Improved performance model matching, expansion and prediction

Abstract

A simulation method is disclosed for simulating the operation of a gas turbine (111). The method includes a global search procedure and an iterative local search procedure for calculating parameters for simulating the operation of the gas turbine (111). The output parameters can also be used to monitor the operation and plan maintenance of the gas turbine (111). Also disclosed is a characterization system for characterizing and simulating the operation of the gas turbine (111).

Description

[Technical Field]

[0001] The present disclosure relates to improved performance model matching, expansion and prediction for simulating and predicting the operation of engines, such as gas turbines, for improved engine monitoring. [Background technology]

[0002] As is known, a gas turbine (also known as a "digital twin") is a rotating machine suitable for converting chemical energy into mechanical energy. It is a machine that is usually used to generate electrical energy or to drive a compressor.

[0003] Generally, a gas turbine comprises a combustion chamber provided with nozzles for injecting the fuel to be burned, the fuel being intended to be burned in the combustion chamber, after which the hot exhaust gases leave the combustion chamber and drive an impeller attached to a shaft, thus providing the mechanical work that is used as required, as mentioned above.

[0004] Modern gas turbines are very complex machines, and therefore, in order to control their operation and optimize their consumption, they are equipped with many sensors that are configured to detect and collect data related to their operation. These data are then collected to enable the realization of a dashboard for the operator to check the operation of the gas turbine in real time. In addition, modern gas turbines are also equipped with processing systems intended to process the data collected by the sensors to realize further processing and optimize the operation of the gas turbine. Summary of the Invention [Problem to be solved by the invention]

[0005] However, in many cases, controlling and monitoring the operation of the system is complex and therefore not always accurate. Also, due to the complexity of the engine, some operating parameters cannot be measured and therefore cannot be properly controlled. Therefore, it is not always possible to perform a constant and accurate diagnosis of a gas turbine, or of an engine in general. This involves approximations in planning maintenance and predicting possible failures, which leads to an increase in the overall operating expenditure (OPEX) for managing the engine.

[0006] Therefore, a method for simulating and predicting the behavior of several operating parameters of a gas turbine would be highly welcome in the art. More generally, it is desirable to provide a method, and an associated system for implementing it, for predicting engine operation with high accuracy and associating patterns of values ​​of correlated operating parameters. [Means for solving the problem]

[0007] The present disclosure relates to performance characterization of gas turbines using detailed physical models that can be used to predict at least unmeasured (or unmeasurable) parameters, track engine performance, and predict engine behavior under hypothetical / what if scenarios.

[0008] The present invention uses data from an automation system that can obtain operational data from on-site monitoring infrastructure (gas turbines). The sensor data is matched to an optimized physics-based model using a novel solver, a program for processing the data. The solver performs local and global searches to find model parameters that best resemble the in-situ sensor measurements. The model parameters are then used to obtain synthetic parameters, which are either unmeasured quantities or simulations under other conditions. These synthetic parameters can then be used to track engine performance (e.g., ISO power) or can be digital redundancy for synthetic / virtual or physical sensors.

[0009] Specifically, the solution encompasses a bundle of model-based methodologies for characterizing and monitoring the performance of field gas turbines. Because the methodologies are based on high-fidelity models, they can also be leveraged to predict gas turbine performance behavior under different environmental and operating conditions. The disclosed method and system objectives can further be offered as a service to customers interested in automated / digital gas turbine performance monitoring, advisory, and simulation.

[0010] Therefore, improved performance model matching, extension and prediction as defined in claim 1 form a particular object of the present invention.

[0011] Preferred embodiments are defined in the dependent claims. [Brief explanation of the drawings]

[0012] A complete understanding of the disclosed embodiments of the present invention and many of the attendant advantages thereof will be readily obtained as the same becomes better understood by reference to the following detailed description when considered in connection with the accompanying drawings. [Figure 1] 1 is a block diagram of a gas turbine operation simulation system according to a first embodiment. [Figure 2] 3 shows a flowchart of a global search procedure of the simulation method according to the first embodiment. [Figure 3] 4 shows a more detailed flowchart of the global search procedure of the simulation method according to the first embodiment. [Figure 4] 3 shows a flowchart of a local search procedure of the simulation method according to the first embodiment. [Figure 5] 1 illustrates a first implementation of the simulation method of the present disclosure. [Figure 6] 1 illustrates a second implementation of the simulation method of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0013] In the various figures, similar parts are designated by the same reference numerals.

[0014] Gas turbines are complex systems, the operation and control of which is technically challenging, taking into account the necessary optimizations required, for example, to reduce pollution, as well as to plan appropriate maintenance that reduces any risk of damage or malfunction. The operation of a gas turbine can be characterized by a large number of physical parameters that must be constantly sensed and monitored in order to properly control and evaluate the operation of the gas turbine. According to one aspect, the present subject matter is directed to an algorithm capable of simulating with high accuracy the operation of a gas turbine, characterized by a large set of said parameters, the evolution and changes of which over time can be determined via a self-updating model, and adapted to (and also predicted) the behavior of the gas turbine so as to plan any possible service or maintenance, as well as to better drive the operation of the gas turbine.

[0015] The present solution therefore relates to a method intended to obtain and use both real and synthetic parameters applicable to simulating different operating conditions of a complex system such as a gas turbine or an engine in general in order to optimize its maintenance and operation.

[0016] Referring now to the drawings, FIG. 1 shows a block diagram of an overall characterization system 1 for characterizing and simulating the operation of a gas turbine (or any other complex machine or engine), which can ideally be divided into two main parts: an infrastructure section 11 and a processing unit U, which comprises an automated performance characterization section 12 and a delivery services section 13.

[0017] The infrastructure section 11 comprises a gas turbine 111 to be controlled, which comprises sensors for detecting operating parameters, the number of which is denoted M, such as the temperature in different parts of the gas turbine 111, the pressure of the gas or exhaust gases, the rotational speed of the rotor, the temperature of the combustor, the pressure of the compressor, the temperature of the compressor, etc.

[0018] The term "parameter" generally contemplates pressure, temperature, or any other measurable parameter that can be detected by sensors included in the gas turbine 111, as well as derived variables and data that cannot be measured directly. Parameters can be obtained over different time frames, such as daily, weekly, or monthly. Each set of parameters is sensed at a specific time to form a record (which can be visualized using a vector). Typically, the number of parameters and data sensed from a gas turbine can range from 30 to 50, although other embodiments or unrelated gas turbines can sense different numbers of parameters and data.

[0019] Considering data over a predefined time interval, N records are collected and NxM records are collected as input.

[0020] The infrastructure section 11 also includes a data recording unit 112 that is hardwired to the sensors of the gas turbine 111. The data recording unit 112 collects sampled and analog-to-digital converted sensor data and signals. Examples of collected data are efficiency, flow capacity and emission coefficients.

[0021] In some embodiments, the data recording unit 112 may be a computer or cloud computing system or mainframe capable of performing calculations and storing data and running software.

[0022] The data recording unit 112 receives data from sensors installed on the gas turbine 111 that is transmitted via the infrastructure.

[0023] The processing unit U may be a computer or a cloud computing system or a mainframe, and may be the same computer or processing means as the data recording unit 112.

[0024] The automated performance characterization section 12 of the processing unit U comprises several computer-implemented modules that implement the automated data processing module 121. In particular, due to connection interruptions, sensor failures, or other unpredictable circumstances, the data may have gaps, delays, and generally non-ideal behavior that may or may not be detected from existing control systems or data infrastructures. The characterization system 1 may execute programs based on methods for automated data processing that have the ability to address remaining data corruption, i.e., outlier detection, filtering, data imputation, resampling, etc., to ensure that the data is suitable for analysis. This data preprocessing is performed by the automated data processing module 121.

[0025] The automated performance characterization section 12 processes the data received or pre-processed by the automated data processing module 121 by means of a physics-based model 122. The module implements a performance characterization method that uses the physics-based model to generate output parameters 123 that characterize the performance of the gas turbine 111.

[0026] Specifically, after undergoing the above-mentioned pre-processing steps performed by the automatic data processing module 121, the data is fed to a physics-based model 122, which uses as input the field data and the output of a physics-based solver model for the gas turbine 111 performance. A processing method processes parameters that characterize the operation of the gas turbine 111. In some embodiments, the method generally compares the field measurement data M and the modeled data, which comparison is used to obtain parameters S for simulating the gas turbine 111, the parameters S being generally represented by the output parameters in step 123.

[0027] In another embodiment of the method implemented in physics-based model 122, the method also adjusts the model parameters H to provide an output that is consistent with the field measurement data, in which case the adjusted model parameters H are considered part of the parameters for subsequent iterations.

[0028] Once the gas turbine 111 has been characterized with parameters obtained and possibly represented in the output parameters 123, these parameters are used to calculate the simulated parameters S as model parameters H mAs diagrammed by the delivery services section 13, which includes additional functional modules that process through the gas turbine 111, the engine performance at different conditions can be input into a model that simulates the engine performance at different conditions, representing the gas turbine 111. In particular, one possible application for the simulation is the calculation of correction parameters (module 131) that represent the performance of the gas turbine 111. In particular, for the gas turbine 111, the ambient and operating conditions significantly affect the performance of the engine, so that in order to detect degradation (loss of power, fuel increase, etc.), the engine power should be corrected to a set of standard conditions (i.e., ISO conditions) so that the corrected performance parameters, as shown in the evaluation performance module 131, can be compared over time.

[0029] Another application is to calculate unmeasured (or non-measurable) parameters (module 133), such as, in the case of a gas turbine 111, power output and / or other parameters that are not directly measured but can be inferred from a simulation model. In this case, the output parameters 123 can be used in conjunction with the model to simulate the unmeasured (measurable) parameters.

[0030] Another application is to track the trending and monitoring of module parameters such as shown in module 134 to track the performance of specific components of the gas turbine 111. The applications mentioned are examples of services that can be provided to monitor, trend and provide expert advice in case of degradation or poor engine performance.

[0031] The characterization in this embodiment is so complete that it can be performed for a fleet of gas turbines 111 that also require fleet optimization.

[0032] The automatic data processing module 121 and physics-based model 122 of the gas turbine 111 are also implemented on a computer in the form of a software program. The output parameters 123 can also be plotted and their statistics evaluated and used to monitor and control the gas turbine 111.

[0033] As mentioned above, data collected from the gas turbine 111 is sensed and processed to provide a simulation of the gas turbine 111 (or engine in general) itself. This is performed by a solver implemented in the physics-based model 122.

[0034] In the following Figures 2, 3 and 4, a global search procedure 2 and a local search procedure 3 are disclosed, and then output parameters of the operation of the gas turbine 111 can be simulated and monitored to derive data of the engine 111 itself and / or unmeasured or immeasurable parameters.

[0035] In this disclosure, batches of data are processed together and the search for a good initial solution for all batches is done only once by a global search procedure 2, and then a local search procedure 3 refines the initial solution for all points (in this case referred to as a local search). For continuous and smooth functions, finding a good starting point is very useful because if the points are close enough and the Jacobian has already been calculated around this starting point, then the local search can be guided without recalculating the Jacobian (i.e., the Jacobian is kept constant), as will be better disclosed below.

[0036] 2, a general flowchart illustrating the operation of the global search procedure 2 of the solver method according to the present disclosure is shown. Specifically, the flowchart shown in FIG. 2 illustrates steps of the solver method that may form the basis of the problem solved by the solver method. In particular, a flowchart of an optimization problem is shown that describes how to make the simulation as close as possible to the actual system (engine or gas turbine) by changing input parameters to the simulation until measured parameters taken from the gas turbine 111 match the simulated parameters.

[0037] The solver method underlying the global search procedure 2 solves an optimization problem where the solution changes slowly over time while minimizing function evaluations. The solution is also adapted to black-box models that do not have clear or explicit formulas (e.g., in the case of cycle decks or models obtained through machine learning techniques).

[0038] Additionally, in the gas turbine 111, the degradation process is typically not a fast process, i.e., degradation of the gas turbine operation takes several weeks. Therefore, when data is processed within a temporal locality (i.e., one week, one day, one month), it is expected that the solutions for each individual record will tend to have similar solutions overall, even when the amount of data is large (i.e., on the order of thousands of records). When a single point (individualized by the value of the record) goes through the optimization process, it is usually possible to start from an initial solution, which may even be far from the actual solution.

[0039] Generally speaking, the methods of the present disclosure are particularly useful when functional assessment is expensive.

[0040] In particular, the solver synthesizes a fictitious record, called a "representative record," constructed from the first old data received for the model. The solver then solves the optimization problem for this representative record at a solution close to all solutions in the dataset. This is reported in Figure 2 as the "representative solution," along with the Jacobian computed around the solution.

[0041] The goal of the solution is to find a set of model parameters that, when input to the model, will make the output of the model as close as possible to measurements taken on the actual gas turbine 111. Typically, this is achieved by solving the model equations or by iterating the parameters in the model until the required convergence is reached.

[0042] The records can be of several types and generally include ambient and operating conditions of the gas turbine 111, which are the so-called "record inputs" or "degrees of freedom" of the model, i.e., the independent variables of the system of equations, which are denoted below by R. These inputs, along with model parameters H of the gas turbine 111, produce system output parameters or dependent variables, which are denoted below by S. In the model, system health (correct operation) is encoded via the model parameters H. The model parameters H may include, by way of example only, efficiency, flow capacity, emission coefficient, etc.

[0043] As generally described above, the general operation of the solver model on which the system is based, i.e., the global search procedure 2, ideally has two parts or main phases. In the first part, all record inputs are input to the model using default model parameters (which may be from design inputs, from previous model iterations, from engineering knowledge, etc.). These record inputs, along with the model, generate simulated output parameters S. The simulated output parameters S are compared with the actual system output parameters S to generate a set of residuals E for all input records R, as better explained below.

[0044] From these input parameter records R and residuals E, a representative record is generated as follows: A "representative input" is generated by averaging the inputs from all records. This is because the parameters R * is not the only way to obtain a "representative input" set of E. Other methods include median or robust averaging techniques such as trimmed mean, winsorized average, weighted average, among others. Similarly, a "representative residual" E * can also be obtained by averaging the residuals E (or using other methods already described). The representative output is calculated as follows:

[0045] Preliminarily, a representative input to the model, R *, and the default model parameters to obtain the output P are the "representative residuals" E * This affects the output P, ​​and the "representative output" P * : is obtained. P * =P+E * (1)

[0046] "Representative record" here means the representative input R * and representative output P * It consists of:

[0047] In this embodiment, only one representative record is synthesized, but it is not limited to only one representative record.

[0048] In other embodiments, the composition of representative records can be extended to generate more than one representative record. This can be done when the solutions are not expected to be close to each other. In this case, the records may be separated using heuristics or clustering techniques such as k-means, and a representative record is obtained for each subpopulation. In these cases, global search procedure 2 and local search procedure 3 can be performed for each representative record and subpopulation.

[0049] 2, which will be commented and analyzed in detail. In step 21, a record containing data from the gas turbine 111 is received. In step 23, only input parameters R that are measured from the gas turbine 111 are selected and read, i.e., their conditions with respect to the parameter's default values ​​are checked. These data are then processed in step 24 by a model function f, which constitutes a model of the system for determining simulated values ​​of the operation of the gas turbine 111. The model function f has as arguments the input measurement parameters R and the model parameters H. At the same time, in step 22, input parameters S that are measured but not necessary for defining the operation and simulation of the gas turbine 111 are selected and read.

[0050] The output of processing step 24 is a set of simulated outputs S' of the gas turbine 111, which are received in step 25 and then compared or differentiated against the actual input parameters S read in step 22, as can be seen in step 26. Then, after comparison step 26, these differences E, called residuals, and the variations of each parameter derived by the comparison between the simulated output parameters S' and the measured output parameters S detected by sensors of the gas turbine 111, are adjusted in solver step 27 to determine the deviation or offset of each simulated parameter with respect to the actual parameter. At this point, these differences are fed back to the model parameters in step 28 to adjust processing step 24. In other words, based on the difference matrix E obtained in processing step 26, updated model parameters are read to feed into the processing of step 24. In addition, the obtained model parameters are reported in reporting step 29 as a solution to the optimization problem with an acceptable comparison between the actual data obtained by sensors of the gas turbine 111 and the simulation data from model simulation step 25, as described above.

[0051] As noted above, Figure 2 is only a broad representation of the data flow. In subsequent figures, the data processing operations are expanded upon to provide further details about the actual processing of the sets of data and parameters obtained from the gas turbine 111 and simulated by the solver.

[0052] Referring now to Figure 3, further processing details are provided. In particular, it can be seen that the input record described above is structured by an NxM matrix, where, as noted above, M is the number of sensed parameters, while N represents the number of available records (or timestamps). In particular, the parameter M is: M=R+S (2) where R are boundary conditions or input parameters that determine the operating point of the gas turbine 111 being monitored or simulated. In effect, R represents the records necessary to specify the operation of the simulated gas turbine 111. The remaining output records S are operational variables that cannot be detected by sensors and therefore cannot be simulated. In general, the "input" matrix referenced in step 23 in both Figures 2 and 3 is an N by R matrix. In this context, the output parameters S represent additional sensed and measured parameters in the system beyond those necessary to fully determine the operating point of the gas turbine 111.

[0053] The "system model" (e.g., a black-box model or a machine learning-based model) embedded in processing step 24 is a function that can be written as follows: f(r,h) (3) where r is a vector of parameters corresponding to the input parameter matrix R, and h is a matrix of model parameters, denoted H for the problem case. The model can also take matrices R and H as follows, including a multi-input configuration to allow multiple inputs simultaneously:

[0054] The system model in step 24, when input into matrices R and H, results in simulated output parameters S'. Initially, the simulated output parameters S' are obtained from the default model parameters H0 or from the model's adjusted coefficients. The outputs, S', are simulated outputs, which (formally) match the output matrix S. The simulated outputs S' are calculated by processing step 24, which obtains the simulated outputs of the system 25. In comparison step 26, the residuals E are calculated. Specifically, the initial residuals for all records are calculated as follows: E=S-S' (4)

[0055] Next, a so-called "representative point" of the system (gas turbine 111) operation is calculated, as will be explained in more detail below. The mean of the residual matrix E is then calculated, which can be a normal mean, median, winsorized mean, etc. (a measure of location). This should be done every N "times", so from an N x S set of parameters, E m We get a 1×S matrix called the matrix.

[0056] Similarly, in step 262, the mean, median, winsorized mean, etc. (also measures of location) are calculated for the input matrix R, to produce a 1×R matrix R m get.

[0057] Next, the time interval detection of the representative point P is calculated as follows: P=f(R m ,H0)+E m (5)

[0058] Referring to the flowchart of Figure 3, the application of the function f(·) is formally performed in step 263, but from a practical standpoint, the simulator performed in step 263 is the same as the simulator performed in system model step 24. As mentioned above, H0 is the initial model parameters, which can be derived from design or testing and is a matrix of parameters. In other words, H0 is simply the initialization of the model parameters, also called health parameters.

[0059] The representative point P is the mean residual E, which represents a 1×S vector in the "space" of parameters. m is a single point (for each instant in time N) with

[0060] A problem solver such as a genetic algorithm, gradient-based, or trust-region solver can then compute the health parameter vector H as follows: m is used to find P=f(R m ,H m ) (6) where H m is called the representative solution, and the averaged residual E m The representative point P can be achieved without correction. The residue is found to be absorbed. This step is the first solver step 27. The representative solution H m is obtained by looping.

[0061] Jacobian J m is approximated for the matrix E in terms of the model parameters H and reported as the representative Jacobian calculated in step 29. This represents how changing each health parameter in H affects the residuals from each component of the matrix S.

[0062] Through global search step 2, H m , P and the Jacobian J m Then, the following partial solution can be obtained around the representative solution: H m is calculated approximately for the matrix E with respect to the matrix H. As mentioned above, the local search procedure 3 m and J m which represents an approximate solution for all of the inputs of the system under analysis (gas turbine 111).

[0063] The local search step 3 of the solver algorithm is shown in Figure 4, and its scope is determined by the H m and J m This is the extent to which we refine the approximate solution given by , and, as mentioned above, find a so-called local search for the solution.

[0064] In general, the local search procedure 3 starts from the representative solution obtained in the global search procedure 2. This representative solution should already be close to the actual solution in all respects, and therefore the local search is performed by solving the Jacobian J m It performs a linear update from the previous step until convergence, or until an excessive number of iterations is reached for each point. The Jacobian does not need to be recomputed for every point, providing a speed gain to the solver.

[0065] Next, referring to FIG. 4 above, here instead of starting with the model parameters first, the local search procedure 3 finds an approximate solution H for all the model parameters. m Similarly, in Figure 3, the same input parameters R (received in step 33) and output parameters S (received in step 32) are considered.

[0066] Therefore, the Jacobian matrix J m This is considered constant (not updated) after obtaining the residual calculated in step 36, i.e., E=S'-S, as in step 261 and equation (3) above. The same solver step, now denoted by reference numeral 37, is performed on the representative Jacobian J m is performed to update the model parameters for all records using the following equations:

number

[0067] In step 38, the representative solution is H m is used.

[0068] The corrected data is fed to the model parameter step 38 to start another iteration, where a new set of simulated output parameters S is calculated in step 34, always by the model function f(·), from which the model parameters H m Calculate a new set of inverse Jacobians

number

[0069] In the first iteration, the local search procedure 3 is m In any case, we start from and there is still a residual E. Therefore, the solution is still an approximation and requires refinement. In any case, at this stage of the simulation method, the available solution is still very close to the current solution and only minor refinement is required. For this reason, the constant Jacobian J obtained in the previous phase or step m It is sufficient to use the following iterations to refine the solution: the residual error E becomes zero or reaches a value within a certain tolerance or threshold, and the refined model parameters H * This is the point-by-point solution to the equation problem that is finally obtained in the reporting step 39. The solution to this problem can be expressed as follows: S=f(R,H * ) (8) where, as described above, R are input parameters, i.e., boundary conditions that determine the operating point of the gas turbine 111 being monitored or simulated, and S are output parameters, sensed parameters that are not essential to obtain the operating point of the gas turbine, or simulated and calculated variables to characterize the gas turbine 111.

[0070] The obtained simulated model parameters H * , which cannot normally be measured and could not be monitored directly or indirectly. The model parameters H also have diagnostic value, so that it is possible to detect possible problems with the gas turbine 111 through analysis of the model parameters H, for example by checking their variations over time and comparing the slope with certain thresholds to predict possible failure of part or all of the gas turbine 111.

[0071] In particular, the model parameters H obtained by the described simulation method * If we continue to use it for diagnostic purposes, such a model parameter H *is a characterization of the gas turbine 111, since it allows the model to match (and thus obtain) real data. However, the purpose of characterizing the system is to characterize the model parameters H * However, it is also possible to "synthesize" or generate new parameters that may be useful to different stakeholders or users. These new parameters generated from the characterized model are called synthesized parameters.

[0072] An example of such a composite parameter is ISO power at full load, which corrects the power output for variations in operating and ambient conditions and therefore constitutes a parameter that depends only on true engine performance, which can be used either for performance tracking or for comparisons between engines within a fleet.

[0073] In addition to ISO power, the computable composite parameters used to evaluate the performance of the gas turbine 111 include ISO heat rate, local rated power, and local rated heat rate. These parameters are designed to represent the health of the system over time, independent of ambient and operating conditions. This allows for recommendations, anomalies to be detected, and problems to be solved. Diagnostic parameters that constitute the explored model parameters H can also be obtained for each subsystem, e.g., each module of the gas turbine 111. In some embodiments, there are diagnostic parameters related to the health of only the axial compressor (which constitutes the model parameters H) and diagnostic parameters related to the health of only the high-pressure turbine. These parameters, referred to as "module health parameters," help solve performance problems by indicating degradation of each module.

[0074] Referring now to Figure 5, the simulated output parameter S' * In particular, in the case of the problem, the input parameters R and the model parameters obtained by the simulation model above, namely the health parameters H* Considering the above, the simulated output S' at the reference condition * is obtained as follows: S' * =f(R,H * ) (8)

[0075] This simulation resulted in a specifically refined assimilated output based on the specific state of the gas turbine 111. The simulated output parameters S' * is obtained from step 25 and may be stored in a storage means, for example the storage means of the data recording unit 112, for further processing, such as calculating the synthesis parameters as described above.

[0076] As mentioned in describing FIG. 1, the characterization system 1 also includes a step 133 of calculating unmeasured (or measurable) parameters. In this case, estimation of unmeasured quantities or virtual sensors is performed. The gas turbine 111 may have important unmeasured quantities, such as combustion temperature, power output, fuel consumption, emissions, interstage pressure / temperature, bleed pressure / temperature, exhaust flow rate, among others. In this case, the refined model parameters H * is used together with the model function f(·) to simulate these unknown quantities that may be by-products of the simulation.

[0077] In another embodiment, virtual sensor redundancy / evaluation is performed. In this case, as described above, the gas turbine 111 is considered to have sensors that are prone to failure. In this case, the model can be used to provide redundancy for existing sensors or to evaluate whether these sensors have failed.

[0078] Referring to Figure 6, an implementation of an embodiment is shown, in which the simulated output parameters S' *is then sent to a condition 1331 to check whether each of the variables or parameters has already been measured, if not it is read as a virtual sensor in step 1332, if it has been measured it is a reason for sensor redundancy 1333. The calculated data is distorted for further processing, for example to the storage means of the data recording unit 112.

[0079] What-if scenario predictions are an enabler for providing new data-based services, including production optimization, emissions minimization, and maintenance optimization, and can be input into further process optimization schemes.

[0080] The characterization output of the algorithm (i.e., the map scalar) has fine-grained diagnostic value as performance losses can be assigned to specific modules, and also allows for more targeted maintenance / corrective actions.

[0081] While aspects of the present invention have been described in terms of various specific embodiments, it will be apparent to those skilled in the art that many modifications, changes, and omissions are possible without departing from the spirit and scope of the claims. Additionally, unless otherwise specified herein, the order or sequence of any process or method steps may be varied or rearranged according to alternative embodiments.

[0082] Reference will now be made in detail to the embodiments of the present disclosure, one or more examples of which are illustrated in the drawings. Each example is provided by way of explanation of the disclosure, not as a limitation of the disclosure. Indeed, it will be apparent to those skilled in the art that various modifications and variations can be made in the present disclosure without departing from the scope or spirit of the disclosure. References throughout this specification to "an embodiment" or "one embodiment" or "some embodiments" mean that a particular feature, structure, or characteristic described in connection with an embodiment is included in at least one embodiment of the disclosed subject matter. Thus, the appearances of the phrases "in one embodiment," "in one embodiment," or "in some embodiments" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0083] When presenting elements of various embodiments, the articles "a," "an," "the," and "said" are intended to mean that there are one or more of the elements. The terms "comprising," "including," and "having" are intended to be inclusive and mean that there may be additional elements other than the listed elements.

Claims

1. 1. A simulation method for simulating the operation of an engine, such as a gas turbine, the engine comprising a plurality of sensors for sensing a corresponding plurality of data representative of measured input parameters, the measured input parameters comprising: an input parameter record for defining the engine's operational and output parameters; The method comprises: A. performing a global search procedure to determine a solution for model parameters, thereby enabling operation of said engine to be simulated; B. performing a local search procedure to calculate refined model parameters for simulating the operation of the engine; C. simulating the output parameters based on the refined model parameters and the input parameters to check the operation of the engine; D. using the output parameters to derive data and / or unmeasured or non-measurable parameters of the engine; said global search procedure comprising: A1. Receiving the data representing the measured parameters from the gas turbine; A2. Selecting said input parameter records from said measured parameter records of said gas turbine; A3. Receiving model parameter default model parameters; A4. Processing the input parameters obtained from default model parameters by a functional model f(·) of the engine to obtain simulated output parameters of the engine; A5. Receiving the simulated output parameters of the engine; A6. Selecting an output parameter record as other measured parameters received from said engine; A7. Calculating residuals as the difference between the simulated output parameters and the output parameters; A8. A step of determining a representative residual parameter of the residuals and a representative input parameter of the input parameters; A9. Determining a representative point based on the representative residual parameters, the representative input parameters, and the default model parameters; A10. Solving an optimization problem to obtain a representative solution for said model parameters, wherein said representative point can be achieved without correction of mean residuals.

2. the local search procedure: B1. Receiving the data representing the measured parameters from the gas turbine; B2. Selecting the input parameter records from the measured parameter records of the gas turbine; B3. Processing the input parameters and the representative solution of the model parameters with a functional model f(·) of the engine to obtain simulated output parameters of the engine; B4. Calculating residuals as the difference between the simulated output parameters and the output parameters obtained by solving for the model parameters; B5. Calculating a new set of model parameters based on the residuals obtained in the previous calculation step; and iteratively repeating steps B3-B5 to obtain the refined model parameters for simulating operation of the engine.

3. the global search procedure includes calculating an approximated Jacobian for the residuals relative to the model parameters as a representative Jacobian; The method of claim 2 , wherein the local search procedure includes the step of calculating a new set of model parameters based on the residuals obtained in a previous calculation step, and is also based on the Jacobian.

4. The step of calculating the refined model parameters is carried out iteratively according to the following formula until obtaining the refined model parameters: [Equation 1] H ｉ＋１ =H ｉ +ΔH where: [Equation 2] The method of claim 3 , wherein: is the inverse Jacobian.

5. The method (2) of claim 1, wherein the step of determining the representative residual is calculated by the mean, median, or winsorized mean.

6. The step of determining the representative points is performed according to the following formula: P=f(R ｍ ,H ０ )+E ｍ The method of claim 1 , wherein the function f(·) is a functional model of the engine.

7. The method of claim 1 , wherein the optimization problem is based on a genetic algorithm, gradient-based, trust region, or the like.

8. said step of using said output parameters comprising: a sub-step of calculating a resultant parameter characterizing said engine with respect to the environmental parameters in which said engine operates, in order to detect anomalies and manage them to solve problems; and storing the synthesis parameters.

9. The method of claim 8 , wherein the composite parameters include ISO power at full load, ISO heat rate, local rated power, and / or local rated heat rate.

10. said step of using said output parameters comprising:

2. The method of claim 1, further comprising the sub-step of performing virtual sensor redundancy / evaluation by estimating unmeasured quantities or virtual sensors, wherein the refined model parameters are used together with a model function f(·) to simulate unknown parameters to determine virtual sensor results, or have the parameters measured by actual sensors of the engine, to check the operation of the actual sensors, so as to have sensor redundancy.

11. The method of claim 1 , wherein the data is detected along a time interval, and N data points are collected for each of the measured input parameters during the time interval.

12. 1. A characterization system for characterizing and simulating operation of a gas turbine, comprising: an infrastructure section, a gas turbine to be controlled, the gas turbine having a plurality of sensors detecting a plurality of parameters detected at different time intervals; an infrastructure section comprising: a data recording unit connected to the sensors of the gas turbine and configured to collect data and signals detected by the sensors; A processing unit, the processing unit comprising: an automated performance characterization section, 1. A physics-based model configured to execute a simulation method for simulating the operation of an engine, such as a gas turbine, by simulating output parameters based on refined model parameters and input parameters from the sensors of the engine, the engine comprising a plurality of sensors sensing a corresponding plurality of data representing measured input parameters, the measured input parameters comprising input parameter records for defining the operation and output parameters of the engine, the method comprising: A. performing a global search procedure to determine a solution for model parameters, thereby enabling operation of said engine to be simulated; B. performing a local search procedure to calculate refined model parameters for simulating the operation of the engine; C. simulating the output parameters based on the refined model parameters and the input parameters to check the operation of the engine; D. using the output parameters to derive data and / or unmeasured or non-measurable parameters of the engine; a delivery services section for deriving data and / or unmeasured or non-measurable parameters of the engine using the output parameters; Equipped with said global search procedure comprising: A1. Receiving the data representing the measured parameters from the gas turbine; A2. Selecting said input parameter records from said measured parameter records of said gas turbine; A3. Receiving model parameter default model parameters; A4. Processing the input parameters obtained from default model parameters by a functional model f(·) of the engine to obtain simulated output parameters of the engine; A5. Receiving the simulated output parameters of the engine; A6. Selecting an output parameter record as other measured parameters received from said engine; A7. Calculating residuals as the difference between the simulated output parameters and the output parameters; A8. A step of determining a representative residual parameter of the residuals and a representative input parameter of the input parameters; A9. Determining a representative point based on the representative residual parameters, the representative input parameters, and the default model parameters; A10. Solving an optimization problem to obtain a representative solution of the model parameters, wherein the representative point can be achieved without correcting the mean residual; system.

13. The system of claim 12 , wherein the automated performance characterization section comprises an automated data processing module configured to correct possible corruption of the data taken from the sensors of the gas turbine.

14. The delivery service section includes: a correction parameter calculation module for detecting engine performance degradation; an unmeasurable parameter calculation module; and a module for trending and monitoring module parameters.

15. The system of claim 12 , wherein the output of the physics-based model is used to perform optimization procedures such as maintenance optimization to improve engine performance, reliability, availability, emissions, and calculation of key performance indicators.

Images