CN110441065A - Gas turbine online test method and device based on LSTM - Google Patents
Gas turbine online test method and device based on LSTM Download PDFInfo
- Publication number
- CN110441065A CN110441065A CN201910601564.3A CN201910601564A CN110441065A CN 110441065 A CN110441065 A CN 110441065A CN 201910601564 A CN201910601564 A CN 201910601564A CN 110441065 A CN110441065 A CN 110441065A
- Authority
- CN
- China
- Prior art keywords
- gas turbine
- data
- lstm
- value
- training
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M15/00—Testing of engines
- G01M15/14—Testing gas-turbine engines or jet-propulsion engines
-
- 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
Abstract
The invention belongs to power plant safety control system field more particularly to a kind of gas turbine online test methods and device based on LSTM.It is characterized by comprising following steps: data acquisition；Normalized；Feature extraction；Training LSTM abnormality detection model；Detection model after needing the data predicted to be input to training is obtained model predication value, makes the difference and seek absolute value predicted value and sensor measured value by abnormal on-line checking, if absolute value has been more than given threshold values, then determining to be abnormal.The present invention is suitable for being spaced and postpone relatively long critical event in processing and predicted time sequence, and the analysis suitable for time series is fitted.The hiding feature in the thought of deep learning and the data information of the next automatic selecting extraction device sensor detection of technology is made full use of, and then realizes the online abnormality detection based on the real-time measuring point data of gas turbine.Present invention acquisition data volume is big, and analytical error is small, and early warning result accuracy rate is high.
Description
Technical field
The invention belongs to power plant safety control system field more particularly to a kind of gas turbine on-line checkings based on LSTM
Method and apparatus.
Background technique
Gas turbine is one of currently the most important dynamic power machine, has in fields such as aviation, electric power, ships and widely answers
With.Gas turbine structure is complicated and working environment is severe, is easy to cause all kinds of failures, if cannot find and repair in time, meeting
Seriously affect the safety and reliability of its operation.As gas turbine application amount increases, pass of the people to its working condition
Note is also more and more, and the shutdown once gas turbine breaks down will affect power system stability, cause huge economic loss,
Even influence national economy stable development, schedule has also just been put on for the research of turbine engine failure.
In the prior art, two classes can be substantially divided into for the detection method of gas turbine exception.One kind is using machine
The method for managing model establishes physical model by macroscopic property based on gas and thermodynamic principles, and with model calculating
Every KPI index of gas turbine, and compared with measured value.If measured value and theoretical value have relatively large deviation, then it is assumed that
Gas turbine exists abnormal.When the main problem of mechanism model is to establish analysis model with physics principle, exist a large amount of
Hypotheses and simplified condition, be not suitable for truth under complication system.
With the equipment detection data obtained in equipment running process increasingly multiplicity and equipment self structure and operation
Environment becomes increasingly complex, and fully understands the operation mechanism of equipment and the relevant data characteristics difficulty of extract equipment malfunction
Higher and higher, another kind of abnormality detection technology attempts the method for maintenance data analysis and machine learning to establish mathematics mould thus
Type, automation, the intelligentized mapping relations found between data characteristics and abnormal patterns, promotes the accurate of method for detecting abnormality
Property.Common machine learning method includes fuzzy logic, support vector machines (SVM), artificial neural network etc..Machine learning
Method is capable of the data of abundant mined information itself, realizes data-driven to the greatest extent, reduces human intervention.However, due to
Gas turbine is shorter in the time that thermoelectricity is able to apply, and fault type is complicated various and repetition case is less, and failure mechanism is difficult to
It fully understands, results in and be difficult accurately to extract and define the correlated characteristic for abnormality detection, it is difficult to guarantee abnormality detection
Accuracy.
Chinese patent 201410745943.7 discloses a kind of adaptive quantity sub-neural network steam turbine failure trend prediction
Method.The method improves three layers of traditional BP neural network model, introduces quantum nerve network, right in input layer
Different historical datas carry out the analysis of trend contribution, reinforce latest data to the influence power of trend, increase input layer to output layer
Be directly connected to weigh, excitation function is adaptively adjusted according to signal characteristic in output layer, to improve convergence rate and precision of prediction；
The method for introducing adaptive learning efficiency, to improve convergence rate.This method has good reliability and robustness, is to solve
The key technology research of steam turbine failure trend prediction can be widely used in steam turbine failure trend prediction.Existing for it
Deficiency is that prediction error is larger, has larger impact to subsequent control.
Summary of the invention
In view of the problems of the existing technology, small, the high combustion based on LSTM of accuracy rate that the present invention provides a kind of errors
Gas-turbine online test method, and the device using this method.
The invention is realized in this way a kind of gas turbine online test method based on LSTM, it is characterised in that: including
Following steps:
Step 1: data acquisition obtains real-time measuring point data from the monitoring of software of gas turbine；
Step 2: normalized, the dimension of unified above-mentioned each measuring point data；
Step 3: feature extraction, the correlation between data set after calculating normalized are special to data after normalization
Sign carries out linear change, by data from high-dimensional space reflection to low dimensional space, reduces data dimension, extracts main spy
Sign；
Step 4: abnormality detection model training, using step 3 treated data as training dataset, training LSTM is different
Normal detection model；
Step 5: the detection model after needing the data predicted to be input to training it is pre- to be obtained model by abnormal on-line checking
Predicted value and sensor measured value are made the difference and seek absolute value by measured value, if absolute value has been more than given threshold values, then determining to occur
It is abnormal.
The dimension of uniform data in the step two is calculated flat in initial data using Z-score standardized method
Mean μ and standard deviation sigma carry out the normalization of data, and formula is as follows:
Data fit mean value after the processing of this method is 0, the standardized normal distribution that standard deviation is 1.
The step three the following steps are included:
Step 3-1: correlation matrix is calculated, the data { X of N number of measuring point is inputted1,X2,...,XN, it is counted according to formula (2)
The Pearson correlation coefficient between every two measuring point is calculated, correlation matrix is formed,
Wherein rij (i, j=1,2 ..., n) indicates original vector Xi, XjThe related coefficient of correlation degree；
Wherein, rijFor one-dimensional vector XiAnd XjRelated coefficient, XikIndicate one-dimensional vector XiIn k-th of element, XjkTable
Show one-dimensional vector XjIn k-th of element, indicate one-dimensional vector XiAverage value, indicate one-dimensional vector XjAverage value, calculate public
Formula are as follows:
Step 3-2: calculating characteristic value and feature vector, first solution characteristic equation | λ E-R |=0 (wherein E is unit vector,
R indicates correlation matrix) characteristic value is found out, and sort by size, corresponding eigenvalue λ is then found out respectivelyi(i=1,2,
... feature vector U n)i(i=1,2 ..., n), calculate principal component matrix Y according to following formula,
Step 3-3: calculating the information contribution rate of each feature vector, calculates eigenvalue λi(i=1,2 ..., n) accumulative
Variance contribution ratio CPV (cumulative percent variance), formula is as follows:
The step four the following steps are included:
Step 4-1: establishing neural network model, the data acquisition system after giving feature extraction:WhereinFor the sensor data vector of one group of input, yiFor instruction
Prediction sensor values label when practicing, training LSTM neural network model make:
Wherein M indicates that network model, Func indicate what network model learntTo the mapping relations of y；
Step 4-2: Select Error function selects mean square error error function, is defined as follows shown:
Step 4-3: determining activation primitive, using tahn activation primitive, control will be exported in the range of [0,1],
To protect and control cell state；Softsign activation primitive is used in the input of LSTM neuron；
Step 4-4: training neural network model, in the initialization weight stage of neural network, in [- 0.08,0.08]
All weight parameters of equality initialization in range allow neural network model to remember all memories in the training initial stage,
It is 1.0 that LSTM, which is arranged, to forget the initial bias value of door, and the initial value of input gate and out gate is the random floating-point on [0,1] section
Numerical value, then using stochastic gradient descent SGD training network, learning rate 0.001, decay factor 0.95, training pattern 50
Wheel, and 10 wheel after each round learning rate all multiplied by decay factor 0.95.
A kind of gas turbine on-line measuring device based on LSTM for installing the above method, which is characterized in that including electrical
The memory, processor, I/O equipment and the warning device that are stored with above method and realize program of connection, the connection installation of I/O equipment
The computer and/or network of the monitoring of software of gas turbine access and obtain real-time measuring point data.
The processor passes through wireless transmission connection handset user end.
Advantages of the present invention and good effect are as follows:
The present invention is based on LSTM technologies to construct abnormality detection model, is suitable for being spaced and prolonging in processing and predicted time sequence
Critical event relatively long late, the analysis suitable for time series are fitted.The thought and technology for making full use of deep learning are come
The hiding feature in the data information of extracting device sensor detection is automatically selected, and then realizes and is based on the real-time measuring point of gas turbine
The online abnormality detection of data.Present invention acquisition data volume is big, and analytical error is small, and early warning result accuracy rate is high.
Detailed description of the invention
Fig. 1 is flow diagram of the invention；
Fig. 2 is neural network schematic diagram of the invention；
Fig. 3 is tanh the and softsign function schematic diagram of the embodiment of the present invention；
Fig. 4 is the gas turbine blower delivery temperature outlier detection effect of the embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
Application principle of the invention is explained in detail with reference to the accompanying drawing.
Embodiment 1:
As shown in Figure 1-3, the gas turbine exception online test method and device based on LSTM proposed, including following step
It is rapid:
Step 1: data acquisition.Gas turbine is mainly made of the big component of compressor, combustion chamber and turbine three.From combustion gas
Main measuring point data is selected in the monitoring of software of turbine, the input as LSTM neural network is analyzed.Measuring point data includes:
The position GT IGV (angle), the position GT IGV, air humidity, gas-turbine compressor intake air temperature, gas turbine inlet air filtration
Device pressure difference measuring point (1) and measuring point (2), gas turbine inlet air filtration device internal pressure, gas turbine inlet air static pressure, GT IGV
Set -1 measuring point (1) and measuring point (2), gas-turbine compressor intake air temperature measuring point (1), measuring point (2) and measuring point (3), gas turbine
Compressor outlet air themperature, gas-turbine combustion chamber housing pressure measuring point (1), measuring point (2) and measuring point (3), combustion engine fuel A are female
Pipe pressure, combustion engine fuel guide main-piping pressure, combustion engine main fuel main-piping pressure, combustion engine fuel top main-piping pressure, gas turbine
Combustion chamber bypasses valve position, gas turbine blades channel maximum temperature, gas turbine blades channel minimum temperature, gas turbine leaf
The variation of piece channel maximum temperature, the variation of gas turbine blades channel minimum temperature, combustion engine delivery temperature maximum value, combustion engine exhaust temperature
Spend minimum value.
There are multiple setting multi-measuring points in the above acquisition data, these measuring points do not have essential distinction, are surveyed with multiple sensors
Same achievement data still has correct reading, redundant measurement after guaranteeing a certain sensor error.
Step 2: data normalization.Initial data is calculated using Z-score standardized method for the dimension of uniform data
In average value mu and standard deviation sigma carry out the normalization of data, formula is as follows:
Data fit mean value after the processing of this method is 0, the standardized normal distribution that standard deviation is 1.
Step 3: feature extraction.The correlation of initial data, while the method based on principal component analysis are calculated, to original number
According to feature carry out linear change, main feature is extracted from initial data.It is asked in statistical analysis of the research towards multivariable
When topic, variable is more, and calculation amount and the complexity for increasing problem analysis are bigger.Accordingly it is desirable to carrying out quantitative analysis
During, it finds and extracts key variables.The variable being related to is less, but the information content for including is enough.Principal component point
Analysis utilizes the thought of dimensionality reduction, by constructing the appropriate linear combination of original index, generates a series of mutual not linearly related synthesis
Property index, therefrom selects a few new overall target, and make them as much as possible containing contained by original index
Information goes the information for explaining original data with less index.Concrete methods of realizing is incited somebody to action by a series of mathematic(al) manipulation
One group of given correlated variables changes into another group of incoherent variable by linear transformation, these new variables according to variance successively
The sequence arrangement successively decreased.It keeps total side of variable constant in mathematic(al) manipulation, makes the first variable that there is maximum variance, referred to as the
One principal component, bivariate variance time is big, and uncorrelated with the first variable, referred to as Second principal component,.
The step of Feature Extraction Method based on principal component analysis, is as follows:
Step 3-1: correlation matrix is calculated.Input the data { X of N number of measuring point1,X2,...,XN, it is counted according to formula (2)
The Pearson correlation coefficient between every two measuring point is calculated, correlation matrix is formed.Wherein rij (i, j=1,2 ..., n) is indicated
Original vector Xi, XjThe related coefficient of correlation degree.
Wherein, rijFor one-dimensional vector XiAnd XjRelated coefficient, XikIndicate one-dimensional vector XiIn k-th of element, XjkTable
Show one-dimensional vector XjIn k-th of element, indicate one-dimensional vector XiAverage value, indicate one-dimensional vector XjAverage value, calculate public
Formula are as follows:
Step 3-2: characteristic value and feature vector are calculated.Characteristic equation is solved first | λ E-R |=0 (wherein E is unit vector,
R indicates correlation matrix) characteristic value is found out, and sort by size, corresponding eigenvalue λ is then found out respectivelyi(i=1,
2 ..., feature vector U n)i(i=1,2 ..., n).Principal component matrix Y is calculated according to following formula.
Step 3-3: calculating the information contribution rate of each feature vector, calculates eigenvalue λi(i=1,2 ..., n) accumulative
Variance contribution ratio CPV (cumulative percent variance), formula is as follows:
Step 4: Fault Model of the training based on LSTM.Based on the feature that preceding step extracts, training is based on LSTM
The abnormality detection model of neural network.
Step 4-1: neural network model is established.Feature extraction based on step 3 is as a result, carry out mould using supervised learning
Type training.Data acquisition system after given feature extraction:Wherein
For the sensor data vector of one group of input, yiPrediction sensor values label when to train.Training LSTM as shown in Figure 2
Neural network model makes:
Wherein M indicates that network model, Func indicate what network model learntTo the mapping relations of y.
Step 4-2: Select Error function.Neural network model carries out numerical value after needing to learn sensing data
The operating status of prediction and then auxiliary judgment equipment, thus mean square error (Mean Squared Error) error function is selected,
Shown in it is defined as follows:
Step 4-3: activation primitive is determined.The activation primitive of door control unit in LSTM neural network activates letter using tahn
Number, control will be exported in the range of [0,1], to protect and control cell state；There are also in the input of LSTM neuron
It waits and uses softsign activation primitive.Tahn activation primitive and softsign activation primitive are as shown in figure 3, its definition is as follows respectively
Shown in face formula (8), (9).
Step 4-4: training neural network model.In the initialization weight stage of neural network, in [- 0.08,0.08]
All weight parameters of equality initialization in range allow model to remember all memories in the training initial stage, and LSTM is arranged
The initial bias value for forgeing door is 1.0, and the initial value of input gate and out gate is the random floating point value on [0,1] section.So
Afterwards using micro- batch stochastic gradient descent training network, learning rate 0.001, decay factor 0.95.Training pattern 50 is taken turns, and
Each round learning rate is all multiplied by decay factor 0.95 after 10 wheels.
Step 5: abnormal on-line prediction.Give one group of input dataPredicted value is exported according to model MCalculate prediction
Difference d between value and true value:
It sets a threshold epsilon to be used to judge equipment state, thinks that equipment is in normal operating condition if d≤ε, it is on the contrary
It is abnormal then to think that equipment occurs, carries out early warning.
Fig. 4 illustrates the abnormality detection effect of gas turbine blower delivery temperature measuring point.As Fig. 4 red circle marks institute
Show, lower section lines representative model predicted value, top lines represent measured value.It can be seen from the figure that model predication value and actual measurement
There is bigger deviations between value, it is believed that equipment is abnormal here, to carry out early warning.
Embodiment 2:
A kind of gas turbine on-line measuring device based on LSTM for installing the above method, being stored with including electrical connection
Above method realizes memory, processor, I/O equipment and the warning device of program, the prison of I/O equipment connection installation gas turbine
The computer and/or network for surveying software, access and obtain real-time measuring point data.
Processor passes through wireless transmission connection handset user end.Pass through handheld device remote monitor early warning.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (8)
1. a kind of gas turbine online test method based on LSTM, it is characterised in that: the following steps are included:
Step 1: data acquisition obtains real-time measuring point data from the monitoring of software of gas turbine；
Step 2: normalized, the dimension of unified above-mentioned each measuring point data；
Step 3: feature extraction, calculate normalized after data set between correlation, to data characteristics after normalization into
Row linear change reduces data dimension, extracts main feature by data from high-dimensional space reflection to low dimensional space；
Step 4: abnormality detection model training, using step 3 treated data as training dataset, training LSTM is examined extremely
Survey model；
Step 5: the detection model after needing the data predicted to be input to training is obtained model prediction by abnormal on-line checking
Value, predicted value and sensor measured value are made the difference and seek absolute value, if absolute value has been more than given threshold values, then determining to occur different
Often.
2. the gas turbine online test method based on LSTM as described in claim 1, which is characterized in that the step two
The dimension of middle uniform data calculates average value mu and standard deviation sigma in initial data using Z-score standardized method to carry out
The normalization of data, formula are as follows:
Data fit mean value after the processing of this method is 0, the standardized normal distribution that standard deviation is 1.
3. the gas turbine online test method based on LSTM as described in claim 1, which is characterized in that the step three
The following steps are included:
Step 3-1: correlation matrix is calculated, the data { X of N number of measuring point is inputted1,X2,...,XN, it is calculated according to formula (2) every
Pearson correlation coefficient between two measuring points forms correlation matrix,
Wherein rij (i, j=1,2 ..., n) indicates original vector Xi, XjThe related coefficient of correlation degree；
Wherein, rijFor one-dimensional vector XiAnd XjRelated coefficient, XikIndicate one-dimensional vector XiIn k-th of element, XjkIndicate one
Dimensional vector XjIn k-th of element, indicate one-dimensional vector XiAverage value, indicate one-dimensional vector XjAverage value, calculation formula
Are as follows:
Step 3-2: calculating characteristic value and feature vector, first solution characteristic equation | λ E-R |=0 (wherein E is unit vector, R table
Show correlation matrix) characteristic value is found out, and sort by size, corresponding eigenvalue λ is then found out respectivelyi(i=1,2 ..., n)
Feature vector Ui(i=1,2 ..., n), calculate principal component matrix Y according to following formula,
Step 3-3: calculating the information contribution rate of each feature vector, calculates eigenvalue λi=(i=1,2 ..., accumulative side n)
Poor contribution rate CPV (cumulative percent variance), formula is as follows:
4. the gas turbine online test method based on LSTM as described in claim 1, which is characterized in that the step four
The following steps are included:
Step 4-1: establishing neural network model, the data acquisition system after giving feature extraction:WhereinFor the sensor data vector of one group of input, yiFor instruction
Prediction sensor values label when practicing, training LSTM neural network model make:
Wherein M indicates that network model, Func indicate what network model learntTo the mapping relations of y；
Step 4-2: Select Error function selects mean square error error function, is defined as follows shown:
Step 4-3: determining activation primitive, using tahn activation primitive, will export control in the range of [0,1], to protect
Shield and control cell state；Softsign activation primitive is used in the input of LSTM neuron；
Step 4-4: training neural network model, the range in the initialization weight stage of neural network, in [- 0.08,0.08]
Interior all weight parameters of equality initialization allow neural network model to remember all memories in the training initial stage, are arranged
The initial bias value that LSTM forgets door is 1.0, and the initial value of input gate and out gate is the random floating point on [0,1] section
Value, then using stochastic gradient descent SGD training network, learning rate 0.001, decay factor 0.95, training pattern 50
Wheel, and 10 wheel after each round learning rate all multiplied by decay factor 0.95.
5. the gas turbine online test method based on LSTM as described in claim 1, which is characterized in that the step
One, main measuring point data is selected from the monitoring of software of gas turbine, measuring point data includes: the position GT IGV, air humidity, combustion
Gas turbine compressor intake air temperature, gas turbine inlet air filtration device pressure difference measuring point, gas turbine inlet air filtration device internal pressure, combustion
Gas-turbine air admission static pressure, gas-turbine compressor intake air temperature measuring point, gas-turbine compressor outlet air temperature, gas turbine
Burning chamber shell pressure-measuring-point, combustion engine fuel A main-piping pressure, combustion engine fuel guide main-piping pressure, combustion engine main fuel main-piping pressure,
Main-piping pressure, gas-turbine combustion chamber bypass valve position, gas turbine blades channel maximum temperature, combustion gas wheel at the top of combustion engine fuel
Machine blade path minimum temperature, gas turbine blades channel maximum temperature change, gas turbine blades channel minimum temperature changes,
Combustion engine delivery temperature maximum value, combustion engine delivery temperature minimum value.
6. the gas turbine online test method based on LSTM as claimed in claim 5, which is characterized in that the combustion gas wheel
Machine air intake filter pressure difference measuring point, the position GT IGV measuring point, gas-turbine compressor intake air temperature measuring point and gas turbine combustion
Room housing pressure-measuring-point is respectively provided with 2 or more measuring points.
7. a kind of gas turbine on-line measuring device being equipped with described in claim 1-4 based on LSTM, which is characterized in that including
The memory, processor, I/O equipment and the warning device that are stored with above method and realize program of electrical connection, the connection of I/O equipment
The computer and/or network for installing the monitoring of software of gas turbine, access and obtain real-time measuring point data.
8. the gas turbine on-line measuring device according to claim 5 based on LSTM, which is characterized in that the processing
Device passes through wireless transmission connection handset user end.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910601564.3A CN110441065B (en) | 2019-07-04 | 2019-07-04 | Gas turbine on-line detection method and device based on LSTM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910601564.3A CN110441065B (en) | 2019-07-04 | 2019-07-04 | Gas turbine on-line detection method and device based on LSTM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110441065A true CN110441065A (en) | 2019-11-12 |
| CN110441065B CN110441065B (en) | 2022-02-08 |
Family
ID=68428472
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910601564.3A Active CN110441065B (en) | 2019-07-04 | 2019-07-04 | Gas turbine on-line detection method and device based on LSTM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110441065B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110926532A (en) * | 2019-11-29 | 2020-03-27 | 四川省生态环境科学研究院 | Digital monitoring system of city raise dust based on big data |
| CN110987448A (en) * | 2019-12-05 | 2020-04-10 | 潍柴动力股份有限公司 | Engine air inlet state monitoring method, device and equipment |
| CN111272436A (en) * | 2020-02-11 | 2020-06-12 | 上海交通大学 | Gas turbine NOx prediction emission monitoring system and method |
| CN111275288A (en) * | 2019-12-31 | 2020-06-12 | 华电国际电力股份有限公司十里泉发电厂 | XGboost-based multi-dimensional data anomaly detection method and device |
| CN111931420A (en) * | 2020-08-07 | 2020-11-13 | 合肥工业大学 | Gas turbine fault prediction method based on nuclear regeneration Hilbert space |
| CN111985610A (en) * | 2020-07-15 | 2020-11-24 | 中国石油大学(北京) | System and method for predicting pumping efficiency of oil pumping well based on time sequence data |
| CN112000015A (en) * | 2020-09-07 | 2020-11-27 | 华北电力大学 | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on LSTM and bio-excitation neural network |
| CN112668200A (en) * | 2021-01-06 | 2021-04-16 | 西安理工大学 | Method and system for analyzing safety of special equipment |
| WO2021098384A1 (en) * | 2019-11-18 | 2021-05-27 | 中国银联股份有限公司 | Data abnormality detection method and apparatus |
| CN113033663A (en) * | 2021-03-26 | 2021-06-25 | 同济大学 | Automatic container terminal equipment health prediction method based on machine learning |
| CN113657628A (en) * | 2021-08-20 | 2021-11-16 | 武汉霖汐科技有限公司 | Industrial equipment monitoring method and system, electronic equipment and storage medium |
| CN114115197A (en) * | 2021-11-24 | 2022-03-01 | 浙大城市学院 | Gas turbine state maintenance decision making system, method and storage medium |
| WO2022048168A1 (en) * | 2020-09-03 | 2022-03-10 | 上海上讯信息技术股份有限公司 | Training method and device for failure prediction neural network model |
| CN114184389A (en) * | 2021-11-18 | 2022-03-15 | 中国船舶重工集团公司第七O三研究所无锡分部 | Cooling effectiveness testing method for exhaust, injection and ventilation in marine gas turbine box |
| CN114239708A (en) * | 2021-12-13 | 2022-03-25 | 中国船舶重工集团公司第七0三研究所 | Combustion engine abnormity detection method based on quality control chart theory |
| WO2022062339A1 (en) * | 2020-09-23 | 2022-03-31 | 江苏科技大学 | System and method for controlling air valve of variable air volume cabin unit |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550700A (en) * | 2015-12-08 | 2016-05-04 | 国网山东省电力公司电力科学研究院 | Time series data cleaning method based on correlation analysis and principal component analysis |
| CN105719094A (en) * | 2016-01-27 | 2016-06-29 | 刘冰 | State evaluation method of power transmission equipment |
| CN107044349A (en) * | 2017-04-17 | 2017-08-15 | 清华大学深圳研究生院 | A kind of breakdown in the motor Forecasting Methodology |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
| CN108398271A (en) * | 2017-02-07 | 2018-08-14 | 山东云舜智能科技有限公司 | A kind of engine state monitor method based on vibration and nerual network technique |
| CN109086888A (en) * | 2018-09-05 | 2018-12-25 | 北京机械设备研究所 | Automobile engine failure judgment method and its device based on voice recognition |
| CN109555566A (en) * | 2018-12-20 | 2019-04-02 | 西安交通大学 | A kind of turbine rotor method for diagnosing faults based on LSTM |
-
2019
- 2019-07-04 CN CN201910601564.3A patent/CN110441065B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550700A (en) * | 2015-12-08 | 2016-05-04 | 国网山东省电力公司电力科学研究院 | Time series data cleaning method based on correlation analysis and principal component analysis |
| CN105719094A (en) * | 2016-01-27 | 2016-06-29 | 刘冰 | State evaluation method of power transmission equipment |
| CN108398271A (en) * | 2017-02-07 | 2018-08-14 | 山东云舜智能科技有限公司 | A kind of engine state monitor method based on vibration and nerual network technique |
| CN107044349A (en) * | 2017-04-17 | 2017-08-15 | 清华大学深圳研究生院 | A kind of breakdown in the motor Forecasting Methodology |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
| CN109086888A (en) * | 2018-09-05 | 2018-12-25 | 北京机械设备研究所 | Automobile engine failure judgment method and its device based on voice recognition |
| CN109555566A (en) * | 2018-12-20 | 2019-04-02 | 西安交通大学 | A kind of turbine rotor method for diagnosing faults based on LSTM |
Non-Patent Citations (1)
| Title |
|---|
| 唐贵基等: "基于ＡＬＩＦ－ＨＴ的汽轮发电机组转子故障诊断", 《动力工程学报》 * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021098384A1 (en) * | 2019-11-18 | 2021-05-27 | 中国银联股份有限公司 | Data abnormality detection method and apparatus |
| CN110926532A (en) * | 2019-11-29 | 2020-03-27 | 四川省生态环境科学研究院 | Digital monitoring system of city raise dust based on big data |
| CN110987448A (en) * | 2019-12-05 | 2020-04-10 | 潍柴动力股份有限公司 | Engine air inlet state monitoring method, device and equipment |
| CN111275288A (en) * | 2019-12-31 | 2020-06-12 | 华电国际电力股份有限公司十里泉发电厂 | XGboost-based multi-dimensional data anomaly detection method and device |
| CN111275288B (en) * | 2019-12-31 | 2023-12-26 | 华电国际电力股份有限公司十里泉发电厂 | XGBoost-based multidimensional data anomaly detection method and device |
| CN111272436A (en) * | 2020-02-11 | 2020-06-12 | 上海交通大学 | Gas turbine NOx prediction emission monitoring system and method |
| CN111985610A (en) * | 2020-07-15 | 2020-11-24 | 中国石油大学(北京) | System and method for predicting pumping efficiency of oil pumping well based on time sequence data |
| CN111931420A (en) * | 2020-08-07 | 2020-11-13 | 合肥工业大学 | Gas turbine fault prediction method based on nuclear regeneration Hilbert space |
| CN111931420B (en) * | 2020-08-07 | 2023-09-29 | 合肥工业大学 | Gas turbine fault prediction method based on nuclear regeneration Hilbert space |
| WO2022048168A1 (en) * | 2020-09-03 | 2022-03-10 | 上海上讯信息技术股份有限公司 | Training method and device for failure prediction neural network model |
| CN112000015A (en) * | 2020-09-07 | 2020-11-27 | 华北电力大学 | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on LSTM and bio-excitation neural network |
| CN112000015B (en) * | 2020-09-07 | 2022-08-05 | 华北电力大学 | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on LSTM and bio-excitation neural network |
| WO2022062339A1 (en) * | 2020-09-23 | 2022-03-31 | 江苏科技大学 | System and method for controlling air valve of variable air volume cabin unit |
| CN112668200B (en) * | 2021-01-06 | 2023-08-29 | 西安理工大学 | Special equipment safety analysis method and system |
| CN112668200A (en) * | 2021-01-06 | 2021-04-16 | 西安理工大学 | Method and system for analyzing safety of special equipment |
| CN113033663A (en) * | 2021-03-26 | 2021-06-25 | 同济大学 | Automatic container terminal equipment health prediction method based on machine learning |
| CN113657628A (en) * | 2021-08-20 | 2021-11-16 | 武汉霖汐科技有限公司 | Industrial equipment monitoring method and system, electronic equipment and storage medium |
| CN114184389A (en) * | 2021-11-18 | 2022-03-15 | 中国船舶重工集团公司第七O三研究所无锡分部 | Cooling effectiveness testing method for exhaust, injection and ventilation in marine gas turbine box |
| CN114184389B (en) * | 2021-11-18 | 2023-08-15 | 中国船舶重工集团公司第七O三研究所无锡分部 | Cooling effectiveness inspection method for exhaust injection ventilation in marine gas turbine box |
| CN114115197A (en) * | 2021-11-24 | 2022-03-01 | 浙大城市学院 | Gas turbine state maintenance decision making system, method and storage medium |
| CN114239708A (en) * | 2021-12-13 | 2022-03-25 | 中国船舶重工集团公司第七0三研究所 | Combustion engine abnormity detection method based on quality control chart theory |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110441065B (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110441065A (en) | Gas turbine online test method and device based on LSTM | |
| CN106709662B (en) | Power equipment operation condition division method | |
| CN110738274A (en) | nuclear power device fault diagnosis method based on data driving | |
| US11840998B2 (en) | Hydraulic turbine cavitation acoustic signal identification method based on big data machine learning | |
| CN108268905A (en) | A kind of Diagnosis Method of Transformer Faults and system based on support vector machines | |
| CN111780800B (en) | Method and system for monitoring, isolating and reconstructing sensor fault | |
| CN110618610A (en) | Gas turbine NOx standard exceeding early warning method and device based on LSTM deep learning | |
| Loboda et al. | Gas turbine fault diagnosis using probabilistic neural networks | |
| CN110348150A (en) | A kind of fault detection method based on dependent probability model | |
| CN104536290A (en) | Soft measuring method and system based on kernel principal component analysis and radial basis function neural network | |
| Ren et al. | A new reconstruction-based auto-associative neural network for fault diagnosis in nonlinear systems | |
| CN109085805A (en) | A kind of industrial process fault detection method based on multi-sampling rate Factor Analysis Model | |
| Loboda | Neural networks for gas turbine diagnosis | |
| Gu et al. | Singular value diagnosis in dam safety monitoring effect values | |
| Loboda | Gas turbine condition monitoring and diagnostics | |
| CN116383636A (en) | Coal mill fault early warning method based on PCA and LSTM fusion algorithm | |
| CN112036087A (en) | Multi-strategy fused nuclear power key equipment fault diagnosis method and system | |
| CN115375026A (en) | Method for predicting service life of aircraft engine in multiple fault modes | |
| Loboda et al. | Neural networks for gas turbine fault identification: multilayer perceptron or radial basis network? | |
| Zhang et al. | System-level anomaly detection for nuclear power plants using variational graph auto-encoders | |
| CN114046816B (en) | Sensor signal fault diagnosis method based on lightweight gradient lifting decision tree | |
| CN111563685A (en) | Power generation equipment state early warning method based on auto-associative kernel regression algorithm | |
| CN116204825A (en) | Production line equipment fault detection method based on data driving | |
| CN114997309A (en) | Water feed pump fault early warning method and device | |
| Ren et al. | Development of input training neural networks for multiple sensor fault isolation |
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 |