CN113110044A - Intelligent BIT design method for heavy-duty gas turbine control system controller module based on Elman neural network and SVM - Google Patents
Intelligent BIT design method for heavy-duty gas turbine control system controller module based on Elman neural network and SVM Download PDFInfo
- Publication number
- CN113110044A CN113110044A CN202110330069.0A CN202110330069A CN113110044A CN 113110044 A CN113110044 A CN 113110044A CN 202110330069 A CN202110330069 A CN 202110330069A CN 113110044 A CN113110044 A CN 113110044A
- Authority
- CN
- China
- Prior art keywords
- neural network
- data
- svm
- output
- controller module
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000013461 design Methods 0.000 title claims abstract description 19
- 238000003745 diagnosis Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000012706 support-vector machine Methods 0.000 claims description 47
- 230000006870 function Effects 0.000 claims description 25
- 238000010606 normalization Methods 0.000 claims description 10
- 238000003062 neural network model Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 2
- 238000011478 gradient descent method Methods 0.000 claims description 2
- 238000003672 processing method Methods 0.000 claims description 2
- 238000000926 separation method Methods 0.000 claims description 2
- 238000013507 mapping Methods 0.000 claims 1
- 230000001537 neural effect Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012938 design process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses an intelligent BIT design method of a heavy-duty gas turbine controller module based on an Elman neural network and an SVM, which comprises the steps of collecting historical operating data of the heavy-duty gas turbine controller module; data standardization processing, namely determining input and output of the Elman neural network according to a time sequence; designing and training an Elman neural network aiming at the single characteristic parameter of the controller module; inputting the processed data into a trained Elman neural network to obtain an output result, and transmitting the output result to the SVM for secondary classification; and obtaining the diagnosis result of the current characteristic parameter, and indicating the normal or fault through the switching value to finish the intelligent BIT diagnosis of the controller module. The invention improves the safety and reliability of the controller module of the heavy-duty gas turbine control system, and reduces the false alarm rate by extracting the information in the historical data of the controller module to establish the intelligent BIT model.
Description
Technical Field
The invention relates to an intelligent BIT design method of a heavy-duty gas turbine control system controller module, in particular to a BIT design method of a heavy-duty gas turbine control system controller module based on an Elman neural network and an SVM.
Background
Along with the development of scientific technology, the control system functions of a plurality of devices are gradually improved, the reliability is further improved, however, the control system electronic devices are difficult to avoid abnormal conditions and faults are difficult to find, in order to monitor the state of the control system in time and ensure that the devices are in a normal operation state, the control system is generally internally provided with a BIT technology, and the state monitoring of the control system is realized.
At present, the domestic heavy gas turbine is in a continuously developed state, the BIT technology of a controller module in a control system judges whether monitoring data are abnormal or not through threshold diagnosis or simple logic to determine the state of the controller module, the defect of high false alarm rate exists, and under the background of big data, the intelligent BIT introducing an advanced data processing mode becomes the focus of attention for improving the identification precision of the BIT.
With the advent of the artificial intelligence era, data-driven-based approaches have made breakthrough progress in many fields. The method mainly comprises the steps of acquiring information in mass data by using a data mining method, and establishing a reliable model to realize the identification of a target state. The method which is widely concerned and used most at present is the neural network, and due to the good nonlinear fitting capacity and generalization capacity of the neural network, the neural network applied to the BIT field can effectively reduce the false alarm rate, improve the stability of the heavy-duty gas turbine control system, improve the intelligent level of the controller module and reduce the maintenance cost.
Disclosure of Invention
The invention aims to provide an intelligent BIT design method of a heavy-duty gas turbine controller module based on an Elman neural network and an SVM. The intelligent monitoring method has the advantages that the real-time monitoring of the state of the controller module of the heavy gas turbine is realized from the data driving angle, the initial diagnosis of BIT is realized by utilizing an Elman neural network, and the running state of the module is determined by utilizing the high-efficiency classification characteristic of the SVM.
In order to achieve the aim, the BIT design method of the heavy-duty gas turbine controller module based on Elman and SVM of the invention comprises the following specific steps:
step 1: collecting historical data of a single analog quantity characteristic state parameter operated by a controller module in a specified time period, and collecting data of a continuous time sequence and a real operation state of a current state parameter;
step 2: carrying out normalization processing on the data, and determining the input and the output of the Elman neural network by combining the real running state;
and step 3: determining a training set and a testing set of the model by adopting a dynamic training mode according to time sequence recursion;
and 4, step 4: designing an Elman neural network according to the characteristic state parameters, determining the data running state by inputting time sequence data, and training and testing the network by using the time sequence data until the training error of the model is less than a set value;
and 5: after the model in the step 4 is trained, transmitting the output of the Elman neural network to an SVM (support vector machine) as input for secondary classification, and determining the state of time series data;
step 6: and (5) regarding the two classification results of the SVM in the step 5, wherein 0 represents normal, 1 represents fault, and the SVM output result of the time series data is used as the diagnosis result of the intelligent BIT.
And further, training the Elman neural network model well according to the historical data set of the selected characteristic parameters, classifying the Elman neural network model by SVM, judging whether the diagnosis result of the BIT reaches a preset diagnosis accuracy rate, if the diagnosis result does not reach the preset BIT fault accuracy rate, acquiring the historical data of the characteristic state parameters and the state data of the controller module again, and repeating the steps 1-6 until the diagnosis result reaches the preset diagnosis accuracy rate.
Further, the single feature state parameter time series data of the controller module respectively obtain 1000 groups, wherein the data ratio of the training test set to the test data is 4: 1.
Further, the output of the Elman neural network model of each characteristic state parameter is transmitted to the SVM for secondary classification.
Further, the output result of the SVM for each time in the input is 0 and 1, which respectively indicate that the state of the current characteristic parameter is normal or fault.
Compared with the traditional BIT, the invention has the following advantages:
according to the method, an Elman neural network model is established by utilizing a large amount of historical data generated by a heavy-duty gas turbine controller module in the operation process, and the precision of a diagnosis result is further improved by utilizing SVM classification. The method has the advantages that the method has a good effect of identifying abnormal values, extracts effective information of diagnosis results and time of BIT by using the Elman neural network, improves diagnosis precision by analyzing the relation between the diagnosis results of the previous moment and the current moment, integrates a time factor into an over-limit diagnosis mode, namely an alarm diagnosis mode, of conventional BIT, can effectively reduce false alarm rate, improve intelligence level of a gas turbine controller module, and improve reliability of equipment operation.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the methods of the prior art, a brief description of the drawings, which are needed to describe the embodiments or the prior art, is provided below.
FIG. 1 is a schematic diagram of the design process of the present invention.
FIG. 2 is a diagram of a host computer interface of a controller module under study.
Fig. 3 is a schematic diagram of data flow in the present invention.
Fig. 4 shows the input and output of the Elman neural network described in the present invention.
Fig. 5 is a schematic diagram of data processing of the Elman neural network according to the present invention.
FIG. 6 is a diagram illustrating SVM data processing according to the present invention.
Detailed Description
The design method of the present invention is described in detail below with reference to the accompanying drawings, and the present embodiment is implemented on the premise of the technical solution of the present invention, and provides a detailed implementation manner and a specific operation process, but the application scope of the present invention is not limited to the following embodiment. Other cases of applying the design method of the present invention without creative efforts by those skilled in the art belong to the protection scope of the present invention.
In the description of the present embodiment, it should be noted that the terms "step 1", "step 2", "step 3", "step 4", "step 5", "step 6", "first", "second", "third", and "fourth" are used for descriptive purposes only and are not to be construed as indicating or implying importance.
The invention provides a design method of an intelligent BIT (BIT based on an Elman neural network and a SVM) controller module of a heavy-duty gas turbine control system. The features of the embodiments described herein are provided to enable developers to achieve particular goals, and for those skilled in the art, the present invention may be incorporated by reference to enable various other forms of development and make various changes and modifications. It will be understood by those skilled in the art that, unless defined separately, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
As shown in FIG. 1, the overall flow of the intelligent BIT design method of the heavy-duty gas turbine control system controller module based on the Elman neural network and the SVM is shown. The method comprises the following specific steps:
step 1: time series data is collected for a single characteristic state parameter of controller module operation over a specified time period, the parameter being selectable from network state, temperature, CPU load, inner layer load, etc., to represent the operational state of a heavy duty gas turbine controller module, and fault information is collected.
Step 2: and determining time sequence data with the input of the Elman neural network as a state parameter, outputting the time sequence data as the state of the current parameter, and normalizing the data.
And step 3: and determining a training set and a testing set of the model by adopting a dynamic training mode according to the recursion of the time sequence.
And 4, step 4: and designing an Elman neural network according to the characteristic state parameters, determining the running state of the current parameters by inputting time series data, and training and testing the network by using the time series data of the monitoring parameters until the training error of the model is smaller than a set value.
And 5: after the model in the step 4 is trained, the output of the Elman neural network is transmitted to the SVM as input for secondary classification, each time sequence corresponds to a classification result, and the running state of the current state parameter based on the time sequence is determined.
Step 6: and (5) outputting the time sequence of the SVM generated based on the time sequence data as a BIT diagnosis result according to the classification result of the SVM in the step 5, wherein 0 represents normal and 1 represents fault, and uploading the result to an upper computer for displaying.
Further, the historical data of the controller module is state parameters of which the conventional BIT diagnosis mode is threshold diagnosis or simple logic diagnosis, the state parameters are arranged according to a time sequence and comprise characteristic state parameter historical data capable of reflecting the state of the controller module, a matrix U represents time sequence data of the state parameters, U is a matrixnData representing the nth time series, i.e. U ═ U1,u2,…un]The matrix O represents the diagnostic data of the state parameters at different times, which data are generally the switching value data, OnIndicates the diagnostic result of the nth time series, i.e. O ═ O1,o2,…on]。
Further, the normalization processing method adopts u ═ u-umin)/(umax-umin) Processing historical data, wherein u' represents data after normalization processing, u represents data before normalization processing, and u represents data after normalization processingminRepresents the minimum value of the characteristic state parameter used, umaxRepresenting the maximum value of the characteristic state parameter used.
Further, the specific method for training the Elman neural network by using the training set of the historical data is that the mean square error is used as a loss function, the learning rate alpha is set, the initial weight and the threshold of the neural network are determined in a random mode, the maximum iteration times and the target error of the network are set, the weight is updated by a gradient descent method, and the training process of the Elman neural network is as follows:
first, for historical data of a characteristic state parameter, the diagnosis result of the next time series is determined by using the previous 5 time series, namely ut-5,ut-4,ut-3,ut-2,ut-1As input to the Elman neural network, otAs the output of the neural network.
In the second place, the first place is,if ut-5,ut-4,ut-3,ut-2,ut-1As input to the neural network, otAs the output of the neural network, the next set of training data is ut-4,ut-3,ut-2,ut-1,utAs input to the neural network, ot+1As the output of the neural network, the training of the neural network is analogized in turn to achieve the effect of reflecting the dynamic characteristics.
And thirdly, inputting training data to an input layer of the Elman, transmitting the training data to a hidden layer for processing, and transmitting the output of the hidden layer to a receiving layer to be used as a delay operator so as to achieve the purpose of memorizing time data. The main calculation process of the Elman neural network is as follows:
elman neural network error discriminant function:
o (k) represents the output of the model, d (k) represents the ideal output.
In the following calculation process, t represents the number of neural network cycles.
Hidden layer:
Hnet(t+1)＝∑w[u(t),c(t)]
h(t+1)＝f(Hnet(t+1))
hnet (t +1) denotes the input to the hidden layer of the Elman neural network, h (t +1) denotes the output of the hidden layer, and w ═ w1,w2]，w1Representing the weight between the input layer and the hidden layer, w2Represents the weight between the accepting layer and the hidden layer, u (t) represents the output of the input layer, c (t) represents the output of the accepting layer, and f represents the Sigmoid function.
Carrying out layer bearing:
Cnet(t+1)＝f(h(t))
cnet (t +1) represents the input to the socket layer, h (t) represents the output of the hidden layer, and f represents the Sigmoid function.
An output layer:
Onet(t+1)＝∑w3h(t+1)
ot＝f(Onet(t+1))
otrepresenting the output of the output layer, w3Representing the weight between the hidden layer and the output layer, h (t +1) representing the output of the hidden layer, and f representing the Sigmoid function.
Fourthly, the output result of the Elman neural network is taken as input and transmitted to an SVM classifier, and the output result of the SVM is the diagnosis result of the current data, and the method is realized as follows:
let the relaxation variable constraints of the sample points be:
in the formula oiRepresenting the i-th feature vector, yiIs oiV denotes the normal vector of the hyperplane, b denotes the intercept of the hyperplane, ξiThe relaxation variable for the ith sample point is represented.
Selecting a target function of soft interval penalty parameters in the non-linear support vector machine SVM:
wherein w represents a normal vector of the hyperplane, C represents a penalty parameter, the penalty for misclassification is larger when the value of C is larger, and the penalty for misclassification is smaller when the value of C is smaller, and xiiThe relaxation variable for the ith sample point is represented, and G (w, C) represents the objective function.
Setting a relaxation variable ξiThe soft interval penalty parameter is 0.001, the soft interval penalty parameter is 0.01, the original features are mapped to a high-order feature space by using a Gaussian radial basis function, and a two-classification recognition model is constructed, in the embodiment, but not limited to, the Gaussian radial basis function is selected, and the formula is as follows:
where x denotes a certain point in the feature space, z denotes the center of the feature function, D (x, z) denotes the euclidean distance from x to z, and σ denotes the width of the kernel function, which defines the radial range of action of the kernel function.
Aiming at the input of the SVM and the diagnosis state of the current original data, the iterative training frequency of the SVM is set to 300, when the iterative frequency reaches 300 times or meets a KKT condition (Carlo needs-Kuen-Tack condition), the training is finished, an identification model obtained by training is selected, an optimal separation hyperplane of a normal state and a fault state is searched through the risky SVM, and the formula is as follows:
sign denotes a sign operation, N denotes the total number of sample data,
the optimal Lagrange multiplier, x, representing the ith sample pointiDenotes the ith sample point, the yiClass label, D (x), representing the ith sample pointiAnd x) represents xiEuclidean distance to x, b*Represents the optimal hyperplane intercept, and f (x) represents the optimal hyperplane.
As shown in fig. 2, the following characteristic parameter data, namely, the CPU temperature, the load, the inner layer load, and the like, can be monitored by the upper computer controller module of the heavy-duty gas turbine.
As shown in fig. 3, the flow of characteristic state data in the present invention is shown. Firstly, diagnosing next time sequence data by using the historical data of the characteristic states and data of every 5 time sequences, establishing an Elman neural network model, inputting and outputting the Elman neural network model as shown in figure 4, transmitting the result of the Elman neural network to an SVM for secondary classification, determining the state of each moment corresponding to the time sequence data, obtaining an accurate switching value diagnosis result, and realizing the intelligent BIT design of a controller module based on the Elman-SVM.
As shown in fig. 5, the Elman neural network model mainly comprises an input layer, a hidden layer, a supporting layer and an output layer, wherein the layers are connected through a weight value, and the next time sequence diagnosis result is predicted mainly by using data of the first 5 time sequences of characteristic state parameters.
As shown in fig. 6, the classification principle of the SVM finds a hyperplane in the feature space, and implements secondary classification of Elman data, which is mainly used for accurately outputting an intelligent BIT diagnosis result.
The above embodiments are only used for illustrating the technical solution of the present invention, and are not limited thereto; the present invention is described in detail with respect to the above examples, as will be appreciated by those of ordinary skill in the art; the method can be implemented by partially or completely replacing the details of the method according to the actual situation for different case implementations, and the modifications and the replacements do not make the essence of the corresponding technical solution depart from the scope of the technical solution designed by the invention.
The invention provides a design method of intelligent BIT of a heavy-duty gas turbine control system controller module based on an Elman neural network model and an SVM, and the specific case introduced herein explains the principle and implementation of the invention and is only used for explaining the principle and implementation mode of the invention; in view of the above, it will be apparent to those skilled in the art that various modifications can be made in the embodiments and applications without departing from the spirit and scope of the invention.
Claims (7)
1. An intelligent BIT design method for a heavy-duty gas turbine control system controller module based on an Elman neural network and an SVM is characterized by comprising the following specific steps of:
step 1: acquiring historical data of a single analog quantity state parameter operated by a controller module in a specified time period, and acquiring data of a continuous time sequence and a real operation state of current data;
step 2: carrying out normalization processing on the data, and determining the input and the output of the Elman neural network by combining the real running state of the data;
and step 3: determining a training set and a testing set of the model by adopting a dynamic training mode according to time sequence recursion;
and 4, step 4: designing an Elman neural network aiming at the characteristic state parameters, determining the state of the analog quantity parameters by inputting time sequence data, and training and testing the network by using the time sequence data until the training error of the model is less than a preset threshold value;
and 5: after the model in the step 4 is trained, transmitting the output of the Elman neural network to an SVM (support vector machine) as input for secondary classification, and determining the state of time series data;
step 6: and (5) regarding the classification result of the SVM in the step 5, wherein 0 represents normal, 1 represents fault, and the output result of the SVM aiming at each time series data is taken as the diagnosis result of the BIT at the current moment.
2. The intelligent BIT design method for the controller module of the heavy gas turbine control system based on the Elman neural network and the SVM as claimed in claim 1, wherein the historical data of the controller module are state parameters when a conventional BIT diagnosis mode is a threshold diagnosis mode or a simple logic diagnosis mode, are sorted according to a time sequence and comprise all characteristic state parameter historical data capable of reflecting the state of the controller module, and a matrix U represents the time sequence data of the state parameters, U represents the time sequence data of the state parametersnData representing the nth time series, i.e. U ═ U1,u2,…un]The matrix O represents the diagnostic data of the state parameters at different times, which data are generally the switching value data, OnIndicates the diagnostic result of the nth time series, i.e. O ═ O1,o2,…on]。
3. The intelligent BIT design method of the controller module of the heavy-duty gas turbine control system based on the Elman neural network and the SVM as claimed in claim 1, wherein in the step 2, the normalization processing method of the characteristic state parameters specifically comprises the following steps:
u' represents data after normalization, u represents data before normalization, and u represents data after normalizationminRepresents the minimum value, u, of the characteristic state parametermaxRepresenting the maximum value of the characteristic state parameter.
4. The intelligent BIT design method of the controller module of the heavy-duty gas turbine based on the Elman neural network and the SVM as claimed in claim 1, wherein in the step 3, the Elman neural network comprises an input layer, an implicit layer, a reception layer and an output layer, and the Elman neural network is trained as follows:
setting a learning rate alpha by taking a mean square error as a loss function, determining initial weight and threshold of the neural network in a random mode, setting maximum iteration times and target errors of the network, updating the weight by adopting a gradient descent method, and training the Elman neural network as follows:
(1) determining the diagnosis result of the next time sequence by using the former 5 time sequences aiming at the historical data of a certain characteristic state parameter, namely ut-5,ut-4,ut-3,ut-2,ut-1As input to the Elman neural network, otAs an output of the neural network;
(2) if ut-5,ut-4,ut-3,ut-2,ut-1As input to the neural network, otAs the output of the neural network, the next set of training data is ut-4,ut-3,ut-2,ut-1,utAs input to the neural network, ot+1As the output of the neural network, training the neural network in sequence to achieve the effect of accurately reflecting the dynamic characteristics;
(3) training data is input to an input layer of the Elman and then transmitted to a hidden layer, and output of the hidden layer is transmitted to a receiving layer to serve as a delay operator, so that the purpose of memorizing time data is achieved. The main calculation process of the Elman neural network is as follows:
elman neural networkError discriminant function:
in the following calculation process t represents the number of neural network cycles,
hidden layer: hnet (t +1) ═ Σ w [ u (t), c (t) ]
h(t+1)＝f(Hnet(t+1))
Hnet (t +1) represents the input of the hidden layer of the neural network, h (t +1) represents the output of the hidden layer, and w ═ w1,w2]，w1Representing the weight between the input layer and the hidden layer, w2Representing the weight between the accepting layer and the hidden layer, u (t) representing the output of the input layer, c (t) representing the output of the accepting layer, and f representing the Sigmoid function;
carrying out layer bearing: cnet (t +1) ═ f (h (t))
Cnet (t +1) represents the input of the accepting layer, h (t) represents the output of the hidden layer, and f represents the Sigmoid function;
an output layer: onet (t +1) ═ Σ w3h(t+1)
ot＝f(Onet(t+1))
otRepresenting the output of the output layer, w3Representing the weight between the hidden layer and the output layer, h (t +1) representing the output of the hidden layer, and f representing the Sigmoid function.
5. The intelligent BIT design method of the controller module of the heavy duty gas turbine based on the Elman neural network and the SVM as claimed in claim 1, wherein in the step 4, when the model training error is smaller than a preset threshold value, the Elman neural network model training is completed.
6. The method for designing the intelligent BIT of the controller module of the heavy-duty gas turbine control system based on the Elman neural network and the SVM as claimed in claim 1, wherein in the step 5, the output layer value of the Elman neural network is used as input to be transmitted to the SVM classifier, the output result is the intelligent BIT diagnosis result of the current data, and the processing process of the output result of the Elman neural network by the SVM is realized as follows:
(1) let the relaxation variable constraints of the sample points be:
yi(v·oi+b)≥1-ξi
oirepresenting the i-th feature vector, yiIs oiV denotes the normal vector of the hyperplane, b denotes the intercept of the hyperplane, ξiA relaxation variable representing the ith sample point;
(2) selecting a target function of soft interval penalty parameters in the non-linear support vector machine SVM:
w represents a normal vector of the hyperplane, C represents a penalty parameter, when the value of C is larger, the penalty for misclassification is larger, and when the value of C is smaller, the penalty for misclassification is smaller, and xiiRepresents the relaxation variable for the ith sample point, and G (w, C) represents the objective function;
(3) setting a relaxation variable ξiAnd soft interval punishment parameters, mapping the original features to a high-order feature space by using a Gaussian radial basis function, and constructing a two-classification recognition model, wherein the Gaussian radial basis function is selected but not limited in the embodiment, and the formula is as follows:
x represents a certain point in the feature space, z represents the center of the feature function, D (x, z) represents the Euclidean distance from x to z, and sigma represents the width of the kernel function and is used for limiting the radial action range of the kernel function;
(4) aiming at the input of the SVM and the diagnosis state of the current original data, the iterative training times of the SVM are set, when the iterative times reach a specified value or meet a KKT condition (Carrocon-Kuen-Tack condition), the training is finished, an identification model obtained by the training is selected, the optimal separation hyperplane of a normal state and a fault state is searched through the risky SVM, and the formula is as follows:
sign denotes a sign operation, N denotes the total number of sample data,
the optimal Lagrange multiplier, x, representing the ith sample pointiDenotes the ith sample point, yiClass label, D (x), representing the ith sample pointiAnd x) represents xiEuclidean distance to x, b*Represents the optimal hyperplane intercept, and f (x) represents the optimal hyperplane.
7. The intelligent BIT design method of the controller module of the heavy duty gas turbine control system based on the LSTM and the bio-excitation neural network as claimed in claim 1, wherein in the step 6, the output results 0 and 1 of the SVM aiming at the data at each moment respectively represent the normal and fault states of the current data, and the information in the results is interpreted to be used as the final diagnosis result of the intelligent BIT to be uploaded to the upper computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110330069.0A CN113110044A (en) | 2021-03-29 | 2021-03-29 | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on Elman neural network and SVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110330069.0A CN113110044A (en) | 2021-03-29 | 2021-03-29 | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on Elman neural network and SVM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113110044A true CN113110044A (en) | 2021-07-13 |
Family
ID=76712618
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110330069.0A Pending CN113110044A (en) | 2021-03-29 | 2021-03-29 | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on Elman neural network and SVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113110044A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116662761A (en) * | 2023-06-28 | 2023-08-29 | 广州发展南沙电力有限公司 | Fuel gas power station important parameter early warning method and system based on data analysis |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
| CN107817404A (en) * | 2017-11-18 | 2018-03-20 | 广西电网有限责任公司电力科学研究院 | A kind of Portable metering automatization terminal trouble-shooter and its diagnostic method |
| CN109934335A (en) * | 2019-03-05 | 2019-06-25 | 清华大学 | High-speed railway track switch method for diagnosing faults based on interacting depth study |
| CN112000084A (en) * | 2020-09-07 | 2020-11-27 | 华北电力大学 | Intelligent BIT design method of controller module based on 1D-CNN and GRU-SVM |
| 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 |
-
2021
- 2021-03-29 CN CN202110330069.0A patent/CN113110044A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
| CN107817404A (en) * | 2017-11-18 | 2018-03-20 | 广西电网有限责任公司电力科学研究院 | A kind of Portable metering automatization terminal trouble-shooter and its diagnostic method |
| CN109934335A (en) * | 2019-03-05 | 2019-06-25 | 清华大学 | High-speed railway track switch method for diagnosing faults based on interacting depth study |
| CN112000084A (en) * | 2020-09-07 | 2020-11-27 | 华北电力大学 | Intelligent BIT design method of controller module based on 1D-CNN and GRU-SVM |
| 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 |
Non-Patent Citations (1)
| Title |
|---|
| 李公法: "《人工智能与计算智能及其应用》", 30 April 2020 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116662761A (en) * | 2023-06-28 | 2023-08-29 | 广州发展南沙电力有限公司 | Fuel gas power station important parameter early warning method and system based on data analysis |
| CN116662761B (en) * | 2023-06-28 | 2024-05-14 | 广州发展南沙电力有限公司 | Fuel gas power station important parameter early warning method and system based on data analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Data-driven bearing fault identification using improved hidden Markov model and self-organizing map | |
| CN109766992B (en) | Industrial control abnormity detection and attack classification method based on deep learning | |
| JP2023042527A (en) | Data drive active power distribution network abnormal state sensing method and system | |
| CN109800875A (en) | Chemical industry fault detection method based on particle group optimizing and noise reduction sparse coding machine | |
| CN108445752B (en) | Random weight neural network integrated modeling method for self-adaptively selecting depth features | |
| CN109145516B (en) | Analog circuit fault identification method based on improved extreme learning machine | |
| Di Maio et al. | Ensemble-approaches for clustering health status of oil sand pumps | |
| Jin et al. | Detecting and diagnosing incipient building faults using uncertainty information from deep neural networks | |
| CN110321940B (en) | Aircraft telemetry data feature extraction and classification method and device | |
| CN111638707B (en) | Intermittent process fault monitoring method based on SOM clustering and MPCA | |
| CN112000015B (en) | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on LSTM and bio-excitation neural network | |
| CN111046961B (en) | Fault classification method based on bidirectional long-time and short-time memory unit and capsule network | |
| Kazemi et al. | A hybrid method for estimating the process change point using support vector machine and fuzzy statistical clustering | |
| CN110598902A (en) | Water quality prediction method based on combination of support vector machine and KNN | |
| CN112504682A (en) | Chassis engine fault diagnosis method and system based on particle swarm optimization algorithm | |
| CN113225346A (en) | Network operation and maintenance situation assessment method based on machine learning | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| CN111507365A (en) | Confidence rule automatic generation method based on fuzzy clustering | |
| CN116109039A (en) | Data-driven anomaly detection and early warning system | |
| CN114609994A (en) | Fault diagnosis method and device based on multi-granularity regularization rebalance incremental learning | |
| CN113110044A (en) | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on Elman neural network and SVM | |
| Cui et al. | Intelligent fault quantitative identification via the improved deep deterministic policy gradient (DDPG) algorithm accompanied with imbalanced sample | |
| CN114578234A (en) | Lithium ion battery degradation and capacity prediction model considering causality characteristics | |
| CN116719831A (en) | Standard database establishment and update method for health monitoring | |
| WO2023231374A1 (en) | Semi-supervised fault detection and analysis method and apparatus for mechanical device, terminal, and medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210713 |