CN114066269A - Running risk assessment technology for old assets - Google Patents
Running risk assessment technology for old assets Download PDFInfo
- Publication number
- CN114066269A CN114066269A CN202111373047.9A CN202111373047A CN114066269A CN 114066269 A CN114066269 A CN 114066269A CN 202111373047 A CN202111373047 A CN 202111373047A CN 114066269 A CN114066269 A CN 114066269A
- Authority
- CN
- China
- Prior art keywords
- risk
- data
- equipment
- classification
- risk assessment
- 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
- 238000012502 risk assessment Methods 0.000 title claims abstract description 46
- 238000005516 engineering process Methods 0.000 title claims abstract description 9
- 238000000034 method Methods 0.000 claims abstract description 63
- 238000004364 calculation method Methods 0.000 claims abstract description 27
- 238000000342 Monte Carlo simulation Methods 0.000 claims abstract description 21
- 238000011156 evaluation Methods 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 18
- 230000008569 process Effects 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims description 42
- 238000005070 sampling Methods 0.000 claims description 24
- 238000005457 optimization Methods 0.000 claims description 21
- 239000013598 vector Substances 0.000 claims description 21
- 230000009977 dual effect Effects 0.000 claims description 15
- 238000012423 maintenance Methods 0.000 claims description 15
- 238000004088 simulation Methods 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 10
- 230000036541 health Effects 0.000 claims description 10
- 238000012706 support-vector machine Methods 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 8
- 230000009471 action Effects 0.000 claims description 6
- 238000004519 manufacturing process Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 238000000926 separation method Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 3
- 230000008571 general function Effects 0.000 claims description 3
- 238000012886 linear function Methods 0.000 claims description 3
- 230000007774 longterm Effects 0.000 claims description 3
- 230000010355 oscillation Effects 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 3
- 238000006467 substitution reaction Methods 0.000 claims description 3
- 230000006872 improvement Effects 0.000 abstract description 5
- 230000008439 repair process Effects 0.000 abstract description 2
- 238000011160 research Methods 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 4
- 238000012614 Monte-Carlo sampling Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000012149 noodles Nutrition 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Artificial Intelligence (AREA)
- Educational Administration (AREA)
- Computing Systems (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides an operation risk assessment technology for old assets, which comprises the following steps: step S1, selecting the loss load amount as a risk assessment index based on the index calculation process and the actual application scene factor; step S2, performing state evaluation on the related data sources of the power information system to acquire the real-time fault rate of the power system equipment, and forming an accurate data source for online risk evaluation; and step S3, performing risk assessment based on the Monte Carlo method, and generating regional real-time fault rate. Step S4, risk calculation is carried out by using a Monte Carlo method, risk value prediction is carried out based on a least square product, and corresponding risk sample data are obtained by setting system element data; then, training of improving the LSSVM is carried out by adopting power grid risk sample data to obtain a learning machine and corresponding parameters for online risk assessment, and real-time system data are input into the learning machine to calculate system risk; the invention can provide scientific basis and method support for the decision of the existing technical improvement and major repair project.
Description
Technical Field
The invention relates to the technical field of data analysis of power systems, in particular to an operation risk assessment technology for old assets.
Background
The power grid company is a natural enterprise with dense assets and technology, has a large number of power grid physical assets and a large variety, and has high management complexity of investment, operation and maintenance. For enterprises with intensive assets, the performance of the enterprises is directly and closely linked with the condition and the use efficiency of the assets, the benefits of a power grid company mainly come from the stable and continuous operation of equipment and are closely related with the cost control of the equipment, and the asset utilization level and the reasonable scale have important influence on the production operation and the operation development of the enterprises. The safe and reliable level of the power grid operation is obviously restricted by the states of asset health and the like, the enterprise operation profit level is influenced by the asset utilization level, the power grid structure is increasingly complex, and hidden dangers such as multiple faults, cascading faults and the like exist at all times, which may cause serious influence on the system. The existing risk library acquires a power grid fault set by adopting an enumeration method, is limited to the huge number of large power grid elements, generally only considers the on-off condition (N-1) of a single device or the on-off condition (N-1-1) of the device in the same component group, but the on-off of the device in an actual system has strong uncertainty, different groups of devices can quit operation, the combination number of multiple faults is extremely large, and the analysis is difficult by using a common method. Therefore, the failure scene generation mode of the risk library does not accord with the equipment failure occurrence condition of the actual power system, and the method has limitation.
The old equipment is a general term for equipment with actual service life greater than the old age standard of the asset. Different types of equipment also have different old age standards, and have no definite national standard or enterprise standard definition. The method is characterized in that the overall quality level, the operation environment and the operation and maintenance level of equipment assets are comprehensively analyzed, the age standards of old assets of three types of equipment, namely transmission lines, transformation equipment and distribution equipment are preliminarily set according to various parameters such as the design life of the assets and the average service life of the assets, and the national supervision and policy requirements are combined, namely, the power transmission lines are more than or equal to 25 years, the transformation equipment is more than or equal to 20 years, the distribution lines and the equipment are more than or equal to 16 years, the communication lines and the equipment are more than or equal to 8 years, the automatic control equipment and instruments are more than or equal to 8 years, production management tools are more than or equal to 10 years, the transportation equipment is more than or equal to 10 years, auxiliary production equipment and instruments are more than or equal to 10 years, houses are more than or equal to 25 years, and buildings are more than or equal to 25 years.
The invention mainly aims at the transformer equipment to carry out research and discussion and develops the rapid algorithm research of the risk assessment of the power system. The method comprises the steps of firstly, carrying out risk evaluation calculation by using a Monte Carlo method, and researching an improvement method of a Monte Carlo sampling process and network topology rapid analysis, load flow calculation, line out-of-limit judgment and rapid optimal load subtraction in an outcome analysis process so as to accurately reflect power grid risk information. The Monte Carlo method and the machine learning method are combined, the Monte Carlo method is used for generating accurate risk samples, the machine learning method is used for off-line training, the risk assessment result can be rapidly calculated in the actual application process, and a decision basis for maintenance and operation is provided for operators.
Disclosure of Invention
The invention provides an old asset operation risk assessment technology which can provide scientific basis and method support for decision of the existing technical improvement and major repair project.
The invention adopts the following technical scheme.
An operation risk assessment technology for old assets belongs to an operation risk assessment method for old assets of power system transformer equipment, and comprises the following steps:
step S1, selecting risk indexes based on an index calculation process and actual application scene factors, wherein the risk evaluation indexes are loss load quantities used for expressing power grid risk values, and the power grid risk values are used for deducing a risk optimization scheme easy to understand and execute;
step S2, obtaining the real-time failure rate of the power system equipment by performing state evaluation on the data sources of the power information system data, the environmental data, the equipment historical data and the overhaul data to form an accurate data source for online risk assessment;
and step S3, performing risk assessment based on the Monte Carlo method, and generating the regional real-time fault rate of the equipment in the external environment by combining the equipment state and the specific influence of external factors on the power grid.
Step S4, risk calculation is carried out by using a Monte Carlo method, risk value prediction is carried out based on a least square product, and corresponding risk sample data are obtained by setting system element data; and then training an improved LSSVM by adopting the power grid risk sample data to obtain a learning machine and corresponding parameters for the online risk assessment of the power system, and inputting real-time system data to the learning machine to calculate the corresponding system risk.
In step S1, the event types related to the grid risk value include personal casualties, equipment damage, misoperation, production site accidents, construction site accidents, load loss, substation voltage loss, unplanned shutdown of important equipment, grid oscillation, incorrect action of a secondary system, and failure of a dispatching communication system;
the load loss amount is a load loss index and is used for resolving the consequences of misoperation, secondary system incorrect action and unplanned shutdown of important equipment.
In step S2, the influencing factors of the failure rate of the power system device include a device load, a maintenance condition, an operation duration, and an operation environment, and the state evaluation parameters of the power system are selected from device delivery information, historical statistical data, maintenance records, a current load, and the operation environment; the factory information of the equipment is fixed data; historical statistical data and maintenance records belong to long-term updating data; the current load and the operation environment belong to short-term update data.
In step S2, two methods are used to obtain the failure rate, specifically:
the method A comprises the steps that when the power system equipment has a large amount of historical operating data, the equipment fault rate is used as a function related to time change, function parameters are obtained through historical data fitting, and the equipment fault rate corresponding to the historical operating data is obtained through data statistics;
and B, based on the current state information of the equipment, calculating to obtain a total health degree index, and then establishing a function of the equipment fault rate on the change of the equipment health degree, so as to obtain equipment fault rate data related to the equipment health degree.
The accurate data source for online risk assessment in step S2 is an offline risk library;
optimizing risk calculations for the offline risk library by evaluating the impact of multiple fault conditions in step S3; and the regional real-time fault rate of the equipment in the external environment is generated by combining the equipment state and the specific influence of external factors on the power grid.
Step S3 specifically includes the following steps:
step S31, constructing a simple and applicable probability model or random model according to the problem in the risk calculation, making the solution of the problem correspond to the required characteristics of the random variable in the model, wherein the required characteristics comprise probability, mean value or variance, and the characteristic parameters of the constructed model are matched with the risk calculation problem or the power system;
and step S32, generating random numbers on a computer according to the distribution of each random variable in the model, and realizing a sufficient number of random numbers required by one simulation process. Generally, uniformly distributed random numbers are generated, then random numbers which obey preset distribution are generated, and then a random simulation test is carried out;
s33, designing and selecting a proper sampling method according to the characteristics of the probability model and the distribution characteristics of the random variables, and sampling each random variable, wherein the sampling comprises direct sampling, layered sampling, related sampling or important sampling;
and step S34, carrying out simulation test and calculation according to the established model to obtain a random solution of the problem.
And step S35, carrying out statistical analysis on the simulation test result, and giving a probability solution of the problem and an accuracy estimation of the solution.
In step S3, using the monte carlo method for state evaluation and risk assessment, the probability of a loss of load event can be replaced using a frequency approximation, expressed as:
PL1, L/N formula I;
in the formula, PLRepresenting the probability of a loss of load event; l represents the frequency of occurrence of the load loss event; n represents the total number of samples;
the risk consequences of a loss of load event are expressed as:
SLindicating a risk consequence; pL(i) Representing the off-load amount of the ith sampling scene; psIndicating a reference capacity; the final risk of loss of load is expressed as:
RL＝PL×SLand (5) formula III.
The power system is a large-scale power system, in the risk value prediction of step S4, after the element data of the system device is set, the system state is sampled by the monte carlo method, and the average value of the sampled samples is used to approximate the substitution expectation value, so as to obtain the given sample set of the grid risk sample data.
In step S4, grid risk sample data is given as { (x) for the sample seti,yi) I ═ 1,2, ·, l }, where x isi∈Rn，yi∈{+1,-1}；
When the improved LSSVM is adopted to train power grid risk sample data, a support vector machine is used for solving nonlinear function optimization, specifically, a decision function for expressing a classification hyperplane is constructed, and then the sample data is classified;
the classification hyperplane is expressed as WX + b being 0 by a formula;
the classification hyperplane can ensure the classification precision and simultaneously maximize blank areas on two sides of the hyperplane so as to realize the optimal classification of linear separable problems; the linear divisible refers to dividing sample points belonging to different classes by one or several straight lines;
when the classification meta-platform meets the constraint condition, the expression is yi(WX + b) -1 is more than or equal to 0, i is 1,2, l is the fourth formula, and the training set is linearly separable. Wherein W is a weight vector, X is an input vector, and b belongs to R as a bias;
and p represents the distance between the hyperplane and the nearest sample, the optimal solution of the hyperplane is the hyperplane with the maximum separation edge, the derivation process is to determine w and b when p is maximum, and the prediction function with the best popularization capability obtained by the method is expressed as:
y ═ f (x) ═ sgn (WX + b) formula five;
wherein sgn (·) is a sign function, so that the classification error rate of the original sample is predicted to be minimum;
after the power grid risk sample data is subjected to geometric analysis, the distance from any point in a sample space to the optimal super platform is
Simplifying to obtain g (x) r | | | W0||＝W0X+b0A formula seven;
normalize the function of equation seven such that | g (x)0) 1, obtaining a support vector which is closest to a decision surface of the classification hyperplane and plays a leading role in the SVM;
the algebraic distance of the support vector to the optimal hyperplane is
The maximization of the separation edge of the hyperplane is equivalent to minimizing the norm W of the weight vector, so that the classification hyperplane with the minimum W is the optimal hyperplane;
the learning machine is a classifier algorithm which converts an optimization problem of solving an optimal classification surface by a maximum interval method into a dual problem of the optimization problem, so that an original classification problem is solved by solving a relatively simple dual problem;
in step S4, when the hyperplane is optimized, the Lagrange optimization method and Wofle dual theory are used to convert the problem into its dual problem, i.e. the maximized general function, expressed as a formula
Wherein alpha isiLagrange multipliers corresponding to the samples i;
solving the above problem to obtain the optimal classification function is:
wherein nsv is the number of support vectors, b is a classification threshold, and can be obtained by taking the median value from any pair of support vectors in two classes;
when the relaxation term xi is introducediWhen the generalized classification surface is realized to solve the linear inseparable condition of the training sample, the minimum misclassification sample and the maximum classification interval are considered in a compromise way;
for sample non-linearity problem, it can be converted into linearity problem in some high-dimensional exchange space by non-linear transformation, and then the most is sought in this high-dimensional spaceClassifying the noodles; since there is only an inner product operation (x) between samplesi·xj) The method is related, so that only inner product operation is needed in a high-dimensional space, and the inner product operation can be realized through a function in an original space;
according to the Hilbert-schmidt principle, as long as the kernel function K (x)i·xj) If the Mercer condition is satisfied, it corresponds to the inner product in a certain swap space; with kernel function K (x) satisfying the Mercer conditioni·xj) Instead of inner products in the formula to achieve linear classification after nonlinear transformation, the optimal classification function becomes:
the method is used for solving a density estimation problem and a linear operator equation problem in the online risk assessment of the power system. Introducing a relaxation variable xiiAnd
the following optimization problem is constructed:
wherein the constant C is a penalty coefficient;
the dual space optimization problem is as follows:
wherein alpha isi,
Is a Lagrange multiplier;
solving the problem to obtain the optimal Lagrange multiplier alphai,
Thus, a fitting function is obtained:
the LSSVM of step S4 adds the least square linear system to the support vector machine, converts the training of the SVM into the solution of a linear equation set to accelerate the training of the SVM and improve the accuracy of the model, specifically: for a sample set S of size ll＝{(xi,yi) I ═ 1,2, …, l }, where x isi,yi∈RnThe least squares support vector machine is learned by machine according to the data sample set Sl-m∪{xl-m+1,…,xmObtaining a predicted value according to the change rule of the S, mapping the S to a high-dimensional feature space through nonlinear mapping, and solving the nonlinear problem in a low-dimensional space by applying a linear function problem in the high-dimensional feature space.
The invention can effectively improve the calculation efficiency of online risk assessment. For the asset intensive enterprises such as the electric power enterprises, the rationality of operation and maintenance investment decision can be further improved by scientific and reasonable equipment operation risk assessment, the capital investment scale of technical improvement and overhaul can be reasonably determined, and the accurate investment of the capital of the power grid enterprises can be realized, so that the current complex internal and external operation situation can be coped with, and the enterprise operation development target can be realized.
The invention develops the rapid algorithm research of the risk assessment of the power system. The method comprises the steps of firstly, carrying out risk evaluation calculation by using a Monte Carlo method, and researching an improvement method of a Monte Carlo sampling process and network topology rapid analysis, load flow calculation, line out-of-limit judgment and rapid optimal load subtraction in an outcome analysis process so as to accurately reflect power grid risk information. The Monte Carlo method and the machine learning method are combined, the Monte Carlo method is used for generating accurate risk samples, the machine learning method is used for off-line training, the risk assessment result can be rapidly calculated in the actual application process, and a decision basis for maintenance and operation is provided for operators.
The technical method provided by the invention combines related research results and sample data to evaluate the operation risk as a research object, and provides a rapid calculation method for the risk and sensitivity of the power grid combining a Monte Carlo method and a support vector machine for combining the power grid aiming at the defects that the current risk evaluation method of the power grid lacks consideration on the operation mode and the external environment of the power grid, is difficult to analyze multiple faults and is difficult to apply on line and the wide application of the power information system at the present stage, so that the rapid and accurate risk evaluation analysis can be performed on the power grid, and the rapid calculation method can be further applied to weak link analysis and equipment operation and maintenance grade division.
Drawings
The invention is described in further detail below with reference to the following figures and detailed description:
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is an optimal hyperplane schematic of an SVM;
fig. 3 is a schematic diagram of the structure of the SVM.
Detailed Description
As shown in the figure, the operation risk assessment technology for the old assets belongs to an operation risk assessment method for the old assets of power system transformer equipment, and comprises the following steps:
step S1, selecting risk indexes based on an index calculation process and actual application scene factors, wherein the risk evaluation indexes are loss load quantities used for expressing power grid risk values, and the power grid risk values are used for deducing a risk optimization scheme easy to understand and execute;
step S2, obtaining the real-time failure rate of the power system equipment by performing state evaluation on the data sources of the power information system data, the environmental data, the equipment historical data and the overhaul data to form an accurate data source for online risk assessment;
and step S3, performing risk assessment based on the Monte Carlo method, and generating the regional real-time fault rate of the equipment in the external environment by combining the equipment state and the specific influence of external factors on the power grid.
Step S4, risk calculation is carried out by using a Monte Carlo method, risk value prediction is carried out based on a least square product, and corresponding risk sample data are obtained by setting system element data; and then training an improved LSSVM by adopting the power grid risk sample data to obtain a learning machine and corresponding parameters for the online risk assessment of the power system, and inputting real-time system data to the learning machine to calculate the corresponding system risk.
In step S1, the event types related to the grid risk value include personal casualties, equipment damage, misoperation, production site accidents, construction site accidents, load loss, substation voltage loss, unplanned shutdown of important equipment, grid oscillation, incorrect action of a secondary system, and failure of a dispatching communication system;
the load loss amount is a load loss index and is used for resolving the consequences of misoperation, secondary system incorrect action and unplanned shutdown of important equipment.
In step S2, the influencing factors of the failure rate of the power system device include a device load, a maintenance condition, an operation duration, and an operation environment, and the state evaluation parameters of the power system are selected from device delivery information, historical statistical data, maintenance records, a current load, and the operation environment; the factory information of the equipment is fixed data; historical statistical data and maintenance records belong to long-term updating data; the current load and the operation environment belong to short-term update data.
In step S2, two methods are used to obtain the failure rate, specifically:
the method A comprises the steps that when the power system equipment has a large amount of historical operating data, the equipment fault rate is used as a function related to time change, function parameters are obtained through historical data fitting, and the equipment fault rate corresponding to the historical operating data is obtained through data statistics;
and B, based on the current state information of the equipment, calculating to obtain a total health degree index, and then establishing a function of the equipment fault rate on the change of the equipment health degree, so as to obtain equipment fault rate data related to the equipment health degree.
The accurate data source for online risk assessment in step S2 is an offline risk library;
optimizing risk calculations for the offline risk library by evaluating the impact of multiple fault conditions in step S3; and the regional real-time fault rate of the equipment in the external environment is generated by combining the equipment state and the specific influence of external factors on the power grid.
Step S3 specifically includes the following steps:
step S31, constructing a simple and applicable probability model or random model according to the problem in the risk calculation, making the solution of the problem correspond to the required characteristics of the random variable in the model, wherein the required characteristics comprise probability, mean value or variance, and the characteristic parameters of the constructed model are matched with the risk calculation problem or the power system;
and step S32, generating random numbers on a computer according to the distribution of each random variable in the model, and realizing a sufficient number of random numbers required by one simulation process. Generally, uniformly distributed random numbers are generated, then random numbers which obey preset distribution are generated, and then a random simulation test is carried out;
s33, designing and selecting a proper sampling method according to the characteristics of the probability model and the distribution characteristics of the random variables, and sampling each random variable, wherein the sampling comprises direct sampling, layered sampling, related sampling or important sampling;
and step S34, carrying out simulation test and calculation according to the established model to obtain a random solution of the problem.
And step S35, carrying out statistical analysis on the simulation test result, and giving a probability solution of the problem and an accuracy estimation of the solution.
In step S3, using the monte carlo method for state evaluation and risk assessment, the probability of a loss of load event can be replaced using a frequency approximation, expressed as:
PL1, L/N formula I;
in the formula, PLRepresenting the probability of a loss of load event; l represents loss of loadThe frequency of occurrence of events; n represents the total number of samples;
the risk consequences of a loss of load event are expressed as:
SLindicating a risk consequence; pL(i) Representing the off-load amount of the ith sampling scene; psIndicating a reference capacity; the final risk of loss of load is expressed as:
RL＝PL×SLand (5) formula III.
The power system is a large-scale power system, in the risk value prediction of step S4, after the element data of the system device is set, the system state is sampled by the monte carlo method, and the average value of the sampled samples is used to approximate the substitution expectation value, so as to obtain the given sample set of the grid risk sample data.
In step S4, grid risk sample data is given as { (x) for the sample seti,yi) I ═ 1,2, ·, l }, where x isi∈Rn，yi∈{+1,-1}；
When the improved LSSVM is adopted to train power grid risk sample data, a support vector machine is used for solving nonlinear function optimization, specifically, a decision function for expressing a classification hyperplane is constructed, and then the sample data is classified;
the classification hyperplane is expressed as WX + b being 0 by a formula;
the classification hyperplane can ensure the classification precision and simultaneously maximize blank areas on two sides of the hyperplane so as to realize the optimal classification of linear separable problems; the linear divisible refers to dividing sample points belonging to different classes by one or several straight lines;
when the classification meta-platform meets the constraint condition, the expression is yi(WX + b) -1 is more than or equal to 0, i is 1,2, l is the fourth formula, and the training set is linearly separable. Wherein W is a weight vector, X is an input vector, and b belongs to R as a bias;
and p represents the distance between the hyperplane and the nearest sample, the optimal solution of the hyperplane is the hyperplane with the maximum separation edge, the derivation process is to determine w and b when p is maximum, and the prediction function with the best popularization capability obtained by the method is expressed as:
y ═ f (x) ═ sgn (WX + b) formula five;
wherein sgn (·) is a sign function, so that the classification error rate of the original sample is predicted to be minimum;
after the power grid risk sample data is subjected to geometric analysis, the distance from any point in a sample space to the optimal super platform is
Simplifying to obtain g (x) r | | | W0||＝W0X+b0A formula seven;
normalize the function of equation seven such that | g (x)0) 1, obtaining a support vector which is closest to a decision surface of the classification hyperplane and plays a leading role in the SVM;
the algebraic distance of the support vector to the optimal hyperplane is
The maximization of the separation edge of the hyperplane is equivalent to minimizing the norm W of the weight vector, so that the classification hyperplane with the minimum W is the optimal hyperplane;
the learning machine is a classifier algorithm which converts an optimization problem of solving an optimal classification surface by a maximum interval method into a dual problem of the optimization problem, so that an original classification problem is solved by solving a relatively simple dual problem;
in step S4, when the hyperplane is optimized, the Lagrange optimization method and Wofle dual theory are used to convert the problem into its dual problem, i.e. the maximized general function, expressed as a formula
Wherein alpha isiLagrange multipliers corresponding to the samples i;
solving the above problem to obtain the optimal classification function is:
wherein nsv is the number of support vectors, b is a classification threshold, and can be obtained by taking the median value from any pair of support vectors in two classes;
when the relaxation term xi is introducediWhen the generalized classification surface is realized to solve the linear inseparable condition of the training sample, the minimum misclassification sample and the maximum classification interval are considered in a compromise way;
for the sample nonlinear problem, the sample nonlinear problem can be converted into a linear problem in a certain high-dimensional exchange space through nonlinear transformation, and then an optimal classification surface is sought in the high-dimensional space; since there is only an inner product operation (x) between samplesi·xj) The method is related, so that only inner product operation is needed in a high-dimensional space, and the inner product operation can be realized through a function in an original space;
according to the Hilbert-schmidt principle, as long as the kernel function K (x)i·xj) If the Mercer condition is satisfied, it corresponds to the inner product in a certain swap space; with kernel function K (x) satisfying the Mercer conditioni·xj) Instead of inner products in the formula to achieve linear classification after nonlinear transformation, the optimal classification function becomes:
the method is used for solving a density estimation problem and a linear operator equation problem in the online risk assessment of the power system. Introducing a relaxation variable xiiAnd
the following optimization problem is constructed:
wherein the constant C is a penalty coefficient;
the dual space optimization problem is as follows:
wherein alpha isi,
Is a Lagrange multiplier;
solving the problem to obtain the optimal Lagrange multiplier alphai,
Thus, a fitting function is obtained:
the LSSVM of step S4 adds the least square linear system to the support vector machine, converts the training of the SVM into the solution of a linear equation set to accelerate the training of the SVM and improve the accuracy of the model, specifically: for a sample set S of size ll＝{(xi,yi) I ═ 1,2, …, l }, where x isi,yi∈RnThe least squares support vector machine is learned by machine according to the data sample set Sl-m∪{xl-m+1,…,xmObtaining a predicted value according to the change rule of the S, mapping the S to a high-dimensional feature space through nonlinear mapping, and solving the nonlinear problem in a low-dimensional space by applying a linear function problem in the high-dimensional feature space.
Claims (9)
1. An operation risk assessment technology for old assets belongs to an operation risk assessment method for old assets of power system transformer equipment, and is characterized in that: the method comprises the following steps:
step S1, selecting risk indexes based on an index calculation process and actual application scene factors, wherein the risk evaluation indexes are loss load quantities used for expressing power grid risk values, and the power grid risk values are used for deducing a risk optimization scheme easy to understand and execute;
step S2, obtaining the real-time failure rate of the power system equipment by performing state evaluation on the data sources of the power information system data, the environmental data, the equipment historical data and the overhaul data to form an accurate data source for online risk assessment;
step S3, performing risk assessment based on the Monte Carlo method, and generating a regional real-time fault rate of the equipment in an external environment by combining the equipment state and the specific influence of external factors on the power grid;
step S4, risk calculation is carried out by using a Monte Carlo method, risk value prediction is carried out based on a least square product, and corresponding risk sample data are obtained by setting system element data; and then training an improved LSSVM by adopting the power grid risk sample data to obtain a learning machine and corresponding parameters for the online risk assessment of the power system, and inputting real-time system data to the learning machine to calculate the corresponding system risk.
2. The operational risk assessment technique for old assets according to claim 1, characterized in that: in step S1, the event types related to the grid risk value include personal casualties, equipment damage, misoperation, production site accidents, construction site accidents, load loss, substation voltage loss, unplanned shutdown of important equipment, grid oscillation, incorrect action of a secondary system, and failure of a dispatching communication system;
the load loss amount is a load loss index and is used for resolving the consequences of misoperation, secondary system incorrect action and unplanned shutdown of important equipment.
3. The operational risk assessment technique for old assets according to claim 1, characterized in that: in step S2, the influencing factors of the failure rate of the power system device include a device load, a maintenance condition, an operation duration, and an operation environment, and the state evaluation parameters of the power system are selected from device delivery information, historical statistical data, maintenance records, a current load, and the operation environment; the factory information of the equipment is fixed data; historical statistical data and maintenance records belong to long-term updating data; the current load and the operation environment belong to short-term update data.
4. The operational risk assessment technique for old assets according to claim 3, characterized in that: in step S2, two methods are used to obtain the failure rate, specifically:
the method A comprises the steps that when the power system equipment has a large amount of historical operating data, the equipment fault rate is used as a function related to time change, function parameters are obtained through historical data fitting, and the equipment fault rate corresponding to the historical operating data is obtained through data statistics;
and B, based on the current state information of the equipment, calculating to obtain a total health degree index, and then establishing a function of the equipment fault rate on the change of the equipment health degree, so as to obtain equipment fault rate data related to the equipment health degree.
5. The operational risk assessment technique for old assets according to claim 1, characterized in that: the accurate data source for online risk assessment in step S2 is an offline risk library;
optimizing risk calculations for the offline risk library by evaluating the impact of multiple fault conditions in step S3; and the regional real-time fault rate of the equipment in the external environment is generated by combining the equipment state and the specific influence of external factors on the power grid.
6. The operational risk assessment technique for old assets according to claim 5, characterized in that: step S3 specifically includes the following steps:
step S31, constructing a simple and applicable probability model or random model according to the problem in the risk calculation, making the solution of the problem correspond to the required characteristics of the random variable in the model, wherein the required characteristics comprise probability, mean value or variance, and the characteristic parameters of the constructed model are matched with the risk calculation problem or the power system;
and step S32, generating random numbers on a computer according to the distribution of each random variable in the model, and realizing a sufficient number of random numbers required by one simulation process. Generally, uniformly distributed random numbers are generated, then random numbers which obey preset distribution are generated, and then a random simulation test is carried out;
s33, designing and selecting a proper sampling method according to the characteristics of the probability model and the distribution characteristics of the random variables, and sampling each random variable, wherein the sampling comprises direct sampling, layered sampling, related sampling or important sampling;
and step S34, carrying out simulation test and calculation according to the established model to obtain a random solution of the problem.
And step S35, carrying out statistical analysis on the simulation test result, and giving a probability solution of the problem and an accuracy estimation of the solution.
7. The operational risk assessment technique for old assets according to claim 6, characterized in that: in step S3, using the monte carlo method for state evaluation and risk assessment, the probability of a loss of load event can be replaced using a frequency approximation, expressed as:
PL1, L/N formula I;
in the formula, PLRepresenting the probability of a loss of load event; l represents the occurrence of a loss of load eventThe frequency of (2); n represents the total number of samples;
the risk consequences of a loss of load event are expressed as:
SLindicating a risk consequence; pL(i) Representing the off-load amount of the ith sampling scene; psIndicating a reference capacity;
the final risk of loss of load is expressed as:
RL＝PL×SLand (5) formula III.
8. The operational risk assessment technique for old assets according to claim 7, wherein: the power system is a large-scale power system, in the risk value prediction of step S4, after the element data of the system device is set, the system state is sampled by the monte carlo method, and the average value of the sampled samples is used to approximate the substitution expectation value, so as to obtain the given sample set of the grid risk sample data.
9. The operational risk assessment technique for old assets according to claim 8, wherein: in step S4, grid risk sample data is given as { (x) for the sample seti,yi) I ═ 1,2, ·, l }, where x isi∈Rn，yi∈{+1,-1}；
When the improved LSSVM is adopted to train power grid risk sample data, a support vector machine is used for solving nonlinear function optimization, specifically, a decision function for expressing a classification hyperplane is constructed, and then the sample data is classified;
the classification hyperplane is expressed as WX + b being 0 by a formula;
the classification hyperplane can ensure the classification precision and simultaneously maximize blank areas on two sides of the hyperplane so as to realize the optimal classification of linear separable problems; the linear divisible refers to dividing sample points belonging to different classes by one or several straight lines;
when the classification meta-platform meets the constraint condition, the expression is yi(WX + b) -1 is more than or equal to 0, i is 1,2, l is the fourth formula, and the training set is linearly separable. Wherein W is a weight vector, X is an input vector, and b belongs to R as a bias;
and p represents the distance between the hyperplane and the nearest sample, the optimal solution of the hyperplane is the hyperplane with the maximum separation edge, the derivation process is to determine w and b when p is maximum, and the prediction function with the best popularization capability obtained by the method is expressed as:
y ═ f (x) ═ sgn (WX + b) formula five;
wherein sgn (·) is a sign function, so that the classification error rate of the original sample is predicted to be minimum;
after the power grid risk sample data is subjected to geometric analysis, the distance from any point in a sample space to the optimal super platform is
Simplifying to obtain g (x) r | | | W0||＝W0X+b0A formula seven;
normalize the function of equation seven such that | g (x)0) 1, obtaining a support vector which is closest to a decision surface of the classification hyperplane and plays a leading role in the SVM;
the algebraic distance of the support vector to the optimal hyperplane is
The maximization of the separation edge of the hyperplane is equivalent to minimizing the norm W of the weight vector, so that the classification hyperplane with the minimum W is the optimal hyperplane;
the learning machine is a classifier algorithm which converts an optimization problem of solving an optimal classification surface by a maximum interval method into a dual problem of the optimization problem, so that an original classification problem is solved by solving a relatively simple dual problem;
in step S4, when the hyperplane is optimized, the Lagrange optimization method and Wofle dual theory are used to convert the problem into its dual problem, i.e. the maximized general function, expressed as a formula
Wherein alpha isiLagrange multipliers corresponding to the samples i;
solving the above problem to obtain the optimal classification function is:
wherein nsv is the number of support vectors, b is a classification threshold, and can be obtained by taking the median value from any pair of support vectors in two classes;
when the relaxation term xi is introducediWhen the generalized classification surface is realized to solve the linear inseparable condition of the training sample, the minimum misclassification sample and the maximum classification interval are considered in a compromise way;
for the sample nonlinear problem, the sample nonlinear problem can be converted into a linear problem in a certain high-dimensional exchange space through nonlinear transformation, and then an optimal classification surface is sought in the high-dimensional space; since there is only an inner product operation (x) between samplesi·xj) The method is related, so that only inner product operation is needed in a high-dimensional space, and the inner product operation can be realized through a function in an original space;
according to the Hilbert-schmidt principle, as long as the kernel function K (x)i·xj) If the Mercer condition is satisfied, it corresponds to the inner product in a certain swap space; with kernel function K (x) satisfying the Mercer conditioni·xj) Instead of inner products in the formula to achieve linear classification after nonlinear transformation, the optimal classification function becomes：
The method is used for solving a density estimation problem and a linear operator equation problem in the online risk assessment of the power system. Introducing a relaxation variable xiiAnd
the following optimization problem is constructed:
wherein the constant C is a penalty coefficient;
the dual space optimization problem is as follows:
wherein alpha isi,
Is a Lagrange multiplier;
solving the problem to obtain the optimal Lagrange multiplier alphai,
Thus, a fitting function is obtained:
the LSSVM of step S4 adds the least square linear system to the support vector machine, converts the training of the SVM into the solution of a linear equation set to accelerate the training of the SVM and improve the accuracy of the model, specifically: for a sample set S of size ll＝{(xi,yi) I ═ 1,2, …, l }, where x isi,yi∈RnThe least squares support vector machine is learned by machine according to the data sample set Sl-m∪{xl-m+1,…,xmObtaining a predicted value according to the change rule of the S, mapping the S to a high-dimensional feature space through nonlinear mapping, and solving the nonlinear problem in a low-dimensional space by applying a linear function problem in the high-dimensional feature space.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111373047.9A CN114066269A (en) | 2021-11-19 | 2021-11-19 | Running risk assessment technology for old assets |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111373047.9A CN114066269A (en) | 2021-11-19 | 2021-11-19 | Running risk assessment technology for old assets |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114066269A true CN114066269A (en) | 2022-02-18 |
Family
ID=80278415
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111373047.9A Pending CN114066269A (en) | 2021-11-19 | 2021-11-19 | Running risk assessment technology for old assets |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114066269A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116777086A (en) * | 2023-08-24 | 2023-09-19 | 北京理工大学 | Predictive maintenance method and system for intelligent steel structure production line based on multimode data |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632310A (en) * | 2013-11-19 | 2014-03-12 | 中国南方电网有限责任公司 | Risk evaluation method for operation of large power grid |
| CN104574220A (en) * | 2015-01-30 | 2015-04-29 | 国家电网公司 | Power customer credit assessment method based on least square support vector machine |
| CN106600138A (en) * | 2016-12-09 | 2017-04-26 | 贵州大学 | Secondary equipment risk assessment method |
| CN108846489A (en) * | 2018-07-02 | 2018-11-20 | 天泽信息产业股份有限公司 | A kind of large scale equipment health Evaluation method based on SVM |
| CN109242223A (en) * | 2018-11-26 | 2019-01-18 | 武汉理工光科股份有限公司 | The quantum support vector machines of city Public Buildings Fire Risk is assessed and prediction technique |
| CN109523155A (en) * | 2018-11-13 | 2019-03-26 | 国网甘肃省电力公司电力科学研究院 | A kind of power grid risk assessment method of Monte Carlo and least square method supporting vector machine |
| US20190392360A1 (en) * | 2017-04-04 | 2019-12-26 | Korea Electric Power Corporation | Power transformer asset management device and method therefor |
| CN111507509A (en) * | 2020-03-27 | 2020-08-07 | 东南大学 | Risk assessment method for extreme events of power system |
-
2021
- 2021-11-19 CN CN202111373047.9A patent/CN114066269A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632310A (en) * | 2013-11-19 | 2014-03-12 | 中国南方电网有限责任公司 | Risk evaluation method for operation of large power grid |
| CN104574220A (en) * | 2015-01-30 | 2015-04-29 | 国家电网公司 | Power customer credit assessment method based on least square support vector machine |
| CN106600138A (en) * | 2016-12-09 | 2017-04-26 | 贵州大学 | Secondary equipment risk assessment method |
| US20190392360A1 (en) * | 2017-04-04 | 2019-12-26 | Korea Electric Power Corporation | Power transformer asset management device and method therefor |
| CN108846489A (en) * | 2018-07-02 | 2018-11-20 | 天泽信息产业股份有限公司 | A kind of large scale equipment health Evaluation method based on SVM |
| CN109523155A (en) * | 2018-11-13 | 2019-03-26 | 国网甘肃省电力公司电力科学研究院 | A kind of power grid risk assessment method of Monte Carlo and least square method supporting vector machine |
| CN109242223A (en) * | 2018-11-26 | 2019-01-18 | 武汉理工光科股份有限公司 | The quantum support vector machines of city Public Buildings Fire Risk is assessed and prediction technique |
| CN111507509A (en) * | 2020-03-27 | 2020-08-07 | 东南大学 | Risk assessment method for extreme events of power system |
Non-Patent Citations (1)
| Title |
|---|
| 赵会茹;苏婕;戴杰超;: "我国风力发电项目投资风险评价――基于支持向量机模型", 技术经济, no. 12, 25 December 2012 (2012-12-25) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116777086A (en) * | 2023-08-24 | 2023-09-19 | 北京理工大学 | Predictive maintenance method and system for intelligent steel structure production line based on multimode data |
| CN116777086B (en) * | 2023-08-24 | 2023-10-31 | 北京理工大学 | Predictive maintenance method and system for intelligent steel structure production line based on multimode data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106951695A (en) | Plant equipment remaining life computational methods and system under multi-state | |
| CN109214708B (en) | Electric power system risk assessment method based on cross entropy theory optimization support vector machine | |
| CN108694470B (en) | Data prediction method and device based on artificial intelligence | |
| CN109523155B (en) | Power grid risk assessment method of Monte Carlo and least square support vector machine | |
| CN110929927A (en) | Artificial intelligence prediction model construction method applying power distribution network universe big data | |
| CN116031888A (en) | Dynamic load prediction-based power flow optimization method, system and storage medium | |
| CN115882456B (en) | Power control method and system based on large-scale power grid tide | |
| CN114386753A (en) | Equipment risk comprehensive analysis early warning method based on main transformer load condition | |
| CN104077493A (en) | Method for constructing state evaluation index system of electric relaying protection system | |
| CN110826228A (en) | Regional power grid operation quality limit evaluation method | |
| CN114066269A (en) | Running risk assessment technology for old assets | |
| CN114429248A (en) | Transformer apparent power prediction method | |
| Atamanyuk et al. | Predictive control of electrical equipment reliability on the basis of the non-linear canonical model of a vector random sequence | |
| Roland et al. | Reliability prediction of Port Harcourt electricity distribution network using NEPLAN | |
| Akula et al. | Comparision of traditional and fuzzy failure mode and effects analysis for smart grid electrical distribution systems | |
| CN107239880B (en) | Power grid technical improvement fund demand measurement and calculation model based on equipment life statistical analysis | |
| Nakawiro et al. | A hybrid method for voltage stability constrained optimal reactive power dispatch | |
| CN111428344B (en) | Construction method of degradation analysis model of electromechanical equipment of expressway | |
| Cui et al. | A frequency domain approach to predict power system transients | |
| Sun et al. | Reliability prediction of distribution network based on PCA-GA-BP neural network | |
| Wang et al. | Identifying correlation between planning strategy and reliability of distribution networks based on machine learning | |
| Yang et al. | Lasso-GBDT-based Investment Forecasting for Distribution Transformer Replacement Projects | |
| Li et al. | A decision-making mechanism of network risk control based on grey relation | |
| Yang et al. | Investigation on data-driven life prediction methods | |
| Li et al. | Demand Forecast of Weapon Equipment Spare Parts Based on Improved Gray-Markov Model |
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 |