CN115689353A - Hydropower station speed regulation system health assessment method based on GS-SVM algorithm - Google Patents
Hydropower station speed regulation system health assessment method based on GS-SVM algorithm Download PDFInfo
- Publication number
- CN115689353A CN115689353A CN202211346071.8A CN202211346071A CN115689353A CN 115689353 A CN115689353 A CN 115689353A CN 202211346071 A CN202211346071 A CN 202211346071A CN 115689353 A CN115689353 A CN 115689353A
- Authority
- CN
- China
- Prior art keywords
- index
- model
- data
- health
- hydropower station
- 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
Images
Abstract
A hydropower station speed regulation system health assessment method based on a GS-SVM algorithm comprises the following steps: obtaining an initial sample set of the hydropower station speed regulating system according to historical operating data and speed regulating system simulation data of the speed regulating system, and constructing a corresponding health state label set; modeling the initial sample set by adopting an analytic hierarchy process, processing model data according to the degradation degree and weight quantization, realizing differential influence of index data on the equipment per se under multiple levels, and establishing a high-efficiency training sample set; taking the characteristic attributes of the samples in the high-efficiency training sample set as input, and taking the health states corresponding to the samples as output, and constructing a support vector machine model; determining the optimal combination of penalty factors and kernel functions by a cross validation method based on a grid search method, optimizing a support vector machine model, and establishing a GS-SVM prediction model; the new environment and different model parameters of the evaluation model are considered, and the health evaluation model is updated; and acquiring real-time data of the speed regulating system to complete the evaluation of the health state of the hydropower station, and acquiring a real-time health evaluation result. The invention improves the stability and the economy of the hydropower station operation.
Description
Technical Field
The invention belongs to the technical field of health assessment of a hydropower station speed regulating system, and particularly relates to a health assessment method of the hydropower station speed regulating system based on a GS-SVM algorithm.
Background
The hydraulic turbine speed regulating system has no replaceable status in the safe and stable operation process of the hydropower station, is one of important control devices of the hydropower station, and has the functions of not only maintaining the stability of the rotating speed of a unit, but also starting, stopping, grid connection, load increase and reduction and the like of the unit. Along with the increase of the working duration, the frequency of unit faults is increased, the generating capacity and the comprehensive performance of the unit can be weakened, and most hydropower stations still adopt scheduled maintenance and after-repair, so that the situations of insufficient maintenance and excessive maintenance often occur, and the economic benefit and the normal operation of the hydropower stations are influenced. In order to ensure the normal operation of the water turbine generator set, the health state prediction research on the water turbine speed regulating system is of great significance.
The existing health assessment method of the speed regulating system is mainly based on two angles of mechanism analysis and data driving, and has the following defects and difficulties:
1. mechanism modeling is difficult, and more assumptions and simplification are carried out, so that the precision and the accuracy of the model are reduced;
2. the traditional data driving method has various limitations when being directly applied to real-time health assessment of a speed regulating system, and does not consider all aspects influencing the health of the speed regulating system or only carries out health assessment on a part of devices;
3. traditional data-driven analysis mining is based on historical persisted data only, and real-time updating of the model is not considered.
Disclosure of Invention
In view of the technical problems in the background art, the method for evaluating the health of the hydropower station speed regulation system based on the GS-SVM algorithm can help the hydropower station staff to take preventive control measures in time, and improve the stability and the economy of the operation of the hydropower station.
In order to solve the technical problems, the invention adopts the following technical scheme to realize:
a hydropower station speed regulation system health assessment method based on a GS-SVM algorithm comprises the following steps:
the method comprises the following steps: obtaining an initial sample set of the hydropower station speed regulating system according to historical operating data and speed regulating system simulation data of the speed regulating system, and constructing a corresponding health state label set;
step two: modeling the initial sample set by adopting an analytic hierarchy process, processing model data according to the degradation degree and weight quantization, realizing differential influence of index data on the equipment per se under multiple levels, and establishing a high-efficiency training sample set;
step three: taking the characteristic attributes of the samples in the high-efficiency training sample set as input, and taking the health states corresponding to the samples as output, and constructing a support vector machine model; determining the optimal combination of penalty factors and kernel functions by a cross validation method based on a grid search method, optimizing a support vector machine model, and establishing a GS-SVM prediction model;
step four: the new environment and different model parameters of the evaluation model are considered, and the health evaluation model is updated;
step five: and acquiring real-time data of the speed regulating system to complete the evaluation of the health state of the hydropower station, and acquiring a real-time health evaluation result.
Further, in the first step, an initial sample set is constructed according to historical operating data and simulation data obtained through a visual simulation tool, the initial sample set is divided into four label sets according to the health state, and the four label sets correspond to scores of different segments.
Further, in the second step, the step of modeling by the analytic hierarchy process is: the running state of the water turbine speed regulating system is divided to finally form three-layer indexes: the first level is a target layer index, namely the finally obtained state score of the water turbine speed regulating system; the second level is the indexes of the component layer, including the performance of the oil system, the performance of the starting working condition, the performance of the stopping working condition, the performance of the load shedding working condition, the performance of the loaded working condition and the historical overhaul performance; and the third level is an index layer which comprises each specific monitoring index or a performance index value calculated by the monitoring index.
Furthermore, in the second step, the magnitude and dimension of each index of the speed regulation system are normalized by comparing the real-time operation data of the system, the specified data range and the operation data during the fault and introducing the relative degradation degree, so that the problem that the performance index is possibly seriously deviated is solved; the degradation degree calculation method includes a more preferable type with a larger degradation degree and a more preferable type with a smaller degradation degree.
Further, in step two, the method for quantizing the weight is as follows: calculating a subjective weight vector of a diagnosis index by using a nine-scale method, and calculating to obtain a combined weight vector by combining an objective weight vector calculated by a CRITIC method, wherein the method comprises the following steps of:
1. and establishing a subjective weight quantization matrix according to a nine-scale method:
1) Establishing an index judgment matrix, if a certain index contains n lower layer factors X = (X) ij ) nn ，A 2 ,…,A n Let i, j belong to 1, 2, \ 8230, n, then x ij Reflecting factor A i Ratio A j The importance of (c); otherwise x ji ＝1/x ij Reflecting factor A j Ratio A i Then the judgment matrix X = (X) is obtained ij ) n×n 。
2) Determining a relative weight coefficient;
(1) multiplying elements of each row of the judgment matrix to obtain a corresponding matrix: x' = (X) ij ') n×1 ；
(2) Performing square extraction on each row of the obtained matrix for n times;
(3) carrying out normalization processing on the matrix to obtain relative weight coefficients of all factors of the layer;
2. establishing an objective weight quantization matrix according to a CRITIC method:
1) Standardization process, if there are x objects on the upper layer, the lower layer has y index parameters, r ij The value of the jth index parameter representing the ith object is as follows:
the forward direction index is as follows:
negative direction indexes are as follows:
2) Determining a relative weight coefficient:
(1) calculating the variation intensity;
(2) calculating a correlation coefficient;
(3) calculating the objective weight of the jth index parameter;
finally according to the subjective weight coefficient w i ', objective weight coefficient w i And obtaining a combined weight coefficient.
Further, in the third step, a support vector machine prediction model is established, and the steps are as follows:
1) The known efficient training sample set is: t = { (d) 1 * ,s 1 ),(d 2 * ,s 2 ),L,(d n * ,s n )}，d * Representing the characteristic attribute of the sample, and s is the health state score value of the equipment;
2) Solving the optimization problem for the support vector machine model:
in the formula alpha i 、α j Is Lagrange multiplier coefficient, c is penalty factor;
3) Mapping the low-dimensional data points to a high-dimensional space using a gaussian kernel function:
in the formula: g is an optimizable kernel function parameter, and sigma is a kernel function width parameter, and has important influence on determining a support vector of the classification hyperplane;
4) Constructing an optimal discriminant function to obtain a support vector machine model:
further, in step three, the determining the optimal combination of penalty factor and kernel function by cross validation method based on the grid search method comprises the following steps:
1) Setting the value range of the penalty factor c as (2) -10 -2 7 ) And the value range of the kernel function g (2) -10 -2 3 ) In 2 of k K is an integer, a two-dimensional grid is established, and the optimal grid point (c) is found 1 ,g 1 ) As an initial parameter pair;
2) Assuming a penalty factor of c 1 Unchanged, kernel function at g 1 Around with 2 0.5 The variation finds the grid points (c) where the best convergence is 1 ,g 2 )；
3) Assume a kernel function of g 2 Invariance with penalty factor of c 1 Around with 2 0.5 The variation finds the grid points (c) where the best convergence is 2 ,g 2 )；
4) The optimal combination meets the optimal precision requirement, and the optimization is finished
And obtaining a speed regulating system health prediction model based on the efficient training sample set through the steps.
Further, the decomposition step in step four is as follows:
1) Using the real-time data to verify a real-time health assessment model validity; model training is based on historical data and therefore must be re-validated in the latest environment; after trial, the model must be evaluated according to the test result and feedback; when the prediction result does not meet the requirement, retraining according to the updated historical data;
2) Under the background of big data, the collected data volume is continuously increased, and the model is retrained through more data, so that the problem of overfitting of the model is suppressed, and the precision and the generalization capability of the model are improved.
Further, in the fifth step, deploying the GS-SVM health prediction model to a monitoring platform, extracting the three-layer index real-time data, and inputting the three-layer index real-time data into the prediction model to obtain a health assessment score.
This patent can reach following beneficial effect:
1. the comprehensive evaluation method based on the multiple indexes established by the analytic hierarchy process avoids the one-sidedness problem of the evaluation result of the single state quantity, and has the characteristics of high accuracy and high practicability.
2. The invention adopts a combined weight method, avoids interference of subjective factors and has higher reliability.
3. The grid search method-support vector machine algorithm provided by the invention combines the advantages of the grid search method and the support vector machine, can improve the prediction speed of the evaluation model, and simultaneously improves the prediction precision of the evaluation model.
4. The health assessment model constructed by the invention is updated based on different environments of equipment, and the model can be endowed with better generalization capability.
Drawings
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a system diagram of a health assessment indicator for a governor system of the present invention;
FIG. 3 is a comparison graph of the importance of the nine-scale method of the present invention;
FIG. 4 is a flow chart of the GS-SVM algorithm optimization of the present invention.
Detailed Description
Example 1:
the preferable scheme is as shown in fig. 1 to 4, and the hydropower station speed regulation system health assessment method based on the GS-SVM algorithm comprises the following steps:
the method comprises the following steps: obtaining an initial sample set of the hydropower station speed regulating system according to historical operating data and speed regulating system simulation data of the speed regulating system, and constructing a corresponding health state label set;
step two: modeling the initial sample set by adopting an analytic hierarchy process, processing model data according to the degradation degree and weight quantization, realizing the differential influence of index data on the equipment per se under multiple levels, and establishing an efficient training sample set;
step three: taking the characteristic attributes of the samples in the efficient training sample set as input, and taking the health states corresponding to the samples as output, and constructing a support vector machine model; determining the optimal combination of penalty factors and kernel functions by a cross validation method based on a grid search method, optimizing a support vector machine model, and establishing a GS-SVM prediction model;
step four: the new environment and different model parameters of the evaluation model are considered, and the health evaluation model is updated;
step five: and acquiring real-time data of the speed regulating system to complete the evaluation of the health state of the hydropower station, and acquiring a real-time health evaluation result.
Preferably, in step one, an initial sample set is constructed according to historical operating data and simulation data obtained by a visual simulation tool (simulink), and the initial sample set is divided into four label sets according to health states, wherein the four label sets respectively have excellent scores (10-8), good scores (8-6), medium scores (6-4) and extremely poor scores (4-2).
Preferably, in the first step, the health diagnosis result of the historical data of the speed regulating system refers to the survey result of the expert questionnaire. A speed regulation system health evaluation index system, as shown in fig. 2, wherein the first-stage index is a target layer index, namely a finally obtained state score of the water turbine speed regulation system; the second level is a component layer index which comprises six parts of oil system performance, starting working condition performance, stopping working condition performance, load shedding working condition performance, on-load working condition performance and historical overhaul performance; the third level is an index layer, and the oil system performance comprises oil pressure, oil level and oil pump efficiency; the starting working condition performance comprises rotating speed rising time, machine frequency overshoot, guide vane opening time and adjusting time; the shutdown working condition performance comprises draft tube inlet pressure, guide vane closing time and volute water pressure rising time; the load shedding working condition performance comprises rotating speed adjusting time, servomotor motionless time, rotating speed swinging times and a rotating speed peak value; the on-load working condition performance comprises load response lag time and load response stabilization time; the historical overhaul performance comprises the operation age and the overhaul frequency.
The specific index parameters are mainly obtained from three aspects:
(1) The system is obtained from a hydraulic turbine speed regulator, and a modern microcomputer speed regulator can obtain some set operation parameters and operation states, so that real-time operation data closely related to system health performance evaluation can be obtained from the set operation parameters and the operation states;
(2) Acquiring a field data set under the existing monitoring system of the hydropower station;
(3) And if the monitoring data meeting the requirements cannot be acquired, selecting a proper sensor, and accessing the signal of the sensor to the existing hydropower station monitoring system so as to acquire and display the required data sample.
Preferably, in step two, the step of modeling by the analytic hierarchy process is: the running state of the water turbine speed regulating system is divided to finally form three layers of indexes: the first level is a target layer index, namely the finally obtained state score of the water turbine speed regulating system; the second level is the indexes of the component layer, including the performance of the oil system, the performance of the starting working condition, the performance of the stopping working condition, the performance of the load shedding working condition, the performance of the loaded working condition and the historical overhauling performance; and the third level is an index layer which comprises each specific monitoring index or a performance index value calculated by the monitoring index.
Preferably, in the second step, the magnitude and dimension of each index of the speed regulation system are normalized by comparing the real-time operation data of the system, the specified data range and the operation data during the fault and introducing the relative degradation degree, so that the problem that the performance index is possibly seriously deviated is solved; the degradation degree calculation method includes a more preferable type with a larger degradation degree and a more preferable type with a smaller degradation degree.
The degradation degree normalization method is more intuitive to show the relationship between the index parameters and the health degree than a standard normalization method: the more preferable type and the less preferable type with a higher degree of degradation are:
1) The more the degree of deterioration is, the more excellent the model is, the deterioration expression is:
in the formula: u is a normalized numerical value; u shape i Raw data which is an index; u shape max Is an ideal value of the index parameter; u shape min The lower limit of the standard value when the index is operated safely.
2) The more excellent the deterioration degree is, the more excellent the deterioration expression is:
in the formula: u' max Operating the upper limit of the standard value for the index when the index is safe; u' min Is an ideal value of the index parameter.
Preferably, in step two, the method for quantizing the weight is: calculating a subjective weight vector of a diagnosis index by using a nine-scale method, and calculating to obtain a combined weight vector by combining an objective weight vector calculated by a CRITIC method, wherein the method comprises the following steps of:
1. and establishing a subjective weight quantization matrix according to a nine-scale method:
1) Establishing an index judgment matrix, and if a certain index contains n lower layer factors X = (X) ij ) nn ，A 2 ,…,A n Let i, j belong to 1, 2, \ 8230, n ij Reflecting factor A i Ratio A j The importance of (c); otherwise x ji ＝1/x ij Reflecting factor A j Ratio A i Then the judgment matrix X = (X) is obtained ij ) n×n 。
2) Determining a relative weight coefficient;
(1) Multiplying elements of each row of the judgment matrix to obtain a corresponding matrix X' = (X) ij ') n×1 ：
(2) And (3) performing n-time evolution on each row of the matrix obtained by the formula (3):
(3) The matrix obtained by the formula (4) is normalized to obtain the relative weight coefficient of each factor of the layer:
2. establishing an objective weight quantization matrix according to a CRITIC method:
(1) Standardization process, if there are x objects on the upper layer, the lower layer has y index parameters, r ij The value of the jth index parameter representing the ith object is as follows:
the forward direction index is as follows:
negative direction index:
(2) Determining relative weight coefficients
(1) Calculating the variation intensity:
in the formula
The average value of the j-th term is represented,
the standard deviation of the j-th item is shown.
(2) Calculating a correlation coefficient
Correlation coefficient c between j index and h index in y index parameters jh Comprises the following steps:
(3) calculating objective weight of jth index parameter
Obtaining the subjective weight coefficient w of the ith object according to the formula (5) i ' obtaining an objective weighting factor w of the ith object according to equation (10) i The combining weight coefficient is:
assuming that a total of N different speed governing system samples are collected, an index initial sample set D = { D = is formed 1 ,d 2 ,Ld i ,Ld N }. Each speed regulating system sample corresponds to n three-level indexes but has different index values, and the degradation degree value of the n three-level indexes of the ith sample is obtained according to the degradation degree calculation formula to form a sample set d i ＝{d i1 ,d i2 ,Ld ij ,Ld in F, if the three-level index d is present ij Corresponding combining weight is w ij Then, the characteristic attribute d of the sample of the ith speed regulating system is obtained i * ＝{d i1 * ,d i2 * ,Ld ij * ,Ld in * }，d ij * ＝d ij ×w ij 。
Preferably, in step three, a support vector machine prediction model is established, and the steps are as follows:
1) The known efficient training sample set is: t = { (d) 1 * ,s 1 ),(d 2 * ,s 2 ),L,(d n * ,s n )}，d * Representing the characteristic attribute of the sample, and s is the health state score value of the equipment;
2) Solving the optimization problem for the support vector machine model:
in the formula of alpha i 、α j Is Lagrange multiplier coefficient, and c is penalty factor;
3) Mapping the low-dimensional data points to a high-dimensional space using a gaussian kernel function:
in the formula: g is an optimizable kernel function parameter, and sigma is a kernel function width parameter, which has an important influence on determining the support vector of the classification hyperplane;
4) Constructing an optimal discriminant function to obtain a support vector machine model:
the popularization capability, generalization level, prediction precision and running speed of the support vector machine prediction model are influenced by a penalty factor c and a kernel function parameter g.
Preferably, in step three, the determining the optimal combination of the penalty factor and the kernel function based on the grid search method, as shown in fig. 4, specifically includes the following steps:
1) Setting the value range of the penalty factor c as (2) -10 -2 7 ) And the value range of the kernel function g (2) -10 -2 3 ) In 2 of k The range where k is an integer establishes a two-dimensional grid, and finds the grid point (c) where convergence is optimal 1 ,g 1 ) As an initial parameter pair;
2) Assuming a penalty factor of c 1 Unchanged, kernel function at g 1 Around with 2 0.5 The variation finds the grid point (c) where the convergence is optimal 1 ,g 2 )；
3) Assume a kernel function of g 2 Invariance with penalty factor of c 1 Around with 2 0.5 The variation finds the grid points (c) where the best convergence is 2 ,g 2 )；
4) And obtaining the optimal parameter combination in all the grid points, and finishing the optimization.
And obtaining an optimal speed regulating system health prediction model according to the optimal combination of the penalty factor c and the kernel function parameter g.
Preferably, the decomposition step at step four is as follows:
1) Using the real-time data to verify the real-time health assessment model validity; model training is based on historical data and therefore must be re-validated in the latest environment; after trial, the model must be evaluated according to the test result and feedback; when the prediction result does not meet the requirement, retraining according to the updated historical data;
2) Under the background of big data, the collected data volume is continuously increased, and the model is retrained through more data, so that the problem of overfitting of the model is suppressed, and the precision and the generalization capability of the model are improved.
For example: considering the new environment, different model parameters, where the evaluation model is located: such as the operating environment at the time, the size of the parameters of the device, the aging degree of the system, etc. Model training is mainly based on historical data and therefore must be re-validated in the latest environment. After trial, the model must be evaluated based on the test results and feedback. And when the prediction result does not meet the requirement, the training can be carried out again according to the updated historical data. Under the background of big data, the collected data volume is continuously increased, and the model can be retrained through more data, so that the problem of overfitting of the model is suppressed, and the precision and the generalization capability of the model are improved.
Preferably, in the fifth step, the GS-SVM health prediction model is deployed to a monitoring platform, and the three-layer index real-time data described herein is extracted and then input to the prediction model to obtain a health assessment score. And displaying the state monitoring data and the prediction result to experts and field operating personnel for further analysis and mining, so as to realize more effective and more scientific predictive maintenance decision from the entity to the data and then to the entity.
The above-described embodiments are merely preferred embodiments of the present invention, and should not be construed as limiting the present invention, and the scope of the present invention is defined by the claims, and equivalents including technical features described in the claims. I.e., equivalent alterations and modifications within the scope hereof, are also intended to be within the scope of this invention.
Claims (9)
1. A hydropower station speed regulation system health assessment method based on a GS-SVM algorithm is characterized by comprising the following steps:
the method comprises the following steps: obtaining an initial sample set of the hydropower station speed regulating system according to historical operating data and speed regulating system simulation data of the speed regulating system, and constructing a corresponding health state label set;
step two: modeling the initial sample set by adopting an analytic hierarchy process, processing model data according to the degradation degree and weight quantization, realizing the differential influence of index data on the equipment per se under multiple levels, and establishing an efficient training sample set;
step three: taking the characteristic attributes of the samples in the high-efficiency training sample set as input, and taking the health states corresponding to the samples as output, and constructing a support vector machine model; determining the optimal combination of penalty factors and kernel functions by a cross validation method based on a grid search method, optimizing a support vector machine model, and establishing a GS-SVM prediction model;
step four: the new environment and different model parameters of the evaluation model are considered, and the health evaluation model is updated;
step five: and acquiring real-time data of the speed regulating system to complete the evaluation of the health state of the hydropower station, and acquiring a real-time health evaluation result.
2. The hydropower station speed regulation system health assessment method based on the GS-SVM algorithm according to claim 1, characterized in that: in the first step, an initial sample set is constructed according to historical operating data and simulation data obtained through a visual simulation tool, the initial sample set is divided into four label sets according to the health state, and the four label sets correspond to different segmented scores.
3. The hydropower station speed regulation system health assessment method based on the GS-SVM algorithm according to claim 1, characterized in that: in the second step, the modeling by the analytic hierarchy process comprises the following steps: the running state of the water turbine speed regulating system is divided to finally form three-layer indexes: the first level is a target layer index, namely the finally obtained state score of the water turbine speed regulating system; the second level is the indexes of the component layer, including the performance of the oil system, the performance of the starting working condition, the performance of the stopping working condition, the performance of the load shedding working condition, the performance of the loaded working condition and the historical overhauling performance; and the third level is an index layer which comprises each specific monitoring index or a performance index value calculated by the monitoring index.
4. The hydropower station speed regulation system health assessment method based on the GS-SVM algorithm according to claim 1, characterized in that: in the second step, the magnitude and dimension of each index of the speed regulation system are normalized by introducing relative degradation degree through comparing the real-time operation data of the system, the specified data range and the operation data during fault, so that the problem that the performance index is possibly seriously deviated is solved; the degradation degree calculation method includes a more preferable type with a larger degradation degree and a more preferable type with a smaller degradation degree.
5. The hydropower station speed regulation system health assessment method based on the GS-SVM algorithm according to claim 1, characterized in that: in step two, the method for quantizing the weight comprises the following steps: calculating a subjective weight vector of a diagnosis index by using a nine-scale method, and calculating to obtain a combined weight vector by combining an objective weight vector calculated by a CRITIC method, wherein the method comprises the following steps of:
1. and establishing a subjective weight quantization matrix according to a nine-scale method:
1) Establishing an index judgment matrix, if a certain index contains n lower layer factors X = (X) ij ) nn ，A 2 ,…,A n Let i, j belong to 1, 2, \8230, n, then x ij Reflecting factor A i Ratio A j The importance of (c); otherwise x ji ＝1/x ij Reflecting factor A j Ratio A i Then the judgment matrix X = (X) is obtained ij ) n×n 。
2) Determining a relative weight coefficient;
(1) multiplying each row element of the judgment matrix to obtain a corresponding matrix: x' = (X) ij ') n×1 ；
(2) Performing square extraction on each row of the obtained matrix for n times;
(3) carrying out normalization processing on the matrix to obtain relative weight coefficients of all factors of the layer;
2. establishing an objective weight quantization matrix according to a CRITIC method:
1) Standardization process, if there are x objects on the upper layer, the lower layer has y index parameters, r ij The value of the jth index parameter representing the ith object is as follows:
the forward direction index:
negative direction index:
2) Determining a relative weight coefficient:
(1) calculating the variation intensity;
(2) calculating a correlation coefficient;
(3) calculating the objective weight of the jth index parameter;
finally according to the subjective weight coefficient w i ', objective weight coefficient w i And obtaining a combined weight coefficient.
6. The hydropower station speed regulating system health assessment method based on the GS-SVM algorithm according to claim 1, wherein in the third step, a support vector machine prediction model is established, and the steps are as follows:
1) The known efficient training sample set is: t = { (d) 1 * ,s 1 ),(d 2 * ,s 2 ),L,(d n * ,s n )}，d * Representing the characteristic attribute of the sample, and s is the health state score value of the equipment;
2) Solving the optimization problem for the support vector machine model:
in the formula alpha i 、α j Is Lagrange multiplier coefficient, c is penalty factor;
3) Mapping the low-dimensional data points to a high-dimensional space by adopting a Gaussian kernel function:
in the formula: g is an optimizable kernel function parameter, and sigma is a kernel function width parameter, which has an important influence on determining the support vector of the classification hyperplane;
4) Constructing an optimal discriminant function to obtain a support vector machine model:
7. the method for assessing the health of the hydropower station governing system based on the GS-SVM algorithm according to claim 1, wherein in step three, the step of determining the optimal combination of the penalty factor and the kernel function by a cross validation method based on the grid search method comprises the following steps:
1) Setting the value range of the penalty factor c as (2) -10 -2 7 ) And the value range of the kernel function g (2) -10 -2 3 ) In 2 of k K is an integer, a two-dimensional grid is established, and the optimal grid point (c) is found 1 ,g 1 ) As an initial parameter pair;
2) Assuming a penalty factor of c 1 Unchanged, kernel function at g 1 Around with 2 0.5 The variation finds the grid point (c) where the convergence is optimal 1 ,g 2 )；
3) Assume a kernel function of g 2 Invariance, penalty factor in c 1 Around with 2 0.5 The variation finds the grid point (c) where the convergence is optimal 2 ,g 2 )；
4) The optimal combination meets the optimal precision requirement, and the optimization is finished
And obtaining the speed regulating system health prediction model based on the efficient training sample set through the steps.
8. The method for assessing the health of the hydropower station speed regulating system based on the GS-SVM algorithm according to claim 1, wherein the decomposition step in the fourth step is as follows:
1) Using the real-time data to verify the real-time health assessment model validity; model training is based on historical data and therefore must be re-validated in the latest environment; after trial, the model must be evaluated according to the test result and feedback; when the prediction result does not meet the requirement, retraining according to the updated historical data;
2) Under the background of big data, the acquired data volume is continuously increased, and the model is retrained through more data, so that the problem of overfitting of the model is suppressed, and the accuracy and generalization capability of the model are improved.
9. The hydropower station speed regulation system health assessment method based on the GS-SVM algorithm according to claim 3, wherein in the fifth step, a GS-SVM health prediction model is deployed to a monitoring platform, and after the three-layer index real-time data is extracted and processed, the three-layer index real-time data is input into the prediction model to obtain a health assessment score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211346071.8A CN115689353A (en) | 2022-10-31 | 2022-10-31 | Hydropower station speed regulation system health assessment method based on GS-SVM algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211346071.8A CN115689353A (en) | 2022-10-31 | 2022-10-31 | Hydropower station speed regulation system health assessment method based on GS-SVM algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115689353A true CN115689353A (en) | 2023-02-03 |
Family
ID=85045142
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211346071.8A Pending CN115689353A (en) | 2022-10-31 | 2022-10-31 | Hydropower station speed regulation system health assessment method based on GS-SVM algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115689353A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116300661A (en) * | 2023-05-18 | 2023-06-23 | 青岛宇方机器人工业股份有限公司 | On-site data acquisition system based on Internet of things |
| CN117236935A (en) * | 2023-11-10 | 2023-12-15 | 四川大学 | Weight self-adaptive water turbine health state assessment method containing subjective and objective information |
| CN117312972A (en) * | 2023-11-30 | 2023-12-29 | 山西海诚智能制造有限公司 | Method for identifying health state of scraper conveyor speed reducer |
| CN117310546A (en) * | 2023-11-03 | 2023-12-29 | 北京迪赛奇正科技有限公司 | UPS power health management monitoring system |
-
2022
- 2022-10-31 CN CN202211346071.8A patent/CN115689353A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116300661A (en) * | 2023-05-18 | 2023-06-23 | 青岛宇方机器人工业股份有限公司 | On-site data acquisition system based on Internet of things |
| CN116300661B (en) * | 2023-05-18 | 2023-08-01 | 青岛宇方机器人工业股份有限公司 | On-site data acquisition system based on Internet of things |
| CN117310546A (en) * | 2023-11-03 | 2023-12-29 | 北京迪赛奇正科技有限公司 | UPS power health management monitoring system |
| CN117236935A (en) * | 2023-11-10 | 2023-12-15 | 四川大学 | Weight self-adaptive water turbine health state assessment method containing subjective and objective information |
| CN117236935B (en) * | 2023-11-10 | 2024-01-26 | 四川大学 | Weight self-adaptive water turbine health state assessment method containing subjective and objective information |
| CN117312972A (en) * | 2023-11-30 | 2023-12-29 | 山西海诚智能制造有限公司 | Method for identifying health state of scraper conveyor speed reducer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115689353A (en) | Hydropower station speed regulation system health assessment method based on GS-SVM algorithm | |
| Qin et al. | Macroscopic–microscopic attention in LSTM networks based on fusion features for gear remaining life prediction | |
| CN110410282B (en) | SOM-MQE and SFCM (Small form-factor pluggable) based wind turbine generator health state online monitoring and fault diagnosis method | |
| CN110414788B (en) | Electric energy quality prediction method based on similar days and improved LSTM | |
| US20230023931A1 (en) | Hydraulic turbine cavitation acoustic signal identification method based on big data machine learning | |
| CN110007652B (en) | Hydroelectric generating set degradation trend interval prediction method and system | |
| CN103927412B (en) | Instant learning debutanizing tower soft-measuring modeling method based on gauss hybrid models | |
| CN111008504B (en) | Wind power prediction error modeling method based on meteorological pattern recognition | |
| CN111241755A (en) | Power load prediction method | |
| CN111963116B (en) | Intelligent gas field system and method for self-adaption and intelligent analysis decision | |
| CN112901449A (en) | Air compressor system energy consumption optimization method based on machine learning | |
| CN110766320A (en) | Method and device for evaluating operation safety of airport intelligent power grid | |
| CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
| CN116822652A (en) | Subway fault prediction method, subway fault prediction device, electronic equipment, subway fault prediction system and storage medium | |
| CN115238573A (en) | Hydroelectric generating set performance degradation trend prediction method and system considering working condition parameters | |
| CN107634544A (en) | The dynamic power control method and system of fired power generating unit | |
| CN116624343B (en) | Wind turbine generator tower abnormal vibration monitoring and health degree evaluation method and system | |
| CN115689001A (en) | Short-term load prediction method based on pattern matching | |
| CN116020879A (en) | Technological parameter-oriented strip steel hot continuous rolling space-time multi-scale process monitoring method and device | |
| CN111461565A (en) | Power supply side power generation performance evaluation method under power regulation | |
| Saraswat et al. | Prognosis of diesel engine (MBT) using feature extraction techniques: a comparative study | |
| Yang et al. | A new ensemble fault diagnosis method based on k-means algorithm | |
| CN115618752B (en) | Power load control method and system based on Autoformer model | |
| CN116505556B (en) | Wind farm power control system and method based on primary frequency modulation | |
| CN117370870B (en) | Knowledge and data compound driven equipment multi-working condition identification and performance prediction method |
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 |