CN116127831A - Soft measurement method for difficult-to-measure parameters of heavy gas turbine - Google Patents
Soft measurement method for difficult-to-measure parameters of heavy gas turbine Download PDFInfo
- Publication number
- CN116127831A CN116127831A CN202211559748.6A CN202211559748A CN116127831A CN 116127831 A CN116127831 A CN 116127831A CN 202211559748 A CN202211559748 A CN 202211559748A CN 116127831 A CN116127831 A CN 116127831A
- Authority
- CN
- China
- Prior art keywords
- model
- layer
- soft measurement
- gas turbine
- input
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000691 measurement method Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 40
- 238000005259 measurement Methods 0.000 claims abstract description 26
- 230000008569 process Effects 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 8
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 5
- 238000003062 neural network model Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 29
- 238000012360 testing method Methods 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 7
- 238000007476 Maximum Likelihood Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 238000000513 principal component analysis Methods 0.000 claims description 5
- 230000000737 periodic effect Effects 0.000 claims description 3
- 230000001537 neural effect Effects 0.000 claims 1
- 230000003044 adaptive effect Effects 0.000 abstract description 2
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
- G06F17/13—Differential equations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computational Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Algebra (AREA)
- Computing Systems (AREA)
- Computer Hardware Design (AREA)
- Artificial Intelligence (AREA)
- Geometry (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Operations Research (AREA)
- Fluid Mechanics (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Measuring Volume Flow (AREA)
Abstract
The invention provides a soft measurement method for difficult-to-measure parameters of a heavy-duty gas turbine, which relates to the technical field of heavy-duty gas turbines and comprises the following steps: grouping input variables by a principal component analysis method to obtain a plurality of sub-blocks; collecting historical operation data of a power plant; establishing a multi-layer soft measurement model; the multi-layer soft measurement model comprises a first layer instant learning model, a second layer Gaussian process regression model and a third layer depth feedforward neural network model; the invention uses an adaptive soft measurement multi-level modeling method based on instant learning and integrated learning to perform soft measurement on parameters which are difficult to measure. The model combines an integrated learning framework based on variable grouping with an instant learning method, realizes real-time updating of the local model through an instant learning algorithm, and improves the accuracy of the model through integrated learning.
Description
Technical Field
The invention relates to the technical field of heavy-duty gas turbines, in particular to a soft measurement method for difficult-to-measure parameters of a heavy-duty gas turbine.
Background
Over long periods of operation, the performance of the components of the gas turbine gradually decreases, which not only reduces the economy of power plant operation, but also affects reliability and safety. Therefore, monitoring the operating state of the gas turbine in real time is of great importance. At present, the real-time monitoring of the running state of the gas turbine is mainly realized by monitoring parameters such as the efficiency, the gas consumption rate and the like obtained by calculating key parameters of the gas turbine. Therefore, ensuring accurate measurement of key parameters is of great importance for safe and reliable operation of the gas turbine.
Most of key parameters of the gas turbine can be directly measured through the sensor, but due to economic or technical reasons, partial parameters which are difficult to directly measure or ensure the measurement accuracy still exist in the operation process, and the parameters can only be obtained through methods such as offline analysis or empirical formulas, so that data lag is caused, the data accuracy is reduced, and the operation state of the gas turbine is affected to be accurately judged in real time.
In view of the above, it is desirable to provide a method for measuring difficult-to-measure parameters of a gas turbine and accurately measuring the difficult-to-measure parameters.
Disclosure of Invention
Therefore, the invention aims to provide a soft measurement method for the difficult-to-measure parameters of a heavy-duty gas turbine, so as to solve the technical problem that the difficult-to-measure parameters of the conventional heavy-duty gas turbine are difficult to measure.
The invention adopts the following technical means:
a soft measurement method for difficult-to-measure parameters of a heavy gas turbine comprises the following steps:
grouping input variables by a principal component analysis method to obtain a plurality of sub-blocks;
collecting historical operation data of a power plant;
establishing a multi-layer soft measurement model;
the multi-layer soft measurement model comprises a first layer instant learning model, a second layer Gaussian process regression model and a third layer depth feedforward neural network model;
a first layer: the new test sample is brought into a first layer instant learning model, the similarity between the new test sample and the historical operation data of each sub-block is calculated, the similarity is arranged in a descending order, M samples with the highest similarity are selected as similar samples, and similar historical samples are selected for each sub-block;
a second layer: based on the constructed multiple sub-blocks, establishing a corresponding local GPR model aiming at similar sample data of each group of input variables;
inputting similar sample data of the group input variables into the GPR model, and outputting predicted values of a plurality of local regression models;
third layer: selecting a DFNN as an integrated model, wherein the DFNN comprises an input layer, a hidden layer, an output layer and full connection among the layers;
the predictive values of the multiple local regression models are input into the DFNN, 70% of the input data are used for training the DFNN integrated model, and 30% of the input data are used for testing the trained model and obtaining the predictive values of the target parameters.
Further, the multi-layer soft measurement model is used for carrying out variable grouping according to the uncorrelated principal component directions through principal component analysis, so as to quickly establish a plurality of sub-blocks; the first layer model selects similar samples for different sub-blocks based on an instant learning algorithm; the second layer of GPR local regression model is input into a similar sample and output into predicted values of a plurality of local regression models; the third layer DFNN integrated model integrates the prediction of the local model, inputs the prediction values of the second layer multiple local regression models, and outputs the prediction values of the target parameters.
Further, the first layer instant learning model adopts euclidean distance to evaluate similarity between samples, when the euclidean distance is reduced, the similarity between samples is high, and the similarity between samples is as follows:
wherein: x is sample data and subscript i, j represents two different sample data.
Further, the local GPR model is:
y=f(x)+α
wherein x is input data; alpha is 0 as the mean and sigma as the variance n 2 Is white gaussian noise; since the data has been normalized and preprocessed, the output y follows a gaussian distribution of N to (0, C), where C is an N x N symmetric positive definite matrix.
Further, the neuron structure of each layer connecting the input layer and the output layer in the DFNN has the following mapping relationship:
y=f(x,θ 0 )
wherein: x represents input, y represents output, θ 0 Representing the optimal parametric solution for the input-output mapping.
Further, in the local GPR model:
wherein C () is a covariance function including a Constant covariance function, a Linear covariance function, a Periodic covariance function,Covariance function.
wherein :σm 、σ n And d is a super parameter, when i=j, β=1, otherwise β=0;
super parameter set θ= { σ m ,σ n The estimated value of d is calculated by a maximum likelihood method of a log likelihood function of a sample, and the log likelihood function is:
in the maximum likelihood method, the super-parameter set θ is typically obtained by calculating the partial derivative of the log-likelihood function:
For a new sample x n+1 According to the property analysis of the Gaussian process, the test sample and the training sample belong to the same distribution, and the joint distribution characteristics are as follows:
wherein kn+1= [ C (xn+1, x 1), C (xn+1, x 2), …, C (xn+1, xn) ].
The output of the GPR model is:
σ n+1 2 =C(x n+1 ,x n+1 )-K n+1 C -1 K n+1 T 。
the invention also provides a storage medium comprising a stored program, wherein the soft measurement method of the difficult-to-measure parameter of the heavy-duty gas turbine is executed when the program runs.
The invention also provides an electronic device which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and is characterized in that the processor runs and executes the soft measurement method of any of the difficult-to-measure parameters of the heavy-duty gas turbine through the computer program.
Compared with the prior art, the invention has the following advantages:
the invention uses an adaptive soft measurement multi-level modeling method based on instant learning and integrated learning to perform soft measurement on parameters which are difficult to measure. The invention groups the variables through principal component analysis, and can quickly realize the construction of a plurality of input variable groups. According to the uncorrelated principal component directions, the first layer model selects similar samples for different sub-blocks based on an instant learning algorithm. On this basis, a second-layer gaussian process regression local regression model was developed. To integrate predictions of the local model, a third layer deep feed forward neural network integration model was developed to predict parameters that are difficult to measure. The model combines an integrated learning framework based on variable grouping with an instant learning method, realizes real-time updating of the local model through an instant learning algorithm, and improves the accuracy of the model through integrated learning. Taking the turbine inlet temperature of the gas turbine and the compressor inlet flow rate as examples. The soft measurement model is feasible and effective for predicting and analyzing the difficult-to-measure parameters through research and verification of actual operation data.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to the drawings without inventive effort to a person skilled in the art.
FIG. 1 is a schematic diagram of a multi-layer soft measurement model of the present invention.
FIG. 2 is a diagram of a third layer of the mold structure of the present invention.
FIG. 3 is a graph comparing turbine inlet temperature predictions with true values.
FIG. 4 is a graph comparing predicted values and true values of compressor inlet flow.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
As shown in FIG. 1, the invention provides a soft measurement method for difficult-to-measure parameters of a heavy-duty gas turbine, which is characterized by comprising the following steps:
grouping input variables by a principal component analysis method to obtain a plurality of sub-blocks;
collecting historical operation data of a power plant;
establishing a multi-layer soft measurement model;
the multi-layer soft measurement model comprises a first layer instant learning model, a second layer Gaussian process regression model and a third layer depth feedforward neural network model; the model can quickly establish a plurality of sub-blocks by performing variable grouping according to the uncorrelated principal component directions through principal component analysis. The first level model selects similar samples for different sub-blocks based on an instant learning algorithm. On the basis, a second-level GPR local regression model is developed, and the second-level GPR local regression model is input into a similar sample and output into a predicted value of a target parameter. In order to integrate the predictions of the local models, a third-stage DFNN integrated model is developed, input as predicted values of a plurality of local regression models of the second stage, and output as predicted values of target parameters.
A first layer: the new test sample is brought into a first layer instant learning model, the similarity between the new test sample and the historical operation data of each sub-block is calculated, the similarity is arranged in a descending order, M samples with the highest similarity are selected as similar samples, and similar historical samples are selected for each sub-block;
a second layer: based on the constructed multiple sub-blocks, establishing a corresponding local GPR model aiming at similar sample data of each group of input variables;
inputting similar sample data of the group input variables into the GPR model, and outputting predicted values of a plurality of local regression models;
third layer: selecting a DFNN as an integrated model, wherein the DFNN comprises an input layer, a hidden layer, an output layer and full connection among the layers;
the predictive values of the multiple local regression models are input into the DFNN, 70% of the input data are used for training the DFNN integrated model, and 30% of the input data are used for testing the trained model and obtaining the predictive values of the target parameters.
1. Model construction
For many key parameters in gas turbines that are difficult to measure directly or to guarantee measurement accuracy, a multi-layer soft measurement model of a just-in-time learning-Gaussian process regression-depth feed-forward neural network (JIT-GPR-DFNN) is proposed. A schematic diagram of the modeling method is shown in fig. 1.
The JIT-GPR-DFNN model groups variables through principal component analysis, and can quickly construct a plurality of input variable groups. The first level model selects similar samples for different sub-blocks based on JIT, depending on the direction of the uncorrelated principal components. On this basis, a second-order GPR local regression model was developed. To integrate predictions of the local model, a third level DFNN integrated model was developed to predict refractory parameters. The model combines an integrated learning framework based on variable grouping with an instant learning method, realizes real-time updating of a local model through a JIT modeling method, and improves the accuracy of the model through integrated learning.
1.1 similar sample selection
It is assumed that the test samples have the same distribution characteristics as the historical data. The first level model selects samples similar to the test samples from all historical data based on the JIT algorithm according to different sub-blocks. When a new test sample comes, a similarity measure needs to be determined before selecting a similar sample. Currently, there are many similarity metrics for the JIT algorithm. Common similarity sample selection methods include distance and angle between samples, and correlation between data. The present model uses euclidean distance to evaluate similarity between samples. The smaller the Euclidean distance, the higher the similarity. The calculation formula is as follows:
for a new test sample x new The similarity to each sub-block of historical operating data is calculated and arranged in descending order. And selecting M samples with highest similarity as similar samples, and selecting similar historical samples for each sub-block.
The sub-blocks are a plurality of groups of input variables obtained by grouping the input variables by a principal component analysis method, and are called a plurality of sub-blocks. Irrespective of the nature of heavy gas turbines, the dividing sub-block belongs to a data preprocessing method, and the building process is as follows:
the measured data of the input variables may be represented by an n×m matrix, where n is the data sample size and m is the number of input variables. The output variables are represented by an nxl matrix, l representing the number of output variables. Conventional PCA can be expressed as:
X=TP T +E
Y=TC T +F
wherein X and Y are an input variable matrix and an output variable matrix; t epsilon Rn x k is the principal component matrix, k is the number of principal components selected; p epsilon Rm x k is the load matrix of the input variables; c epsilon Rl x k is the load matrix of the output variables; e and F are residual matrices.
The sub-blocks may be defined per principal component direction based on the uncorrelated behavior of the extracted components. The variables contained in the sub-blocks are determined by their weights. The weights of each principal component are as follows:
wherein ,pij Is the ith element of the jth principal component of the load matrix P. The larger the value of the i-th element, the greater its contribution to the j-th principal component. Thus, the weight values are ordered in descending order to obtain { w ' (j, 1), w ' (j, 2),. Fw ' (j, m) }, the variables contained in each sub-block being determined by the following conditions:
where θ=80% is a custom value.
The variables contained in each sub-block are thus determined and the input variable dataset X is divided into k sub-blocks.
1.2 local regression model
Based on the constructed multiple variable groups, corresponding local GPR models are built for sample data of each group of input variables. The GPR model can be described as:
y=f(x)+α (2)
wherein x is input data, f () is an objective function, α is 0 as the mean, and variance is σ n 2 Is a gaussian white noise of (c). Since the data has been normalized and preprocessed, the output y follows a gaussian distribution of N to (0, C), where C is an N x N symmetric positive definite matrix.
Where C () is a covariance function, with no fixed form. The common covariance function is Constant, linear, periodic,Covariance, etc. The model selects the noise item +.>Covariance function, the calculation formula is as follows:
wherein ,σm 、σ n And d is a superparameter. When i=j, β=1, otherwise β=0. Super parameter set θ= { σ m ,σ n The estimated value of d is calculated by a maximum likelihood method of a log likelihood function of a sample, and the log likelihood function is:
in the maximum likelihood method, the super-parameter set θ is typically obtained by calculating the partial derivative of the log-likelihood function:
For a new sample x n+1 According to the property analysis of the Gaussian process, the test sample and the training sample should belong to the same distribution, and the joint distribution characteristics are as follows:
wherein ,Kn+1 =[C(x n+1 ,x 1 ),C(x n+1 ,x 2 ),…,C(x n+1 ,x n )]。
Thus, the output of the GPR model is:
σ n+1 2 =C(x n+1 ,x n+1 )-K n+1 C -1 K n+1 T (9)
1.3 Integrated model
The model selects DFNN as an integrated model, and the DFNN is a typical deep learning model and is characterized by comprising a plurality of hidden layers. The model structure is shown in fig. 2 and consists of an input layer, a hidden layer, an output layer and full connection among the layers. The hidden layers may be designed into any number of layers depending on the complexity of the model, with the connections between each layer representing the weights of the features. The neuron structure of each layer connecting the input layer and the output layer has the following mapping relation:
y=f(x,θ 0 ) (10)
wherein x and y represent input and output respectively; θ 0 Representing the optimal parametric solution for the input-output mapping.
The output of the local model in 1.2 is taken as the input of the DFNN, the input data is divided into 7:3, 70% of data is used for training the DFNN integrated model, and 30% of data is used for testing the trained model and obtaining a predicted value. Because the DFNN model has strong nonlinear mapping capability, the prediction results of a plurality of models can be well fitted, and therefore integration of the local model is achieved.
The training method of the DFNN integrated model comprises the following steps:
step 1, randomly initializing the weight of each node of the neural network;
step 2 uses the input variable to forward transfer to obtain an output result y';
step 3, calculating an error between the output result y' and the true value y;
step 4, back-propagating the error, and updating the weight of each node;
step 5 repeats this process until the error between the output result y' and the true value y satisfies the stop condition or no longer drops.
2. Model verification
The model is researched by taking the air inlet flow of the air compressor and the temperature of the turbine opening as difficult-to-measure parameters, wherein the input variables are shown in a table 1, and the output variables are the air inlet flow of the air compressor and the temperature of the turbine opening.
Table 1 input variables
The pairs of predicted values and true values of the validation set, such as those shown in fig. 3 and 4, can be seen to be very close to the true values. The absolute values of the average relative errors of the turbine inlet temperature and the air inlet flow of the air compressor are respectively 0.11 percent and 0.21 percent, and the maximum relative errors are respectively 0.81 percent and 4.14 percent. The soft measurement model has good prediction effect, and the JIT-GPR-DFNN soft sensing model can replace off-line analysis to obtain the true value of the variable.
The invention also provides a storage medium comprising a stored program, wherein the program is used for executing the soft measurement method of the difficult-to-measure parameters of the heavy gas turbine when running.
The invention also provides an electronic device which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and is characterized in that the processor runs and executes a soft measurement method of the difficult-to-measure parameters of the heavy-duty gas turbine through the computer program.
In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
In the several embodiments provided in the present application, it should be understood that the disclosed technology content may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, for example, may be a logic function division, and may be implemented in another manner, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-only memory (ROM), a random access memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (9)
1. The soft measurement method for the difficult-to-measure parameters of the heavy-duty gas turbine is characterized by comprising the following steps of:
grouping input variables by a principal component analysis method to obtain a plurality of sub-blocks;
collecting historical operation data of a power plant;
establishing a multi-layer soft measurement model;
the multi-layer soft measurement model comprises a first layer instant learning model, a second layer Gaussian process regression model and a third layer depth feedforward neural network model;
a first layer: the new test sample is brought into a first layer instant learning model, the similarity between the new test sample and the historical operation data of each sub-block is calculated, the similarity is arranged in a descending order, M samples with the highest similarity are selected as similar samples, and similar historical samples are selected for each sub-block;
a second layer: based on the constructed multiple sub-blocks, establishing a corresponding local GPR model aiming at similar sample data of each group of input variables;
inputting similar sample data of the group input variables into the GPR model, and outputting predicted values of a plurality of local regression models;
third layer: selecting a DFNN as an integrated model, wherein the DFNN comprises an input layer, a hidden layer, an output layer and full connection among the layers;
the predictive values of the multiple local regression models are input into the DFNN, 70% of the input data are used for training the DFNN integrated model, and 30% of the input data are used for testing the trained model and obtaining the predictive values of the target parameters.
2. The soft measurement method of difficult-to-measure parameters of a heavy gas turbine according to claim 1, wherein the multi-layer soft measurement model is used for quickly establishing a plurality of sub-blocks by performing variable grouping according to the direction of an uncorrelated principal component through principal component analysis; the first layer model selects similar samples for different sub-blocks based on an instant learning algorithm; the second layer of GPR local regression model is input into a similar sample and output into predicted values of a plurality of local regression models; the third layer DFNN integrated model integrates the prediction of the local model, inputs the prediction values of the second layer multiple local regression models, and outputs the prediction values of the target parameters.
3. The soft measurement method of refractory parameters of a heavy gas turbine according to claim 1, wherein the first layer of instant learning model evaluates similarities between samples using euclidean distances, and when the euclidean distances become smaller, the similarities between samples become higher, and the similarities between samples are:
wherein: x is sample data and subscript i, j represents two different sample data.
4. The method for soft measurement of difficult-to-measure parameters of a heavy gas turbine according to claim 1, wherein the local GPR model is:
y=f(x)+α
wherein x is input data; alpha is 0 as the mean and sigma as the variance n 2 Is white gaussian noise; since the data has been normalized and preprocessed, the output y follows a gaussian distribution of N to (0, C), where C is an N x N symmetric positive definite matrix.
5. The soft measurement method of refractory parameters of a heavy duty gas turbine according to claim 1, wherein the neuronal structure of each layer of the DFNN coupled to the input layer and the output layer has the following mapping relationship:
y=f(x,θ 0 )
wherein: x represents input, y represents output, θ 0 Representing the optimal parametric solution for the input-output mapping.
6. The method for soft measurement of refractory parameters of a heavy duty gas turbine according to claim 4, wherein in said local GPR model:
7. The method for soft measurement of refractory parameters of a heavy duty gas turbine according to claim 6, wherein said method comprisesThe covariance function is:
wherein :σm 、σ n And d is a super parameter, when i=j, β=1, otherwise β=0;
super parameter set θ= { σ m ,σ n The estimated value of d is calculated by a maximum likelihood method of a log likelihood function of a sample, and the log likelihood function is:
in the maximum likelihood method, the super-parameter set θ is typically obtained by calculating the partial derivative of the log-likelihood function:
for a new sample x n+1 According to the property analysis of the Gaussian process, the test sample and the training sample belong to the same distribution, and the joint distribution characteristics are as follows:
wherein kn+1= [ C (xn+1, x 1), C (xn+1, x 2), …, C (xn+1, xn) ].
The output of the GPR model is:
σ n+1 2 =C(x n+1 ,x n+1 )-K n+1 C -1 K n+1 T 。
8. a storage medium comprising a stored program, wherein the program, when run, performs the method of soft measurement of a difficult parameter of a heavy gas turbine according to any one of claims 1 to 7.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor is operative to perform the method for soft measurement of a difficult parameter of a heavy gas turbine as claimed in any one of claims 1 to 7 by means of the computer program.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211559748.6A CN116127831A (en) | 2022-12-06 | 2022-12-06 | Soft measurement method for difficult-to-measure parameters of heavy gas turbine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211559748.6A CN116127831A (en) | 2022-12-06 | 2022-12-06 | Soft measurement method for difficult-to-measure parameters of heavy gas turbine |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116127831A true CN116127831A (en) | 2023-05-16 |
Family
ID=86298283
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211559748.6A Pending CN116127831A (en) | 2022-12-06 | 2022-12-06 | Soft measurement method for difficult-to-measure parameters of heavy gas turbine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116127831A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116933693A (en) * | 2023-09-12 | 2023-10-24 | 华能南京燃机发电有限公司 | Gas turbine performance detection method and device |
-
2022
- 2022-12-06 CN CN202211559748.6A patent/CN116127831A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116933693A (en) * | 2023-09-12 | 2023-10-24 | 华能南京燃机发电有限公司 | Gas turbine performance detection method and device |
CN116933693B (en) * | 2023-09-12 | 2024-02-09 | 华能南京燃机发电有限公司 | Gas turbine performance detection method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ma et al. | Discriminative deep belief networks with ant colony optimization for health status assessment of machine | |
US20230023931A1 (en) | Hydraulic turbine cavitation acoustic signal identification method based on big data machine learning | |
CN108445752B (en) | Random weight neural network integrated modeling method for self-adaptively selecting depth features | |
Wang et al. | Random convolutional neural network structure: An intelligent health monitoring scheme for diesel engines | |
CN116757534A (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
Ayodeji et al. | Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction | |
CN111813084A (en) | Mechanical equipment fault diagnosis method based on deep learning | |
CN111648992B (en) | Gas turbine compressor fault identification early warning method | |
Akpudo et al. | Towards bearing failure prognostics: A practical comparison between data-driven methods for industrial applications | |
CN106896219A (en) | The identification of transformer sub-health state and average remaining lifetime method of estimation based on Gases Dissolved in Transformer Oil data | |
CN115294671A (en) | Air compressor outlet pressure prediction method and prediction system | |
Yin et al. | A new health estimation model for CNC machine tool based on infinite irrelevance and belief rule base | |
Suryo et al. | Improved time series prediction using LSTM neural network for smart agriculture application | |
CN114218872A (en) | Method for predicting remaining service life based on DBN-LSTM semi-supervised joint model | |
Xiao et al. | A noise-boosted remaining useful life prediction method for rotating machines under different conditions | |
CN116127831A (en) | Soft measurement method for difficult-to-measure parameters of heavy gas turbine | |
Huang et al. | Bayesian neural network based method of remaining useful life prediction and uncertainty quantification for aircraft engine | |
Wang et al. | Evaluation and Prediction Method of Rolling Bearing Performance Degradation Based on Attention‐LSTM | |
Cheng et al. | Interactive hybrid model for remaining useful life prediction with uncertainty quantification of bearing in nuclear circulating water pump | |
Deng et al. | A remaining useful life prediction method with automatic feature extraction for aircraft engines | |
CN113126489A (en) | CNN-GRU-BINN-based intelligent BIT design method for heavy-duty gas turbine control system | |
Wenqiang et al. | Remaining useful life prediction for mechanical equipment based on temporal convolutional network | |
CN114637620B (en) | Database system abnormal classification prediction method based on SVM algorithm | |
Zhao et al. | Remaining useful life prediction method based on convolutional neural network and long short-term memory neural network | |
CN115758187A (en) | Coal mine mechanical motor state prediction method based on K-CNN-N-GRU |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |