CN108549757B - Model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method - Google Patents
Model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method Download PDFInfo
- Publication number
- CN108549757B CN108549757B CN201810288855.7A CN201810288855A CN108549757B CN 108549757 B CN108549757 B CN 108549757B CN 201810288855 A CN201810288855 A CN 201810288855A CN 108549757 B CN108549757 B CN 108549757B
- Authority
- CN
- China
- Prior art keywords
- gpr
- model
- prediction
- flow rate
- discharge flow
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000012360 testing method Methods 0.000 claims description 54
- 230000008569 process Effects 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 10
- 150000001875 compounds Chemical class 0.000 claims description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 239000007789 gas Substances 0.000 description 10
- 230000008859 change Effects 0.000 description 6
- 238000013461 design Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 230000001419 dependent effect Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000001747 exhibiting effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 230000010349 pulsation Effects 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000013332 literature search Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004088 simulation 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Control Of Positive-Displacement Pumps (AREA)
Abstract
本发明公开了一种模型自选择的往复式混输泵排出流率预测方法，它包括以下步骤：(1)确定预测模型的输入和输出变量，收集建模样本；(2)建立往复式混输泵排出流率的局部GPR模型；(3)建立往复式混输泵排出流率的加权GPR模型；(4)建立往复式混输泵排出流率的即时GPR模型；(5)基于预测概率信息，自动为每个新的输入样本点选择合适的预测模型；(6)重复步骤(2)至(5)，可从局部GPR、加权GPR和即时GPR模型中，为新的工况下每个输入样本点找到最合适的预测模型，继而得出新的工况下，往复式混输泵的排出流率曲线，本发明基于有限建模样本，实现对混输工况下往复式混输泵排出流率的建模和预测，在工程上容易实施、准确性高。
The invention discloses a model self-selecting method for predicting the discharge flow rate of a reciprocating mixed pump, which comprises the following steps: (1) determining the input and output variables of the prediction model, and collecting modeling samples; (3) Establish a weighted GPR model of the discharge flow rate of the reciprocating mixed pump; (4) Establish an instant GPR model of the discharge flow rate of the reciprocating mixed pump; (5) Based on the predicted probability information, and automatically select a suitable prediction model for each new input sample point; (6) Repeat steps (2) to (5), from the partial GPR, weighted GPR and immediate GPR models, for each new operating condition. Find the most suitable prediction model for each input sample point, and then obtain the discharge flow rate curve of the reciprocating mixed pump under the new working condition. Modeling and prediction of pump discharge flow rate is easy to implement and accurate in engineering.
Description
Technical Field
The invention relates to an important parameter modeling and predicting method for a reciprocating type mixed transportation pump in a design stage, in particular to a model self-selection reciprocating type mixed transportation pump discharge flow rate predicting method suitable for complex frequency-dependent oil-gas mixed transportation working conditions.
Background
The reciprocating type mixing and conveying pump has the functions of a pump and a compressor, and can effectively increase the yield of oil and natural gas in the oil exploitation process. When the reciprocating pump delivers incompressible medium, the piston and the inlet and outlet valves are moved cyclically so that the pump chamber discharge flow rate equals the rate of change of its volume, the pump chamber discharge flow rate curve for one stroke exhibiting a sinusoidal variation. When the reciprocating pump delivers compressible and incompressible mixed media such as oil gas and the like, the open and close hysteresis phenomena of the inlet and outlet valves can be induced by the heterogeneous flow frequency-variable impact load generated in the pump. Due to the compressible and incompressible nature of the gas, the lag in opening of the outlet valve will cause the pump chamber flow rate to rise rapidly to a peak, with the fluctuations then returning to a sinusoidal variation, exhibiting abrupt segments on the discharge flow rate curve; a short and small backflow occurs after the closing delay of the outlet valve, and a backflow section is presented on the discharge flow rate curve. It can be seen that under the condition of oil-gas mixture transportation, the pump cavity discharge flow rate curve of one stroke of the reciprocating pump presents four stages of a zero flow section, a sudden change section, a sine section and a backflow section. The sudden change phenomenon in the flow discharge process of the reciprocating pump is one of the main reasons for intensifying flow pulsation, inducing noise and vibration and reducing pump efficiency. Therefore, the research on the interaction between the discharge flow rate characteristic of the reciprocating pump and the oil-gas mixed transportation working condition has important significance for guiding the design of the mixed transportation pump engineering and ensuring the stable operation of the mixed transportation pump.
The scholars at home and abroad mainly study the flow characteristics of screw-type, vane-type and gear-type mixed delivery pumps on the basis of a mechanism model and a Computational Fluid Dynamics (CFD) simulation technology. As a newly used mixed transportation pump type, research on the flow characteristics of a reciprocating oil-gas mixed transportation pump is rarely carried out. The instantaneous flow of the pump under several mixed transportation working conditions is obtained by means of a CFD modeling tool only in Zhang Shengchang and the like without considering any volume loss, and the flow pulsation characteristics of the pump are researched. In general, most of the above studies on various multiphase pumps do not consider leakage and energy loss of the pumps, and the reduction of the in-pump flow pattern into homogeneous flow is not enough to describe the flow characteristics of the multiphase pumps which are frequency-variant and complicated. Meanwhile, mechanism models describing various mixed delivery pumps are complex, and parameters such as valve clearance instantaneous pressure, temperature, flow coefficient and the like in the models are difficult to obtain from experiments, so that the engineering application of the models is limited. In addition, the division of dynamic meshes, selection of multiphase flow and turbulence models, user-defined functions and other CFD modeling processes are highly dependent on the experience of the research and designer. Therefore, it is necessary to establish a reciprocating type multiphase pump discharge flow rate model with strong universality and high accuracy to adapt to complex frequency-dependent oil-gas multiphase working conditions.
In recent years, data-driven empirical models have been used to predict variables with large measurement lags in nonlinear industrial processes without substantial knowledge of complex internal phenomena and without much experience on the part of designers. These advantages can solve the above mechanism and CFD modeling problem simultaneously, and provide a new method for modeling and predicting the discharge flow rate of the multiphase pump.
Under different mixing and transportation working conditions, the discharge flow rate curve presents different process characteristics; under the same mixed transportation working condition, the zero flow section, the abrupt change section, the sine section and the backflow section of the discharge flow rate curve also present different local characteristics, and the discharge flow rate of the abrupt change section presents the characteristics of rapid nonlinear change and the like. In addition, the few multi-phase flow meter products on the market have the limitations of high manufacturing cost, measurement lag and the like, and the prediction performance is very difficult to improve by acquiring a large amount of reliable modeling data from experiments. By considering the factors, the characteristics of the whole complex process can be more completely described by using the prediction uncertainty information provided by a Gaussian Process Regression (GPR) model and adopting an adaptive hybrid modeling method. However, literature search revealed that the GPR model was not used to predict the discharge flow rate of reciprocating compound pumps.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provides a model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method suitable for the complex frequency-dependent oil-gas mixed transportation working condition.
Aiming at the defects and shortcomings existing in the mechanism and the CFD modeling process, the invention provides a self-adaptive modeling and predicting method utilizing GPR prediction uncertainty, which can automatically select a proper prediction model for each test point based on limited samples and improve the prediction accuracy and reliability of the discharge flow rate.
The model self-selection reciprocating type multiphase pump discharge flow rate prediction method comprises the following steps:
(1) determining input variables and output variables of a prediction model, collecting a modeling sample set S, and classifying the sample set S to obtain M sample subsets, namely S ═ (S)1，…，sm)TM is 1, …, M; the mth subset of samples is denoted as
Wherein N ismRepresents the m-th sampleSample number of the subset;
(2) partial GPR model for establishing discharge flow rate of reciprocating type mixed delivery pump
Performing learning training on each sample subset independently, and establishing a partial GPR prediction model of the discharge flow rate;
according to GPR definition, m local model GPRmThe output of (d) is expressed as:
in the formula, CmRepresents a covariance matrix whose i row and j column elements are represented as:
in the formula, xm，idDenotes xm，iThe d-th element of (1); j, then δm，ij1, otherwise δ m，ij0; d is the training sample point xm，iThe input dimension of (a); thetam＝[am，0，am，1，vm，0，wm，1，…，wm，d，bm]TIs the model coefficient;
using formula (1) and formula (2), M partial GPR models complete off-line modeling, defined as GPRm，m＝1，…，M；
Testing sample set for new input
Where T represents the number of newly input test sample sets, NtNumber of samples, x, representing the t-th new input test sample sett，iDenotes xtThe ith sample point; GPRmFor xt，iPredicted output of (2)
Sum variance
Respectively, as follows:
in the formula (I), the compound is shown in the specification,
representing the covariance between the new input test sample and the training sample; k is a radical ofm，ti＝C(xt，i，xt，i) Is the covariance of the newly input test sample;
for a new input test sample point, calculating and obtaining M online prediction information based on the local GPR model from the formula (3) and the formula (4);
(3) establishing a weighted GPR model of the discharge flow rate of the reciprocating type mixed delivery pump;
based on Bayesian inference, conditional probability P (GPR) is proposedm|xt，i) To GPRmEach sample point x of the model and new input sett，iThe relationship between the two is evaluated; p (GPR)m|xt，i) The calculation is as follows:
in the formula, P (GPR)m) And P (x)t，i|GPRm) Prior probability and conditional probability, respectively; when there is no process prior knowledge, P (GPR)m|xt，i) Expressed as:
for new input test sample point x based on probability analysis methodt，iIn terms of P (GPR)m|xt，i) The larger the GPRmThe more appropriate the model is to predict it;
merging the M partial GPR model pairs xt，iProbability information of prediction, weighting prediction values of GPR model
And its variance
The expression is as follows:
(4) establishing an instant GPR model of the discharge flow rate of the reciprocating type mixing delivery pump;
(4.1) from sample set s ═ { xn，ynN1, …, N (N is the total number of samples in the sample set), is a new input test sample point xt，iSelecting a suitable similar modeling sample; definition of
ηt，ni＝exp(-||xn-xt，i||)，n＝1，…，N (9)
Describing a modeled sample point xnAnd new input test sample point xt，iThe similarity relationship of (1); etat，niThe larger the value is, the more similar the relationship between the two is; thus, by setting the appropriate threshold λ, by formula
ηt，ni＞λ (10)
Testing sample points x for new inputst，iSelecting a proper similar modeling sample set;
(4.2) establishing an instant GPR model by using the formula (1) and the formula (2) based on the similar modeling sample set selected by the formula (10); calculating to obtain the sample point x of the instantaneous GPR model from the formula (3) and the formula (4)t，iPredicted value of (2)
And its variance
(5) Automatically selecting a suitable prediction model for each new input test sample point based on the prediction probability information;
(5.1) for each new input test sample point xt，iSelecting a suitable local GPR prediction model;
based on the probability information provided by equation (6), there is a model of Maximum Conditional Probability (MCP), i.e.
MCPt，i＝maxP(GPRm|xt，i)，m＝1，…，M (11)
Testing sample points x for new inputst，iThe most suitable partial GPR prediction model, the corresponding predicted value and the variance thereof are respectively recorded as
And
(5.2) test sample points x for each new input from the local, weighted and immediate GPR modelt，iSelecting a proper prediction model;
the predicted variance can be used to describe the input test sample point xt，iAnd uncertainty of its prediction model; if an inappropriate model is used for the new input test sample xt，iIf prediction is performed, the corresponding variance value is large; based on this, can be selected from
And
in which the model with the smallest variance (MV) is selected, i.e.
I.e. new input test sample point xt，iThe most appropriate predictive model;
(6) and (5) repeating the steps (2) to (5), finding the most appropriate prediction model for each input test sample point under the new working condition from the local GPR model, the weighted GPR model and the instant GPR model, and further obtaining the discharge flow rate curve of the reciprocating type multiphase pump under the new working condition.
The model self-selection reciprocating type multiphase pump discharge flow rate prediction method is characterized in that in the step (5), each input test sample point x is described by using the maximum conditional probability and the prediction variancet,iAnd the uncertainty of its prediction model, so that an appropriate model can be selected for prediction.
By adopting the technology, compared with the prior art, the invention has the following beneficial effects: the model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method provided by the invention is based on a limited modeling sample, realizes modeling and prediction of the discharge flow rate of the reciprocating type mixed transportation pump under the mixed transportation working condition, and is easy to implement in engineering and high in accuracy compared with the complexity of a mechanism model modeling process, the dependence of a CFD modeling process on the experience level of a designer and the like.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
fig. 2a to 2d are comparison of experimental results of a new input sample set with self-selected model prediction results according to the present invention, wherein fig. 2a is a graph of local GPR prediction results of the new input sample set, fig. 2b is a graph of weighted GPR prediction results of the new input sample set, fig. 2c is a graph of immediate GPR prediction results of the new input sample set, and fig. 2d is a graph of self-selected prediction results of the new input sample set model.
Detailed Description
The technical scheme of the invention is further explained by combining the drawings and the embodiment of the specification.
Examples
As shown in fig. 1, a model self-selection method for predicting a flow rate of a reciprocating type mixing pump includes the following steps:
(1) determining input variables and output variables of a prediction model, and collecting modeling samples;
in the embodiment, the inlet pressure P is selected by comprehensively considering the influence factors of the flow rate of the reciprocating type mixing and conveying pump and the design parameters of a prototype of the three-cylinder double-acting reciprocating type mixing and conveying pump for experimentss(0.2-0.4 MPa) and outlet pressure Pd(1.0-3.0 MPa), air content beta (0.1-0.8) and crank angle theta (180 degrees-the end of the discharge process) are input variables; considering that the flow rate of the conventional reciprocating pump can be calculated by a theoretical formula from the flow rate of one pump cavity, different numbers of pump cylinders are usually adopted in practical engineering application, and the flow rate Q of one pump cavity of a prototype is selected as an output variable without loss of generality;
acquiring experimental data under different working conditions from an experimental system; samples at different crank angles for one stroke, i.e. with the same inlet pressure, outlet pressure, gas fraction, pump speed, clearance volume, are grouped into a sample subset. The 15 sample subsets obtained are expressed as S ═ (S)1，…，s15)T9 of them are used as training sample set(s)1，…，s9) The remaining 6 are used as test sample sets(s)10，…，s15) The corresponding working conditions are respectively Ps0.3, 0.35, 0.25, 0.2, 0.4MPa, outlet pressure Pd3.0, 2.5, 3.0, 1.0MPa), gas content β (0.5, 0.3, 0.4, 0.1, 0.7, 0.8), crank angle θ (180 ° end of discharge process); in order to illustrate the advantages of the method, 6 test samples are concentrated, the first 3 working conditions are similar to the training working conditions, and the last three working conditions are different;
(2) partial GPR model for establishing discharge flow rate of reciprocating type mixed delivery pump
Establishing 9 local GPR models off line based on the formulas (1) and (2);
based on the formulas (3) and (4), 9 pieces of local GPR model online prediction information can be obtained respectively, and the test sample set S is subjected to10The predicted value and variance of one sample point of (a);
(3) weighted GPR model for establishing discharge flow rate of reciprocating type mixed delivery pump
Based on equations (5) to (8), a weighted GPR model is established, and S is obtained10The predicted value and variance of one sample point of (a);
(4) and establishing an instant GPR model of the discharge flow rate of the reciprocating type multiphase pump.
Based on the formulae (9) and (10), is S10Selecting a suitable similar modeling sample set from one sample point; establishing an instant GPR model based on the formula (1) and the formula (2); calculating a predicted value and a variance of the instant GPR model to the sample point by using a formula (3) and a formula (4);
(5) automatically selecting a suitable prediction model for each new input sample point based on the prediction probability information;
based on formula (11), is S10Selecting the most suitable local GPR model corresponding to the maximum conditional probability; from the local, weighted and immediate GPR model, S, based on equation (12)10Selecting the most suitable model, i.e. the model corresponding to the minimum prediction variance;
(6) repeating steps (2) to (5) for a test sample set S from the local GPR, weighted GPR and immediate GPR models10Finding the most suitable prediction model for each sample point, and then obtaining S10An exhaust flow rate curve;
(7) repeating steps (2) to (6) can obtain the discharge flow rate curves of the remaining 5 test sample sets.
The prediction results of the 6 test sample sets obtained by the method are compared with the experimental results, and the common index of mean square error (RMSE) is selected as an evaluation standard. For the tth test sample set, the RMSE evaluation criteria are defined as follows:
the RMSE index is a non-negative number, and the smaller the value, the better the prediction effect. The comparison results are shown in table 1.
Table 1 prediction performance of the method of the invention for a set of test samples
As can be seen from the results in Table 1, based on limited modeling samples, the method (i.e., the prediction method based on model self-selection) of the invention can well capture the characteristic information of the first 3 test samples (which are similar to the training condition); main characteristic information of the last 3 test samples (different from the training working condition) can be well captured; compared with a single local GPR, a weighted GPR and an instant GPR prediction model which are directly used, the method can better capture the characteristic information of each test sample set and obtain better prediction performance.
From test sample set S10The detailed prediction results (as shown in fig. 2a to 2d) show that the prediction effect of the local GPR prediction model in the ascending section of the mutation section is better, the prediction effect of the weighted GPR prediction model in the vicinity of the peak value of the mutation section is better, the prediction effect of the immediate GPR prediction model in the zero flow section and the sinusoidal section is better, and the model self-selection method better integrates the advantages of the three prediction models and has the best prediction effect. Therefore, the method automatically selects a proper prediction model for each test sample point by utilizing the probability information provided by the GPR, can better extract the characteristic information in the sample, and improves the overall prediction precision.
Finally, it takes only a few minutes to complete the online prediction of the 6 test sample sets based on the modeling data for the 15 conditions provided by the experiment. Under the same computing resource condition, the traditional CFD modeling link usually takes more than half a month, and the established CFD model is not necessarily accurate and is not necessarily suitable for a test set under a new working condition.
Therefore, the model self-selection prediction method can provide more accurate model and prediction for the discharge flow rate of the reciprocating type mixing and conveying pump. In addition, the simple and reliable implementation method can reduce the design complexity, reduce the design cost, save the modeling time and provide an effective auxiliary means for the design of the current reciprocating type mixing and transporting pump.
The embodiments described in this specification are merely illustrative of implementations of the inventive concept and the scope of the present invention should not be considered limited to the specific forms set forth in the embodiments but rather by the equivalents thereof as may occur to those skilled in the art upon consideration of the present inventive concept.
Claims (2)
1. A model self-selecting reciprocating type multiphase pump discharge flow rate prediction method comprises the following steps:
(1) determining input variables and output variables of a prediction model, collecting a modeling sample set S, classifying the sample set S to obtain M sample subsets, namely S ═ (S ═1,…,Sm)TM is 1, …, M; the mth subset of samples is denoted as
Wherein N ismRepresenting the number of samples of the mth sample subset;
(2) partial GPR model for establishing discharge flow rate of reciprocating type mixed delivery pump
Performing learning training on each sample subset independently, and establishing a partial GPR prediction model of the discharge flow rate;
according to GPR definition, m local model GPRmThe output of (d) is expressed as:
in the formula, CmRepresents a covariance matrix whose i row and j column elements are represented as:
in the formula, xm,idDenotes xm,iThe d-th element of (1); i is equal to j, and j is equal to,then deltam,ij1, otherwise δm,ij0; d is the training sample point xm,iThe input dimension of (a); thetam＝[am,0,am,1,vm,0,wm,1,…,wm,d,bm]TIs the model coefficient;
using formula (1) and formula (2), M partial GPR models complete off-line modeling, defined as GPRm,m＝1,…,M；
Testing sample set for new input
Where T represents the number of newly input test sample sets, NtNumber of samples, x, representing the t-th new input test sample sett,iRepresents XtThe ith sample point; GPRmFor xt,iPredicted output of (2)
Sum variance
Respectively, as follows:
in the formula (I), the compound is shown in the specification,
representing the covariance between the new input test sample and the training sample; k is a radical ofm,ti＝C(xt,i,xt,i) Is the covariance of the newly input test sample;
for a new input test sample point, calculating and obtaining M online prediction information based on the local GPR model from the formula (3) and the formula (4);
(3) establishing a weighted GPR model of the discharge flow rate of the reciprocating type mixed delivery pump;
based on Bayesian inference, conditional probability P (GPR) is proposedm|xt,i) To GPRmEach sample point x of the model and new input sett,iThe relationship between the two is evaluated; p (GPR)m|xt,i) The calculation is as follows:
in the formula, P (GPR)m) And P (x)t,i|GPRm) Prior probability and conditional probability, respectively; when there is no process prior knowledge, P (GPR)m|xt,i) Expressed as:
for new input test sample point x based on probability analysis methodt,iIn terms of P (GPR)m|xt,i) Where M is 1, …, the larger M, the more GPRmThe more appropriate the model is to predict it;
merging the M partial GPR model pairs xt,iProbability information of prediction, weighting prediction values of GPR model
And its variance
The expression is as follows:
(4) establishing an instant GPR model of the discharge flow rate of the reciprocating type mixing delivery pump;
(4.1) from sample set S ═ { xn,ynN is 1, …, where N is the total number of sample sets and is the new input test sample point xt,iSelecting a suitable similar modeling sample; definition of
ηt,ni＝exp(-||xn-xt,i||),n＝1,…,N (9)
Describing a modeled sample point xnAnd new input test sample point xt,iThe similarity relationship of (1); etat,niThe larger the value is, the more similar the relationship between the two is; thus, by setting the appropriate threshold λ, by formula
ηt,ni＞λ (10)
Testing sample points x for new inputst,iSelecting a proper similar modeling sample set;
(4.2) establishing an instant GPR model by using the formula (1) and the formula (2) based on the similar modeling sample set selected by the formula (10); calculating to obtain the sample point x of the instantaneous GPR model from the formula (3) and the formula (4)t,iPredicted value of (2)
And its variance
(5) Automatically selecting a suitable prediction model for each new input test sample point based on the prediction probability information;
(5.1) for each new input test sample point xt,iSelecting a suitable local GPR prediction model;
based on the probability information provided by equation (6), there is a model of maximum conditional probability MCP, i.e.
MCPt,i＝maxP(GPRm|xt,i),m＝1,…,M (11)
Testing sample points x for new inputst,iBest fit of the Chinese traditional medicineThe appropriate local GPR prediction model, the corresponding predicted values and their variances are recorded as
And
(5.2) test sample points x for each new input from the local, weighted and immediate GPR modelt,iSelecting a proper prediction model;
the prediction variance is used to describe the input test sample point xt,iAnd uncertainty of its prediction model; if an inappropriate model is used for the new input test sample xt,iIf prediction is performed, the corresponding variance value is large; based on this, can be selected from
And
in which the model with the minimum variance MV is selected, i.e.
I.e. new input test sample point xt,iThe most appropriate predictive model;
(6) and (5) repeating the steps (2) to (5), finding the most appropriate prediction model for each input test sample point under the new working condition from the local GPR model, the weighted GPR model and the instant GPR model, and further obtaining the discharge flow rate curve of the reciprocating type multiphase pump under the new working condition.
2. The model self-selecting reciprocating multiphase pump discharge flow rate prediction method of claim 1, wherein in step (5), each input test sample point x is described by the proposed maximum conditional probability and prediction variancet,iAnd the uncertainty of its predictive model,so that a suitable model is selected for prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810288855.7A CN108549757B (en) | 2018-04-03 | 2018-04-03 | Model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810288855.7A CN108549757B (en) | 2018-04-03 | 2018-04-03 | Model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108549757A CN108549757A (en) | 2018-09-18 |
| CN108549757B true CN108549757B (en) | 2021-10-26 |
Family
ID=63513887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810288855.7A Active CN108549757B (en) | 2018-04-03 | 2018-04-03 | Model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108549757B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446741B (en) * | 2018-12-21 | 2023-04-18 | 浙江工业大学 | Modeling and predicting method for instantaneous temperature characteristic of pump cavity of mixed transportation pump |
| CN113553673B (en) * | 2021-07-21 | 2023-03-21 | 浙江工业大学 | Centrifugal pump efficiency prediction method based on data-driven modeling |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103884075A (en) * | 2014-01-06 | 2014-06-25 | 浙江工业大学 | Computational fluid dynamics and energy prediction hybrid based greenhouse energy-saving control method |
| CN104699894A (en) * | 2015-01-26 | 2015-06-10 | 江南大学 | JITL (just-in-time learning) based multi-model fusion modeling method adopting GPR (Gaussian process regression) |
| CN105913078A (en) * | 2016-04-07 | 2016-08-31 | 江南大学 | Multi-mode soft measurement method for improving adaptive affine propagation clustering |
| WO2017188501A1 (en) * | 2016-04-29 | 2017-11-02 | 경희대학교 산학협력단 | Method for recovering original signal in reduced complexity ds-cdma system |
| CN107451101A (en) * | 2017-07-21 | 2017-12-08 | 江南大学 | It is a kind of to be layered integrated Gaussian process recurrence soft-measuring modeling method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090150126A1 (en) * | 2007-12-10 | 2009-06-11 | Yahoo! Inc. | System and method for sparse gaussian process regression using predictive measures |
| CN104778298B (en) * | 2015-01-26 | 2017-09-19 | 江南大学 | Gaussian process regression soft sensor modeling method based on EGMM |
| CN106056127A (en) * | 2016-04-07 | 2016-10-26 | 江南大学 | GPR (gaussian process regression) online soft measurement method with model updating |
-
2018
- 2018-04-03 CN CN201810288855.7A patent/CN108549757B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103884075A (en) * | 2014-01-06 | 2014-06-25 | 浙江工业大学 | Computational fluid dynamics and energy prediction hybrid based greenhouse energy-saving control method |
| CN104699894A (en) * | 2015-01-26 | 2015-06-10 | 江南大学 | JITL (just-in-time learning) based multi-model fusion modeling method adopting GPR (Gaussian process regression) |
| CN105913078A (en) * | 2016-04-07 | 2016-08-31 | 江南大学 | Multi-mode soft measurement method for improving adaptive affine propagation clustering |
| WO2017188501A1 (en) * | 2016-04-29 | 2017-11-02 | 경희대학교 산학협력단 | Method for recovering original signal in reduced complexity ds-cdma system |
| CN107451101A (en) * | 2017-07-21 | 2017-12-08 | 江南大学 | It is a kind of to be layered integrated Gaussian process recurrence soft-measuring modeling method |
Non-Patent Citations (2)
| Title |
|---|
| Auto-Switch Gaussian Process Regression-Based Probabilistic Soft Sensors for Industrial Multigrade Processes with Transitions;Yi Liu 等;《Industrial & Engineering Chemistry Research》;20150413;第54卷(第18期);5037–5047 * |
| 基于即时学习的复杂非线性过程软测量建模及应用;袁小锋;《中国博士学位论文全文数据库 信息科技辑》;20170815(第8期);I140-24 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108549757A (en) | 2018-09-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ping et al. | Introducing machine learning and hybrid algorithm for prediction and optimization of multistage centrifugal pump in an ORC system | |
| CN107918709B (en) | Prediction method for transient opening height of one-way valve for multiphase mixed transportation pump | |
| CN106677763B (en) | Dynamic integrated modeling-based oil well working fluid level prediction method | |
| Pei et al. | Optimization on the impeller of a low-specific-speed centrifugal pump for hydraulic performance improvement | |
| CN104481496B (en) | Fault diagnosis method of sucker-rod pump well | |
| Deng et al. | Actively exploring informative data for smart modeling of industrial multiphase flow processes | |
| Vicario et al. | Meta‐models in computer experiments: Kriging versus artificial neural networks | |
| CN108549757B (en) | Model self-selection reciprocating type mixed transportation pump discharge flow rate prediction method | |
| CN109446741B (en) | Modeling and predicting method for instantaneous temperature characteristic of pump cavity of mixed transportation pump | |
| CN108510120B (en) | Mixed modeling and prediction method for discharge flow rate of reciprocating type mixed transmission pump | |
| Lin et al. | Neural architecture design for gpu-efficient networks | |
| Chen et al. | Knowledge-based turbomachinery design system via a deep neural network and multi-output Gaussian process | |
| Al-Rbaihat et al. | Performance evaluation of supersonic flow for variable geometry radial ejector through CFD models based on DES-turbulence models, GPR machine learning, and MPA optimization | |
| Deng et al. | Active learning for modeling and prediction of dynamical fluid processes | |
| Deng et al. | Hybrid model for discharge flow rate prediction of reciprocating multiphase pumps | |
| CN108694293B (en) | An Active Modeling and Prediction Method for Flow Rate of Reciprocating Mixed Pump | |
| CN113283039B (en) | Engine exhaust system optimization method, device, medium and electronic equipment | |
| CN113379103B (en) | Prediction method of pump equipment internal flow field based on reduced order model | |
| Chu et al. | Rapid modeling method for performance prediction of centrifugal compressor based on model migration and SVM | |
| CN113343390B (en) | Engine linearization modeling method based on neural network nonlinear model | |
| CN116401755A (en) | CNN-LSTM-Attention network structure-based solid rocket engine total thrust prediction method | |
| Chen et al. | “No Free Lunch” in Neural Architectures? A Joint Analysis of Expressivity, Convergence, and Generalization | |
| Deng et al. | Special probabilistic prediction model for temperature characteristics of dynamic fluid processes | |
| Ortikkhuzhayev et al. | Gas dynamics modelling of a self-acting reciprocating compressor discharge valve in hyperworks CFD | |
| Razaaly et al. | Uncertainty Quantification of an ORC turbine blade under a low quantile constrain |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |