CN108022058B - Wind turbine state reliability assessment method - Google Patents

Wind turbine state reliability assessment method Download PDF

Info

Publication number
CN108022058B
CN108022058B CN201810052286.6A CN201810052286A CN108022058B CN 108022058 B CN108022058 B CN 108022058B CN 201810052286 A CN201810052286 A CN 201810052286A CN 108022058 B CN108022058 B CN 108022058B
Authority
CN
China
Prior art keywords
state
wind turbine
degradation
performance
probability
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810052286.6A
Other languages
Chinese (zh)
Other versions
CN108022058A (en
Inventor
李建兰
张旭然
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huazhong University of Science and Technology
Original Assignee
Huazhong University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huazhong University of Science and Technology filed Critical Huazhong University of Science and Technology
Priority to CN201810052286.6A priority Critical patent/CN108022058B/en
Publication of CN108022058A publication Critical patent/CN108022058A/en
Application granted granted Critical
Publication of CN108022058B publication Critical patent/CN108022058B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • 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
    • G06Q10/20—Administration of product repair or maintenance
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Physics & Mathematics (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Game Theory and Decision Science (AREA)
  • Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
  • Wind Motors (AREA)

Abstract

The invention belongs to the field of wind turbines, and discloses a wind turbine state reliability assessment method. The method can be used for reliability analysis of mechanical equipment with degradation characteristics, avoids the dependence of the traditional reliability analysis method on large sample data, realizes the state reliability evaluation of the wind turbine under the condition of small samples, and can provide reference for the safe operation of the similar equipment and provide basis for the optimization of maintenance strategies.

Description

Wind turbine state reliability assessment method
Technical Field
The invention belongs to the field of wind turbines, and particularly relates to a method for evaluating the state reliability of a wind turbine.
Background
The mechanical parts of the wind turbine have performance degradation characteristics, the reliability of the mechanical parts of the wind turbine is gradually reduced along with the time, and accurate evaluation of the reliability is an important basis for developing maintenance optimization. The existing equipment reliability evaluation method is mainly based on statistics of large samples, reliability analysis of the equipment is based on a large amount of test or operation data, and a fault time distribution rule of the equipment is obtained through statistical analysis. However, for the wind turbine which is developed on a large scale only in recent years, the running time is not long, the monitoring system is incomplete, and the difficulty of lacking sufficient statistical data is faced at present, so that the reliability data of the wind turbine is lacking. And the wind turbines are large-scale equipment, the mechanical structure is complex, and the service lives of the wind turbines in different regions are different due to different meteorological resource conditions and different maintenance modes of different wind fields. Therefore, the reliability evaluation method for large sample statistics is not suitable for large wind turbines.
Disclosure of Invention
Aiming at the defects or the improvement requirements of the prior art, the invention provides a wind turbine state reliability assessment method which can be used for reliability analysis of mechanical equipment with degradation characteristics, avoids the dependence of the traditional reliability analysis method on large sample data, realizes state reliability assessment of the equipment under the condition of small samples, and can provide reference for safe operation of similar equipment and provide basis for maintenance strategy optimization.
In order to achieve the above object, according to the present invention, there is provided a wind turbine state reliability assessment method, comprising:
1) preprocessing the performance index degradation data of the wind turbine: rejecting performance index data with obvious errors, and selecting the performance index degradation data of the wind turbine in the degradation period as the performance index degradation data;
2) applying the hidden Markov model to the polymorphic reliability assessment of the wind turbine: taking the health state of the wind turbine as the hidden state of a hidden Markov model, taking the performance index degradation data of the wind turbine entering the degradation period obtained in the step 1) as an observable value, defining a state transition probability matrix by adopting the transition relation between the health states, and taking the mapping relation between the health states and the performance indexes as an observed value probability matrix; improving the state transition probability matrix based on the degradation rule of the performance index degradation data entering the degradation period obtained in the step 1) to obtain the state transition probability matrix with the performance degradation characteristics, thereby realizing the reliability evaluation of the wind turbine.
Preferably, in step 1), the specific process of exception data elimination is as follows:
defining a performance degradation rate:
Figure BDA0001552687720000021
wherein (t)1,y1),(t2,y2),…,(ti,yi),…,(tn,yn) N groups of wind turbine performance index data, y, collected at equal time intervalsiIs tiMeasured value of performance index data at time, i is more than or equal to 1 and less than or equal to n, ynMonitoring data when the wind turbine is in fault shutdown; zetaiRepresenting the performance degradation rate of the ith section;
the criteria for outliers are:
Figure BDA0001552687720000022
wherein, Δ ζiIn order to achieve a relative increase in the rate of performance degradation,
Figure BDA0001552687720000023
the sequence delta zeta is relative increment of performance degradation rate ═ delta zeta1,Δζ2,…,Δζn-2]Average value of (1), UiIs Δ ζiσ is the standard deviation of the degradation rate sequence Δ ζ, Δ ζiReflecting the direction of change of the performance degradation when Δ ζiTo account for the increased deterioration in wind turbine performance, Δ ζ is usediThe performance degradation trend of the wind turbine is weakened;
if the inverse of the residual error of the performance degradation rate increment is greater than 3 times the standard deviation, i.e., Δ ζ is satisfied simultaneouslyi< 0 and-UiIf the data is more than 3 sigma, the performance degradation rate is increased sharply and negatively, and the performance index degradation data at the point is considered abnormal and should be removed.
Preferably, in step 1), an absolute threshold method is adopted to determine whether the wind turbine enters a degradation period, and the specific process is as follows:
(t1,y1),(t2,y2),…,(ti,yi),…,(tn,yn) N groups of wind turbine performance index data collected at equal time intervals, where yiIs tiThe measured value of the wind turbine performance index data at the moment, i is more than or equal to 1 and less than or equal to n, and when the performance index data yiAnd when the degradation threshold value is exceeded, the wind turbine is considered to enter a degradation period, namely:
yi≥y* (3)
wherein, y*The performance degradation threshold value of the wind turbine performance index is represented, and the performance index degradation data (t) entering the degradation period is obtainedd,yd)…，(tj,yj)…，(tn,yn) Wherein d is more than 1 and less than n, and j is more than or equal to d and less than or equal to n.
Preferably, a degradation rate method is adopted to judge whether the wind turbine enters a degradation period, and the specific process is as follows:
define degradation rate ζ of only two successive measuring pointsiAnd ζi+1When the performance degradation rate exceeds a specified performance degradation rate threshold, the performance degradation interval is considered to be entered, namely the criterion is satisfied:
(ζi≥ζ*)∧(ζi+1≥ζ*)＝1 (4)
therein, ζ*Obtaining performance index degradation data (t) entering degradation period as performance degradation rate thresholdd,yd)…，(tj,yj)…，(tn,yn) Wherein d is more than 1 and less than n, j is more than or equal to d and less than or equal to n, and i is more than or equal to 1 and less than or equal to n.
Preferably, the hidden markov model of the wind turbine in step 2) is:
λ＝{SOπ(td)AB} (5)
wherein S ═ S1,S2,…,SN]The state of health of the wind turbine is divided into four different states of normal, abnormal, degraded and fault, namely S ═ S1,S2,S3,S4]；
O＝[O(td),O(td+1),…,O(tj),…,O(tn)]Is an observable sequence, and O (t)j)∈{Gm}，GmIs the performance index degradation data (t) of the wind turbine in the degradation period in the step 1)d,yd)…，(tj,yj)…，(tn,yn) Corresponding performance level, wherein1＜d＜n，d≤j≤n，Gm＝[1,2,…,M]And M represents the number of levels of the observable performance indicator classification. Dividing the performance index data of the wind turbine into M (M is 5) grades by taking the operation threshold value as a boundary, and expressing the M grades as Gm＝[1,2,3,4,5]At each time tjThe performance level of (c) is obtained by dividing equally over the entire degradation period:
Figure BDA0001552687720000041
A＝(apq)N×Nis a state transition probability matrix, apq＝P(Sq|Sp) Indicating that the wind turbine is in a hidden state SpTransition to hidden state SqProbability of (B) ═ Bqm)N×MAs a probability matrix of observed values, bqm＝P(Ot＝Gm|Sq) Indicating that the wind turbine is in a hidden state SqIn the case of (2) the performance index degradation data is at level GmProbability of (d), pi (t)d)＝[π(td)S1,π(td)S2,...,π(td)SN](N is 4) is the initial time t of the wind turbinedThe hidden state probability distribution vector represents the probability that the wind turbine is in different hidden states at the initial moment.
Preferably, in step 2), the state transition probability matrix a of the hidden markov model is:
Figure BDA0001552687720000042
in the formula, the state transition probability matrix A is a 4 x 4 matrix and is obtained by training through a Balum-welch algorithm, and the sum of the probabilities of each state transitioning to all the states is 1. In which the state transition can only be shifted towards the degradation direction, i.e. p > q, apq0 and due to the gradual change and continuity of the wind turbine degradation process, a can only jump between itself and the next degraded state, i.e. p < q-1pq＝0。
Preferably, the data (t) is degraded for the performance index entering the degradation periodd,yd),…(tj,yj),…(tn,yn) Performing dimensionless processing to obtain a performance degradation function of the wind turbine:
Figure BDA0001552687720000051
in the formula, f (t)j) As a dimensionless performance degradation function, yminIs (t)d,yd),…,(tj,yj),…(tn,yn) Minimum value of (1) yd，ymaxIs a maximum value, i.e. a fault point value yn。
Introducing a performance degradation function into a state transition matrix A of an equation (7) to construct a state transition probability matrix with time correlation, and reflecting the influence of the performance degradation of the equipment on the state transition probability, namely:
Figure BDA0001552687720000052
in formula (II), a'pq(tj) Are all time-dependent degradation functions, tjAnd performance index degradation data sequence (t)d,yd),…(tj,yj),…(tn,yn) The time of the process is kept consistent.
Preferably, the order state transition probability transition function has the same characteristic parameter, state S, as the equipment performance degradation function1I.e. the state transition probability matrix for the normal state has the following form:
Figure BDA0001552687720000053
in the formula (I), the compound is shown in the specification,1denotes S1State-time state transition probability transition function f*(tj) And a dimensionless performance degradation function f (t)j) The offset of (2); a is12，a23，a34，a44Obtained by the formula (7).
According to hidden Markov modelIterative computation of simultaneous punctuation, when tUTime S2Probability of state pi (t)U)S2Greater than S1Probability of state pi (t)U)S1When the wind turbine enters the state S2If "abnormal", then from tUTime of day start state S1All remain in the state of S1Jump to S2Transition probability level of time a'12＝f*(tU-1) And state S2State transition probability of (a)23And a22The wind turbine is in state S according to the change of the performance degradation function along with time2I.e. the state transition probability matrix for an abnormal state has the following form:
Figure BDA0001552687720000061
in the formula (I), the compound is shown in the specification,2denotes S2State-time state transition probability transition function f*(tj) And a dimensionless performance degradation function f (t)j) The offset of (2);
when t isVTime state S3Probability of (t) ([ pi ])V)S3Greater than state S2Probability of (t) ([ pi ])V)S2When the wind turbine enters the state S3"degenerate" is then from tVTime of day on, state S2Is maintained in a state of S2Jump to S3Transition probability level of time a'23＝f*(tV-1) And state S3State transition probability of (a)33And a34The wind turbine is in the state S when the wind turbine begins to change along with the time according to the performance degradation function3I.e. the state transition probability matrix of the degraded state has the following form:
Figure BDA0001552687720000062
when t isWTime state S4Probability of (t) ([ pi ])W)S4Greater than state S3Probability of (t) ([ pi ])W)S3When the wind turbine enters a 'fault' state, the t is the timeWTime of day start S3The state probability is kept at S3Jump to S4Transition probability level of time a'34＝f*(tW-1) And state S4State transition probability of (a)44Is always 1 because of S4The state is only jumped to the wind turbine, and the wind turbine is in the state S4I.e. the state transition probability matrix for the fault state has the following form:
Figure BDA0001552687720000063
preferably, t is computed from the improved hidden Markov modeljThe probability of the state probability vector at the moment corresponding to each operation state of the wind turbine at different moments is as follows:
Figure BDA0001552687720000064
when the wind turbine is in a fault state, the designated function is lost, so that the probability that the wind turbine is in the fault state is the unreliability degree of the wind turbine, and the reliability degree is as follows:
R(tj)S4＝1-π(tj)S4 (15)
in the formula, pi (t)j)S4The probability that the wind turbine is in a fault state at each moment is shown, and j is more than or equal to d and less than or equal to n.
In general, compared with the prior art, the above technical solution contemplated by the present invention can achieve the following beneficial effects:
1) the invention takes the performance degradation rate increment sequence of the characteristic parameters as an analysis object, and judges that the data point of which the opposite value of the performance degradation rate increment residual error is more than 3 times of the standard deviation is an abnormal value according to the pauta criterion, thereby eliminating the measurement abnormal point which obviously violates the performance degradation rule of the equipment and being beneficial to obtaining accurate data.
2) The method obtains the performance degradation track through performance index degradation data analysis, provides an equipment state transition rule based on performance degradation track description, establishes a hidden Markov state transition matrix with performance degradation characteristics, obtains an equipment state transition probability matrix with time correlation, and finally obtains the equipment state reliability through improving the state probability calculation of the hidden Markov model.
Drawings
FIG. 1 is a diagram of a hidden Markov model;
FIG. 2 is a state probability diagram derived from an improved hidden Markov model based on performance degradation;
FIG. 3 is a graph of reliability derived from an improved hidden Markov model based on performance degradation;
FIG. 4 is a flow chart of a reliability evaluation based on an improved hidden Markov model of performance degradation;
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
According to the method, a device performance degradation rule is extracted according to performance index degradation data of the wind turbine, a state transition matrix with degradation characteristics is obtained by improving hidden Markov according to the degradation rule, the state reliability of the wind turbine is obtained by improving the state probability of the hidden Markov model, and the reliability analysis of the wind turbine component based on small sample data is realized.
1. Data processing method
1.1 Performance indicator degradation data preprocessing
The correct selection of performance index degradation data is a prerequisite for accurate reliability assessment. However, due to the reasons of production field management, operator working errors and the like, the problem of local errors of performance index degradation data is often caused. Therefore, in order to smoothly conduct reliability evaluation, it is necessary to eliminate the performance of the significant error.
1.1.1 Exception data culling
In the process of measuring performance data, some abnormal values are inevitably generated due to objective or subjective reasons, so that a large error occurs in an analysis result, and therefore the abnormal values need to be removed before performance degradation analysis is performed on the data. The invention provides a method for judging an abnormal value based on a performance degradation rate.
Defining a performance degradation rate:
Figure BDA0001552687720000081
wherein (t)1,y1),(t2,y2),…,(ti,yi),…,(tn,yn) N groups of wind turbine performance index data, y, collected at equal time intervalsiIs tiMeasured value of performance index data at time, i is more than or equal to 1 and less than or equal to n, ynMonitoring data when the wind turbine is in fault shutdown; zetaiRepresenting the performance degradation rate of the ith section;
the criteria for outliers are:
Figure BDA0001552687720000091
in the formula (2), Δ ζiIn order to achieve a relative increase in the rate of performance degradation,
Figure BDA0001552687720000092
the sequence delta zeta is relative increment of performance degradation rate ═ delta zeta1,Δζ2,…,Δζn-2]Average value of (1), UiIs Δ ζiσ is the standard deviation of the degradation rate increment sequence Δ ζ, Δ ζiReflecting the direction of change of the performance degradation when Δ ζiTo account for the increased deterioration in wind turbine performance, Δ ζ is usediThe performance degradation trend of the wind turbine is weakened;
according to pauta's criterion, a data point in the sequence is considered an outlier when its absolute value of the residual error is greater than 3 standard deviations. For a wind turbine with performance degradation characteristics, when the wind turbine is operated to be close to a failure, the performance is rapidly degraded, which often results in a performance degradation rate Δ ζiA sharp positive increase in which there is a possibility that the residual error in the rate of performance degradation is greater than 3 times the standard deviation (Δ ζ)i＞0andUi> 3 σ), but this is reasonable. However, if the performance degradation rate is increased sharply and negatively with the time, the degradation rule of the wind turbine is violated, and the phenomenon is often caused by the measurement error of an operator. Therefore, if the inverse value of the residual error of the performance degradation rate is larger than 3 times the standard deviation (Δ ζ)i＜0and-Ui> 3 σ), indicating that a sharp negative increase in the rate of performance degradation has occurred at this time, the data is considered abnormal and should be rejected.
1.1.2 Performance indicator degradation data selection
A typical mechanical device performance degradation process is the bathtub curve. At the initial stage of equipment operation, the fault rate is high due to equipment running-in, and rapidly drops after a period of time, and the equipment enters the normal working period, wherein the fault rate is low and the performance is stable. Over time, the failure rate increases and the equipment enters a degradation period due to problems such as abrasion, fatigue, aging and the like caused by the internal and external environments of the equipment parts. From the analysis of the reliability, the equipment performance in the degradation period is degraded continuously along with the time, and the degradation rule of the equipment performance is truly reflected. And the data volume of the two stages is large, valuable information cannot be provided for the performance degradation rule of the equipment, and even the existence of the two stages can weaken the information of degradation data in the degradation period. Thus, the present invention will perform performance degradation performance analysis using only the data of the degradation period.
The method for judging the deterioration period of the wind turbine comprises two methods: absolute threshold methods and degradation rate methods. The absolute threshold method is mainly characterized in that a performance parameter degradation threshold is set according to data or production experience provided by a manufacturer, and when the performance parameter exceeds the degradation threshold, the wind turbine is considered to enter a degradation period, namely:
yi≥y* (3)
in the formula (3), y*A performance degradation threshold representing a characteristic performance indicator of the wind turbine. The absolute threshold method is suitable for devices that degrade continuously and at a relatively slow rate.
The degradation rate rule determines the degradation period based on the change in the degradation rate of the performance parameter. In order to prevent misjudgment caused by measurement errors, the condition that the degradation rate of two continuous measuring points exceeds a specified performance degradation rate threshold value is defined, and the condition that the degradation rate of two continuous measuring points enters a performance degradation interval is considered to meet the criterion:
(ζi≥ζ*)∧(ζi+1≥ζ*)＝1 (4)
in the formula (4), ζ*Is a performance degradation rate threshold. The degradation rate method is suitable for devices that degrade relatively quickly. The performance of the equipment is stable in a stable period, the performance degradation rate is very low, and the degradation rate shows a remarkable increasing trend when the equipment enters a degradation period. Practice proves that the actual performance parameters can judge that the equipment enters the degradation period as long as any criterion of an absolute threshold value method and a degradation rate method is met. From the sequence (t) based on the above method1,y1),(t2,y2),…,(ti,yi)…(tn,yn) Selecting the data entering the deterioration period as (t)d,yd),…,(tj,yj),…(tn,yn) And d is more than 1 and less than n.
2. Reliability assessment using hidden Markov models based on performance degradation
Reliability is a key index widely recognized in an energy system, and has an important guiding function on safe and economic operation of modern power industry. The traditional Reliability Model is a Binary State Reliability Model (Binary State Reliability Model), which only allows the system to experience two health states of normal and failure. In practice, however, many systems present intermediate transition states that are clearly distinguishable from these two extremes, and therefore a Multi-State Reliability Model (Multi-State System Reliability Model) should be established that includes multiple states, allowing a finite number of different health states for the System or its components. For a polymorphic system, reliability represents the ability of the system to operate above a certain performance level for a period of time.
For a wind turbine, performance indexes of equipment, such as temperature, pressure, flow and the like, can be directly monitored on site, and the health condition of the wind turbine cannot be directly observed. Therefore, the Hidden Markov Model (Hidden Markov Model) is applied to the multi-state reliability evaluation of the wind turbine, the health state and the performance index of the wind turbine are respectively used as the Hidden state and the observable value of the Hidden Markov Model, the transition relation between the health states defines a state transition probability matrix, and the mapping relation between the health state and the performance index is used as an observation value probability matrix, so that the reliability evaluation of the wind turbine is realized.
The hidden Markov model of the wind turbine is as follows:
λ＝{SOπ(td)AB} (5)
in formula (5), S ═ S1,S2,…,SN]The hidden states which are not measurable by the HMM represent different health states in the process that the wind turbine gradually degrades from normal to failure, and N is the number of the hidden states. O ═ O (t)d),O(td+1),…,O(tj),…,O(tn)]Is an observable sequence, is a wind turbine performance index that can be obtained through measurement or testing of instruments and meters, and O (t)j)∈{Gm}，GmIs the degradation data (t) of the performance index in the deterioration period of the wind turbined,yd)…，(tj,yj)…，(tn,yn) Corresponding performance class, Gm＝[1,2,…,M]And M represents the number of levels of the observable performance indicator classification. A ═ apq)N×NIs a state transition probability matrix, apq＝P(Sq|Sp) Indicating that the wind turbine is in a hidden state SpTransition to hidden state SqProbability of (B) ═ Bqm)N×MAs a probability matrix of observed values, bqm＝P(Ot＝Gm|Sq) Indicating that the wind turbine is in a hidden state SqIn the case of (2) the performance index degradation data is at level GmProbability of (d), pi (t)d)＝[π(td)S1,π(td)S2,...,π(td)SN](N is 4) is the initial time t of the wind turbinedAnd the hidden state probability distribution vector represents the probability that the equipment is in different hidden states at the initial moment.
2.2.1 Observation sequence and hidden State
The observable performance indexes of the wind turbine comprise performance parameters such as vibration, temperature and power which can be quantitatively measured through monitoring equipment. Because the measured values of the performance indexes are real number sequences, if the measured values are brought into hidden Markov for calculation, huge dimensionality of the probability matrix of the observed values is caused. Thus, the performance indicators may be processed to reduce the dimensionality of the observation probability matrix. The performance index of the wind turbine is divided into M (M is 5) grades Gm＝[1,2,3,4,5]And reflecting the degradation condition of the performance index degradation data:
Figure BDA0001552687720000121
the hidden state reflects the health state of the wind turbine, can not be directly measured, can only be perceived through an observable sequence, and can reflect the mapping relation between the hidden state and an observable value through an observed value probability matrix in hidden Markov. The invention divides the health state of the wind turbine into four grades of normal, abnormal, degraded and fault, namely S ═ S1,S2,S3,S4](N＝4)。
2.2.2 State transition probability matrix with Performance degradation characteristics
A typical hidden markov model is a time-sequential probabilistic model describing the states of a process by discrete random variables, and a state transition matrix of the model represents the probability of interconversion between the states, and the probability value of each state is only related to the previous state, so the state transition matrix is usually a fixed probability matrix trained based on statistical data. However, for the wind turbine, the conversion between the states has special directionality due to the characteristic of performance deterioration, namely, the states sequentially change along the directions of normal, abnormal, degraded and fault, and the health stateThe transition only occurs between the state itself and the adjacent state, and the state can only jump in one direction towards the degradation direction, but not in reverse direction, i.e. the state of the wind turbine can not be changed from failure to normal without maintenance, as shown in fig. 1. In the figure S1、S2、S3、S4The 4 states of the wind turbine, namely normal, abnormal, degraded and failed, are respectively represented.
According to the wind turbine performance degradation characteristics shown in FIG. 1, the hidden Markov state transition probability matrix A can be simplified as follows:
Figure BDA0001552687720000131
the state transition probability matrix A of the formula (7) is a 4 x 4 matrix, which represents the jump probability of the operation state of the wind turbine among 4 states of normal, abnormal, degraded and fault, and the sum of the probability of each state transitioning to all the states is 1. Due to the unidirectional nature of the device degradation, the matrix A is an upper triangular matrix (p > q, a)pq0), indicating that its state transition can only be shifted toward degradation. However, due to the gradual nature and continuity of the wind turbine degradation process, a state can only jump between itself and the next degraded state. Thus, in the state transition probability matrix (p < q-1, a)pq＝0)。
For traditional hidden Markov, the parameter estimation problem of the model can be solved by adopting Baum-Welch algorithm, namely, by using initial parameter lambda of the model0＝(A0,B0,π(td) Continuously iterating and training until the parameter precision requirement is met, and obtaining a state transition matrix A and an observation value matrix B. However, each element in the state transition matrix a of the conventional hidden markov is a fixed state probability and does not change with time.
Aiming at the performance degradation characteristic of mechanical equipment, the invention improves the traditional hidden Markov, introduces a performance degradation function into a state transition matrix A, constructs a state transition probability with time correlation, and reflects the influence of the equipment performance degradation on the state transition probability, namely:
Figure BDA0001552687720000132
in the formula (8), a'pq(tj) Are time dependent degradation functions.
The level of the performance index can reflect the degree of equipment performance degradation, when the value is lower, the equipment performance state is normal, and when the value is increased, the equipment performance state is reduced. Therefore, the equipment performance degradation trend can be described by the numerical variation track of the equipment performance index.
Degradation data (t) for performance indicators entering degradation periodd,yd),…,(tj,yj),…(tn,yn) Performing dimensionless processing to obtain a performance degradation function of the equipment:
Figure BDA0001552687720000141
in the formula, f (t)j) As a dimensionless performance degradation function, yminIs (t)d,yd),…,(tj,yj),…(tn,yn) Minimum value of (1) yd，ymaxIs a maximum value, i.e. a fault point value yn。
The migration of the health state of the equipment is related to the performance degradation degree of the equipment, and for each state, the equipment jumps to the next worse state with higher probability only when the performance index value reaches a certain level, namely the performance degradation is accumulated to a certain degree, and the equipment state jump probability can be considered as a variable and has a change trend similar to the performance degradation of the equipment. Therefore, the invention adopts the dimensionless performance degradation function to describe the variation trend of the equipment state jump probability.
The state transition matrix of the conventional HMM is a fixed probability matrix, which is an average state transition probability statistically obtained from vibration data. Therefore, the fixed probability a of the next state transition in the state transition probability matrix of the conventional HMM can be madep(p+1)AsImproving corresponding transition probability variable sequences { a'p(p+1)(tj) And will be the approximate average of
Figure BDA0001552687720000142
Endowing each state transition probability variable sequence { a 'in the state transition probability matrix'p(p+1)(tj) The initial value of. Let the state transition probability transition function have the same characteristic parameters as the bearing performance degradation function, and the state transition probability matrix of the 1 st state (normal) has the following form:
Figure BDA0001552687720000143
in the formula (10), the first and second groups,1denotes S1State-time state transition probability transition function f*(tj) And a dimensionless performance degradation function f (t)j) The amount of offset of (c).
Iterative computation of different time points according to a hidden Markov model when t isUTime of day device is at S2Probability pi (t) of state, i.e. abnormal stateU)S2Greater than S1Probability of state pi (t)U)S1When the health state of the equipment is considered to enter abnormal, the health state of the equipment is judged to be abnormal from tUState S in the time-onset state transition probability matrix1Is maintained in state S1Jump S2Transition probability level of time a'12＝f*(tU-1) And state S2State transition probability of (a)23And a22The wind turbine is in state S according to the change of the performance degradation function along with time2I.e. the state transition probability matrix for an abnormal state has the following form:
Figure BDA0001552687720000151
in the formula (12)2Denotes S2State transition probability transition function f*(tj) And a dimensionless performance degradation function f (t)j) Is offset fromAmount of the compound (A).
When t isVThe time of day device is in state S3Probability of (degeneration). pi. (t)V)S3Greater than state S2Probability of (anomaly) pi (t)V)S2When the health state of the wind turbine is considered to enter 'degradation', tVState S in the time-onset state transition probability matrix1The transition probability of (2) is maintained as a'12＝f*(tU-1) Invariable, S2Is in transition probability holding state S2Jump S3Transition probability level of time a'23＝f*(tV-1) And state S3State transition probability of (a)33And a34The wind turbine is in the state S when the wind turbine begins to change along with the time according to the performance degradation function3I.e. the state transition probability matrix of the degraded state has the following form:
Figure BDA0001552687720000152
when t isWAt the moment that the wind turbine is in the state S4I.e. probability of fault condition pi (t)W)S4Greater than state S3I.e. probability of degenerate state pi (t)W)S3At the moment, the health state of the wind turbine is considered to enter the fault, and then the t is carried outWState S in the time-onset state transition probability matrix1The transition probability of (2) is maintained as a'12Invariable, S2The transition probability of (2) is maintained as a'23，S3State probability remains at S3Jump S4Transition probability level of time a'34＝f*(tW-1) And state S4State transition probability of (a)44Is always 1 because of S4The state is only jumped to the wind turbine, and the wind turbine is in the state S4I.e. the state transition probability matrix for the fault state has the following form:
Figure BDA0001552687720000161
2.2.3 wind turbine State reliability assessment
The state probability vector of each moment can be calculated according to the initial distribution and the state transition probability, and the probability of the corresponding wind turbine in different states is as follows:
Figure BDA0001552687720000162
from pi (t) at each momentj) The probability of the wind turbine being in different states at each moment based on the improved hidden Markov model shown in FIG. 2 can be obtained.
When a wind turbine is in a fault state, it is meant to lose its designated function. Therefore, the probability of the wind turbine being in the fault state is the unreliability, and the reliability of the wind turbine is as follows:
R(tj)S4＝1-π(tj)S4 (15)
in the formula (15), π (t)j)S4Representing the probability of the wind turbine being in a fault state, denoted by R (t)j) The reliability of the wind turbine shown in FIG. 3 may be plotted over time.
In summary, the wind turbine condition reliability assessment process is shown in FIG. 4.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (6)

1. A wind turbine state reliability assessment method is characterized by comprising the following steps:
1) preprocessing the performance index degradation data of the wind turbine: rejecting performance index data with obvious errors, and selecting the performance index data of the wind turbine in the deterioration period as performance index degradation data;
2) applying the hidden Markov model to the polymorphic reliability assessment of the wind turbine: taking the health state of the wind turbine as the hidden state of a hidden Markov model, taking the performance index degradation data of the wind turbine entering the degradation period obtained in the step 1) as an observable value, defining a state transition probability matrix by adopting the transition relation between the health states, and taking the mapping relation between the health states and the performance indexes as an observed value probability matrix; improving the state transition probability matrix based on the degradation rule of the performance index degradation data entering the degradation period obtained in the step 1) to obtain a state transition probability matrix with performance degradation characteristics, so as to realize reliability evaluation of the wind turbine;
the state transition probability matrix A of the hidden Markov model is as follows:
Figure FDA0002633563140000011
wherein the state transition probability matrix A is a 4 x 4 matrix trained by Balum-welch algorithm, and the sum of the probabilities of each state transitioning to all states is 1, wherein the state transitions can only transition towards the degradation direction, i.e. p > q, apq0 and due to the gradual change and continuity of the wind turbine degradation process, a can only jump between itself and the next degraded state, i.e. p < q-1pq＝0；
Degradation data (t) for performance indicators entering degradation periodd,yd),…(tj,yj),…(tn,yn) Performing dimensionless processing to obtain a performance degradation function of the wind turbine:
Figure FDA0002633563140000012
in the formula, f (t)j) As a dimensionless performance degradation function, yminIs (t)d,yd),…,(tj,yj),…(tn,yn) Minimum value of (1) yd，ymaxIs a maximum value, i.e. a fault point value yn，1≤i≤n，1＜d＜n，d≤j≤n；
Introducing a performance degradation function into a state transition matrix A of an equation (7) to construct a state transition probability matrix with time correlation, and reflecting the influence of the performance degradation of the equipment on the state transition probability, namely:
Figure FDA0002633563140000021
in formula (II), a'pq(tj) Are all time-dependent degradation functions, tjAnd performance index degradation data sequence (t)d,yd),…(tj,yj),…(tn,yn) The time of the operation is kept consistent;
making the state transition probability transition function have the same characteristic parameter as the equipment performance degradation function, state S1I.e. the state transition probability matrix for the normal state has the following form:
Figure FDA0002633563140000022
in the formula (I), the compound is shown in the specification,1denotes S1State-time state transition probability transition function f*(tj) And a dimensionless performance degradation function f (t)j) The offset of (2); a is12，a23，a34，a44Obtained by formula (7);
iterative computation of different time points according to a hidden Markov model when t isUTime S2Probability of state pi (t)U)S2Greater than S1Probability of state pi (t)U)S1When the wind turbine enters the state S2If "abnormal", then from tUTime of day start state S1All remain in the state of S1Jump to S2Transition probability level of time a'12＝f*(tU-1) And state S2State transition probability of (a)23And a22The wind turbine is in state S according to the change of the performance degradation function along with time2I.e. the state transition probability matrix for an abnormal state has the following form:
Figure FDA0002633563140000023
in the formula (I), the compound is shown in the specification,2denotes S2State-time state transition probability transition function f*(tj) And a dimensionless performance degradation function f (t)j) The offset of (2);
when t isVTime state S3Probability of (t) ([ pi ])V)S3Greater than state S2Probability of (t) ([ pi ])V)S2When the wind turbine enters the state S3"degenerate" is then from tVTime of day on, state S2Is maintained in a state of S2Jump to S3Transition probability level of time a'23＝f*(tV-1) And state S3State transition probability of (a)33And a34The wind turbine is in the state S when the wind turbine begins to change along with the time according to the performance degradation function3I.e. the state transition probability matrix of the degraded state has the following form:
Figure FDA0002633563140000031
when t isWTime state S4Probability of (t) ([ pi ])W)S4Greater than state S3Probability of (t) ([ pi ])W)S3When the wind turbine enters a 'fault' state, the t is the timeWTime of day start S3The state probability is kept at S3Jump to S4Transition probability level of time a'34＝f*(tW-1) And state S4State transition probability of (a)44Is always 1 because of S4The state is only jumped to the wind turbine, and the wind turbine is in the state S4I.e. the state transition probability matrix for the fault state has the following form:
Figure FDA0002633563140000032
2. the method for evaluating the reliability of the state of the wind turbine as claimed in claim 1, wherein in the step 1), the specific process of abnormal data elimination is as follows:
defining a performance degradation rate:
Figure FDA0002633563140000033
wherein (t)1,y1),(t2,y2),…,(ti,yi),…,(tn,yn) N groups of wind turbine performance index data, y, collected at equal time intervalsiIs tiMeasured value of performance index data at time, i is more than or equal to 1 and less than or equal to n, ynMonitoring data when the wind turbine is in fault shutdown; zetaiRepresenting the performance degradation rate of the ith section;
the criteria for outliers are:
Figure FDA0002633563140000041
wherein, Δ ζiIn order to achieve a relative increase in the rate of performance degradation,
Figure FDA0002633563140000042
the sequence delta zeta is relative increment of performance degradation rate ═ delta zeta1,Δζ2,…,Δζn-2]Average value of (1), UiIs Δ ζiσ is the standard deviation of the degradation rate sequence Δ ζ, Δ ζiReflecting the direction of change of the performance degradation when Δ ζiTo account for the increased deterioration in wind turbine performance, Δ ζ is usediThe performance degradation trend of the wind turbine is weakened;
if the inverse of the residual error of the performance degradation rate increment is greater than 3 times the standard deviation, i.e., Δ ζ is satisfied simultaneouslyi< 0 and-UiWhen the value is more than 3 sigma, the performance degradation rate is sharp at the timeAnd if the data is increased negatively, the degraded data of the performance index at the current point is considered to be abnormal, and the data is rejected.
3. The method for evaluating the state reliability of the wind turbine as claimed in claim 1, wherein in the step 1), an absolute threshold method is adopted to judge whether the wind turbine enters a degradation period, and the specific process is as follows:
(t1,y1),(t2,y2),…,(ti,yi),…,(tn,yn) N groups of wind turbine performance index data collected at equal time intervals, where yiIs tiThe measured value of the wind turbine performance index data at the moment, i is more than or equal to 1 and less than or equal to n, and when the performance index data yiAnd when the degradation threshold value is exceeded, the wind turbine is considered to enter a degradation period, namely:
yi≥y* (3)
wherein, y*The performance degradation threshold value of the wind turbine performance index is represented, and the performance index degradation data (t) entering the degradation period is obtainedd,yd)…，(tj,yj)…，(tn,yn) Wherein d is more than 1 and less than n, and j is more than or equal to d and less than or equal to n.
4. The method for evaluating the state reliability of the wind turbine as claimed in claim 1, wherein a degradation rate method is used for judging whether the wind turbine enters a degradation period, and the specific process is as follows:
define degradation rate ζ of only two successive measuring pointsiAnd ζi+1When the performance degradation rate exceeds a specified performance degradation rate threshold, the performance degradation interval is considered to be entered, namely the criterion is satisfied:
(ζi≥ζ*)∧(ζi+1≥ζ*)＝1 (4)
therein, ζ*Obtaining performance index degradation data (t) entering degradation period as performance degradation rate thresholdd,yd)…，(tj,yj)…，(tn,yn) D is more than 1 and less than n, j is more than or equal to d and less than or equal to n, i is more than or equal to 1 and less than or equal to n, zetaiExpress ith sectionThe rate of degradation is enabled.
5. The method of claim 3, wherein the hidden Markov models of the wind turbine in step 2) are:
λ＝{SOπ(td)AB} (5)
wherein S ═ S1,S2,…,SN]The state of health of the wind turbine is divided into four different states of normal, abnormal, degraded and fault, namely S ═ S1,S2,S3,S4]；
O＝[O(td),O(td+1),…,O(tj),…,O(tn)]Is an observable sequence, and O (t)j)∈{Gm}，GmIs the performance index degradation data (t) of the wind turbine in the degradation period in the step 1)d,yd)…，(tj,yj)…，(tn,yn) Corresponding performance grades, wherein d is more than 1 and less than n, d is more than or equal to j and less than or equal to n, Gm＝[1,2,…,M]M represents the number of grades divided by the observable performance index, and the wind turbine performance index data is divided into M grades by taking the operation threshold as a boundary, and is represented as Gm＝[1,2,3,4,5]At each time tjThe performance level of (c) is obtained by dividing equally over the entire degradation period:
Figure FDA0002633563140000051
A＝（apq）N×Nis a state transition probability matrix, apq＝P(Sq|Sp) Indicating that the wind turbine is in a hidden state SpTransition to hidden state SqProbability of (B) ═ Bqm)N×MAs a probability matrix of observed values, bqm＝P(Ot＝Gm|Sq) Indicating that the wind turbine is in a hidden state SqIn the case of (2) the performance index degradation data is at level GmProbability of (d), pi (t)d)＝[π(td)S1,π(td)S2,...,π(td)SN]For the initial moment t of the wind turbinedThe hidden state probability distribution vector represents the probability that the wind turbine is in different hidden states at the initial moment.
6. The wind turbine state reliability assessment method according to claim 1, wherein t is calculated according to an improved hidden Markov modeljThe probability vector of the state at the moment, corresponding to the probability of each state of the wind turbine at different moments, is as follows:
Figure FDA0002633563140000061
when the wind turbine is in a fault state, the designated function is lost, so that the probability that the wind turbine is in the fault state is the unreliability degree of the wind turbine, and the reliability degree is as follows:
R(tj)S4＝1-π(tj)S4 (15)
in the formula, pi (t)j)S4The method is used for representing the probability of the wind turbine in the fault state at each moment, and dividing the health state of the wind turbine into four different states of normal, abnormal, degraded and fault, namely S ═ S1,S2,S3,S4]，d≤j≤n，1＜d＜n。
CN201810052286.6A 2018-01-19 2018-01-19 Wind turbine state reliability assessment method Active CN108022058B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810052286.6A CN108022058B (en) 2018-01-19 2018-01-19 Wind turbine state reliability assessment method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810052286.6A CN108022058B (en) 2018-01-19 2018-01-19 Wind turbine state reliability assessment method

Publications (2)

Publication Number Publication Date
CN108022058A CN108022058A (en) 2018-05-11
CN108022058B true CN108022058B (en) 2020-12-08

Family

ID=62075185

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810052286.6A Active CN108022058B (en) 2018-01-19 2018-01-19 Wind turbine state reliability assessment method

Country Status (1)

Country Link
CN (1) CN108022058B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109145128B (en) * 2018-08-15 2021-06-29 东北大学 Online visualization method for large-scale data
US11639711B2 (en) * 2018-08-31 2023-05-02 Siemens Aktiengesellschaft Methods and systems for performance loss estimation of single input systems
CN109657187B (en) * 2018-11-13 2023-04-28 广东电网有限责任公司 10kV cable line state evaluation method
CN111046458B (en) * 2019-11-14 2022-07-01 北京航空航天大学 Metal roof performance degradation model based on Markov chain and method thereof
US10748091B1 (en) 2020-01-16 2020-08-18 Applied Underwriters, Inc. Forecasting digital reservoir controller
CN112799890B (en) * 2020-12-31 2022-10-14 南京航空航天大学 Bus SEU-resistant reliability modeling and evaluating method
CN112783676B (en) * 2020-12-31 2022-09-20 南京航空航天大学 Storage system soft error resistance reliability modeling and evaluating method based on Z language
CN113887055A (en) * 2021-10-11 2022-01-04 西安因联信息科技有限公司 Centrifugal fan performance degradation evaluation method and system based on operation parameters

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102096770A (en) * 2011-01-25 2011-06-15 上海交通大学 Network-based performance degradation assessment method of rotary machine
CN103034170A (en) * 2012-11-27 2013-04-10 华中科技大学 Numerical control machine tool machining performance prediction method based on intervals
CN103472392A (en) * 2013-09-06 2013-12-25 海特韦尔流体控制（河南）有限公司 Service life prediction method for electric actuating mechanism
CN106227906A (en) * 2016-05-20 2016-12-14 广州韵脉质量技术服务有限公司 A kind of appraisal procedure of the intelligent manufacturing equipment reliability analyzed based on performance degradation

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9817881B2 (en) * 2013-10-16 2017-11-14 Cypress Semiconductor Corporation Hidden markov model processing engine

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102096770A (en) * 2011-01-25 2011-06-15 上海交通大学 Network-based performance degradation assessment method of rotary machine
CN103034170A (en) * 2012-11-27 2013-04-10 华中科技大学 Numerical control machine tool machining performance prediction method based on intervals
CN103472392A (en) * 2013-09-06 2013-12-25 海特韦尔流体控制（河南）有限公司 Service life prediction method for electric actuating mechanism
CN106227906A (en) * 2016-05-20 2016-12-14 广州韵脉质量技术服务有限公司 A kind of appraisal procedure of the intelligent manufacturing equipment reliability analyzed based on performance degradation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Health Condition Monitoring of Machines Based on Hidden Markov Model and Contribution Analysis;Jianbo Yu;《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》;20120831;第61卷(第8期);全文 *
航空发动机气路部件退化仿真及健康评估;李润国;《中国优秀硕士学位论文全文数据库 工程科技II辑》;20160215(第2期);全文 *

Also Published As

Publication number Publication date
CN108022058A (en) 2018-05-11

Similar Documents

Publication Publication Date Title
CN108022058B (en) Wind turbine state reliability assessment method
CN108335021B (en) Method for evaluating state reliability of wind turbine and maintenance decision optimization
Butler et al. Exploiting SCADA system data for wind turbine performance monitoring
CN110362048A (en) Blower critical component state monitoring method and device, storage medium and terminal
CN112683535B (en) Bearing life prediction method based on multi-stage wiener process
CN112418517B (en) Prediction method for freeze injury short-term early warning
CN109344610B (en) Method and device for detecting sequence attack
Le et al. Multi-branch Hidden semi-Markov modeling for RUL prognosis
CN111504647A (en) AR-MSET-based performance degradation evaluation method for rolling bearing
CN112598144A (en) CNN-LSTM burst fault early warning method based on correlation analysis
CN112418306A (en) Gas turbine compressor fault early warning method based on LSTM-SVM
CN112734201B (en) Multi-equipment overall quality evaluation method based on expected failure probability
CN111563685B (en) Power generation equipment state early warning method based on auto-associative kernel regression algorithm
CN111192163B (en) Generator reliability medium-short term prediction method based on wind turbine generator operating data
CN117471346A (en) Method and system for determining remaining life and health status of retired battery module
CN112329272A (en) Residual life fitting prediction method integrating photovoltaic module degradation and life data
CN110794683A (en) Wind power gear box state evaluation method based on deep neural network and kurtosis characteristics
US11339763B2 (en) Method for windmill farm monitoring
CN113654651A (en) Strong robust signal early degradation feature extraction and equipment running state monitoring method
Kumar et al. Gas turbine engine operational data analysis for anomaly detection: Statistical vs. neural network approach
CN114783550B (en) Artificial intelligence-based new material online evaluation method and system
CN113627088B (en) Machine performance degradation evaluation method and system based on gene programming and data fusion
CN111709191B (en) Gear performance degradation evaluation method based on AR model and dictionary learning
CN110543153B (en) Method for detecting transition condition fault of multi-condition industrial process
CN117494026B (en) Method, system and storage medium for positioning icing fault of power transmission line under cold and tidal weather

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant