Surge margin prediction system and method

Abstract

The present application discloses a surge boundary prediction system and method, belonging to the field of adaptive prediction technology, the method comprising: obtaining a multi-dimensional surge boundary database, collecting real-time surge influence parameters during compressor operation to construct a multi-dimensional real-time working condition vector, performing multi-level similarity calculation with experimental working condition vectors in the database, matching real-time surge boundaries, and dynamically updating according to parameter variation and set threshold; after matching to the boundary, setting an abnormality evaluation period, calculating the Hotelling statistic to evaluate parameter abnormalities, if there is a joint abnormality, obtaining the parameter contribution degree, marking the abnormal source, outputting the level, dynamically compensating the abnormal parameters to obtain the corrected real-time surge boundary; finally, according to the corrected boundary, obtaining the initial surge margin according to the real-time monitoring parameters, adaptively adjusting according to the change trend, realizing surge early warning, aiming to solve the problems of inaccurate prediction of the compressor surge boundary of the aero-engine and untimely warning in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of adaptive prediction, and more particularly to a surge margin prediction system and method. BACKGROUND

[0002] During the operation of an aero gas turbine engine compressor, compressor surge margin prediction is a key technology to ensure its safe and stable operation. At present, the industry mainly uses steady-state characteristic map method, steady CFD simulation, empirical formula model, single parameter threshold monitoring and rule-based control strategy. Among them, the steady-state characteristic map method relies on test data to draw the surge boundary line, but it is only suitable for steady-state conditions and cannot cover the dynamic process within the flight envelope; steady CFD simulation can simulate the critical point of flow field separation, but it cannot capture non-steady-state phenomena, resulting in prediction deviation; the empirical formula model is based on simplified assumptions and is not suitable for complex conditions; single parameter threshold monitoring has a high false alarm rate due to the neglect of multi-physical field coupling effects; and the rule-based control strategy lacks adaptability and is prone to failure in extreme conditions.

[0003] The existing technology generally has insufficient dynamic response capability, weak multi-parameter collaborative analysis capability, poor model generalization, difficulty in balancing real-time performance and accuracy, and lack of interpretability. These problems lead to a significant decrease in the prediction accuracy and reliability of existing prediction methods in fast transient processes, extreme environments or new working conditions, especially the inability to achieve real-time dynamic correction of the surge margin, and the inability of the surge margin adjustment strategy to adapt to parameter change trends and multi-physical field coupling effects, thereby leading to a lag in surge warning or misjudgment, which cannot meet the demand for safe operation of modern aero engines under all conditions. Therefore, in order to overcome these limitations, the present application proposes a surge margin prediction system and method. SUMMARY

[0004] In view of the deficiencies of the prior art, the purpose of the present application is to provide a surge margin prediction system and method, which solves the problems of deviation caused by uncorrected parameter abnormalities in the prediction of the surge margin of an aero compressor, and inaccurate and timely adjustment of the surge margin, and improves the prediction accuracy and timeliness of the surge margin.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] A surge margin prediction system, comprising:

[0007] A multi-dimensional surge margin database is obtained, real-time surge influencing parameters of the compressor are collected to construct a multi-dimensional real-time working condition vector, multi-level similarity calculation is performed with experimental working condition vectors in the multi-dimensional surge margin database, and real-time surge margin is matched; and an update time threshold and an adaptive change amount threshold are configured, and the real-time surge margin is dynamically updated according to the change amount of the real-time surge influencing parameters;

[0008] When the real-time surge boundary is matched, an abnormality evaluation period is configured, the surge influence parameter values in the abnormality evaluation period are obtained, and the Hotelling statistics of each set of surge influence parameter values is calculated, the surge influence parameters are evaluated for abnormality to determine whether there is a joint abnormality situation;

[0009] If there is a joint abnormality situation, the Hotelling statistics of the surge influence parameter vector of the joint abnormality situation is calculated, the contribution degrees of each surge influence parameter are obtained to mark the abnormality source and divide the abnormality level, the corrected surge influence parameter corresponding to the abnormality source is obtained and dynamic compensation is performed, and a corrected real-time working condition vector is constructed to obtain a corrected real-time surge boundary;

[0010] According to the corrected real-time surge boundary, the initial surge margin is obtained through the collected real-time surge monitoring parameter values, and the initial surge margin is adaptively adjusted based on the change trend of the real-time surge monitoring parameter values to perform surge early warning.

[0011] Specifically, the specific steps of obtaining the real-time surge boundary of the compressor include:

[0012] Based on the collected surge influence parameter values, a real-time working condition vector is constructed, and similarity calculation is performed with the experimental working condition vectors in the multi-dimensional surge boundary database;

[0013] A full matching threshold is configured, if there is an experimental working condition vector in the multi-dimensional surge boundary database with a similarity greater than the full matching threshold with the real-time working condition vector, the experimental working condition vector is selected as the target working condition vector according to the similarity, and the surge boundary corresponding to the target working condition vector in the multi-dimensional surge boundary database is taken as the real-time surge boundary of the compressor;

[0014] Otherwise, set the extreme value of the surge influence parameter, perform similarity calculation on the collected surge influence parameter values and the extreme value of the surge influence parameter, measure the similarity by calculating the relative deviation of each surge influence parameter and the corresponding extreme value, and select the target surge influence parameter;

[0015] The experimental working condition vectors containing the target surge influence parameter values are filtered out from the multi-dimensional surge boundary database to construct an experimental working condition vector set;

[0016] If there is no value in the multi-dimensional surge boundary database that is the same as the target surge influence parameter value, the experimental working condition vector corresponding to the parameter value closest to the target surge influence parameter value is selected to construct the experimental working condition vector set.

[0017] Specifically, the specific steps of obtaining the real-time surge boundary of the compressor further include:

[0018] The real-time working condition vector is calculated for similarity with the set of experimental working condition vectors, a working condition vector with the highest similarity is selected as the target working condition vector according to the similarity, and the surge boundary corresponding to the target working condition vector in the multi-dimensional surge boundary database is taken as the real-time surge boundary of the compressor;

[0019] An update time threshold is configured, and the real-time change amount of the surge influence parameter is calculated within the update time threshold after the real-time surge boundary is obtained; a change amount threshold of adaptive update is set; the real-time change amount of the surge influence parameter is compared with the corresponding change amount threshold to determine whether the surge boundary update is triggered.

[0020] Specifically, the specific steps of the abnormality evaluation of the surge influence parameter include:

[0021] The historical surge influence parameter values of the compressor are obtained, the covariance matrix of the surge influence parameter is calculated, and the mean vector of the surge influence parameter is constructed;

[0022] An abnormality evaluation period is configured, and when the real-time surge boundary is matched, the surge influence parameter values within the abnormality evaluation period from the current time are obtained, and a time series surge influence parameter vector is constructed;

[0023] For each group of surge influence parameter vectors in the time series surge influence parameter vector, the Hotelling statistics of the surge influence parameter vector relative to the mean vector and the covariance matrix of the surge influence parameter are calculated;

[0024] A confidence threshold is configured, the Hotelling statistics range is set according to the confidence threshold, the Hotelling statistics value calculated for each group of surge influence parameter vectors is compared with the set Hotelling statistics range, and it is determined whether the current surge influence parameter vector has a joint abnormality.

[0025] Specifically, the specific steps of the abnormality evaluation of the surge influence parameter further include:

[0026] The Hotelling statistics of the surge influence parameter vector determined to have a joint abnormality is decomposed, and the contribution degree of each surge influence parameter to the abnormality degree is calculated;

[0027] A contribution threshold is configured, and it is determined whether the surge influence parameter is marked as an abnormal source according to the contribution degree calculated for each surge influence parameter;

[0028] After the abnormality evaluation period ends, the number of surge influence parameter vectors with a joint abnormality, the type of abnormal source, and the number of occurrences of the same abnormal source in the period are counted, and the counted data is input into a fuzzy logic classifier to output an abnormality level of the abnormality degree of the surge influence parameter of the current compressor.

[0029] Specifically, the specific steps of modifying the real-time surge boundary include:

[0030] An abnormality level of the abnormality degree of the compressor surge influence parameter is obtained, including a low-risk level, a medium-risk level, and a high-risk level; when the abnormality degree of the compressor surge influence parameter is the medium-risk level or the high-risk level, a boundary modification strategy is triggered;

[0031] If the boundary modification strategy is triggered, a surge influence parameter marked as an abnormality source is obtained as a modified surge influence parameter;

[0032] For each modified surge influence parameter, a mean value and a standard deviation in an abnormality evaluation period thereof are calculated, and a difference between the mean value in the abnormality evaluation period and a corresponding real-time surge influence parameter value is calculated as a compensation amount of the modified surge influence parameter.

[0033] Specifically, the specific steps of modifying the real-time surge boundary further include:

[0034] A confidence weighted coefficient is configured by constructing an exponential function of the compensation amount of the modified surge influence parameter and the standard deviation in the abnormality evaluation period thereof, and the modified surge influence parameter is dynamically compensated in combination with the compensation amount of the modified surge influence parameter;

[0035] After the dynamic compensation of all the modified surge influence parameters, a modified real-time working condition vector is constructed, and a modified real-time surge boundary of the compressor is obtained based on a multi-dimensional surge boundary database.

[0036] Specifically, the specific steps of adaptively adjusting the initial surge margin for surge early warning include:

[0037] According to the real-time surge boundary, a key parameter value corresponding to the real-time surge boundary is selected from a real-time surge monitoring parameter value, and a real-time key parameter vector is constructed;

[0038] According to a real-time surge boundary model, a theoretical value of each key parameter value at the surge boundary is obtained, and a real-time theoretical key parameter vector is constructed;

[0039] A distance measurement method is used to calculate a distance between the real-time key parameter vector and the real-time theoretical key parameter vector as the initial surge margin;

[0040] A prediction period is configured, a time series model is established for each surge monitoring parameter value, a predicted value of each surge monitoring parameter in the prediction period is fitted, and a difference between the predicted value and the current value of each surge monitoring parameter is calculated;

[0041] A margin adjustment threshold is configured, and a surge monitoring parameter having a difference between the predicted value and the current value greater than the margin adjustment threshold is selected as a margin adjustment parameter;

[0042] Configure a regulation period, calculate the change rate of the margin adjustment parameter value in the regulation period, set a regulation coefficient, based on the initial surge margin, combine the maximum absolute value of the change rate of the margin adjustment parameter value and its sign characteristics, adaptively regulate the initial surge margin to obtain a regulated surge margin;

[0043] Set a surge early warning threshold, when the regulated surge margin is less than the surge early warning threshold, perform surge early warning.

[0044] A surge boundary prediction method, comprising the following steps:

[0045] Obtain a multi-dimensional surge boundary database, collect real-time surge influence parameters of a compressor to construct a multi-dimensional real-time working condition vector, perform multi-level similarity calculation on the experimental working condition vector in the multi-dimensional surge boundary database, and match the real-time surge boundary;

[0046] And configure an update time threshold and an adaptive change amount threshold, dynamically update the real-time surge boundary according to the change amount of the real-time surge influence parameter;

[0047] When the real-time surge boundary is matched, configure an abnormality evaluation period, obtain the surge influence parameter values in the abnormality evaluation period and calculate the Hotteling statistics of each group of surge influence parameter values, perform abnormality evaluation on the surge influence parameters to determine whether there is a joint abnormality situation;

[0048] If there is a joint abnormality situation, calculate the Hotteling statistics of the surge influence parameter vector of the joint abnormality situation, obtain the contribution degree of each surge influence parameter, mark the abnormality source and divide the abnormality level, obtain the corrected surge influence parameter corresponding to the abnormality source and perform dynamic compensation, construct a corrected real-time working condition vector to obtain a corrected real-time surge boundary;

[0049] According to the corrected real-time surge boundary, obtain an initial surge margin through the collected real-time surge monitoring parameter values, and adaptively adjust the initial surge margin based on the change trend of the real-time surge monitoring parameter values to perform surge early warning.

[0050] The beneficial effects of the present application are:

[0051] This application effectively addresses core issues in existing aero-engine compressor surge prediction, such as large boundary model deviations, delayed early warning responses, and poor adaptability to complex operating conditions, through multi-level similarity matching for dynamic surge boundary correction and a margin-adaptive adjustment strategy based on parameter change rates. Specifically, it significantly improves the alignment between boundary predictions and actual operating conditions by dynamically correcting surge boundaries through real-time anomaly assessment and parameter compensation; enhances the timeliness and accuracy of early warnings by combining multi-physics parameter change trend analysis and fuzzy logic classification; expands robustness in complex scenarios such as unsteady states and high altitudes by employing extreme value matching and dynamic threshold update mechanisms; and achieves adaptive compensation for compressor aging drift and optimization of the safe operating range through online learning and multi-parameter collaborative analysis. This system comprehensively overcomes the technical bottlenecks of traditional methods in dynamic response, multi-parameter coupled analysis, and model generalization, providing a reliable guarantee for the safe operation of aero-engines under all operating conditions. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the surge boundary prediction system of the present invention;

[0053] Figure 2 This is a flowchart illustrating the specific steps involved in obtaining the real-time surge boundary of the compressor according to the present invention.

[0054] Figure 3 This is a flowchart illustrating the specific steps of the present invention for anomaly assessment of surge influence parameters;

[0055] Figure 4 A flowchart illustrating the specific steps involved in correcting the real-time surge boundary in this invention;

[0056] Figure 5 The flowchart shows the specific steps of adaptively adjusting the initial surge margin and performing surge warning according to the present invention. Detailed Implementation

[0057] Please see Figure 1 This embodiment introduces a surge boundary prediction system, including: simulating full-condition operation through a variable geometry test bench in an experimental environment, synchronously adjusting surge influencing parameters, and collecting surge monitoring parameters in real time through a sensor array; triggering surge event recording based on the joint discrimination criteria of surge monitoring parameters, and establishing a multi-dimensional surge boundary database.

[0058] Specifically, surge-affecting parameters include: rotational speed, flow rate, pressure ratio, inlet temperature, inlet pressure, and atmospheric humidity;

[0059] Different rotational speeds will change the airflow velocity, pressure distribution, and other factors inside the compressor, which will affect the occurrence of surge. When the rotational speed deviates from the design value by a large margin, it is easy to cause blade stall, which will trigger surge. When the flow rate is too low, the airflow inside the compressor becomes unstable, which is prone to separation and backflow, increasing the risk of surge; when the flow rate is too high, the compressor may also be in an overload state, which also affects surge. The pressure ratio reflects the degree of gas compression in the compressor. Both too high or too low pressure ratios can disrupt the stable flow of air in the compressor. If the pressure ratio suddenly increases, it may cause the separation of airflow on the blade surface to intensify, prompting the occurrence of surge. Changes in inlet temperature will affect the physical properties of gas, such as density and viscosity. Higher inlet temperatures will reduce the density of the gas, and under the same flow rate and rotational speed, the stress on the blades will change, which may cause the surge boundary to move. Fluctuations in inlet pressure will affect the intake conditions of the compressor. When the inlet pressure decreases, the compressibility of the gas increases, and the stability of the airflow deteriorates, which is prone to trigger surge. Changes in humidity will affect the physical properties of air, such as density and specific heat capacity. In a high humidity environment, the water vapor content in the air increases, which may change the flow characteristics of the airflow and have some impact on surge.

[0060] In particular, the surge monitoring parameters include: pressure parameters, flow parameters, temperature parameters, vibration parameters, rotational speed parameters, and acoustic emission parameters.

[0061] The pressure parameters include inlet pressure, outlet pressure, and pressure between stages of blades. Changes in pressure can directly reflect the stability of airflow in the compressor. Real-time monitoring of flow data can help determine the working state of the compressor. Temperature parameters include inlet temperature, outlet temperature, and temperature changes at key positions, all of which are closely related to surge. The vibration parameters of the compressor during operation can reflect the internal stress and airflow stability. Excessive vibration amplitude or abnormal vibration frequency may be a manifestation of surge, as surge will cause unstable vibration of the compressor structure. Accurate monitoring of the rotational speed parameters of the compressor ensures that it operates within a stable range. Sudden changes or instability in rotational speed can lead to the occurrence of surge. When surge occurs, the compressor will produce special acoustic signals. By monitoring the frequency, intensity, and other characteristics of acoustic emission parameters, the occurrence of surge can be predicted in advance.

[0062] In particular, the joint discrimination criteria for surge monitoring parameters include: pressure drop criterion, flow fluctuation criterion, vibration overrun criterion, and acoustic emission characteristic criterion.

[0063] The pressure sudden drop criterion refers to that when the outlet pressure drops by more than a preset pressure threshold within a preset time, the surge is determined to occur. The flow fluctuation criterion refers to that if the flow fluctuates periodically and the fluctuation amplitude exceeds a preset fluctuation threshold, and the fluctuation frequency is within a preset fluctuation range, the surge is determined to occur. The vibration overrun criterion refers to that when the vibration amplitude exceeds a preset vibration threshold, and the vibration frequency abnormally changes beyond a preset frequency range, the surge is determined to occur. The acoustic emission feature criterion refers to monitoring the main frequency and energy distribution of the acoustic emission signal. When the main frequency of the acoustic emission signal deviates, and the energy of a specific frequency band increases by more than a preset energy threshold, the surge is determined to occur.

[0064] Specifically, the multi-dimensional surge boundary database includes experimental basis data, surge event information, a set of characteristic parameters, a multi-dimensional surge boundary model, and metadata information.

[0065] The experimental basis data includes a working condition parameter record, records the environmental conditions at each experiment, that is, the specific set value and real-time change curve of the surge influencing parameter, and further includes real-time acquisition data of various types of surge monitoring parameters, which are stored in the form of time series, providing a basis for analyzing the dynamic changes of various parameters during the surge occurrence. The surge event information includes the specific time of surge occurrence, the duration, the instantaneous value of each monitoring parameter at the time of occurrence, and the change trend. The set of characteristic parameters refers to the characteristic parameters extracted from various types of surge monitoring parameters for joint discrimination according to multiple surge monitoring parameters, including the amplitude, frequency, and phase of pressure pulsation, the change rate of flow, and the frequency spectrum characteristics of vibration including the main frequency, secondary main frequency, and energy distribution. The multi-dimensional surge boundary model is based on the experimental basis data to establish a surge boundary model under different surge influencing parameters, describing the quantitative relationship between the surge boundary and each surge influencing parameter. The metadata information records the detailed information of the experimental equipment, including the model, performance index, accuracy level, and the relevant information of the experimental personnel, such as name, professional background, and operation experience.

[0066] When the compressor is in a running state, the real-time surge influencing parameters and real-time surge monitoring parameters of the compressor are acquired through the sensor array configured for the compressor; and the real-time surge influencing parameters are matched with the multi-dimensional surge boundary database through multi-level similarity measurement to acquire the real-time surge boundary of the compressor.

[0067] Please refer to Figure 2 , preferably, the specific steps of acquiring the real-time surge boundary of the compressor include:

[0068] Based on the values of the acquired surge influencing parameters, each surge influencing parameter is arranged in a certain order to construct a multi-dimensional vector as a real-time working condition vector; a basis is provided for subsequent comparison with historical working conditions, so that the surge boundary prediction can be based on comprehensive consideration of multiple parameters, avoiding one-sidedness of single parameter analysis.

[0069] The constructed real-time operating condition vector and the experimental operating condition vector in the multi-dimensional surge boundary database are calculated for similarity, and the similarity calculation can adopt the Euclidean distance, cosine similarity, etc. The similarity degree of the real-time operating condition and the historical experimental operating condition is quantified, the operating condition similar to the current compressor operating state can be quickly located from the massive data in the database, thereby providing a reference for determining the real-time surge boundary and improving the accuracy and efficiency of the surge boundary prediction.

[0070] A full match threshold is configured. If there is an experimental operating condition vector in the multi-dimensional surge boundary database with a similarity to the real-time operating condition vector greater than the full match threshold, that is, the difference of each corresponding component is within a very small error range, then the experimental operating condition vector is selected as the target operating condition vector according to the similarity size, and the surge boundary corresponding to the target operating condition vector in the multi-dimensional surge boundary database is taken as the real-time surge boundary of the compressor. The data accumulated by the historical experiments are used to quickly and accurately determine the surge boundary under the current operating condition, reduce complex calculation and uncertainty, provide a reliable boundary reference for safe operation of the compressor, and effectively predict the distance of the compressor from the surge state under the current operating condition.

[0071] If there is no experimental operating condition vector with a similarity to the real-time operating condition vector greater than the full match threshold, then the extreme values of the surge influence parameters are set according to the compressor performance. The extreme values of the surge influence parameters represent the maximum and minimum values that the surge influence parameters may reach under different operating conditions of the compressor. By considering the extreme cases of the parameters, the analysis range is widened, which helps to reasonably predict the surge boundary under complex or rare operating conditions and ensures the integrity and adaptability of the surge boundary prediction.

[0072] The values of the collected surge influence parameters and the extreme values of the surge influence parameters are calculated for similarity one by one. The relative deviation of each surge influence parameter from the corresponding extreme value is calculated to measure the similarity, and the surge influence parameter with the maximum similarity is selected as the target surge influence parameter. Focusing on the key influencing factors, the parameter with the greatest impact on the surge boundary is quickly determined under complex operating conditions, which provides a basis for subsequent screening of related historical operating condition vectors and improves the pertinence and effectiveness of the surge boundary prediction under special operating conditions.

[0073] The experimental operating condition vectors containing the target surge influence parameter values are selected from the multi-dimensional surge boundary database to construct an experimental operating condition vector set.

[0074] If there is no value in the multi-dimensional surge margin database that is the same as the target surge influence parameter value, an experimental operating condition vector corresponding to the parameter value closest to the target surge influence parameter value is selected to construct an experimental operating condition vector set. The vector set is constructed to narrow the analysis range, making subsequent calculations more targeted, improving the accuracy and efficiency of surge margin prediction, and better fitting the current compressor operating state based on historical operating conditions related to the current key parameters.

[0075] The real-time operating condition vector is calculated for similarity with the experimental operating condition vector set, and the experimental operating condition vector with the highest similarity is selected as the target operating condition vector based on the similarity size. The surge margin corresponding to the target operating condition vector in the multi-dimensional surge margin database is taken as the real-time surge margin of the compressor. In order to find a relatively reasonable surge margin based on historical data and the current operating condition when the initial matching fails, ensure that the surge margin prediction can be performed under various operating conditions, and enhance the robustness of the prediction system.

[0076] An update time threshold is configured, and the real-time change amount of the surge influence parameter is calculated within the update time threshold after obtaining the real-time surge margin. The real-time surge influence parameter is monitored. The real-time change amount of the surge influence parameter is continuously monitored within the update time threshold to timely capture the dynamic changes of the compressor operating state.

[0077] According to the compressor operating condition, an adaptive update change amount threshold is set, and the real-time change amount of the surge influence parameter is compared with the corresponding change amount threshold. If the real-time change amount of the surge influence parameter is greater than the corresponding change amount threshold, the surge margin update is triggered, i.e. the real-time surge influence parameter is matched with the multi-dimensional surge margin database by multi-level similarity measurement to obtain the real-time surge margin of the compressor. Otherwise, the cumulative duration after obtaining the real-time surge margin is counted, and if it is equal to the update time threshold, the surge margin update is triggered. The prediction result is more consistent with the real-time operating condition, the real-time performance and accuracy of the surge margin prediction are improved, and the occurrence of surge is effectively prevented.

[0078] Preferably, the specific steps of adaptive update of the change amount threshold include:

[0079] During the operation of the compressor, the change of the surge influence parameter is tracked in real time, a sliding window data buffer is established, and for each surge influence parameter, the difference operation is performed on the surge influence parameter values in the sliding window data buffer to obtain the real-time change amount. The change amount is obtained in real time through the sliding window and the difference operation, providing real-time and accurate data basis for threshold update.

[0080] In the initial stage of the compressor start, an initial change threshold matrix is set for each surge influencing parameter, and the initial change threshold is determined according to the compressor design manual; the threshold is updated through absolute change triggering and trend triggering, including:

[0081] The sensitivity coefficient is configured, and the threshold is updated when the real-time change of any one of the surge influencing parameters is greater than the product of the change threshold of the surge influencing parameter at the previous moment and the sensitivity coefficient. The sensitivity coefficient is in the range of 0.05 to 0.15, and is adjusted according to the specific operating characteristics of the compressor and the sensitivity requirement of the surge monitoring.

[0082] The trend interval K and the trend coefficient are configured, and the threshold is updated when the change of the surge influencing parameter in the continuous K sampling values presents a monotonically increasing or monotonically decreasing trend, and the cumulative change exceeds the product of the threshold of the surge influencing parameter at the previous moment and the trend coefficient. This trend triggering mechanism can capture the potential trend of parameter change, and adjust the threshold in advance to adapt to the change of the operating state of the compressor.

[0083] The evaluation period Q is set, and the threshold updating process is started when the threshold is updated, the same condition data in the past Q operating periods is extracted, the change sequence of each surge influencing parameter is constructed, and the statistical characteristics of each surge influencing parameter are counted, including the mean, the standard deviation, the kurtosis and the skewness.

[0084] Based on the statistical characteristics of the surge influencing parameters and the threshold of the surge influencing parameters before the threshold is updated, the threshold is updated using an adaptive weight allocation strategy, that is:

[0085] ;

[0086] Wherein, is the threshold of the first surge influencing parameter after the threshold is updated, is the threshold of the first surge influencing parameter before the threshold is updated, is the mean of the first surge influencing parameter, is the variance of the first surge influencing parameter, is the historical threshold retention coefficient, in the range of 0.4 to 0.6, representing the retention degree of the threshold at the previous moment, is the new data weight coefficient, in the range of 0.6 to 0.8, representing the importance of new data in the threshold update; is the confidence factor, which is dynamically adjusted according to the kurtosis of the surge influencing parameter. For example, if the kurtosis is greater than 3.5, it means that the distribution presents a sharp peak, and at this time 1.5; if the kurtosis is less than or equal to 2.5, the data distribution is relatively flat, 0.8; otherwise, the data distribution is approximately normal, 1; this dynamic adjustment can more reasonably determine the threshold value according to the distribution characteristics of the data.

[0087] When the real-time surge boundary is matched, the surge influencing parameter is abnormally evaluated, and the real-time surge boundary is corrected according to the abnormal evaluation result;

[0088] Please refer to Figure 3 , preferably, the specific steps of abnormally evaluating the surge influencing parameter include:

[0089] Collect the historical surge influencing parameter values of the compressor during the past normal operation, which includes the rotating speed, flow, pressure ratio, inlet temperature, inlet pressure and atmospheric humidity, etc. Calculate the covariance matrix of the surge influencing parameter, which is used to describe the mutual relationship and fluctuation between the surge influencing parameters, and calculate the mean value of the historical surge influencing parameter value respectively, to build the mean value vector of the surge influencing parameter, reflecting the average level of the surge influencing parameter under normal operation state;

[0090] Configure the abnormal evaluation period, the length of the abnormal evaluation period needs to be determined according to the operation characteristics of the compressor and the actual application scene. For the compressor with frequent changes in operating state, a shorter evaluation period can be set.

[0091] When the real-time surge boundary is matched, start to obtain the surge influencing parameter values within the abnormal evaluation period from the current time, take the evaluation period as the time span, arrange each group of surge influencing parameter values corresponding to each time in turn, and build the time series surge influencing parameter vector.

[0092] For each group of surge influencing parameter vector in the time series surge influencing parameter vector, calculate the holtlin statistics of the group of surge influencing parameter vector relative to the mean value vector and the covariance matrix of the surge influencing parameter. The holtlin statistics can comprehensively reflect the deviation degree of each group of surge influencing parameter vector from the mean value vector of the parameter under the historical normal working condition. The calculation formula of the holtlin statistics is:

[0093] ;

[0094] Wherein, is the holtlin statistics value of the th group of surge influencing parameter vector, is the th group of surge influencing parameter vector, is the mean value vector of the surge influencing parameter, is the covariance matrix of the surge influencing parameter;

[0095] Configure confidence thresholds, such as 95% or 99%, based on actual needs and the statistical characteristics of historical data, and set the range of the Hotelling statistic according to the confidence thresholds.

[0096] For each group of surge influence parameter vectors, the calculated Hotling statistic value is compared with the set Hotling statistic range. If the Hotling statistic value of the surge influence parameter vector exceeds the Hotling statistic range, it is determined that the current surge influence parameter vector has a joint anomaly. This indicates that the combination of parameters is significantly different from the parameter combination pattern during historical normal operation, which may indicate an abnormality in the compressor's operating state.

[0097] The Hotelling statistic of the surge influence parameter vector indicating the presence of joint anomalies is decomposed, and the contribution of each surge influence parameter to the degree of anomaly is calculated, i.e.:

[0098] ;

[0099] in, It is the first The first group of surge influence parameter vectors The contribution of each surge-influencing parameter to the degree of anomaly. It is the first The first group of surge influence parameter vectors One surge-affecting parameter value, It is the first in the mean vector of surge influence parameters One surge-affecting parameter value, It is the first inverse of the covariance matrix. OK;

[0100] Configure a contribution threshold. For each surge impact parameter, if the calculated contribution is greater than the contribution threshold, then the surge impact parameter is marked as an anomaly source, meaning it plays a major role in the current parameter anomaly.

[0101] After the anomaly assessment period ends, the number of surge influence parameter vectors, the type of anomaly source, and the frequency of identical anomaly sources in cases of joint anomalies within that period are statistically analyzed. This statistical data is then input into a pre-built fuzzy logic classifier. The fuzzy logic classifier processes the input data by defining a series of fuzzy rules and membership functions. For example, the fuzzy rules can be set as follows: if there are a large number of vectors with joint anomalies and a certain key anomaly source occurs frequently, a high anomaly level is output; if there are few anomaly vectors, and the anomaly source is a non-key parameter with infrequent occurrences, a low anomaly level is output. In this way, multiple factors are comprehensively considered to ultimately output an anomaly level reflecting the degree of anomaly in the current compressor surge influence parameters, providing a basis for subsequent boundary correction and other operations.

[0102] Please see Figure 4 Preferably, the specific steps for correcting the real-time surge boundary include:

[0103] Obtain the anomaly level of the compressor surge impact parameters, including low-risk, medium-risk, and high-risk levels;

[0104] For low-risk levels, only relevant logs are recorded. At this time, the abnormal parameters have little impact on the surge boundary, and no boundary correction is performed.

[0105] For the medium-risk and high-risk level trigger boundary correction strategy, based on the abnormality level of the compressor surge influence parameters, offset compensation is performed on the surge influence parameters corresponding to each abnormal source to construct a corrected real-time operating condition vector to match the corrected real-time surge boundary.

[0106] Preferably, the specific steps for offset compensation of the surge influence parameters corresponding to each anomaly source include:

[0107] Obtain surge impact parameters marked as abnormal sources, and use them as correction surge impact parameters;

[0108] For each corrected surge impact parameter, calculate its mean and standard deviation within the abnormal assessment period. The difference between the mean within the abnormal assessment period and the corresponding real-time surge impact parameter value is used as the compensation amount for the corrected surge impact parameter.

[0109] By constructing an exponential function of the compensation amount and the standard deviation of the corrected surge influence parameter within its anomaly assessment period, and configuring a confidence weighting coefficient, parameter compensation can be performed more reasonably. Combined with the compensation amount of the corrected surge influence parameter, dynamic compensation is then applied to the corrected surge influence parameter.

[0110] ;

[0111] in, It is the first after dynamic compensation One corrected surge effect parameter value It is the first The real-time surge effect parameter value corresponding to the corrected surge effect parameter value. It is the first The compensation amount for each correction surge effect parameter, It is the first time during the abnormal assessment period The standard deviation of the corrected surge effect parameter;

[0112] Dynamic compensation is performed on all corrected surge impact parameters to construct a corrected real-time operating condition vector. Based on a multi-dimensional surge boundary database, the corrected real-time surge boundary of the compressor is obtained.

[0113] Based on real-time surge monitoring parameter values ​​and real-time surge boundaries, the initial surge margin is obtained, and the initial surge margin is adaptively adjusted based on the changing trend of surge monitoring parameter values ​​to provide surge early warning.

[0114] Please see Figure 5 Preferably, the specific steps for adaptively adjusting the initial surge margin and providing surge warning include:

[0115] Based on the real-time surge boundary, key parameter values ​​corresponding to the real-time surge boundary are selected from the real-time surge monitoring parameter values, and the key parameters are sorted according to the parameter order of the surge boundary to construct a real-time key parameter vector;

[0116] Based on the real-time surge boundary model, the theoretical values ​​of each key parameter at the surge boundary are obtained to construct a real-time theoretical key parameter vector;

[0117] Distance metrics, including Euclidean distance or Mahalanobis distance, are used to calculate the distance between the real-time critical parameter vector and the real-time theoretical critical parameter vector, which is used as the initial surge margin.

[0118] Configure the prediction period. The length of the prediction period is determined according to the compressor operating characteristics and actual needs. For compressors with frequent changes in operating conditions, it can be set to 5 seconds. For relatively stable compressors, it can be set to 30 seconds. Establish a time series model for each surge monitoring parameter value. Use methods such as autoregressive integral moving average model or long short-term memory network to fit the predicted value of each surge monitoring parameter within the prediction period, and calculate the difference between the predicted value and the current value of each surge monitoring parameter.

[0119] Configure a margin adjustment threshold, filter out surge monitoring parameters whose difference between the predicted value and the current value is greater than the margin adjustment threshold, and mark them as margin adjustment parameters;

[0120] Configure the control period, calculate the rate of change of the margin adjustment parameter value within the control period, and adaptively adjust the initial surge margin based on the rate of change, i.e.:

[0121] ;

[0122] in, It is the surge margin after adaptive adjustment. It is the initial surge margin. This is the control coefficient, which is set according to the degree of influence of the margin adjustment parameters on the surge boundary and the compressor's operational safety requirements. Its value ranges from 0.1 to 0.5. It is the first The rate of change of each margin adjustment parameter It is the number of margin adjustment parameters. It is a symbolic function. The sign of the rate of change for adjusting the parameter value to obtain the margin of maximum rate of change.

[0123] Set a surge warning threshold and determine in real time whether the surge margin is less than the surge warning threshold. When the surge margin is less than the surge warning threshold, a surge warning is issued and the warning information is released in multiple ways, such as displaying a conspicuous warning icon and text prompt on the monitoring system interface, and sending SMS or email notifications to relevant operators so that operators can quickly and accurately obtain the warning information and take corresponding measures.

[0124] Example 2

[0125] A surge boundary prediction method includes the following steps:

[0126] Step S1: Obtain a multi-dimensional surge boundary database;

[0127] Step S2: Collect real-time surge impact parameters of the compressor to construct a multi-dimensional real-time operating condition vector, and perform multi-level similarity calculation with the experimental operating condition vector in the multi-dimensional surge boundary database to match the real-time surge boundary;

[0128] Step S3: Configure the update time threshold and adaptive change threshold, and dynamically update the real-time surge boundary based on the real-time surge influence parameter change.

[0129] Step S4: When a real-time surge boundary is matched, configure the anomaly assessment period, obtain the surge impact parameter values ​​within the anomaly assessment period, and calculate the Hotling statistic for each group of surge impact parameter values. Perform anomaly assessment on the surge impact parameters to determine whether there are joint anomalies.

[0130] Step S5: If there are joint anomalies, calculate the Hotling statistic of the surge influence parameter vector of the joint anomalies, obtain the contribution of each surge influence parameter, mark the anomaly source and classify the anomaly level, obtain the corrected surge influence parameter corresponding to the anomaly source and perform dynamic compensation, construct the corrected real-time operating condition vector to obtain the corrected real-time surge boundary.

[0131] Step S6: Based on the corrected real-time surge boundary, obtain the initial surge margin by collecting the real-time surge monitoring parameter values, and adaptively adjust the initial surge margin based on the changing trend of the real-time surge monitoring parameter values ​​to provide a surge warning.

[0132] Preferably, the specific steps for obtaining the real-time surge boundary of the compressor include:

[0133] Based on the collected surge impact parameter values, a real-time operating condition vector is constructed, and its similarity with the experimental operating condition vector in the multi-dimensional surge boundary database is calculated.

[0134] A full match threshold is configured, and if the similarity between the experimental operating condition vector and the real-time operating condition vector in the multi-dimensional surge boundary database is greater than the full match threshold, the experimental operating condition vector is selected as the target operating condition vector according to the similarity, and the surge boundary corresponding to the target operating condition vector in the multi-dimensional surge boundary database is taken as the real-time surge boundary of the compressor.

[0135] Otherwise, an extreme value of the surge influence parameter is set, the similarity between the collected value of the surge influence parameter and the extreme value of the surge influence parameter is calculated, the relative deviation of each surge influence parameter from the corresponding extreme value is calculated to measure the similarity, and a target surge influence parameter is selected;

[0136] An experimental operating condition vector set is constructed by selecting the experimental operating condition vector containing the target surge influence parameter value from the multi-dimensional surge boundary database.

[0137] If there is no value in the multi-dimensional surge boundary database that is the same as the target surge influence parameter value, an experimental operating condition vector corresponding to the parameter value closest to the target surge influence parameter value is selected to construct the experimental operating condition vector set.

[0138] The similarity between the real-time operating condition vector and the experimental operating condition vector set is calculated, the experimental operating condition vector with the highest similarity is selected as the target operating condition vector according to the similarity, and the surge boundary corresponding to the target operating condition vector in the multi-dimensional surge boundary database is taken as the real-time surge boundary of the compressor.

[0139] An update time threshold is configured, and within the update time threshold after obtaining the real-time surge boundary, the real-time change amount of the surge influence parameter is calculated, an adaptive update change amount threshold is set, and the real-time change amount of the surge influence parameter is compared with the corresponding change amount threshold to determine whether to trigger the surge boundary update.

[0140] Preferably, the specific steps of the abnormality assessment of the surge influence parameter include:

[0141] The historical surge influence parameter values of the compressor are obtained, the covariance matrix of the surge influence parameter is calculated, and the mean vector of the surge influence parameter is constructed.

[0142] An abnormality assessment period is configured, and when the real-time surge boundary is matched, the surge influence parameter values within the abnormality assessment period from the current time are obtained, and a time series surge influence parameter vector is constructed.

[0143] The Hotelling statistic of each group of surge influence parameter vectors in the time series surge influence parameter vector relative to the mean vector and the covariance matrix of the surge influence parameter is calculated.

[0144] A confidence threshold is configured, and a range of Hotelling statistics is set according to the confidence threshold. The Hotelling statistic value calculated for each set of surge influence parameter vectors is compared with the set range of Hotelling statistics, and it is determined whether the current surge influence parameter vector has a joint abnormal situation.

[0145] The Hotelling statistic of the surge influence parameter vector determined to have a joint abnormal situation is decomposed, and the contribution degree of each surge influence parameter to the abnormality degree is calculated.

[0146] A contribution threshold is configured, and it is determined whether to mark the surge influence parameter as an abnormal source according to the contribution degree calculated for each surge influence parameter.

[0147] After the abnormality evaluation period ends, the number of surge influence parameter vectors with a joint abnormal situation, the type of abnormal source, and the number of occurrences of the same abnormal source in the period are counted. The counted data is input into a fuzzy logic classifier, and an abnormality level of the abnormality degree of the compressor surge influence parameters is output.

[0148] Preferably, the specific steps of modifying the real-time surge boundary include:

[0149] The abnormality level of the abnormality degree of the compressor surge influence parameters includes a low-risk level, a medium-risk level, and a high-risk level. When the abnormality degree of the compressor surge influence parameters is a medium-risk level or a high-risk level, a boundary modification strategy is triggered.

[0150] If the boundary modification strategy is triggered, the surge influence parameter marked as an abnormal source is obtained as a modified surge influence parameter.

[0151] For each modified surge influence parameter, the mean value and the standard deviation in the abnormality evaluation period are calculated. The difference between the mean value in the abnormality evaluation period and the corresponding real-time surge influence parameter value is calculated as the compensation amount of the modified surge influence parameter.

[0152] A confidence weighting coefficient is configured by constructing an exponential function of the compensation amount of the modified surge influence parameter and the standard deviation in the abnormality evaluation period. The modified surge influence parameter is dynamically compensated in combination with the compensation amount of the modified surge influence parameter.

[0153] After all the modified surge influence parameters are dynamically compensated, a modified real-time operating condition vector is constructed, and a multi-dimensional surge boundary database is matched to obtain a modified real-time surge boundary of the compressor.

[0154] Preferably, the specific steps of adaptively adjusting the initial surge margin for surge warning include:

[0155] Based on the real-time surge boundary, key parameter values ​​corresponding to the real-time surge boundary are selected from the real-time surge monitoring parameter values ​​to construct a real-time key parameter vector;

[0156] Based on the real-time surge boundary model, obtain the theoretical value of each key parameter at the surge boundary, and construct a real-time theoretical key parameter vector.

[0157] A distance metric method is used to calculate the distance between the real-time key parameter vector and the real-time theoretical key parameter vector, which is used as the initial surge margin.

[0158] Configure the prediction period, establish a time series model for each surge monitoring parameter value, fit the predicted value of each surge monitoring parameter within the prediction period, and calculate the difference between the predicted value and the current value of each surge monitoring parameter.

[0159] Configure a margin adjustment threshold, filter out surge monitoring parameters whose difference between the predicted value and the current value is greater than the margin adjustment threshold, and mark them as margin adjustment parameters;

[0160] Configure the control period, calculate the rate of change of the margin adjustment parameter value within the control period, set the control coefficient, and based on the initial surge margin, combine the maximum absolute value and sign characteristics of the rate of change of the margin adjustment parameter to adaptively control the initial surge margin and obtain the controlled surge margin.

[0161] Set a surge warning threshold. When the adjusted surge margin is less than the surge warning threshold, a surge warning will be issued.

[0162] Working principle and its effects:

[0163] The surge boundary prediction system constructs a multi-dimensional surge boundary database, collects compressor operating parameters in real time and builds an operating condition vector, dynamically determines the real-time surge boundary through multi-level similarity matching, and corrects the boundary model by combining parameter anomaly assessment and dynamic compensation mechanisms. Finally, it achieves surge early warning based on the adaptive adjustment of the margin between the corrected boundary and real-time monitored parameters. Specifically, the system first obtains the initial boundary by matching historical operating conditions through similarity calculation, then identifies parameter anomalies through Hotling statistic analysis, and dynamically compensates and corrects the boundary for abnormal parameters; simultaneously, it adjusts the surge margin based on parameter change rate and trend analysis, combined with control coefficients and sign characteristics, to trigger the early warning. This method overcomes the limitations of traditional steady-state models, significantly improving the dynamic adaptability and early warning accuracy of surge boundary prediction through real-time data-driven boundary correction and multi-parameter collaborative analysis, effectively solving the core problems of large boundary model deviation, delayed early warning response, and poor adaptability to complex operating conditions in existing technologies.

[0164] Among them, the multi-stage similarity matching mechanism ensures that the similar historical working condition can still be quickly located under the non-steady state condition through the full matching threshold screening and the extreme value interpolation strategy; the Hotteling statistics and the contribution degree decomposition realize the quantitative diagnosis of the multi-parameter joint anomaly, and the fuzzy logic classification output anomaly level is combined to provide a decision basis for boundary correction; the dynamic compensation strategy optimizes the parameter compensation amount through the reliability weighting coefficient, so that the corrected boundary is more suitable for the actual running state; the margin self-adaptive adjustment is based on the sign and amplitude characteristics of the parameter change rate, dynamically adjusts the early warning threshold, and identifies the potential risk in advance. Overall, through online learning and incremental updating, the compressor aging drift compensation, multi-physical field coupling analysis and extreme condition robustness optimization are realized, which provides a reliable guarantee for the safe operation of the aero-engine under all working conditions.

[0165] The above only describes the preferred embodiments of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments only. Any technical solutions falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application shall also be considered as the protection scope of the present application.

Claims

1. A surge margin prediction system, characterized by, The application relates to a method for acquiring a real-time surge boundary of a compressor. The method comprises the following steps: acquiring a multi-dimensional surge boundary database, collecting real-time surge influence parameter values of the compressor, constructing a multi-dimensional real-time working condition vector, and performing multi-level similarity calculation on experimental working condition vectors in the multi-dimensional surge boundary database to match the real-time surge boundary; and configuring an update time threshold and an adaptive change threshold, and dynamically updating the real-time surge boundary according to a change amount of the real-time surge influence parameter values; when the real-time surge boundary is matched, an abnormality evaluation period is configured, the surge influence parameter values in the abnormality evaluation period are acquired, and the Hotelling statistics of each group of surge influence parameter values are calculated, the surge influence parameters are abnormally evaluated, and it is determined whether a joint abnormality exists; if the joint abnormality exists, the Hotelling statistics of a surge influence parameter vector of the joint abnormality are calculated, the contribution degrees of the surge influence parameters are acquired, the abnormal sources are marked, and the abnormal levels are divided, the corrected surge influence parameters corresponding to the abnormal sources are acquired, and dynamic compensation is performed, a corrected real-time working condition vector is constructed, and a corrected real-time surge boundary is acquired; according to the corrected real-time surge boundary, the initial surge margin is acquired through the collected real-time surge monitoring parameter values, and the initial surge margin is adaptively adjusted based on the change trend of the real-time surge monitoring parameter values, and surge early warning is performed; the specific steps of abnormally evaluating the surge influence parameters comprise the following steps: acquiring historical surge influence parameter values of the compressor, calculating a covariance matrix of the surge influence parameters, and constructing a mean value vector of the surge influence parameters; an abnormality evaluation period is configured, the surge influence parameter values in the abnormality evaluation period are acquired, a time sequence surge influence parameter vector is constructed, and the Hotelling statistics of each group of surge influence parameter vectors in the time sequence surge influence parameter vector are calculated; a confidence threshold is configured, the Hotelling statistics range is set according to the confidence threshold, the Hotelling statistics value calculated for each group of surge influence parameter vectors is compared with the set Hotelling statistics range, and it is determined whether the current surge influence parameter vector has a joint abnormality; the Hotelling statistics of the surge influence parameter vector determined to have the joint abnormality are decomposed, and the contribution degrees of each surge influence parameter to the abnormality degree are calculated; a contribution threshold is configured, and it is determined whether the surge influence parameter is marked as an abnormal source according to the contribution degree calculated for each surge influence parameter; after the abnormality evaluation period ends, the number of surge influence parameter vectors having the joint abnormality, the types of abnormal sources and the number of appearances of the same abnormal source in the period are counted, and the counted data are input into a fuzzy logic classifier, and an abnormality level of the abnormality degree of the surge influence parameters of the current compressor is output.

2. The surge margin prediction system of claim 1, wherein, The specific steps of acquiring the real-time surge boundary of the compressor comprise the following steps: a real-time working condition vector is constructed based on the collected surge influence parameter values, and similarity calculation is performed on the experimental working condition vectors in the multi-dimensional surge boundary database; The full match threshold is configured, if the similarity between the experimental working condition vector and the real-time working condition vector in the multi-dimensional surge boundary database is greater than the full match threshold, the experimental working condition vector is selected as the target working condition vector according to the similarity, and the surge boundary corresponding to the target working condition vector in the multi-dimensional surge boundary database is taken as the real-time surge boundary of the compressor; Otherwise, the extreme value of the surge influence parameter is set, the similarity between the collected value of the surge influence parameter and the extreme value of the surge influence parameter is calculated, the relative deviation of each surge influence parameter from the corresponding extreme value is calculated to measure the similarity, and the target surge influence parameter is selected; The experimental working condition vector set is constructed by screening the experimental working condition vector containing the target surge influence parameter value from the multi-dimensional surge boundary database. If there is no value in the multi-dimensional surge boundary database that is the same as the target surge influence parameter value, the experimental working condition vector corresponding to the parameter value closest to the target surge influence parameter value is selected to construct the experimental working condition vector set.

3. The surge margin prediction system of claim 2, wherein, The specific steps for obtaining the real-time surge boundary of the compressor further include: The similarity between the real-time working condition vector and the experimental working condition vector set is calculated, the experimental working condition vector with the highest similarity is selected as the target working condition vector according to the similarity, and the surge boundary corresponding to the target working condition vector in the multi-dimensional surge boundary database is taken as the real-time surge boundary of the compressor; The real-time change amount of the surge influence parameter is calculated within the update time threshold after obtaining the real-time surge boundary, the change amount threshold of adaptive update is set, and the real-time change amount of the surge influence parameter is compared with the corresponding change amount threshold to determine whether the surge boundary update is triggered.

4. The surge boundary prediction system of claim 1, wherein, The specific steps for modifying the real-time surge boundary include: The abnormality level of the compressor surge influence parameter abnormality degree is obtained, including the low risk level, the medium risk level and the high risk level; when the compressor surge influence parameter abnormality degree is the medium risk level or the high risk level, the boundary modification strategy is triggered; If the boundary modification strategy is triggered, the surge influence parameter marked as the abnormal source is obtained as the modified surge influence parameter; For each modified surge influence parameter, the mean value and the standard deviation in the abnormal evaluation period are calculated, and the difference between the mean value in the abnormal evaluation period and the real-time surge influence parameter value is calculated as the compensation amount of the modified surge influence parameter.

5. The surge boundary prediction system of claim 4, wherein, The specific steps for modifying the real-time surge boundary further include: The compensation amount of the modified surge influence parameter and the standard deviation in the abnormal evaluation period are constructed to configure the credibility weighting coefficient, and the modified surge influence parameter is dynamically compensated by combining the compensation amount of the modified surge influence parameter; After dynamically compensating all the modified surge influence parameters, the modified real-time working condition vector is constructed, the multi-dimensional surge boundary database is matched, and the modified real-time surge boundary of the compressor is obtained.

6. The surge boundary prediction system of claim 1, wherein, The specific steps for adaptively adjusting the initial surge margin and performing surge warning include: According to the real-time surge boundary, the key parameter value corresponding to the real-time surge boundary is selected from the real-time surge monitoring parameter value to construct the real-time key parameter vector; According to the real-time surge margin model, the theoretical value of each key parameter value at the surge margin is obtained, and a real-time theoretical key parameter vector is constructed; The distance between the real-time key parameter vector and the real-time theoretical key parameter vector is calculated as the initial surge margin by using the distance measurement method; The prediction period is configured, and the time series model of each surge monitoring parameter value is established to fit the predicted value of each surge monitoring parameter in the prediction period, and the difference between the predicted value and the current value of each surge monitoring parameter is calculated; The margin adjustment threshold is configured, and the surge monitoring parameters whose difference between the predicted value and the current value is greater than the margin adjustment threshold are selected and marked as margin adjustment parameters; The control period is configured, and the change rate of the margin adjustment parameter value is calculated in the control period, the control coefficient is set, and the initial surge margin is adaptively adjusted based on the initial surge margin, combined with the maximum absolute value of the change rate of the margin adjustment parameter value and its sign characteristics, to obtain the adjusted surge margin; The surge warning threshold is set, and when the adjusted surge margin is less than the surge warning threshold, the surge warning is performed.

7. A method for surge margin prediction, implemented based on a surge margin prediction system according to any one of claims 1-6, characterized in that, The method comprises the following steps: Obtain a multi-dimensional surge boundary database; Collect real-time surge influence parameters of the compressor to construct a multi-dimensional real-time working condition vector, and perform multi-level similarity calculation on the experimental working condition vector in the multi-dimensional surge boundary database to match the real-time surge boundary; Configure an update time threshold and an adaptive change amount threshold, and dynamically update the real-time surge boundary according to the change amount of the real-time surge influence parameters; When the real-time surge boundary is matched, configure an abnormality evaluation period, obtain the surge influence parameter values in the abnormality evaluation period, calculate the Hotelling statistics of each group of surge influence parameter values, and perform abnormality evaluation on the surge influence parameters to determine whether there is a joint abnormality situation; If the joint abnormality situation exists, calculate the Hotelling statistics of the surge influence parameter vector of the joint abnormality situation, obtain the contribution degree of each surge influence parameter, mark the abnormal source and divide the abnormality level, obtain the corrected surge influence parameter corresponding to the abnormal source and perform dynamic compensation, and construct a corrected real-time working condition vector to obtain a corrected real-time surge boundary; According to the corrected real-time surge boundary, the initial surge margin is obtained by collecting the real-time surge monitoring parameter values, and the initial surge margin is adaptively adjusted based on the change trend of the real-time surge monitoring parameter values to perform surge warning.

Images