency assessments.

[0071] After completing the above process, the real-time operating data is successfully converted into equivalent operating parameters for the corresponding standard environmental conditions, providing complete input for the calculation of energy efficiency deviation based on the theoretical energy efficiency reference surface in step S105.

[0072] Furthermore, the nonlinear machine learning model, which inputs real-time collected environmental parameters, actual operating performance parameters, and air quality parameters into a trained system, converts the operating performance parameters and air quality parameters under non-standard environments into equivalent operating performance parameters under corresponding standard environmental conditions, including: The real-time collected temperature, relative humidity and atmospheric pressure are converted into thermal properties to obtain specific enthalpy and absolute humidity. These are then combined with real-time operating performance parameters and air quality parameters according to the time of collection to form real-time input data consistent with the training input format. The real-time input data is processed using the same normalization method as the training phase, so that the numerical range of the real-time input data is consistent with the training input, and a normalized input that can be directly recognized by the model is obtained. The normalized input is fed into the trained nonlinear machine learning model to obtain the model output representing the offset of operating performance parameters and air quality parameters under the current environmental conditions. The real-time operating performance parameters and air quality parameters are compensated based on the offset output by the model to obtain preliminary equivalent operating performance parameters under standard environmental conditions. Physical consistency verification is performed on the preliminary equivalent operating performance parameters. By comparing the physical relationship between the inlet specific enthalpy, the outlet specific enthalpy and the compression ratio, the compensation amount that does not conform to the physical boundary of the compressed air system is adjusted back to obtain the final equivalent operating performance parameters.

[0073] Before inputting real-time collected environmental parameters, operational performance parameters, and air quality parameters into a trained nonlinear machine learning model, these parameters need to undergo thermophysical property conversion and format standardization to ensure that the input data format is completely consistent with the data used by the model during the training phase. Thermophysical property conversion refers to calculating the specific enthalpy and absolute humidity of air based on temperature, relative humidity, and atmospheric pressure. Specific enthalpy is the heat content per unit mass of air, which can be obtained using an enthalpy-humidity chart or its corresponding mathematical formula; absolute humidity describes the actual mass of water vapor per unit volume of air. For example, when the air temperature is 30°C, the relative humidity is 60%, and the atmospheric pressure is 101.3 kPa, the partial pressure of water vapor can be calculated first using commonly used air thermodynamic property relationships, then the saturated water vapor density can be calculated, and finally the absolute humidity and specific enthalpy can be calculated. Those skilled in the art can obtain the required thermophysical property values ​​simply by consulting an air enthalpy-humidity chart, without the need for additional complex tools. After the conversion is completed, these thermophysical parameters are combined with real-time operating performance parameters, such as power consumption, output air volume and exhaust pressure, as well as air quality parameters such as dew point, oil content and particulate matter concentration, according to the collection timestamp, to form real-time input data containing all relevant physical quantities.

[0074] To ensure that real-time input data can be correctly identified by the trained nonlinear machine learning model, it must undergo normalization processing that is completely consistent with the training phase. Normalization refers to adjusting parameters with different dimensions and numerical ranges to the same or comparable scale to prevent a single parameter from dominating the model's calculation results due to excessively large values. Normalization can be achieved through common linear scaling methods, such as subtracting the mean of the training data and then dividing by the standard deviation of the training data. If the power consumption is scaled by dividing by 100 per kilowatt-hour during the training phase, the power consumption in the real-time input data must also undergo the same scaling processing to maintain the uniformity of the data space. The normalized data constitutes a set of inputs with a consistent numerical range and stable structure, which can be directly used as the model's input vector.

[0075] After the normalized input is fed into the trained nonlinear machine learning model, the model generates offset predictions for each dimension of the input vector based on the patterns of environmental factors and air quality changes affecting operational performance learned during the training phase. The offset represents the theoretical deviation trend of the corresponding operational performance parameters or air quality parameters from standard environmental conditions under the current environmental conditions. For example, when the ambient temperature and air humidity increase, the model may predict a negative offset for the output air volume, indicating that the actual output air volume under the same load will be lower than under standard environmental conditions. The model's calculation of the offset is based on the multidimensional nonlinear relationship established during training; therefore, this step only requires inputting the normalized data into the model to obtain the offset, without the need for manually setting additional rules.

[0076] After obtaining the offset, it needs to be used to compensate for real-time operating performance parameters and air quality parameters to obtain preliminary equivalent values ​​under standard environmental conditions. The compensation method involves adding or subtracting the offset item by item according to the correspondence between parameters. For example, if the model predicts an offset of +0.05 MPa for exhaust pressure, indicating that current environmental conditions increase the exhaust pressure relative to standard environmental conditions, then the compensated exhaust pressure should be the measured exhaust pressure minus 0.05 MPa. Those skilled in the art can achieve compensation using the simplest linear superposition method. By performing compensation item by item for all operating performance parameters and air quality parameters, the so-called preliminary equivalent operating performance parameters can be obtained.

[0077] To ensure that the compensated parameters do not violate the physical laws of the compressed air system, a physical consistency check is required on the preliminary equivalent operating performance parameters. This check is based on the fundamental energy conservation principle of the compressor, which states that there is a fixed physical relationship between the inlet specific enthalpy, the outlet specific enthalpy, and the compression ratio. For example, the higher the compression ratio, the higher the outlet specific enthalpy should be. If the compensated result shows an unreasonable phenomenon where the outlet specific enthalpy is lower than the inlet specific enthalpy, it indicates that the compensation amount is too large or the direction is incorrect, requiring a correction. Correction can be achieved by reducing the offset percentage, for example, by multiplying the offset by 0.8 or 0.5, so that the compensated parameters once again meet the physical boundary requirements. The check can be performed on different parameters; for example, the output air volume should not increase inversely with increasing compression ratio. The parameters after physical consistency check are the final equivalent operating performance parameters, which can be used in subsequent energy efficiency assessment steps.

[0078] Step S105: Map the equivalent operating parameters to the theoretical energy efficiency reference surface, and obtain the energy efficiency assessment results reflecting the deviation of system energy efficiency and the degree of performance degradation by calculating the distance vector between the parameter points and the reference surface.

[0079] In step S105, the equivalent operating parameters obtained in step S104 need to be mapped and compared with the theoretical energy efficiency reference surface constructed in step S102. By calculating the distance vector between the equivalent operating parameter points and the theoretical reference surface, an energy efficiency assessment result that can reflect the degree of deviation of system energy efficiency and the degree of performance degradation can be obtained.

[0080] In practice, equivalent operating parameters are a set of operating data that have undergone environmental compensation, representing the theoretical performance of the system under standard environmental conditions. These equivalent parameters typically include equivalent unit energy consumption, equivalent output gas volume, equivalent pressure stability, and equivalent air quality indicators. For example, after compensation in step S104, a certain operating point may become "equivalent energy consumption 3.2 kWh / m³". 3 Equivalent gas volume 5.0 m³ 3 / min, equivalent pressure fluctuation 0.04 MPa, equivalent dew point −10℃. This set of data represents the assumed operating state of the system under standard environmental conditions.

[0081] The theoretical energy efficiency benchmark is the "ideal energy efficiency boundary" constructed in step S102 under standard conditions. It is essentially a multi-dimensional surface, where each location provides the theoretically optimal energy consumption value under different pressure levels and gas flow ranges. For example, for an output pressure of 0.7 MPa and a flow rate of 5 m³ / s... 3 At the operating point of / min, the theoretical energy efficiency benchmark may give "theoretical optimal unit energy consumption of 2.95 kWh / m". 3 The reference surface includes parameters such as "maximum permissible pressure fluctuation of 0.03 MPa" and "air quality deviation limit." This reference surface can be constructed using interpolation, spline surfaces, or machine learning models to ensure that it has a clear reference value under each operating condition.

[0082] In this step, "mapping to the theoretical energy efficiency reference surface" means finding a reference reference point corresponding to the equivalent operating parameters. For example, when the equivalent operating point is "pressure 0.7 MPa, gas volume 5 m³ / s",... 3 When the value is " / min", the theoretically optimal operating point under the same pressure and gas volume conditions is found directly based on the reference surface. If the equivalent parameters happen to fall between the calculation points on the reference surface, the corresponding reference value can be estimated by linear interpolation or cubic spline interpolation. Those skilled in the art can complete this operation simply by pre-setting the reference surface table or interpolation function in the system.

[0083] The distance vector is used to represent the degree to which equivalent operating parameters deviate from the theoretical optimum. For example, when the theoretical optimum energy consumption per unit is 2.95 kWh / m³. 3 The equivalent unit energy consumption is 3.20 kWh / m³. 3The energy efficiency deviation can then be expressed as the difference between the two, which is 0.25 kWh / m³. 3 If the theoretical upper limit of pressure fluctuation is 0.03 MPa, while the actual equivalent pressure fluctuation is 0.04 MPa, then the deviation is 0.01 MPa. For air quality, if the theoretically permissible dew point is −12℃ and the equivalent operating dew point is −10℃, then the deviation is 2℃. The vector formed by these differences is the distance vector. Those skilled in the art can understand this as "obtaining a list of differences through item-by-item comparison."

[0084] To further enhance the intuitiveness of the assessment, the distance vector can be summarized as a single energy efficiency deviation. For example, weight values ​​can be set based on the degree of influence of each parameter on energy efficiency, and the deviations of each parameter can be combined according to their weights to obtain a comprehensive deviation index. For example, the weight of energy consumption deviation could be set to 0.6, the weight of pressure fluctuation deviation to 0.3, and the weight of air quality deviation to 0.1. Then the comprehensive deviation index can be obtained by simple weighted summation. The weight settings can be determined based on the company's management experience or equipment maintenance needs; this technical solution does not limit the specific weight values.

[0085] Ultimately, the distance vector or comprehensive deviation index constitutes the energy efficiency assessment result, characterizing the degree of deviation of the equipment from the theoretical energy efficiency benchmark. A small deviation indicates that the equipment is operating close to its ideal state; a larger deviation usually signifies system performance degradation, such as increased internal leakage, decreased compression efficiency, aging filter components, or deteriorating air quality. Engineers can use the degree of deviation to determine whether maintenance, parameter adjustment, repair, or optimization of operating strategies are necessary.

[0086] Through the above process, this step achieves a quantitative comparison between equivalent operating parameters and theoretical energy efficiency benchmarks. The evaluation results have physical consistency, interpretability, and comparability, making the environmental compensation energy efficiency evaluation of this invention feasible and of engineering value.

[0087] Furthermore, the process of mapping equivalent operating parameters to a theoretical energy efficiency reference surface, and obtaining energy efficiency assessment results reflecting the system's energy efficiency deviation and performance degradation by calculating the distance vector between parameter points and the reference surface, includes: Each equivalent operating parameter is organized into standardized operating points according to three dimensions: pressure, flow rate, and power consumption, so that the operating points have clear coordinates in three-dimensional space; The operating point is placed in the coordinate system of the theoretical energy efficiency reference surface. By performing local interpolation on the reference surface at the corresponding pressure and flow positions, the reference energy consumption value of the operating point under the theoretical optimal conditions is obtained, and the theoretical energy efficiency reference point is formed. The three-dimensional difference between the operating point and the theoretical energy efficiency benchmark is decomposed into components along the energy consumption direction, along the pressure direction and along the flow direction, and a difference vector representing the source of efficiency and performance deviation is constructed based on this. The difference vector is normalized according to the acceptable offset range defined in the thermodynamic constraint set, so that the offset of each component is characterized by the same dimension, and the normalized vector is used as the energy efficiency deviation vector of the system under standard environment. The magnitude of the energy efficiency deviation vector is calculated and used to measure the overall deviation of the system from the theoretical energy efficiency benchmark. The main sources of degradation are determined based on the direction of the deviation vector, thereby obtaining an energy efficiency assessment result that reflects the degree of energy efficiency deviation and performance degradation of the system.

[0088] When evaluating the energy efficiency of equivalent operating parameters, these parameters must first be organized into standardized operating points that can be located on the theoretical energy efficiency benchmark. Equivalent operating parameters refer to the power consumption, output gas volume, and exhaust pressure obtained through environmental compensation in the preceding steps, corresponding to standard environmental conditions. To ensure spatial comparability of these parameters during calculation, they need to be constructed as operating points in a three-dimensional coordinate system, based on pressure, flow rate, and power consumption. For example, if the equivalent exhaust pressure at a certain moment is 0.75 MPa and the equivalent output gas volume is 4.2 m³ / s... 3 If the flow rate is / min and the equivalent power consumption is 18.5 kW, then this point can be expressed as P=(0.75, 4.2, 18.5), where the three coordinate axes correspond to pressure, flow rate and power consumption, respectively. This three-dimensional representation can be used for benchmark comparison, interpolation and vector analysis in subsequent steps.

[0089] After representing the operating point as three-dimensional coordinates, it needs to be mapped to the coordinate system of the theoretical energy efficiency reference surface. The theoretical energy efficiency reference surface is a three-dimensional surface constructed under standard environmental conditions based on the principle of isentropic compression and historical best energy efficiency samples. It provides a corresponding minimum theoretical energy consumption value for each set of pressure and flow rate, and therefore is completely consistent with the three-dimensional coordinate system. Those skilled in the art can view this three-dimensional surface as a mathematical function of energy consumption varying with pressure and flow rate, but its value is obtained through interpolation of actual data, not derived from purely theoretical formulas. When the pressure and flow rate of the operating point fall between discrete nodes of the theoretical energy efficiency reference surface, local interpolation calculations are required. For example, when the pressure at a certain operating point is 0.75 MPa and the flow rate is 4.2 m³ / s... 3When the reference surface provides energy consumption values ​​for four nodes (0.7, 4.0), (0.7, 4.5), (0.8, 4.0), and (0.8, 4.5), the theoretical minimum energy consumption value corresponding to that point can be obtained using bilinear interpolation. The interpolated theoretical energy efficiency reference point can be expressed as B = (0.75, 4.2, E_min), where E_min is the interpolated theoretical minimum energy consumption value.

[0090] To determine the degree of deviation between the actual operating state and the theoretical optimal state, it is necessary to calculate the three-dimensional difference between the operating point and the theoretical energy efficiency benchmark, and then decompose this difference directionally. The three-dimensional difference is calculated by subtracting the corresponding coordinates of the benchmark from the three coordinates of the operating point. For example, if the operating point is P=(0.75,4.2,18.5) and the benchmark is B=(0.75,4.2,16.8), then the difference vector is Δ=(0,0,1.7), indicating that the operating state is consistent with the benchmark state in the pressure and flow dimensions, but the power consumption dimension is 1.7 kW higher. In general, the difference vector may not be zero in any of the three dimensions; therefore, it needs to be decomposed into three components along the energy consumption direction, the pressure direction, and the flow direction to identify the source of the energy efficiency deviation. For example, if the difference is Δ=(0.03, -0.1, 2.5), it indicates a slight increase in pressure, a slight decrease in flow, and a significantly higher power consumption; the effects in each direction need to be reflected in the evaluation.

[0091] To ensure comparability of offsets across different dimensions, the difference vector needs to be normalized based on the aforementioned set of thermodynamic constraints. The thermodynamic constraint set defines the allowable offset range for each dimension; for example, pressure offset should not exceed ±0.1 MPa, and flow rate offset should not exceed ±0.5 m. 3 / min, energy consumption deviation should not exceed ±3 kW. Normalization can be achieved using simple scaling, such as dividing pressure deviation by 0.1, flow deviation by 0.5, and energy consumption deviation by 3, so that each dimension falls within the range of -1 to 1 after normalization. The normalized deviation vector directly reflects the proportion of each deviation relative to the allowable range in that dimension; the closer the value is to 1 or -1, the further that dimension is from the physically reasonable range.

[0092] After obtaining the normalized deviation vector, its magnitude needs to be calculated to assess the overall energy efficiency deviation. The magnitude represents the overall length of the vector and is an important indicator of the comprehensive deviation of the operating state from the theoretical energy efficiency benchmark. If the normalized deviation vector is V=(v1,v2,v3), its magnitude can be obtained by squaring and taking the square root. For example, if V=(0.3, -0.5, 0.9), then the magnitude is sqrt(0.3). 2+ (-0.5) 2 + 0.9 2 A magnitude of approximately 1.088 indicates a relatively high degree of overall deviation. Those skilled in the art can consider the magnitude as a quantifiable risk indicator of the current energy efficiency status of the compressed air system. Furthermore, by analyzing the direction of the deviation vector, the main sources of performance degradation can be identified. For example, deviations primarily concentrated in the energy consumption dimension indicate a decline in system efficiency, while deviations primarily concentrated in the flow dimension may suggest pipe blockage or leakage. Ultimately, through magnitude and directionality analysis, both the system's energy efficiency deviation and performance degradation level can be obtained simultaneously, providing a basis for subsequent maintenance and optimization.

[0093] Furthermore, the three-dimensional difference between the operating point and the theoretical energy efficiency benchmark is decomposed into components along the energy consumption direction, the pressure direction, and the flow direction, respectively. A difference vector representing the source of efficiency and performance deviation is then constructed based on this, including: The differences between the operating point and the theoretical energy efficiency benchmark point are calculated in three dimensions: energy consumption, pressure, and flow rate. The differences are then split into energy consumption difference components, pressure difference components, and flow rate difference components along the three-dimensional coordinate axis to form an initial component set. The initial component set is associated with the compression ratio sensitivity parameter corresponding to the pressure level of the operating point. By applying an adjustment coefficient that matches the compression ratio sensitivity to each component, the component variation amplitude can accurately reflect the energy efficiency sensitivity of the actual equipment in different pressure ranges, thus obtaining the directional component set after sensitivity correction. Based on the numerical variation trend of each component in the directional component set, the variation direction of the energy consumption component, pressure component and flow component near the operating point is calculated, and the variation direction is used to construct the directional weight that can characterize the offset attribute. The directional component and the directional weight are combined into the offset attribute enhancement component. The offset attribute enhancement components are uniformly converted according to the offset scale required for energy efficiency evaluation, so that the three types of components, energy consumption, pressure and flow rate, express the degree of performance offset on the same scale. The converted three types of components are combined into a difference vector in coordinate order to characterize the source of deviation of the current operating point in the dimensions of energy consumption, pressure and flow rate.

[0094] When analyzing the deviation between the operating point and the theoretical energy efficiency benchmark, it is necessary to decompose the differences between the two in three dimensions: energy consumption, pressure, and flow rate, in order to identify the source of the deviation from the perspective of different operating attributes. In implementation, it is first necessary to clarify that both the operating point and the theoretical energy efficiency benchmark are expressed in three-dimensional coordinates, where the three coordinate values ​​correspond to the energy consumption, pressure, and flow rate values ​​under equivalent operating parameters, respectively. The so-called three-dimensional difference refers to subtracting the energy consumption, pressure, and flow rate of the theoretical energy efficiency benchmark at the same location from the actual operating parameters at the same operating point. For example, when the energy consumption at a certain equivalent operating point is 6.2 kWh / m³... 3 The theoretical energy efficiency benchmark is 5.5 kWh / m³. 3 If the pressure at the operating point is 0.02 MPa higher than the reference pressure, the pressure difference is 0.02; if the flow rate is 0.15 m³ lower than the reference value, the difference in the energy consumption direction is 0.7. 3 If the flow rate is / min, then the difference in flow direction is -0.15. In this way, the initial energy consumption difference component, pressure difference component, and flow rate difference component can be obtained. These three components are combined to form an initial component set, which is used for subsequent offset source determination.

[0095] To ensure these differential components accurately reflect the sensitivity of compressed air equipment across different pressure ranges, a compression ratio sensitivity parameter needs to be introduced, and the initial components need to be corrected for sensitivity. Compression ratio sensitivity is a coefficient used to describe the degree to which the energy efficiency of equipment changes under different pressure conditions. Its value can be obtained by calculating the proportion of power consumption change caused by a unit pressure change based on the measured characteristic curve of the equipment. For example, if a piece of equipment experiences an increase of approximately 0.12 kWh / m³ for every 0.01 MPa increase in pressure near 0.7 MPa... 3 If the energy consumption is such that the compression ratio sensitivity can be defined as 12, then when associating the initial component set with this sensitivity parameter, the energy consumption difference component needs to be multiplied by the sensitivity parameter, and the pressure difference component needs to be scaled inversely according to the sensitivity. This makes the three-dimensional difference components more consistent with the true physical laws after correction. The more sensitive the pressure range, the more significant the changes will be reflected in the difference. Without sensitivity correction, the operating point offsets in different pressure ranges will be treated equally in the evaluation, thus failing to accurately reflect the source of system energy efficiency deviation.

[0096] After obtaining the directional components after sensitivity correction, it is necessary to further determine the direction of change of each component. The direction of change refers to whether the deviation of the component relative to the theoretical value is positively large, negatively small, or close to zero. For example, a positive energy consumption component indicates high operating energy consumption, while a negative value indicates low operating energy consumption; a positive pressure component indicates operating pressure higher than the theoretical value, which may mean reduced system pressure regulation efficiency; a negative flow component indicates flow rate lower than the expected level, which may mean equipment performance degradation or increased pipeline resistance. Utilizing this directional information requires constructing directional weights. These weights amplify the deviation directions that are more important for energy efficiency evaluation and reduce the influence of deviation directions with weaker physical meaning. For example, when the energy consumption deviation direction is consistent with the pressure deviation direction, it indicates a coupled deviation, and the weight of this combined direction should be increased; while when the flow rate deviation direction is opposite to the energy consumption deviation direction, its influence in the overall deviation calculation needs to be reduced. The directional weights can be constructed using an amplitude amplification method. For example, if both the energy consumption and pressure components are positive, the amplitude of the energy consumption direction can be multiplied by 1.3 to reflect its importance in deviation evaluation.

[0097] After obtaining the offset attribute enhancement components, the three components need to be scaled to allow for comparison on the same scale. Since energy consumption, pressure, and flow rate have different dimensions, they cannot be measured within a unified offset space without a standardized scale. For example, if the energy consumption offset is 0.8 kWh / m³... 3 However, the pressure offset is only 0.02 MPa, so their absolute values ​​cannot be directly compared. In implementation, a linear interval mapping can be used to map the three components to a unified interval of -1 to 1 based on their respective allowable offset intervals. For example, the allowable interval for energy consumption offset is ±1.0, for pressure offset it is ±0.05, and for flow rate offset it is ±0.5. Dividing each of the three components by its corresponding maximum value yields a unified scale representation. In this unified scale, all three components are dimensionless values, allowing those skilled in the art to perform vector calculations directly.

[0098] After scaling, the offset components in energy consumption, pressure, and flow direction are combined in the order of energy consumption direction, pressure direction, and flow direction to form the final difference vector, which comprehensively characterizes the source of deviation at the current operating point of the system. The difference vector not only reflects the magnitude of the deviation but also preserves its direction, enabling subsequent energy efficiency deviation calculations to simultaneously capture both the degree and attribute of the deviation. Those skilled in the art can use this to determine from a three-dimensional offset perspective whether the equipment exhibits performance degradation issues such as decreased energy efficiency, abnormal pressure drop, or insufficient gas supply.

[0099] To facilitate understanding of the above-described difference direction decomposition and offset vector construction process by those skilled in the art, a specific example based on the actual operation record of a compressed air system is provided below. Assume that the coordinates of the equivalent operating point obtained after environmental compensation at a certain moment under standard environmental conditions are: energy consumption 6.4 kWh / m². 3 Pressure 0.72 MPa, flow rate 3.85 m³ / s 3 / min; meanwhile, the coordinates of the theoretical energy efficiency benchmark point at the same pressure and flow rate location are: energy consumption 5.7 kWh / m 3 Pressure 0.70 MPa, flow rate 4.00 m³ / s 3 / min. According to the aforementioned definition, the three-dimensional difference between the operating point and the reference point can be directly calculated as the energy consumption direction difference of 0.7, the pressure direction difference of 0.02, and the flow direction difference of -0.15. Here, a positive value indicates that the actual operating parameters are higher than the theoretical values, and a negative value indicates that the actual values ​​are lower than the theoretical values.

[0100] Assuming the compression ratio sensitivity coefficient of the equipment within this pressure range is 10, this means that a pressure change of 0.01 MPa will result in a change in energy consumption per unit flow rate of approximately 0.10 kWh / m³. 3 In this situation, sensitivity correction is needed to the initial directional difference to make the offset more closely match the actual equipment characteristics. Sensitivity correction can be achieved through simple proportional adjustments. For example, multiplying the energy consumption directional difference by a sensitivity coefficient yields a corrected energy consumption component of approximately 7.0; scaling the pressure directional difference inversely according to the sensitivity yields a pressure correction component of approximately 0.002; the sensitivity of the flow rate directional difference has a relatively small impact in this range and can be retained at its original value of -0.15. The three corrected components obtained in this way—7.0, 0.002, and -0.15—more accurately reflect the deviation characteristics of the operating point: energy consumption offset is dominant, pressure offset has a minimal impact, and flow rate offset is slightly negative.

[0101] After obtaining the sensitivity-corrected components, it is necessary to further determine the offset direction and offset attributes of the three components. The energy consumption component is positive and has the largest amplitude, indicating that the equipment has significantly excessive energy consumption at that moment; the pressure component is positive but has a very small amplitude, indicating that the operating pressure is slightly higher than the theoretical value, but will not significantly affect energy efficiency; the flow rate component is negative, indicating that the actual gas production capacity is slightly lower than the theoretical baseline. Based on the combination relationship of the offset directions, directional weights can be constructed. For example, an amplification weight of 1.3 can be applied to the energy consumption component, because the simultaneous occurrence of high energy consumption and low flow rate usually indicates an increased risk of system efficiency decline; the pressure component, due to its limited offset amplitude and lack of significant coupling effect, can be given a weakened weight of 0.8; the flow rate component can be given a moderate weight of 1.1 to reflect its impact on performance degradation. Multiplying these weights by the corrected components yields the enhanced energy consumption offset value of approximately 9.1, the pressure offset value of 0.0016, and the flow rate offset value of -0.165, respectively.

[0102] The aforementioned enhancement components still exhibit dimensional differences, thus requiring unified mapping to the same interval. Assuming the allowable range for energy consumption offset is ±12, pressure offset is ±0.02, and flow rate offset is ±0.5, the enhanced components can be divided by their respective maximum allowable ranges to obtain dimensionless offset components at a unified scale, approximately: 0.76 in the energy consumption direction, 0.08 in the pressure direction, and -0.33 in the flow rate direction. Combining the components in the three directions in a prescribed order yields the final difference vector (0.76, 0.08, -0.33). This vector not only characterizes the degree to which the equipment deviates from the theoretical reference plane at this moment but also indicates that the deviation primarily originates from high energy consumption and low flow rate, with a relatively weak contribution from pressure offset.

[0103] A second embodiment of this application provides an electronic device, the electronic device comprising: processor; The memory is used to store a program, which, when read and executed by the processor, executes the intelligent energy efficiency assessment method for compressed air systems based on environmental compensation provided in the first embodiment of this application.

[0104] The third embodiment of this application provides a computer-readable storage medium storing a computer program thereon. When the program is executed by a processor, it executes the intelligent energy efficiency assessment method for compressed air systems based on environmental compensation provided in the first embodiment of this application.

[0105] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A method for intelligent energy efficiency assessment of compressed air systems based on environmental compensation, characterized in that, include: The system collects the operating performance parameters and air quality parameters of the compressed air system, and simultaneously collects environmental parameters including temperature, humidity and atmospheric pressure. It uses enthalpy-humidity diagrams to establish the thermal property mapping relationship between environmental parameters and air quality parameters, and aligns them in time order to form a quality coupling feature dataset. Based on the principle of isentropic compression and combined with historical best energy efficiency samples, a theoretical energy efficiency benchmark surface corresponding to standard environmental conditions is constructed, and a set of thermodynamic constraints is formed to limit the ideal boundary of operating performance based on the theoretical energy efficiency benchmark surface corresponding to standard environmental conditions. Based on the quality coupling feature dataset and the thermodynamic constraint set, a nonlinear machine learning model is trained to learn the nonlinear mapping relationship between the performance offset caused by changes in environmental parameters and the air quality offset. The real-time collected environmental parameters, actual operating performance parameters, and air quality parameters are input into the trained nonlinear machine learning model, which converts the operating performance parameters and air quality parameters under non-standard environments into equivalent operating performance parameters under corresponding standard environmental conditions. By mapping equivalent operating parameters to a theoretical energy efficiency reference surface and calculating the distance vector between parameter points and the reference surface, an energy efficiency assessment result reflecting the deviation of system energy efficiency and the degree of performance degradation can be obtained.

2. The intelligent energy efficiency assessment method for compressed air systems based on environmental compensation according to claim 1, characterized in that, The system collects operating performance parameters and air quality parameters of the compressed air system, and simultaneously collects environmental parameters including temperature, humidity, and atmospheric pressure. It then uses an enthalpy-humidity diagram to establish a thermal property mapping relationship between environmental parameters and air quality parameters, aligning them chronologically to form a quality-coupled feature dataset, including: At each operating node, multiple sensors for power consumption, output gas volume, pressure, dew point, oil content, and particulate matter are activated, and the internal clocks of the sensors are synchronized with a unified time base to obtain raw data with consistent timestamps. The synchronized ambient temperature, relative humidity, and atmospheric pressure are sequentially converted into absolute humidity content, specific enthalpy, and saturated humidity. The thermal property parameters are then combined with the corresponding air quality parameters to form a first characteristic sequence containing the coupling relationship between thermal property quality. The particulate matter concentration and oil content in the first feature sequence are corrected for deviation. By comparing the average value of the stable segment of adjacent sampling segments, the data difference caused by sensor zero drift is eliminated, and the corrected second feature sequence is obtained. Based on the actual pipeline layout, the difference between ambient temperature and humidity and compressed air outlet temperature is used to calculate the local water vapor condensation amount, and the dew point and absolute moisture content are compensated according to the local water vapor condensation amount to form a third characteristic sequence that reflects the true thermal and humid state. The third feature sequence is interpolated and resampled according to the selected uniform sampling period to align all parameters on the same time axis and organize them in chronological order into a quality coupled feature dataset containing environmental thermal properties, air quality and operational performance.

3. The intelligent energy efficiency assessment method for compressed air systems based on environmental compensation according to claim 1, characterized in that, Based on the isentropic compression principle and combined with historical best energy efficiency samples, a theoretical energy efficiency benchmark surface corresponding to standard environmental conditions is constructed. Based on this benchmark surface, a set of thermodynamic constraints is formed to define the ideal boundary of operational performance, including: Under standard environmental conditions, the theoretical minimum energy consumption of multiple typical operating points is calculated by using the inlet pressure, outlet pressure and thermal properties of the compressed air system as inputs, through isentropic compression relationship. The theoretical minimum energy consumption of each operating point is then combined with the corresponding flow range to form an initial theoretical energy consumption dataset. The initial theoretical energy consumption dataset is compared with the optimal energy efficiency samples selected from historical operation records, and the theoretical minimum energy consumption of each operation point is corrected according to the deviation range that the actual equipment can achieve under optimal operating conditions, to obtain the second energy consumption dataset after equipment characteristic calibration. The continuity of energy consumption values ​​of adjacent operating points in the second energy consumption dataset is checked. By calculating the gas volume gradient and energy consumption gradient between operating points, discrete energy consumption points caused by short-term abnormal operating conditions are removed, and intermediate operating condition values ​​with smooth transition are re-inserted to obtain a continuous and fitable third energy consumption dataset. The third energy consumption dataset is interpolated in two dimensions according to pressure level and flow level to obtain a continuous theoretical energy efficiency reference surface covering the entire operating range. The minimum energy consumption curve corresponding to the same pressure level in the continuous theoretical energy efficiency reference surface is extracted again as the performance limit curve to form a complete ideal performance boundary. Based on the ideal performance boundary, the acceptable upper limit of energy consumption, the upper limit of pressure fluctuation, and the upper limit of air humidity are calculated respectively. The upper limit values ​​are combined with the corresponding optimal values ​​on the continuous theoretical energy efficiency reference surface to form a set of thermodynamic constraint intervals, so that the predicted operating performance shall not be lower than the theoretical limit or higher than the physically achievable range, thus forming a set of thermodynamic constraints that limit the ideal boundary of operating performance.

4. The intelligent energy efficiency assessment method for compressed air systems based on environmental compensation according to claim 1, characterized in that, The nonlinear machine learning model includes: The environmental disturbance coding unit is used to receive environmental thermophysical parameters such as temperature, relative humidity, atmospheric pressure and their specific enthalpy value and absolute humidity obtained by converting them through the enthalpy-humidity diagram, and to obtain an environmental offset vector to characterize the magnitude of environmental disturbance by performing a nonlinear scaling transformation on the rate of change of temperature and humidity through a multi-segment piecewise hyperbolic function. The quality offset deconstruction unit is used to receive air quality parameters such as dew point, oil content, and particulate matter concentration, as well as operating performance parameters such as power consumption, output air volume, and pressure. After aligning the received parameters by time, it inputs them into a residual deconstructor based on local trend analysis. By calculating the difference between the local linear trend and the global average trend of adjacent sampling points, a quality offset vector representing the contribution of air quality offset is obtained. The coupled response inference unit takes the environmental offset vector and the air quality offset vector as joint inputs and performs three types of operations in sequence through a three-stage nonlinear combination mechanism: environmental priority amplification, cross-sensitivity suppression, and energy efficiency correlation enhancement. Among them, environmental priority amplification enhances the sensitivity under high temperature and high humidity conditions through an exponential amplification factor; cross-sensitivity suppression reduces the collinearity effect by performing inverse scaling on the product term of the air quality offset vector and the environmental offset vector; and energy efficiency correlation enhancement forms the coupled offset response vector by introducing the instantaneous change rate of the operating performance parameters into the offset weight. The standard state equivalent restoration unit is used to take the coupled offset response vector and the operating performance parameters as inputs. Through continuous interval piecewise regression, an environmental offset compensation function is established for each operating performance parameter. The environmental offset compensation function adopts different fitting slopes in different flow, pressure and dew point intervals, and the compensated parameters are output as equivalent operating performance parameters under the corresponding standard environmental conditions. The offset consistency correction unit is used to perform physical consistency constraint correction on the equivalent operating performance parameters. By comparing the physical boundary relationship between the inlet specific enthalpy, the outlet specific enthalpy and the compression ratio, it identifies points that violate energy conservation or do not meet the physical constraints of the compressor, and corrects the abnormal points back to the allowable offset range to obtain the final standard state equivalent operating performance parameters.

5. The intelligent energy efficiency assessment method for compressed air systems based on environmental compensation according to claim 1, characterized in that, The nonlinear machine learning model, trained by inputting real-time collected environmental parameters, actual operational performance parameters, and air quality parameters, converts operational performance parameters and air quality parameters under non-standard environments into equivalent operational performance parameters under corresponding standard environmental conditions, including: The real-time collected temperature, relative humidity and atmospheric pressure are converted into thermal properties to obtain specific enthalpy and absolute humidity. These are then combined with real-time operating performance parameters and air quality parameters according to the time of collection to form real-time input data consistent with the training input format. The real-time input data is processed using the same normalization method as the training phase, so that the numerical range of the real-time input data is consistent with the training input, and a normalized input that can be directly recognized by the model is obtained. The normalized input is fed into the trained nonlinear machine learning model to obtain the model output representing the offset of operating performance parameters and air quality parameters under the current environmental conditions. The real-time operating performance parameters and air quality parameters are compensated based on the offset output by the model to obtain preliminary equivalent operating performance parameters under standard environmental conditions. Physical consistency verification is performed on the preliminary equivalent operating performance parameters. By comparing the physical relationship between the inlet specific enthalpy, the outlet specific enthalpy and the compression ratio, the compensation amount that does not conform to the physical boundary of the compressed air system is adjusted back to obtain the final equivalent operating performance parameters.

6. The intelligent energy efficiency assessment method for compressed air systems based on environmental compensation according to claim 1, characterized in that, The process of mapping equivalent operating parameters to a theoretical energy efficiency reference surface, and obtaining energy efficiency assessment results reflecting the system's energy efficiency deviation and performance degradation by calculating the distance vector between parameter points and the reference surface, includes: Each equivalent operating parameter is organized into standardized operating points according to three dimensions: pressure, flow rate, and power consumption, so that the operating points have clear coordinates in three-dimensional space; The operating point is placed in the coordinate system of the theoretical energy efficiency reference surface. By performing local interpolation on the reference surface at the corresponding pressure and flow positions, the reference energy consumption value of the operating point under the theoretical optimal conditions is obtained, and the theoretical energy efficiency reference point is formed. The three-dimensional difference between the operating point and the theoretical energy efficiency benchmark is decomposed into components along the energy consumption direction, along the pressure direction and along the flow direction, and a difference vector representing the source of efficiency and performance deviation is constructed based on this. The difference vector is normalized according to the acceptable offset range defined in the thermodynamic constraint set, so that the offset of each component is characterized by the same dimension, and the normalized vector is used as the energy efficiency deviation vector of the system under standard environment. The magnitude of the energy efficiency deviation vector is calculated and used to measure the overall deviation of the system from the theoretical energy efficiency benchmark. The main sources of degradation are determined based on the direction of the deviation vector, thereby obtaining an energy efficiency assessment result that reflects the degree of energy efficiency deviation and performance degradation of the system.

7. The intelligent energy efficiency assessment method for compressed air systems based on environmental compensation according to claim 3, characterized in that, The process involves performing a continuity check on the energy consumption values ​​of adjacent operating points in the second energy consumption dataset. By calculating the gas volume gradient and energy consumption gradient between operating points, discrete energy consumption points caused by short-term operating condition anomalies are removed, and smooth transition intermediate operating condition values ​​are re-inserted to obtain a continuously fit third energy consumption dataset, including: The running points in the second energy consumption dataset are sorted in ascending order of gas volume, and the gas volume gradient representing the rate of change of gas volume is calculated for adjacent running points after sorting, so as to obtain the first gradient sequence used to judge the stability of the change of running points. The first gradient sequence is compared with the energy consumption changes in the second energy consumption dataset. The energy consumption gradient representing the rate of energy consumption change is calculated for adjacent operating points. The gas volume gradient and the energy consumption gradient are combined to obtain a joint gradient judgment sequence for identifying short-term operating condition anomalies. Based on the joint gradient determination sequence, the operating points that simultaneously satisfy the sudden change in gas volume gradient and the deviation of energy consumption gradient from the continuous trend are marked as discrete energy consumption points. After removing discrete energy consumption points, a local transition interval is constructed based on the relationship between gas volume and energy consumption changes of the operating points before and after removal, and the parameters of the transition interval that need to be supplemented are obtained. Based on the transition interval parameters, multiple intermediate operating points with increasing gas volume and smooth energy consumption change characteristics are generated between the eliminated operating points, and the intermediate operating points are inserted into the corresponding positions so that the gas volume changes monotonically with the sequence and the energy consumption change remains continuous. After inserting intermediate operating points, the operating points are reorganized into a continuous energy consumption sequence according to gas volume order to form a third energy consumption dataset that can reflect the real changing trend of operating conditions and can be used for subsequent interpolation fitting.

8. The intelligent energy efficiency assessment method for compressed air systems based on environmental compensation according to claim 6, characterized in that, The method involves decomposing the three-dimensional difference between the operating point and the theoretical energy efficiency benchmark into components along the energy consumption direction, the pressure direction, and the flow direction, respectively. Based on this, a difference vector representing the source of efficiency and performance deviations is constructed, including: The differences between the operating point and the theoretical energy efficiency benchmark point are calculated in three dimensions: energy consumption, pressure, and flow rate. The differences are then split into energy consumption difference components, pressure difference components, and flow rate difference components along the three-dimensional coordinate axis to form an initial component set. The initial component set is associated with the compression ratio sensitivity parameter corresponding to the pressure level of the operating point. By applying an adjustment coefficient that matches the compression ratio sensitivity to each component, the component variation amplitude can accurately reflect the energy efficiency sensitivity of the actual equipment in different pressure ranges, thus obtaining the directional component set after sensitivity correction. Based on the numerical variation trend of each component in the directional component set, the variation direction of the energy consumption component, pressure component and flow component near the operating point is calculated, and the variation direction is used to construct the directional weight that can characterize the offset attribute. The directional component and the directional weight are combined into the offset attribute enhancement component. The offset attribute enhancement components are uniformly converted according to the offset scale required for energy efficiency evaluation, so that the three types of components, energy consumption, pressure and flow rate, express the degree of performance offset on the same scale. The converted three types of components are combined into a difference vector in coordinate order to characterize the source of deviation of the current operating point in the dimensions of energy consumption, pressure and flow rate.

Images