Gas turbine performance evaluation method based on compressor degradation inversion

Abstract

The present application belongs to the technical field of data processing, and particularly relates to a gas turbine performance evaluation method based on compressor degradation inversion. The method comprises: obtaining gas turbine operation data and health samples, generating a health benchmark characteristic surface and determining a steady-state segment and a natural excitation segment; generating a steady-state thermodynamic residual and a phase response kernel residual to form a dual-domain observation residual; performing thermodynamic closed loop flow decomposition on the dual-domain observation residual to obtain a loop flow residual; generating a dual-domain degradation fingerprint dictionary and an interference companion dictionary and performing interference stripping to obtain an interference stripped residual and an interference stripped degradation dictionary; performing group sparse constraint inversion based on the interference stripped degradation dictionary and a degradation primitive coherence coefficient to obtain compressor degradation variables; stripping the degradation components to output performance loss contributions. The present application can enable steady-state deviation, dynamic response deviation and non-compressor interference to participate in evaluation in a unified processing chain to form a decomposable gas turbine performance evaluation result.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to a gas turbine performance evaluation method based on compressor degradation inversion. Background Technology

[0002] Gas turbine performance evaluation is typically used to determine the unit's operating status, compressor degradation, and sources of performance loss. Existing technologies often employ health baseline curves, steady-state performance deviations, compressor segmentation models, or empirical correction coefficients to compare and analyze operating data such as pressure ratio, temperature, power, and heat consumption, thereby determining whether compressor degradation exists. While these methods have some applicability under single steady-state conditions, in actual gas turbine operation, load conditions, adjustable guide vane opening, inlet pressure loss, exhaust back pressure, and fuel calorific value deviations all simultaneously affect the observation results. Relying solely on steady-state deviations can easily lead to misinterpreting non-compressor disturbances as compressor degradation.

[0003] Furthermore, existing assessment methods typically treat temperature deviations, pressure deviations, and power deviations separately, lacking a unified constraint on the closed-loop relationship of the gas turbine thermodynamic cycle, making it difficult to distinguish between measurement point deviations, operating condition changes, and true cycle residuals. For compressor internal flow capacity degradation, isentropic efficiency degradation, abnormal extraction leakage, and equivalent deviations of adjustable guide vanes, existing methods often use single residuals or single model parameters for fitting, failing to establish the correspondence between degradation mechanisms and steady-state deviations and dynamic response deviations, resulting in unclear competitive and exclusionary relationships among different degradation factors. Moreover, existing results usually present overall performance deviations, failing to explain the individual performance loss contributions of various compressor degradation components, which is detrimental to subsequent operation and maintenance assessments. Summary of the Invention

[0004] This invention provides a gas turbine performance evaluation method based on compressor degradation inversion, which is used to at least solve the problem of how to invert compressor degradation and output the performance loss contribution of each degradation component under the conditions of gas turbine operating condition changes and non-compressor disturbances.

[0005] This invention provides a gas turbine performance evaluation method based on compressor degradation inversion, the method comprising: The process involves acquiring gas turbine operating data and health samples, generating a health baseline characteristic surface based on the health samples, and determining steady-state segments and natural excitation segments based on the gas turbine operating data. Steady-state thermodynamic residuals are generated from the steady-state segments, and phased response kernel residuals are generated from the natural excitation segments, forming a dual-domain observation residual. The dual-domain observation residual is then subjected to thermodynamic closed-loop decomposition to obtain the circulation residual. A dual-domain degradation fingerprint dictionary and an interference adjoint dictionary are generated from compressor degradation primitives and non-compressor interference primitives, respectively. The circulation residual and dual-domain degradation fingerprint dictionary are then stripped based on the interference adjoint dictionary to obtain the interference stripped residual and interference stripped degradation dictionary. The coherence coefficients of degradation primitives are determined based on the interference stripped degradation dictionary, and sparse constraint inversion is performed on the interference stripped residual based on the interference stripped degradation dictionary and degradation primitive coherence coefficients to obtain the compressor degradation variable. The degradation component of the compressor degradation variable is stripped, and the performance loss contribution is output.

[0006] In one possible implementation, generating a healthy baseline characteristic surface includes: determining equivalent operating condition variables and healthy state variables based on healthy samples. The equivalent operating condition variables include corrected speed, adjustable guide vane opening, load state, intake pressure loss, and exhaust back pressure. The healthy state variables include healthy pressure ratio, healthy compressor outlet temperature, healthy equivalent airflow, healthy isentropic efficiency, and healthy gas turbine performance parameters. Based on the equivalent operating condition variables and healthy state variables, a characteristic surface satisfying monotonic constraints and efficiency range constraints is established, and this characteristic surface is determined as the healthy baseline characteristic surface.

[0007] In one possible implementation, generating the phased response kernel residual includes: extracting a control perturbation sequence and an operational response sequence from natural excitation fragments; generating an excitation phase axis based on the cumulative perturbation of the control perturbation sequence; resampling the control perturbation sequence and the operational response sequence to the excitation phase axis; generating a phased response kernel based on the resampled control perturbation sequence and operational response sequence; determining healthy natural excitation fragments from healthy samples; generating a healthy response kernel based on the healthy natural excitation fragments; and generating a phased response kernel residual based on the difference between the phased response kernel and the healthy response kernel.

[0008] In one possible implementation, the dual-domain observation residuals are decomposed into a thermodynamic closed-loop circulation, including: establishing a thermodynamic closed-loop residual map based on the compressor inlet state, compressor outlet state, combustion chamber heat input state, turbine exhaust state, and power generation state; mapping the dual-domain observation residuals to edge residuals in the thermodynamic closed-loop residual map; separating gradient residuals from the edge residuals based on the correlation between nodes and edges; and determining the circulation residuals from the separated edge residuals based on the thermodynamic closed-loop relationship.

[0009] In one possible implementation, a dual-domain degradation fingerprint dictionary is generated from compressor degradation primitives, including: dividing the compression process into multiple degradation base regions according to the compressor's normalized isentropic enthalpy rise coordinates; setting up flow capacity degradation primitives, isentropic efficiency degradation primitives, and extraction leakage anomaly primitives in each degradation base region, and setting up adjustable guide vane equivalent deviation primitives; applying degradation perturbations corresponding to each compressor degradation primitive to the healthy baseline characteristic surface under the same equivalent operating condition variables and the same natural excitation conditions, generating steady-state residual fingerprints and phased response kernel residual fingerprints corresponding to each compressor degradation primitive; associating the steady-state residual fingerprints and phased response kernel residual fingerprints as dual-domain degradation fingerprints, and generating a dual-domain degradation fingerprint dictionary based on each dual-domain degradation fingerprint.

[0010] In one possible implementation, an interference adjoint dictionary is generated from non-compressor interference primitives, including: dividing the non-compressor interference primitives into fuel calorific value deviation primitives, measuring point deviation primitives, intake pressure loss deviation primitives, and exhaust back pressure deviation primitives. Under the same equivalent operating condition variables and the same natural excitation conditions, the gas turbine performance evaluation equipment applies the interference perturbation corresponding to each non-compressor interference primitive to the healthy reference characteristic surface, generating a steady-state residual fingerprint and a phased response kernel residual fingerprint for each non-compressor interference primitive; the steady-state residual fingerprint and the phased response kernel residual fingerprint are associated as interference fingerprints, and an interference adjoint dictionary is generated based on each interference fingerprint.

[0011] In one possible implementation, the circulation residual and the dual-domain degenerate fingerprint dictionary are extracted based on the interference-associated dictionary, including: constructing an interference projection operator based on the interference-associated dictionary and the residual weight matrix; extracting non-compressor interference components from the circulation residual according to the interference projection operator to obtain the interference-extracted residual; and extracting fingerprint components that can be represented by the interference-associated dictionary from the dual-domain degenerate fingerprint dictionary according to the interference projection operator to obtain the interference-extracted degenerate dictionary.

[0012] In one possible implementation, the degenerate primitive coherence coefficient is determined based on the weighted inner product between any two degenerate primitive fingerprints in the interference stripping degeneracy dictionary. The weights of the weighted inner product are determined by the residual weight matrix. The degenerate primitive coherence coefficient is used to indicate the similarity relationship between any two degenerate primitive fingerprints. The group sparse constraint inversion includes: applying coherence suppression constraints to simultaneously activated degenerate primitives based on the interference stripping degeneracy dictionary and the degenerate primitive coherence coefficient, and solving the compressor degenerate variables in combination with the degenerate variable boundary constraints.

[0013] In one possible implementation, after obtaining the compressor degradation variables, the method further includes: dividing the steady-state segment and the natural excitation segment into multiple operating condition subsets; performing group sparse constraint inversion based on some operating condition subsets to obtain candidate degradation variables, and projecting the candidate degradation variables onto the operating condition subsets that did not participate in the group sparse constraint inversion; selecting credible degradation variables from the candidate degradation variables based on the residual reduction after projection and the occurrence frequency of the candidate degradation variables, and using the credible degradation variables as inputs for stripping degradation components.

[0014] In one possible implementation, the degradation components are stripped from the reliable degradation variables and the performance loss contribution is output. This includes: determining the performance of the gas turbine in the degradation state based on the reliable degradation variables; stripping the flow capacity degradation component, isentropic efficiency degradation component, extraction leakage anomaly component, and adjustable guide vane equivalent deviation component from the reliable degradation variables to obtain the post-degradation gas turbine performance of each component; determining the performance loss contribution corresponding to each degradation component based on the performance of the gas turbine in the degradation state and the post-degradation gas turbine performance of each component, and forming a performance evaluation result based on the performance loss contribution.

[0015] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows: By generating a health benchmark characteristic surface based on health samples, the gas turbine operating data can serve as a unified health reference under the same equivalent operating condition variables, enabling steady-state segments and natural excitation segments to participate in subsequent residual generation under the same benchmark.

[0016] By generating steady-state thermodynamic residuals from steady-state segments and phased response kernel residuals from natural excitation segments, steady-state deviations and dynamic response deviations can be simultaneously incorporated into the dual-domain observation residuals, reducing the degradation identification bias caused by relying solely on steady-state data.

[0017] By performing thermodynamic closed-loop decomposition on the dual-domain observation residuals, the circulation residuals can be extracted from the thermodynamic cycle relationship, making the measurement point offset and local state deviations distinguishable from the cyclic closed-loop residuals.

[0018] By generating a dual-domain degradation fingerprint dictionary and an interference-associated dictionary using compressor degradation primitives and non-compressor interference primitives respectively, and then stripping the interference-associated dictionary, the impact of fuel calorific value deviation, measurement point deviation, intake pressure loss deviation, and exhaust back pressure deviation on compressor degradation inversion can be reduced.

[0019] By using the interference-stripped degradation dictionary and degradation primitive coherence coefficients in group sparse constraint inversion, the simultaneous activation of similar degradation primitives can be limited. By stripping the degradation components of the compressor degradation variables, the performance loss contribution corresponding to each degradation component can be output, giving the performance evaluation results a decomposable technical basis. Attached Figure Description

[0020] Figure 1 This is a schematic flowchart of the method of the present invention; Figure 2 This is a graph showing the generation of gas turbine operating data and health baseline characteristics in an embodiment of the present invention; Figure 3 This is a diagram showing the generation of dual-domain observation residuals in an embodiment of the present invention; Figure 4 These are the thermal closure residual diagram and the circulation residual decomposition diagram in the embodiments of the present invention; Figure 5 This is a graph showing the contribution of interference stripping and performance loss in an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of one or more embodiments. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this invention.

[0022] In this embodiment, the compressor degradation inversion uses gas turbine operating data and health samples as processing inputs. The gas turbine operating data comes from the gas turbine control system, performance monitoring system, or historical operating database, and includes at least the compressor inlet state, compressor outlet state, adjustable guide vane opening, load state, inlet pressure loss, and exhaust back pressure. The health samples are used to generate a health baseline characteristic surface. The gas turbine performance evaluation equipment determines steady-state segments and natural excitation segments based on the gas turbine operating data. Steady-state thermodynamic residuals are generated from the steady-state segments, and phased response kernel residuals are generated from the natural excitation segments, forming a dual-domain observation residual. The dual-domain observation residual is decomposed into a circulation residual through thermodynamic closed-loop decomposition. The circulation residual, dual-domain degradation fingerprint dictionary, and interference companion dictionary enter the interference stripping process to obtain the interference stripping residual and the interference stripping degradation dictionary. The gas turbine performance evaluation equipment performs group sparse constraint inversion on the interference stripping residual based on the interference stripping degradation dictionary and the degradation primitive coherence coefficient to obtain the compressor degradation variables, and provides a calculation basis for subsequent stripping degradation components and output performance loss contributions.

[0023] like Figure 1 As shown, a gas turbine performance evaluation method based on compressor degradation inversion includes the following processing steps.

[0024] When acquiring gas turbine operating data, the gas turbine performance evaluation equipment reads the unit's operating records from the gas turbine control system, performance monitoring system, or historical operating database. The gas turbine operating data must at least cover the compressor inlet status, compressor outlet status, adjustable guide vane opening, load status, fuel consumption status, intake pressure loss, and exhaust back pressure.

[0025] Health samples are taken from the operating records of the gas turbine after cleaning, maintenance, or when it has passed performance acceptance. The gas turbine performance evaluation equipment uses the health samples to generate a health baseline characteristic surface, which can characterize the health state variables under different equivalent operating conditions.

[0026] After generating the health baseline characteristic surface, the gas turbine performance evaluation equipment identifies steady-state segments and naturally excited segments from the gas turbine operating data. The steady-state segment is determined by continuous sampling records where the changes in load state, adjustable guide vane opening, compressor inlet state, and compressor outlet state within the same time window all meet the steady-state determination criteria. These criteria are determined based on the fluctuation range of the corresponding variables in the health samples and the sampling accuracy of the measurement points. The naturally excited segment is determined by continuous sampling records where the adjustable guide vane opening, load state, or fuel control records show continuous changes, and the compressor inlet state, compressor outlet state, turbine exhaust state, or power generation state have synchronous responses. The steady-state segment enters the steady-state thermodynamic residual generation stage, while the naturally excited segment enters the phased response kernel residual generation stage, ensuring that subsequent dual-domain observation residuals simultaneously possess both steady-state deviation information and dynamic response information.

[0027] In one embodiment, when the gas turbine performance evaluation equipment generates a health baseline characteristic surface based on a health sample, it extracts equivalent operating condition variables for each operating record in the health sample. These equivalent operating condition variables describe the operating condition position of the gas turbine operating record and include at least corrected speed, adjustable guide vane opening, load condition, intake pressure loss, and exhaust back pressure.

[0028] The corrected rotational speed is used to eliminate the influence of compressor inlet temperature changes on the rotational speed coordinate, and can be calculated using the following formula:

[0029] in, To correct the rotational speed, This refers to the actual speed of the generator unit. As the reference inlet temperature, The compressor inlet temperature is used. The corrected rotational speed is used as the operating condition input for the health baseline characteristic surface, so that health samples under different ambient temperatures can fall into a unified operating condition coordinate system.

[0030] The adjustable guide vane opening is obtained from the adjustable guide vane feedback value in the gas turbine operating data, with the feedback value being preferred over the command value. If the feedback value is missing, it can be corrected using the command value and actuator verification record within the same time window. The load condition is determined by the relationship between the generating power and the rated power; the intake pressure loss is determined by the intake side pressure loss record; and the exhaust back pressure is determined by the exhaust side pressure record or the calculated exhaust back pressure value.

[0031] The gas turbine performance evaluation equipment simultaneously extracts health state variables from the healthy samples. These health state variables characterize the operating state of the gas turbine under equivalent operating condition variables when it is in a healthy state, and include at least the healthy pressure ratio, healthy compressor outlet temperature, healthy equivalent airflow, healthy isentropic efficiency, and healthy gas turbine performance parameters. The healthy pressure ratio is determined by the ratio of the compressor outlet pressure to the compressor inlet pressure; the healthy compressor outlet temperature is obtained from the compressor outlet temperature measuring point; and the healthy equivalent airflow can be obtained from the airflow measuring point or calculated from the mass flow rate closed-loop relationship.

[0032] The healthy isentropic efficiency reflects the energy conversion state of the compressor compression process in a healthy sample and can be calculated using the following formula:

[0033] in, For healthy isentropic efficiency, This refers to the compressor inlet temperature. This refers to the compressor outlet temperature. For health stress ratio, The air adiabatic index is used. The healthy isentropic efficiency is incorporated as a health state variable into the health baseline characteristic surface, and is subsequently used to generate the steady-state thermodynamic residual and the dual-domain degradation fingerprint dictionary.

[0034] After the equivalent operating condition variables and health state variables are paired, the gas turbine performance evaluation equipment establishes operating condition coordinates according to the equivalent operating condition variables and maps the health state variables onto the operating condition coordinates to form candidate characteristic surfaces. Candidate characteristic surfaces can be established using spline fitting, piecewise interpolation, or constrained regression. The specific fitting method can be determined based on the number of health samples, the coverage of operating conditions, and the completeness of measurement points.

[0035] To avoid local fluctuations in the healthy baseline characteristic surface that do not conform to the operating characteristics of the gas turbine, the gas turbine performance evaluation equipment introduces monotonic constraints and efficiency range constraints when establishing the characteristic surface. Monotonic constraints are used to limit the trends of healthy pressure ratio and healthy equivalent airflow with the corrected speed at the same adjustable guide vane opening, and also to limit the trends of healthy equivalent airflow with the adjustable guide vane opening at the same corrected speed. Efficiency range constraints are used to limit the healthy isentropic efficiency to within the normal operating range of the compressor, avoiding unreasonable extrapolation of the characteristic surface due to abnormal healthy samples.

[0036] When an operating record in the healthy sample violates the monotonic constraint or efficiency range constraint, the gas turbine performance evaluation equipment marks that operating record as an abnormal healthy sample and excludes it from the final fitting of the healthy baseline characteristic surface. After eliminating abnormal healthy samples, the gas turbine performance evaluation equipment re-establishes the characteristic surface using the remaining healthy samples and determines the characteristic surface that satisfies the monotonic constraint and efficiency range constraint as the healthy baseline characteristic surface.

[0037] After the health baseline characteristic surface is generated, it is aligned with the steady-state and natural excitation segments in the gas turbine operating data. The steady-state segment queries the corresponding health state variables in the health baseline characteristic surface to generate the steady-state thermodynamic residual. The natural excitation segment determines the healthy natural excitation segment under the same equivalent operating condition variables based on the health baseline characteristic surface and generates a healthy response kernel, which is used to generate the phased response kernel residual. Thus, the health baseline characteristic surface serves as a unified health reference for the subsequent generation of dual-domain observation residuals.

[0038] After receiving the steady-state segment, the natural excitation segment, and the health reference characteristic surface, the gas turbine performance evaluation equipment queries the health state variables under the same equivalent operating condition variables for the steady-state segment, and calculates the difference between the actual operating state in the steady-state segment and the health state variables to generate the steady-state thermodynamic residual.

[0039] For the natural excitation segment, the gas turbine performance evaluation equipment extracts the phased response relationship between the control disturbance and the operating response, and generates the phased response kernel residual.

[0040] Steady-state thermodynamic residuals are used to reflect thermodynamic deviations under stable operating conditions, while phased response kernel residuals are used to reflect dynamic response deviations during control disturbances. The gas turbine performance evaluation equipment stores the steady-state thermodynamic residuals and phased response kernel residuals in association with the segment number, equivalent operating condition variables, and excitation phase axis to form dual-domain observation residuals, and then transfers the dual-domain observation residuals to the thermodynamic closed-loop decomposition stage.

[0041] In one embodiment, the gas turbine performance evaluation equipment extracts a control disturbance sequence and an operating response sequence from the natural excitation segment. The control disturbance sequence consists of continuously changing control quantities within the natural excitation segment, which can be taken from adjustable guide vane opening, load conditions, and fuel control records involved in gas turbine regulation. The operating response sequence consists of operating states that change with the control disturbance within the same natural excitation segment, which can be taken from compressor inlet state, compressor outlet state, turbine exhaust state, and power generation state.

[0042] Before the natural excitation fragment is included in the calculation, the gas turbine performance evaluation equipment performs time alignment on the control disturbance sequence and the operating response sequence. Time alignment is achieved using the sampling timestamps in the gas turbine operating data. When the sampling periods of the control disturbance sequence and the operating response sequence are different, the gas turbine performance evaluation equipment interpolates the lower-sampled sequence according to the higher sampling period, or extracts the higher-sampled sequence according to the lower sampling period. The sampling processing method is determined based on the sampling stability of the gas turbine operating data to ensure that the control disturbance sequence and the operating response sequence are comparable at the same time point.

[0043] The gas turbine performance evaluation equipment generates an excitation phase axis based on the cumulative disturbance of the control disturbance sequence. The excitation phase axis represents the progress of the control disturbance from start to finish within a natural excitation segment, rather than directly using time length to represent the response progress. For natural excitation segments with different durations but similar control disturbance patterns, the excitation phase axis enables the operating response sequence to align along the disturbance progress.

[0044] The excitation phase axis can be generated by the following formula:

[0045] in, For the first The excitation phase value at each sampling location, This is the sampling location number. The total number of sampling locations in the natural excitation segment. To control the number of control variables in the disturbance sequence. For the first The control quantity in the first The change within each sampling interval For the first The normalization scale for each control variable. The normalization scale is determined based on the normal adjustment amplitude of the corresponding control variable in the healthy sample, and is used to eliminate the influence of the difference in dimensions of different control variables on the cumulative disturbance.

[0046] After obtaining the excitation phase axis, the gas turbine performance evaluation equipment resamples the control disturbance sequence and operating response sequence to the excitation phase axis. During resampling, the excitation phase value is between zero and one, where zero represents the start position of the natural excitation segment and one represents the end position of the natural excitation segment. The gas turbine performance evaluation equipment reads the control disturbance sequence and operating response sequence at fixed phase intervals, ensuring that each natural excitation segment generates control disturbance data and operating response data at the same number of phase positions.

[0047] The gas turbine performance evaluation equipment generates a phased response kernel based on the resampled control disturbance sequence and operating response sequence. The phased response kernel describes the transmission relationship of the control disturbance sequence to the operating response sequence along the excitation phase axis. The phased response kernel can be obtained through least-squares fitting or robust regression. When short-term measurement point jumps exist in the natural excitation segment, the gas turbine performance evaluation equipment reduces the influence of the jump sampling position in the fitting, ensuring that the phased response kernel primarily reflects the changes in operating response caused by continuous control disturbances.

[0048] The gas turbine performance evaluation equipment also determines healthy natural excitation segments from the healthy samples. Healthy natural excitation segments need to have similar equivalent operating condition variables and similar control disturbance amplitudes to the current natural excitation segment. The allowable range of difference in equivalent operating condition variables is determined based on the operating condition distribution density and measurement point sampling accuracy of the healthy samples, to avoid the healthy response kernel being generated from samples with excessively large operating condition differences. The gas turbine performance evaluation equipment processes the healthy natural excitation segments using the same excitation phase axis generation method and generates a healthy response kernel based on the processed healthy natural excitation segments.

[0049] The phased response kernel residual is determined by the difference between the phased response kernel and the healthy response kernel, and can be generated by the following formula:

[0050] in, For the excitation phase value The phased response kernel residual at the point, For the excitation phase value Phased response kernel at the location, For the excitation phase value The health response kernel is then formed. After the phased response kernel residual is formed, it is written together with the steady-state thermodynamic residual into the same segment of the dual-domain observation residual, so that the subsequent thermodynamic closed circulation decomposition can utilize both steady-state deviation and dynamic response deviation.

[0051] After receiving the dual-domain observation residuals, the gas turbine performance evaluation equipment writes them into the corresponding thermodynamic closed residual diagram of the gas turbine's thermodynamic cycle. The thermodynamic closed residual diagram is used to represent the energy, pressure, temperature, and power transfer relationships between the compressor inlet state, compressor outlet state, combustion chamber heat input state, turbine exhaust state, and power generation state.

[0052] After the dual-domain observation residuals are incorporated into the thermodynamic closed-loop residual diagram, the gas turbine performance evaluation equipment converts them into edge residuals on each connection relationship. The portion of the edge residuals that can be explained by a single state node offset is separated into gradient residuals, and the remaining portion is further decomposed according to the thermodynamic closed-loop relationship.

[0053] After decomposition, the gas turbine performance evaluation equipment identifies the residuals that cannot be explained by the gradient residuals and are retained along the thermodynamic closed loop as circulating residuals. These circulating residuals then enter the subsequent disturbance-related dictionary stripping stage to reduce the impact of measurement point offsets on the inversion of compressor degradation variables.

[0054] In one embodiment, when the gas turbine performance evaluation equipment establishes a thermodynamic closed residual diagram, it sets the compressor inlet state, compressor outlet state, combustion chamber heat input state, turbine exhaust state, and power generation state as nodes. The compressor inlet state consists of the compressor inlet pressure, compressor inlet temperature, and equivalent airflow; the compressor outlet state consists of the compressor outlet pressure, compressor outlet temperature, and pressure ratio; the combustion chamber heat input state consists of the fuel consumption state and gas turbine performance parameters; the turbine exhaust state consists of the exhaust temperature, exhaust pressure, and exhaust back pressure; and the power generation state consists of the load state and power generation.

[0055] The edges in the thermal closed residual graph represent the thermodynamic relationships between nodes. The edge between the compressor inlet state and the compressor outlet state carries the pressure ratio residual, compressor outlet temperature residual, and equivalent airflow residual. The edge between the compressor outlet state and the combustion chamber heat input state carries the fuel heat input residual. The edge between the combustion chamber heat input state and the turbine exhaust state carries the exhaust temperature residual and exhaust pressure residual. The edge between the turbine exhaust state and the power generation state carries the power balance residual. The edge from the power generation state back to the compressor inlet state carries the environmental correction residual and the mass flow closure residual. Through the above node and edge settings, the dual-domain observation residual can be mapped to the edge residuals in the thermal closed residual graph.

[0056] After the edge residuals are formed, the gas turbine performance evaluation equipment separates the gradient residuals based on the correlation between nodes and edges. The correlation between nodes and edges indicates that when the state of a node shifts, it will simultaneously affect the edge residuals connected to that node. The gas turbine performance evaluation equipment establishes a node-edge correlation matrix based on the direction of the edge residuals and the node connection relationships, and uses the node-edge correlation matrix to estimate the node shift. The gradient residuals can be expressed by the following formula:

[0057] in, For gradient residuals, The node-edge affinity matrix, Node offset. The node-edge correlation matrix is ​​determined by the connectivity between nodes and edges in the thermal closed residual graph, and the node offset is obtained by fitting the edge residual. Gradient residuals are used to represent residual components that can be explained by local offsets of compressor inlet state, compressor outlet state, combustion chamber heat input state, turbine exhaust state, or power generation state.

[0058] After gradient residual separation, the gas turbine performance evaluation equipment subtracts the gradient residual from the edge residuals to form the remaining edge residuals. The remaining edge residuals are no longer preferentially interpreted as measurement point offsets at individual nodes, but are decomposed according to the thermodynamic closed-loop relationship. The thermodynamic closed-loop relationship consists of a closed transmission chain formed by compressor compression, combustion chamber heat input, turbine expansion, power generation output, and environmental correction. The gas turbine performance evaluation equipment establishes a loop correlation matrix based on the closed transmission chain and determines the residual components retained along the closed transmission chain in the remaining edge residuals.

[0059] The circulation residual can be expressed by the following formula:

[0060] in, For circulation residuals, The loop incidence matrix, The loop residuals are the values ​​of the loop correlation matrix, which is determined by the thermodynamic closed loop relationship. The loop residuals are obtained by fitting the residuals of the remaining edges. The circulation residuals are used to represent residual components that cannot be explained by a single node offset but persist in the gas turbine thermodynamic closed loop relationship.

[0061] When a dual-domain observation residual cannot be completely mapped to the edge residual in the thermodynamic closed residual diagram, the gas turbine performance evaluation equipment marks this dual-domain observation residual as a residual to be completed and performs local mapping according to available nodes and edges. The residual to be completed does not participate in the loop residual fitting but is retained in the residual record of the steady-state segment or the natural excitation segment. After completing the thermodynamic closed circulation decomposition, the circulation residual is transferred to the interference-accompanied dictionary stripping stage, and the gradient residual is saved as reference data for the measurement point deviation analysis.

[0062] After receiving the circulating residual, health baseline characteristic surface, steady-state segment, and natural excitation segment, the gas turbine performance evaluation equipment generates a dual-domain degradation fingerprint dictionary according to the compressor degradation primitive and an interference companion dictionary according to the non-compressor interference primitive.

[0063] The dual-domain degradation fingerprint dictionary describes the response patterns of compressor degradation primitives in steady-state thermodynamic residuals and phased response kernel residuals. The interference adjoint dictionary describes the response patterns of non-compressor interference primitives in the same observation domain.

[0064] After generating the two dictionaries, the gas turbine performance evaluation equipment uses the interference adjoint dictionary to construct an interference projection operator, and simultaneously processes the circulation residual and dual-domain degradation fingerprint dictionary based on the interference projection operator to generate the interference stripping residual and interference stripping degradation dictionary. The interference stripping residual and interference stripping degradation dictionary then proceed to the subsequent sparse constraint inversion stage.

[0065] In one embodiment, when the gas turbine performance evaluation equipment generates a dual-domain degradation fingerprint dictionary from the compressor degradation primitive, it first divides the compression process according to the compressor normalized isentropic enthalpy rise coordinate. The compressor normalized isentropic enthalpy rise coordinate is used to represent the relative position of the compression process from the inlet to the outlet. The closer the value is to the inlet side, the closer it is to the front compression region; the closer the value is to the outlet side, the closer it is to the rear compression region.

[0066] The gas turbine performance evaluation equipment divides the compression process into multiple degradation base regions based on the compressor's normalized isentropic enthalpy rise coordinates. The number of degradation base regions is determined by the coverage of healthy samples, the number of compressor measurement points, and the degradation locations that can be reliably distinguished. The number of degradation base regions should not be too small, otherwise the locational differences in flow capacity degradation, isentropic efficiency degradation, and extraction leakage anomalies will be difficult to represent; conversely, the number of degradation base regions should not be too large, otherwise excessively high similarity relationships may arise between degradation primitives in the dual-domain degradation fingerprint dictionary.

[0067] Within each degradation base region, the gas turbine performance evaluation equipment sets up a flow capacity degradation base unit, an isentropic efficiency degradation base unit, and an extraction leakage anomaly base unit. The flow capacity degradation base unit represents the change in healthy equivalent airflow caused by a reduction in compressor flow capacity. The isentropic efficiency degradation base unit represents the deviation of healthy isentropic efficiency caused by changes in the energy conversion state during compression. The extraction leakage anomaly base unit represents the deviation of mass flow closure caused by flow losses in the extraction branch, sealing parts, or interstages. The adjustable guide vane equivalent deviation base unit is not repeatedly set according to the degradation base regions, but is instead included in the dual-domain degradation fingerprint dictionary as a global base unit affecting the relationship between the adjustable guide vane opening and the compressor response.

[0068] Under the same equivalent operating conditions and natural excitation, the gas turbine performance evaluation equipment applies degradation perturbations corresponding to each compressor degradation element to the healthy baseline characteristic surface. The amplitude of the degradation perturbation is determined based on the noise level at the measurement points in the healthy sample and the historical degradation amplitude. The amplitude needs to be greater than the random fluctuations in the healthy sample, but it cannot cause the healthy baseline characteristic surface to enter the nonlinear extrapolation range. For the flow capacity degradation element, the degradation perturbation manifests as a decrease in the healthy equivalent airflow; for the isentropic efficiency degradation element, the degradation perturbation manifests as a decrease in the healthy isentropic efficiency; for the extraction leakage anomaly element, the degradation perturbation manifests as a deviation in the mass flow rate closure relationship; and for the adjustable guide vane equivalent deviation element, the degradation perturbation manifests as a shift in the compressor response relationship corresponding to the adjustable guide vane opening.

[0069] After each compressor degradation unit is subjected to a degradation perturbation, the gas turbine performance evaluation equipment recalculates the steady-state thermodynamic residuals under the same equivalent operating conditions, obtaining a steady-state residual fingerprint; and recalculates the phased response kernel residuals under the same natural excitation conditions, obtaining a phased response kernel residual fingerprint. The steady-state residual fingerprint represents the thermodynamic deviation pattern caused by the compressor degradation unit under stable operating conditions, while the phased response kernel residual fingerprint represents the dynamic response deviation pattern caused by the compressor degradation unit during control perturbation.

[0070] The gas turbine performance evaluation equipment correlates the steady-state residual fingerprint and the phased response kernel residual fingerprint corresponding to the same compressor degradation primitive according to the degradation primitive identifier, degradation base region, and equivalent operating condition variables, forming a dual-domain degradation fingerprint. Each dual-domain degradation fingerprint retains the correspondence between the steady-state residual fingerprint and the phased response kernel residual fingerprint, enabling subsequent sparse constraint inversion to utilize both steady-state deviation and dynamic response deviation. Multiple dual-domain degradation fingerprints are arranged in the order of degradation primitives to generate a dual-domain degradation fingerprint dictionary.

[0071] In one embodiment, when the gas turbine performance evaluation equipment generates an interference adjoint dictionary from non-compressor interference elements, it divides the non-compressor interference elements into fuel calorific value deviation elements, measurement point deviation elements, intake pressure loss deviation elements, and exhaust back pressure deviation elements. The non-compressor interference elements are not used to characterize compressor degradation, but rather to characterize external interferences or measurement offsets that may exhibit similar deviation patterns in the circulation residuals.

[0072] The fuel calorific value deviation element represents the difference between the recorded lower heating value of the fuel and the actual fuel heat input. The measurement point deviation element represents the measurement offset in compressor inlet conditions, compressor outlet conditions, turbine exhaust conditions, or power generation conditions. The intake pressure loss deviation element represents the impact of changes in intake system pressure loss on compressor inlet conditions and equivalent airflow. The exhaust back pressure deviation element represents the impact of changes in exhaust system back pressure on turbine exhaust conditions and power generation conditions.

[0073] Under the same equivalent operating conditions and the same natural excitation, the gas turbine performance evaluation equipment applies disturbances corresponding to each non-compressor disturbance element to the healthy baseline characteristic surface. The amplitude of the disturbance is determined based on the allowable error of the measurement points in the healthy sample, the accuracy of fuel calorific value recording, the historical fluctuation range of intake pressure loss, and the historical fluctuation range of exhaust back pressure. The principle for setting the disturbance is to cover common non-compressor disturbances without representing obvious compressor degradation errors as non-compressor disturbances.

[0074] After each non-compressor disturbance primitive is subjected to a disturbance, the gas turbine performance evaluation equipment generates a corresponding steady-state residual fingerprint and a phased response kernel residual fingerprint. For the fuel calorific value deviation primitive, the steady-state residual fingerprint mainly reflects the deviation between the combustion chamber heat input state and the power generation state. For the measurement point deviation primitive, the steady-state residual fingerprint mainly reflects the edge residuals connected to the node where the measurement point is located. For the inlet pressure loss deviation primitive, the steady-state residual fingerprint mainly reflects the edge residuals from the compressor inlet state to the compressor outlet state. For the exhaust back pressure deviation primitive, the steady-state residual fingerprint mainly reflects the edge residuals from the turbine exhaust state to the power generation state. The phased response kernel residual fingerprint is used to record the response pattern of the above disturbances in the natural excitation segment.

[0075] The gas turbine performance evaluation equipment associates the steady-state residual fingerprint and the phased response kernel residual fingerprint corresponding to the same non-compressor disturbance primitive as a disturbance fingerprint. It then arranges these disturbance fingerprints according to the type of non-compressor disturbance primitive and equivalent operating condition variables to generate a disturbance adjoint dictionary. This dictionary serves as the basis for the disturbance projection operator, used to remove residual components that can be explained by non-compressor disturbance primitives before compressor degradation variable inversion.

[0076] In one embodiment, when the gas turbine performance evaluation equipment strips the circulating residuals and dual-domain degradation fingerprint dictionary based on the interference-associated dictionary, it first constructs an interference projection operator based on the interference-associated dictionary and the residual weight matrix. The residual weight matrix is ​​used to represent the reliability of different residual components in the stripping process. The residual weight matrix can be determined according to the measurement point accuracy, the residual fluctuation range in healthy samples, and the sampling integrity of natural excitation segments; the higher the measurement point accuracy, the smaller the residual fluctuation, and the more complete the sampling, the higher the weight of the corresponding residual component in the residual weight matrix.

[0077] The interference projection operator can be constructed using the following formula:

[0078] in, To interfere with the projection operator, To provide an accompanying dictionary for interference, To interfere with the transposition of the accompanying dictionary, The residual weight matrix is... For regularization parameters, It is the identity matrix. The regularization parameter is determined based on the condition number of the interference adjoint dictionary and the noise level in the healthy samples, and is used to avoid unstable projections when the interference fingerprints in the interference adjoint dictionary are similar.

[0079] The gas turbine performance evaluation equipment uses an interference projection operator to process the circulation residuals, obtaining the non-compressor interference components that can be represented by the interference adjoint dictionary. The non-compressor interference components are then subtracted from the circulation residuals to obtain the interference stripping residuals. The interference stripping residuals can be calculated using the following formula:

[0080] in, To interfere with the stripping residual, It is the identity matrix. To interfere with the projection operator, The circulation residuals are represented by the disturbance stripping residuals. These residuals are used to retain circulation residual components that cannot be explained by non-compressor disturbance primitives and serve as observational inputs for subsequent sparse constraint inversions.

[0081] The gas turbine performance evaluation equipment uses the same interference projection operator to process the dual-domain degenerate fingerprint dictionary. It then extracts fingerprint components that can be represented by the interference adjoint dictionary from the dual-domain degenerate fingerprint dictionary, obtaining the interference-extracted degenerate dictionary. The interference-extracted degenerate dictionary can be calculated using the following formula:

[0082] in, To prevent the degeneration of the dictionary from being stripped away, It is the identity matrix. To interfere with the projection operator, This is a dual-domain degenerate fingerprint dictionary. The interference stripping degenerate dictionary is used to retain the degenerate fingerprints that have been distinguished from non-compressor interference primitives, and together with the interference stripping residuals, it enters the group sparse constraint inversion stage.

[0083] When a disturbance fingerprint in the disturbance adjoint dictionary is similar to the dual-domain degradation fingerprints corresponding to multiple compressor degradation primitives, the gas turbine performance evaluation equipment does not directly delete the corresponding compressor degradation primitive. Instead, it simultaneously processes the circulation residual and the dual-domain degradation fingerprint dictionary through the disturbance projection operator, ensuring that the disturbance stripping residual and the disturbance stripping degradation dictionary are located in the same stripping space. Therefore, the observation input and degradation dictionary used in subsequent sparse constraint inversions have a consistent coordinate relationship, avoiding inversion bias caused by stripping only the circulation residual without stripping the dual-domain degradation fingerprint dictionary.

[0084] After receiving the interference stripping residual and the interference stripping degradation dictionary, the gas turbine performance evaluation equipment calculates the coherence coefficient of the degradation primitives and uses the coherence coefficient of the degradation primitives to restrict similar degradation primitives from participating in the interpretation of the same interference stripping residual at the same time.

[0085] After completing the calculation of the coherence coefficients of the degradation primitives, the gas turbine performance evaluation equipment uses the interference stripping residual as the observation input, the interference stripping degradation dictionary as the inversion dictionary, and performs group sparse constraint inversion based on the coherence coefficients of the degradation primitives to obtain the compressor degradation variables.

[0086] The compressor degradation variable is used to represent the activation status and degradation magnitude of each degradation primitive in the current gas turbine operating data. The compressor degradation variable enters the operating condition subset verification stage, and after projection and screening of candidate degradation variables, a credible degradation variable is formed. The credible degradation variable serves as the input for stripping the degradation components.

[0087] In one embodiment, the gas turbine performance evaluation device calculates the coherence coefficients of degraded primitives based on an interference stripping degradation dictionary. Each column of the degradation primitive fingerprint in the interference stripping degradation dictionary corresponds to a compressor degradation primitive, and the degradation primitive fingerprint includes the steady-state residual fingerprint and the phased response kernel residual fingerprint of the corresponding compressor degradation primitive in the interference stripping space. The gas turbine performance evaluation device takes any two degradation primitive fingerprints and calculates a weighted inner product according to the residual weight matrix to characterize the similarity relationship between the two degradation primitive fingerprints under the current observation conditions.

[0088] The coherence coefficient of the degenerate primitive can be determined by the following formula:

[0089] In the formula, For the first The degenerate primitive fingerprint and the first The degenerate primitive coherence coefficients between degenerate primitive fingerprints To interfere with the stripping of the degenerate dictionary, the first... A degenerate primitive fingerprint, To interfere with the stripping of the degenerate dictionary, the first... A degenerate primitive fingerprint, The residual weight matrix is... For the first Transposition of a degenerate primitive fingerprint For the first The transpose of a degenerate primitive fingerprint. The closer the degenerate primitive coherence coefficient is to one, the more similar the two degenerate primitive fingerprints are; the closer the degenerate primitive coherence coefficient is to zero, the easier it is to distinguish the two degenerate primitive fingerprints.

[0090] When performing sparse constraint inversion using gas turbine performance evaluation equipment, the interference stripping residual is used as the observation input, the interference stripping degradation dictionary is used as the inversion dictionary, and the compressor degradation variables are used as the variables to be solved. The compressor degradation variables are grouped according to the degradation primitive type: flow capacity degradation primitives form a flow capacity degradation variable group, isentropic efficiency degradation primitives form an isentropic efficiency degradation variable group, extraction leakage anomaly primitives form an extraction leakage anomaly variable group, and adjustable guide vane equivalent deviation primitives form an adjustable guide vane equivalent deviation variable group. This grouping method is used to limit the concentrated activation of inversion results in a few degradation types, avoiding the simultaneous generation of scattered responses from multiple irrelevant degradation primitives.

[0091] The sparse constraint inversion can be solved by the following formula:

[0092] In the formula, For compressor degradation variables, To interfere with the stripping residual, To prevent the degeneration of the dictionary from being stripped away, The residual weight matrix is... For the group of sparse constraint coefficients, This is the index of the degenerate variable group. For the first A set of degenerate variables, This is the coherence suppression constraint coefficient. The coherence coefficient of the degenerate primitive. For the first The degenerate variables corresponding to each degenerate primitive. For the first The degradation variables correspond to each degradation primitive. The sparse constraint coefficients are determined based on the number of major degradation types in historical degradation samples and are used to limit the simultaneous activation of irrelevant degradation variable groups. The coherence suppression constraint coefficients are determined based on the distribution of the coherence coefficients of degradation primitives in the interference stripping degradation dictionary and are used to suppress the simultaneous activation of two degradation primitives when their fingerprints are highly similar.

[0093] When solving for compressor degradation variables, the gas turbine performance evaluation equipment also incorporates boundary constraints for these degradation variables. These boundary constraints are determined based on the compressor's safe operating range, the fluctuation range of health samples, the degradation magnitude in maintenance records, and the inversion stability requirements of the gas turbine performance evaluation equipment. Flow capacity degradation variables, isentropic efficiency degradation variables, extraction leakage anomaly variables, and adjustable guide vane equivalent deviation variables each have corresponding upper and lower limits. These boundary constraints are used to prevent the inversion results from exceeding the compressor's physically feasible range. The solution results that satisfy the group sparsity constraints, coherence suppression constraints, and degradation variable boundary constraints are identified as compressor degradation variables and transferred to the operating condition subset verification stage.

[0094] In one embodiment, after obtaining the compressor degradation variables, the gas turbine performance evaluation equipment divides the steady-state segment and the naturally excited segment into multiple operating condition subsets. These operating condition subsets are formed according to equivalent operating condition variables, with the division criteria including load state, adjustable guide vane opening, corrected speed, inlet pressure loss, and exhaust back pressure. During division, steady-state segments and naturally excited segments within the same operating condition subset have similar equivalent operating condition variables, while different operating condition subsets maintain sufficient differences in operating conditions to distinguish the gas turbine's operating state. The number of operating condition subsets is determined based on the number of steady-state segments, the number of naturally excited segments, and the coverage of equivalent operating condition variables, ensuring that each operating condition subset has residual samples that can be used for sparse constraint inversion.

[0095] The gas turbine performance evaluation equipment selects a subset of operating conditions from multiple operating condition subsets for grouped sparse constraint inversion to obtain candidate degenerate variables. These candidate degenerate variables are not directly used as the final result; instead, they are projected onto operating condition subsets that did not participate in the grouped sparse constraint inversion. During projection, the gas turbine performance evaluation equipment calls the disturbance stripping degradation dictionary corresponding to the operating condition subsets that did not participate in the grouped sparse constraint inversion, substitutes the candidate degenerate variables into this disturbance stripping degradation dictionary, generates prediction residuals, and compares the prediction residuals with the disturbance stripping residuals of that operating condition subset.

[0096] The decrease in residual after projection can be determined by the following formula:

[0097] In the formula, This represents the decrease in residual after projection. For the disturbance stripping residuals in the subset of working cases that did not participate in the sparse constraint inversion, For the interference stripping degenerate dictionary in the subset of working cases that did not participate in the group sparse constraint inversion These are candidate degenerate variables. When the decrease in the projected residual is positive and meets the residual decrease criterion, it indicates that the candidate degenerate variable can explain the disturbance stripping residuals in the subset of operating conditions that did not participate in the group sparse constraint inversion.

[0098] The gas turbine performance evaluation equipment repeatedly selects different subsets of operating conditions for grouped sparse constraint inversion and counts the occurrence frequency of each candidate degradation variable. The occurrence frequency is used to represent the stability of the same candidate degradation variable obtained by repeatedly solving it in different combinations of operating condition subsets. The occurrence frequency determination criterion is determined based on the total number of operating condition subsets and the number of available samples; when the number of operating condition subsets is large, the occurrence frequency determination criterion can be increased, and when the number of operating condition subsets is small, the occurrence frequency determination criterion can be decreased to avoid the omission of credible degradation variables due to insufficient samples.

[0099] The gas turbine performance evaluation equipment filters credible degradation variables based on the reduction in residuals after projection and the frequency of occurrence of candidate degradation variables. A candidate degradation variable is determined to be credible if it simultaneously meets both the residual reduction and frequency criteria. A candidate degradation variable is not included in the credible degradation variable list if it only appears in a single operating condition subset, or if the residual reduction after projection onto an operating condition subset not participating in the group sparse constraint inversion is insufficient. After filtering, the credible degradation variables are passed to the degradation component stripping stage to determine the performance loss contribution corresponding to each degradation component.

[0100] After receiving the credible degradation variables, the gas turbine performance evaluation equipment inputs the credible degradation variables into the healthy baseline characteristic surface and the gas turbine performance calculation relationship to determine the gas turbine performance in the degradation state.

[0101] Degraded state gas turbine performance is used to represent the gas turbine performance state under current compressor degradation conditions, serving as a benchmark for the calculation of each degradation component.

[0102] The gas turbine performance evaluation equipment separates the flow capacity degradation component, isentropic efficiency degradation component, extraction leakage anomaly component, and adjustable guide vane equivalent deviation component from the credible degradation variables to obtain the gas turbine performance after the removal of each degradation component.

[0103] The gas turbine performance evaluation equipment calculates the difference between the performance of the gas turbine in the degraded state and the performance of the gas turbine after the separation of each degraded component, determines the performance loss contribution corresponding to each degraded component, and forms the performance evaluation result based on the performance loss contribution.

[0104] In one embodiment, when the gas turbine performance evaluation equipment determines the performance of a degraded gas turbine based on credible degradation variables, it writes each degradation component in the credible degradation variables into the equivalent operating condition variable position corresponding to the health baseline characteristic surface. The credible degradation variables originate from the previous stage of operating condition subset verification and have retained the type and amplitude information of the flow capacity degradation component, isentropic efficiency degradation component, extraction leakage anomaly component, and adjustable guide vane equivalent deviation component. The gas turbine performance evaluation equipment determines the compressor degradation primitive that needs to be applied based on the type information and determines the amount of change of the degradation primitive on the health state variables based on the amplitude information.

[0105] The flow capacity degradation component affects the healthy equivalent airflow rate, causing it to change according to the corresponding degradation amplitude. The isentropic efficiency degradation component affects the healthy isentropic efficiency, causing it to change according to the corresponding degradation amplitude. The extraction leakage anomaly component affects the mass flow rate closure relationship, causing a shift in the flow closure between the compressor inlet state, compressor outlet state, and combustion chamber heat input state. The adjustable guide vane equivalent deviation component affects the mapping relationship between the adjustable guide vane opening and the compressor response, causing corresponding changes in the healthy pressure ratio, healthy equivalent airflow rate, and healthy compressor outlet temperature at the same adjustable guide vane opening.

[0106] After each degradation component is written, the gas turbine performance evaluation equipment recalculates the degradation pressure ratio, degradation compressor outlet temperature, degradation equivalent airflow, degradation isentropic efficiency, and degradation gas turbine performance parameters under the same equivalent operating conditions. The calculated gas turbine performance state is then defined as the degradation state gas turbine performance. The degradation state gas turbine performance can include power generation, fuel consumption, heat consumption, exhaust temperature, and compressor operating margin. The specific selection method for the gas turbine performance parameters is determined based on the reliably obtainable measurement points in the gas turbine operating data and the gas turbine performance evaluation task.

[0107] When stripping degradation components, the gas turbine performance evaluation equipment maintains the equivalent operating condition variables, healthy baseline characteristic surfaces, and unstripped degradation components unchanged, removing only one degradation component to be stripped from the credible degradation variables. For the flow capacity degradation component, the equipment sets its amplitude to zero and retains the isentropic efficiency degradation component, the extraction leakage anomaly component, and the adjustable guide vane equivalent deviation component, calculating the gas turbine performance after stripping the flow capacity degradation component. For the isentropic efficiency degradation component, the extraction leakage anomaly component, and the adjustable guide vane equivalent deviation component, the equipment performs the stripping calculation separately using the same method, ensuring that each stripped gas turbine performance reflects only the change in gas turbine performance after the removal of a single degradation component.

[0108] The performance loss contribution can be determined by the following formula:

[0109] in, For the first The performance loss contribution corresponding to each degradation component This is the sequence number of the degenerate component. For the first Afterburner performance after the stripping of individual degradation components The performance loss contribution represents the change in engine performance relative to the degraded state engine performance after a certain degradation component is removed from the credible degradation variable.

[0110] When the gas turbine performance parameter is power generation, the performance loss contribution is the difference between the power generation after stripping and the power generation in the degraded state. When the gas turbine performance parameter is fuel consumption state or heat consumption state, the performance loss contribution is the difference between the degraded fuel consumption state or degraded heat consumption state and the stripped fuel consumption state or degraded heat consumption state. The gas turbine performance evaluation equipment determines the direction of the difference according to the physical meaning of different gas turbine performance parameters, so that the performance loss contribution maintains a consistent direction of degradation loss.

[0111] Before generating performance evaluation results, the gas turbine performance evaluation equipment performs consistency verification on the gas turbine performance after stripping each degradation component. If, after stripping a degradation component, the gas turbine performance after stripping does not conform to the direction of change of the circulation residual obtained from the thermodynamic closed-loop decomposition, the gas turbine performance evaluation equipment marks the performance loss contribution corresponding to that degradation component as a contribution to be reviewed, and retains the type and amplitude information of that degradation component. If the gas turbine performance after stripping conforms to the direction of change of the circulation residual, the gas turbine performance evaluation equipment writes the performance loss contribution corresponding to that degradation component into the performance evaluation results.

[0112] The performance evaluation results should at least record the performance loss contributions corresponding to the flow capacity degradation component, the isentropic efficiency degradation component, the extraction leakage anomaly component, and the equivalent deviation component of the adjustable guide vanes. The performance evaluation results may also record the gas turbine performance in the degraded state, the gas turbine performance after stripping each degradation component, the type information of the credible degradation variables, the amplitude information of the credible degradation variables, and the contribution markers to be verified. After the performance evaluation results are generated, the process proceeds to the gas turbine operation and maintenance analysis or gas turbine performance evaluation report generation stage.

[0113] This embodiment uses a GT9 gas turbine from a gas-fired power plant as the object to be processed. The gas turbine performance evaluation equipment reads the gas turbine's operating data from the performance monitoring system and selects healthy samples within 72 hours after water washing from the historical operating database. The sampling period for the gas turbine's operating data is 10 seconds, and the number of healthy samples is 4320.

[0114] like Figure 2 As shown, the gas turbine operating data includes compressor inlet status, compressor outlet status, adjustable guide vane opening, load status, intake pressure loss, and exhaust back pressure. Figure 2 The batch GT9-2026-05 represents the gas turbine operating data batch of this embodiment. Both the steady-state segment S-214 and the natural excitation segment E-072 are obtained by segmenting the gas turbine operating data. The steady-state segment S-214 is used to generate the steady-state thermodynamic residual, and the natural excitation segment E-072 is used to generate the phased response kernel residual.

[0115] The healthy sample includes healthy pressure ratio, healthy compressor outlet temperature, healthy equivalent airflow, healthy isentropic efficiency, and healthy gas turbine performance parameters. Figure 2 Sample H-084 is a healthy sample generated from the healthy baseline characteristic surface. The corrected speed of sample H-084 is 0.983, the adjustable guide vane opening is 46.5°, the load condition is 82.7%, the intake pressure loss is 1.21 kPa, the exhaust back pressure is 3.7 kPa, and the healthy pressure ratio is 13.55.

[0116] Gas turbine performance evaluation equipment determines equivalent operating condition variables and health state variables based on healthy samples. Equivalent operating condition variables include corrected engine speed, adjustable guide vane opening, load condition, inlet pressure loss, and exhaust back pressure; health state variables include healthy pressure ratio, healthy compressor outlet temperature, healthy equivalent airflow, healthy isentropic efficiency, and healthy gas turbine performance parameters. The gas turbine performance evaluation equipment establishes characteristic surfaces based on the equivalent operating condition variables and health state variables, and performs monotonicity constraints and efficiency range constraints checks on these characteristic surfaces. Characteristic surfaces that satisfy both monotonicity constraints and efficiency range constraints are determined as healthy baseline characteristic surfaces.

[0117] The corrected rotational speed for steady-state segment S-214 is 0.982, the adjustable guide vane opening is 46.2°, and the load condition is 82.4%. The gas turbine performance evaluation equipment queries the health state variables under the same equivalent operating conditions as steady-state segment S-214 in the health baseline characteristic surface, and calculates the difference between the actual operating state and the health state variables in steady-state segment S-214 to obtain the steady-state thermodynamic residual.

[0118] The adjustable guide vane opening of the naturally excited segment E-072 changes by 1.6°, the load condition changes by 2.1%, and the time window is 13:10-13:18. The gas turbine performance evaluation equipment extracts the control disturbance sequence and operating response sequence from the naturally excited segment E-072, generates the excitation phase axis based on the cumulative disturbance of the control disturbance sequence, and resamples the control disturbance sequence and operating response sequence to the excitation phase axis.

[0119] like Figure 3 As shown, the excitation phase axis ranges from 0 to 1, with the horizontal axis representing the excitation phase axis and the vertical axis representing the response amplitude. The phased response kernel corresponding to the natural excitation segment E-072 has amplitudes of 0.18, 0.39, 0.51, 0.56, and 0.58 at excitation phase values ​​of 0.2, 0.4, 0.6, 0.8, and 1.0, respectively; the healthy response kernel corresponding to the healthy natural excitation segment HE-018 has amplitudes of 0.22, 0.47, 0.62, 0.67, and 0.69 at the same excitation phase values.

[0120] The gas turbine performance evaluation equipment generates phased response kernel residuals based on the difference between the phased response kernel and the healthy response kernel. These residuals include a pressure ratio residual of +0.031, a compressor outlet temperature residual of +0.024, an equivalent airflow residual of -0.027, a phase response gain residual of -0.110, and a phase recovery area residual of -0.083. These residuals are then written into the dual-domain observation residuals, corresponding to the segment S-214+E-072, and enter the thermodynamic closed-loop decomposition stage.

[0121] The gas turbine performance evaluation equipment maps the dual-domain observation residuals to a thermally closed residual diagram. The thermally closed residual diagram includes the compressor inlet state, compressor outlet state, combustion chamber heat input state, turbine exhaust state, and power generation state.

[0122] like Figure 4 As shown, the compressor inlet conditions include P1 of 100.8 kPa and T1 of 302.6 K, the compressor outlet conditions include P2 of 1362 kPa and T2 of 642.8 K, the combustion chamber heat input conditions include fuel heat input of 560 MW, the turbine exhaust conditions include exhaust temperature of 842 K and exhaust back pressure of 3.8 kPa, and the power generation conditions include power of 226.4 MW and load of 82.4%.

[0123] Figure 4 The edge residuals e1 to e5 are obtained by mapping the dual-domain observation residuals. Edge residual e1 is the pressure ratio residual +0.031, edge residual e2 is the heat input residual -0.018, edge residual e3 is the exhaust temperature residual +0.024, edge residual e4 is the power residual -0.026, and edge residual e5 is the mass closure residual +0.019. The gas turbine performance evaluation equipment separates the gradient residual from the edge residuals based on the correlation between nodes and edges; the gradient residual is 0.021. After separating the gradient residual, the gas turbine performance evaluation equipment determines the circulating residual based on the thermodynamic closed-loop relationship; the circulating residual is 0.064, and sends the circulating residual to the interference stripping stage.

[0124] The gas turbine performance evaluation equipment divides the compression process into four degradation base regions according to the compressor's normalized isentropic enthalpy rise coordinates. Within each degradation base region, it sets up degradation elements for flow capacity, isentropic efficiency, and extraction leakage anomalies, along with an equivalent deviation element for adjustable guide vanes, thus forming 13 compressor degradation elements. Under the same equivalent operating condition variables and the same natural excitation conditions, the gas turbine performance evaluation equipment applies degradation perturbations corresponding to each compressor degradation element to the healthy baseline characteristic surface, generating steady-state residual fingerprints and phased response kernel residual fingerprints. These steady-state residual fingerprints and phased response kernel residual fingerprints are then correlated into a dual-domain degradation fingerprint, generating a dual-domain degradation fingerprint dictionary.

[0125] The gas turbine performance evaluation equipment divides non-compressor interference elements into fuel calorific value deviation elements, measuring point deviation elements, inlet pressure loss deviation elements, and exhaust back pressure deviation elements. The interference disturbances for the fuel calorific value deviation elements are set based on the accuracy of the fuel calorific value records; the interference disturbances for the measuring point deviation elements are set based on the allowable error of the measuring points; and the interference disturbances for the inlet pressure loss deviation elements and exhaust back pressure deviation elements are set based on historical fluctuation ranges. After generating steady-state residual fingerprints and phased response kernel residual fingerprints for each non-compressor interference element, the gas turbine performance evaluation equipment correlates these fingerprints with interference fingerprints and generates an interference adjoint dictionary.

[0126] like Figure 5 As shown, the circulation residual is 0.064, the dual-domain degradation fingerprint dictionary contains 13 degradation primitives, and the interference adjoint dictionary contains 4 types of interference primitives. The gas turbine performance evaluation equipment constructs an interference projection operator based on the interference adjoint dictionary and the residual weight matrix. The residual weight matrix is ​​determined according to the measurement point accuracy, the residual fluctuation range of healthy samples, and the sampling integrity of natural excitation fragments. Figure 5 The regularization parameter γ is 0.02, used to limit projection instability caused by similar interference fingerprints in the interference-associated dictionary. The circulation residual of 0.064 is processed by the interference projection operator to obtain the interference stripping residual of 0.041; the dual-domain degenerate fingerprint dictionary is processed by the same interference projection operator to obtain the interference stripping degenerate dictionary.

[0127] The gas turbine performance evaluation equipment calculates the coherence coefficient of the degradation primitives based on the interference stripping degradation dictionary. Figure 5 The degradation element coherence coefficient matrix shows that the degradation element coherence coefficient between the flow capacity degradation element A2 and the isentropic efficiency degradation element E2 is 0.72, the degradation element coherence coefficient between the isentropic efficiency degradation element E2 and the adjustable guide vane equivalent deviation element V is 0.52, and the degradation element coherence coefficient between the extraction leakage anomaly element L3 and the adjustable guide vane equivalent deviation element V is 0.28. The gas turbine performance evaluation equipment uses the interference stripping residual as the observation input and the interference stripping degradation dictionary as the inversion dictionary. Based on the degradation element coherence coefficient, it applies coherence suppression constraints to simultaneously activated degradation elements and performs group sparse constraint inversion by combining degradation variable boundary constraints.

[0128] In this embodiment, the candidate degraded variables obtained from the group sparse constraint inversion include a flow capacity degradation component of 0.032, an isentropic efficiency degradation component of 0.047, an extraction leakage anomaly component of 0.009, and an adjustable guide vane equivalent deviation component of -0.6°. The gas turbine performance evaluation equipment divides the steady-state segment and the natural excitation segment into 5 operating condition subsets, uses 4 of these subsets for group sparse constraint inversion, and projects the candidate degraded variables onto the operating condition subsets that did not participate in the group sparse constraint inversion. In this embodiment, the residual reduction criterion is set to be not less than 0.20, and the occurrence count criterion is set to be not less than 3 times in 5 combinations. The candidate degraded variable appears 4 times in 5 combinations, and the residual reduction after projection is 0.42, which satisfies the occurrence count criterion and the residual reduction criterion, and is therefore determined as a credible degraded variable.

[0129] The gas turbine performance evaluation equipment determines the performance of a gas turbine in a degraded state based on reliable degradation variables. The power generation capacity of the gas turbine in the degraded state is 226.4 MW. The gas turbine performance evaluation equipment separates the flow capacity degradation component, isentropic efficiency degradation component, extraction leakage anomaly component, and adjustable guide vane equivalent deviation component from the reliable degradation variables to obtain the gas turbine performance after removing each degradation component.

[0130] After removing the flow capacity degradation component, the power generation is 230.2 MW, with a performance loss contribution of 3.8 MW. After removing the isentropic efficiency degradation component, the power generation is 232.0 MW, with a performance loss contribution of 5.6 MW. After removing the abnormal pumping leakage component, the power generation is 227.8 MW, with a performance loss contribution of 1.4 MW. After removing the equivalent deviation component of the adjustable guide vanes, the power generation is 227.3 MW, with a performance loss contribution of 0.9 MW.

[0131] Figure 5The performance loss contribution bars in the data show that the performance loss contribution corresponding to the isentropic efficiency degradation component is 5.6 MW, the performance loss contribution corresponding to the flow capacity degradation component is 3.8 MW, the performance loss contribution corresponding to the extraction leakage anomaly component is 1.4 MW, and the performance loss contribution corresponding to the adjustable guide vane equivalent deviation component is 0.9 MW. Based on these performance loss contributions, the gas turbine performance evaluation equipment generates performance evaluation results. The performance evaluation results record the gas turbine performance in the degraded state, the gas turbine performance after stripping each degradation component, the performance loss contribution corresponding to each degradation component, the type information of the reliable degradation variable, and the amplitude information of the reliable degradation variable. The performance evaluation results are then transmitted to the gas turbine operation and maintenance analysis stage.

[0132] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0133] The various embodiments in this invention are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A gas turbine performance evaluation method based on compressor degradation inversion, characterized in that, The method includes: Acquire gas turbine operating data and health samples, generate a health baseline characteristic surface based on the health samples, and determine steady-state segments and natural excitation segments based on the gas turbine operating data; Steady-state thermodynamic residuals are generated from the steady-state segments, and phased response kernel residuals are generated from the natural excitation segments, forming dual-domain observation residuals; Thermodynamic closed-loop decomposition is performed on the dual-domain observation residuals to obtain the circulation residuals; A dual-domain degradation fingerprint dictionary and an interference adjunct dictionary are generated from the compressor degradation primitive and the non-compressor interference primitive, respectively. The circulation residual and the dual-domain degradation fingerprint dictionary are then stripped based on the interference adjunct dictionary to obtain the interference stripping residual and the interference stripping degradation dictionary. Based on the interference stripping degradation dictionary, the degradation primitive coherence coefficients are determined, and based on the interference stripping degradation dictionary and the degradation primitive coherence coefficients, the interference stripping residuals are subjected to group sparse constraint inversion to obtain the compressor degradation variables. The degradation component of the compressor degradation variable is removed, and the performance loss contribution is output.

2. The method according to claim 1, characterized in that, The generation of the health baseline characteristic surface includes: determining equivalent operating condition variables and health state variables based on the health sample. The equivalent operating condition variables include corrected speed, adjustable guide vane opening, load state, intake pressure loss and exhaust back pressure. The health state variables include healthy pressure ratio, healthy compressor outlet temperature, healthy equivalent air flow rate, healthy isentropic efficiency and healthy gas turbine performance parameters. Based on the equivalent operating condition variables and the health state variables, a characteristic surface that satisfies monotonic constraints and efficiency range constraints is established, and the characteristic surface is determined as the health benchmark characteristic surface.

3. The method according to claim 1, characterized in that, The generation of the phased response kernel residual includes: Extract the control perturbation sequence and the running response sequence from the natural excitation fragment, and generate the excitation phase axis based on the cumulative perturbation amount of the control perturbation sequence; The control disturbance sequence and the running response sequence are resampled to the excitation phase axis, and a phased response kernel is generated based on the resampled control disturbance sequence and running response sequence. Determine healthy natural stimulus fragments from the healthy samples, and generate a healthy response kernel based on the healthy natural stimulus fragments; The phased response kernel residual is generated based on the difference between the phased response kernel and the healthy response kernel.

4. The method according to claim 1, characterized in that, The thermal closed-loop decomposition of the dual-domain observation residuals includes: A thermodynamic closed residual diagram is established based on the compressor inlet state, compressor outlet state, combustion chamber heat input state, turbine exhaust state, and power generation state. The dual-domain observation residual is mapped to the edge residual in the thermal closed residual graph; The gradient residual is separated from the edge residual based on the association between nodes and edges, and the circulation residual is determined from the separated edge residual based on the thermal closed loop relationship.

5. The method according to claim 2, characterized in that, The generation of the dual-domain degradation fingerprint dictionary from the compressor degradation primitive includes: The compression process is divided into multiple degradation base regions according to the normalized isentropic enthalpy rise coordinate of the compressor. In each degradation base region, a flow capacity degradation base, an isentropic efficiency degradation base, and an extraction leakage anomaly base are set, and an adjustable guide vane equivalent deviation base is also set. Under the same equivalent operating condition variables and the same natural excitation conditions, the degradation perturbation corresponding to each of the compressor degradation primitives is applied to the health reference characteristic surface respectively, generating the steady-state residual fingerprint and phased response kernel residual fingerprint corresponding to each of the compressor degradation primitives; The steady-state residual fingerprint and the phased response kernel residual fingerprint are associated as a dual-domain degenerate fingerprint, and a dual-domain degenerate fingerprint dictionary is generated based on each of the dual-domain degenerate fingerprints.

6. The method according to claim 2, characterized in that, The generation of the interference accompaniment dictionary from the non-compressor interference primitives includes: The non-compressor interference element is divided into fuel calorific value deviation element, measuring point deviation element, intake pressure loss deviation element and exhaust back pressure deviation element; Under the same equivalent operating condition variables and the same natural excitation conditions, the disturbance perturbation corresponding to each of the non-compressor disturbance primitives is applied to the health reference characteristic surface, and the steady-state residual fingerprint and phased response kernel residual fingerprint corresponding to each non-compressor disturbance primitive are generated. The steady-state residual fingerprint and the phased response kernel residual fingerprint are associated as interference fingerprints, and the interference adjoint dictionary is generated based on each interference fingerprint.

7. The method according to claim 1, characterized in that, The process of stripping the circulating residual and the dual-domain degenerate fingerprint dictionary based on the interference-associated dictionary includes: An interference projection operator is constructed based on the aforementioned interference adjoint dictionary and the residual weight matrix used to represent the confidence level of different residual components; The interference stripping residual is obtained by stripping the non-compressor interference component from the circulation residual using the interference projection operator. The interference stripping degenerate dictionary is obtained by stripping fingerprint components that can be represented by the interference adjoint dictionary from the dual-domain degenerate fingerprint dictionary using the interference projection operator.

8. The method according to claim 7, characterized in that, The degenerate primitive coherence coefficient is determined based on the weighted inner product between any two degenerate primitive fingerprints in the interference stripping degenerate dictionary. The weights of the weighted inner product are determined by the residual weight matrix. The degenerate primitive coherence coefficient is used to indicate the similarity relationship between any two degenerate primitive fingerprints. The set of sparse constraint inversions includes: Based on the interference stripping degradation dictionary and the degradation primitive coherence coefficient, coherence suppression constraints are applied to the simultaneously activated degradation primitives, and the compressor degradation variables are solved by combining degradation variable boundary constraints.

9. The method according to claim 8, characterized in that, After obtaining the compressor degradation variable, the method further includes: The steady-state segment and the natural excitation segment are divided into multiple operating condition subsets; Based on a subset of working conditions, group sparse constraint inversion is performed to obtain candidate degenerate variables, and the candidate degenerate variables are projected onto a subset of working conditions that did not participate in the group sparse constraint inversion. Based on the decrease in residual after projection and the frequency of occurrence of the candidate degenerate variables, credible degenerate variables are selected from the candidate degenerate variables, and the credible degenerate variables are used as inputs for stripping degenerate components.

10. The method according to claim 9, characterized in that, The step of stripping degradation components from the trusted degradation variables and outputting the performance loss contribution includes: The performance of the gas turbine in its degraded state is determined based on the aforementioned reliable degradation variables; The flow capacity degradation component, isentropic efficiency degradation component, extraction leakage anomaly component, and adjustable guide vane equivalent deviation component are respectively removed from the credible degradation variables to obtain the post-combustion engine performance after the degradation components are removed. Based on the performance of the gas turbine in the degraded state and the performance of the gas turbine after stripping each degraded component, the performance loss contribution corresponding to each degraded component is determined, and a performance evaluation result is formed based on the performance loss contribution.

Images