CN115659612A - Method for predicting residual life of offshore wind turbine generator with typhoon influence taken into consideration - Google Patents

Abstract

The invention relates to a method for predicting the remaining life of an offshore wind turbine considering the influence of a typhoon. The state degradation and shock correlation model quantifies the ability to withstand external shocks; based on the time-varying randomness of the performance degradation of the unit performance caused by typhoon random shocks, a time-varying failure rate model of offshore wind turbine components considering the dependence of degradation and shocks is constructed, and based on In this model, under the scenario of typhoon impact, the remaining life prediction model of unit components is constructed, and the particle swarm optimization algorithm is used to optimize the parameters with the minimum prediction error to obtain an optimized remaining life prediction model; based on real-time data, the remaining life prediction model is corrected and optimized to realize the remaining life predict. Compared with the prior art, the present invention has the advantages of high life prediction accuracy and the like.

Description

A Method for Remaining Life Prediction of Offshore Wind Turbine Considering the Impact of Typhoon

technical field

The invention relates to the field of life prediction of offshore wind turbines, in particular to a method for predicting the remaining life of offshore wind turbines considering the influence of typhoons.

Background technique

Offshore wind power is an important force to accelerate my country's realization of the "3060" dual-carbon strategy, and the highly reliable operation of offshore wind turbines is a key factor for the sustainable development of large-scale offshore wind power. Accurately predicting the remaining life of offshore wind turbines has important theoretical significance and engineering application value for reducing the risk of sudden failure and making high-reliability operation and high-efficiency maintenance decisions. The coastal areas where large-scale and clustered development of offshore wind power in my country are severely affected by severe weather such as typhoons. The strong winds and huge waves brought by typhoons have brought huge challenges to the accurate prediction of the remaining life of offshore wind turbines.

The existing research on life prediction of offshore wind turbines mainly has the following two problems: 1) In the past, in the research on the remaining life prediction of wind turbines, it was usually based on the natural degradation process of the wind turbines, assuming that the performance of the turbines gradually deteriorates with time, ignoring the objective Due to the random impact of typhoon weather at sea, it is difficult to accurately describe the evolution law of the actual degradation of the unit under the harsh operating environment at sea. 2) Under the background of multi-source and multi-dimensional operation and maintenance big data at sea, the problems of accurately describing the relationship between random shock and degradation state of units, accurate modeling of unit failure process and uncertainty quantification of remaining life prediction still need to be solved.

Contents of the invention

The purpose of the present invention is to provide a method for predicting the remaining life of offshore wind turbines considering the influence of typhoons, to ensure that the mean square error of prediction is minimized under typhoon weather, to realize the interactive linkage between the state of the unit components and the random degradation model, and to obtain accurate offshore wind turbines. The lifetime prediction value of the wind turbine.

The purpose of the present invention can be achieved through the following technical solutions:

A method for predicting the remaining life of an offshore wind turbine considering the impact of a typhoon, comprising the following steps:

Based on the impact mechanism of typhoon on unit component degradation, the impact of direct impact and indirect impact on component failure rate is analyzed, and an incremental model of component failure rate under typhoon impact is constructed;

Aiming at the problem of quantifying the correlation between component degradation process and impact, combined with the monitoring data of unit operating status, construct a degradation and impact correlation model of component operating status to quantify the ability to withstand external impacts;

Based on the time-varying randomness of performance degradation caused by typhoon random impact, combined with the incremental model of component failure rate and the correlation model between degradation and impact under typhoon impact, a time-varying failure rate model of offshore wind turbine components is constructed considering the dependence of degradation and impact;

Based on the time-varying failure rate model of offshore wind turbine components, in the typhoon impact scenario, the remaining life prediction model of the unit components is constructed, and the particle swarm optimization algorithm is used to optimize the component failure rate increment model and the correlation between degradation and impact under the typhoon impact with the minimum prediction error The parameters of the performance model are used to obtain an optimized remaining life prediction model;

Based on the optimized remaining life prediction model, combined with the typhoon impact monitoring data and unit real-time operating status monitoring data, the optimized remaining life prediction model is corrected and the remaining life prediction is realized.

The construction of the component failure rate incremental model under the impact of the typhoon includes the following steps:

Obtain typhoon data in typhoon-prone areas;

Construct the Batts typhoon wind field model;

Assuming that the number of typhoon shocks encountered by offshore wind turbines during operation obeys the Poission process with a parameter of λ, and the typhoon wind field model remains unchanged before and after the typhoon passes through the wind farm, the Monte Carlo sampling method is used to obtain the probability of each typical typhoon shock, and the establishment Typhoon random impact model;

Based on the random impact model of typhoon, the influence of direct impact and indirect impact on component failure rate is analyzed, and the incremental model of component failure rate under typhoon impact is constructed.

The typhoon random impact model is:

Among them, P(N(t)=n) represents the probability of n times of typhoon impact within (0, t), λ is the arrival rate of typhoon, f _G (t _si ) is the required time for the first arrival of the i-th impact Probability density at time t _si , Γ(i) is the gamma function value; V _Rmax is the gradient wind speed at the maximum wind radius, K is an empirical parameter, the value range is 6.93-6.97, Δp is the typhoon center pressure difference, R _max is the radius of the maximum wind speed, f is a scientific parameter; v is the instantaneous wind speed of the typhoon, V _s is the moving speed of the typhoon; r is the distance from any position of the typhoon to the center of the typhoon; x is an empirical parameter, and its value ranges from 0.5 to 0.7.

The component failure rate incremental model under the impact of typhoon is:

为部件退化与冲击相关性函数，表示部件k在运行状态为

时受到冲击对故障率增量的影响，t_si和t_ei分别表示台风对机组冲击影响的首达时刻和解除时刻，α、ε为部件的台风直接或间接冲击修正系数，v为台风瞬时风速。in,

is the component degradation and impact correlation function, which means that the component k is in the running state of

t _si and t _ei represent the first arrival time and release time of the impact of typhoon on the unit, respectively, α and ε are the typhoon direct or indirect impact correction coefficients of components, and v is the typhoon instantaneous wind speed .

Distinguishing the effects of direct shocks and indirect shocks, the component failure rate increment model under direct shocks is expressed as:

The incremental model of component failure rate under indirect shock is expressed as:

Among them, Δλ _sk1i is the failure rate of external components of the unit caused by the direct impact of typhoon, α _k1 and ε _k1 are the correction coefficients of direct typhoon impact respectively, w is the wind load perpendicular to the surface of the fan, ρ is the air density, and C _s is the shape coefficient , S _α is the projected area on the plane perpendicular to the wind direction; Δλ _sk2i represents the failure rate of the internal components of the unit caused by the indirect typhoon impact, α _k2 and ε _k2 are the correction coefficients of typhoon indirect impact respectively, p is the power of the fan, and C _p is the wind energy Conversion efficiency coefficient, λ _a is the tip speed ratio of the fan, β _a is the pitch angle of the fan blade, and r _a is the radius of the fan blade.

The construction of the degradation and impact correlation model based on the operating state of the components includes the following steps:

Obtain historical operation monitoring data of offshore wind turbines;

Extract the health state characteristics of offshore wind turbine components under different operating conditions through principal component cluster analysis, and divide different component states according to the difference between other states and health states;

According to the component state division results and historical state evolution process under different working conditions, based on the Markov chain state transition process, a stochastic state model is established to describe the historical state transition process of components, and the degradation and impact correlation model is obtained.

The degradation-shock correlation model is expressed as:

R=X'X/ _nq

Y=XU=[Y ₁ ,Y ₂ ,…,Y _q ]

为部件k的运行状态，

为部件退化与冲击相关性函数，μ为状态量修正系数，风机的某一部件相关的SCADA数据标准化处理后的输入变量矩阵X包含q维初始参量，每维参量包含n_q个样本，X'表示X的转置，相关系数矩阵R的特征方程有q个特征值，λ₁≥λ₂≥…≥λ_q，U＝(U₁,U₂,…,U_q)，为特征值对应的特征向量，Y为主成分，按照主成分方差从大到小的顺序进行排序，Y_q为第q主成分，λ_ji为工况j下第i个主成分对应的特征值，δ为累积方差贡献率，每类工况j的主成分分析结果按照累积方差贡献率大于δ选取ω_j个主成分，π(t_g)为监测点t_g时刻风机部件各状态的概率分布，A为状态转移矩阵，Δt为任意时间间隔，t₀为状态转移时间歩长，S_k为各状态量化值向量矩阵。in,

is the operating state of component k,

is the component degradation and impact correlation function, μ is the state quantity correction coefficient, the input variable matrix X after normalization processing of the SCADA data related to a certain component of the fan contains q-dimensional initial parameters, and each dimensional parameter contains n _q samples, X' Represents the transposition of X, the characteristic equation of the correlation coefficient matrix R has q eigenvalues, λ ₁ ≥λ ₂ ≥…≥λ _q , U=(U ₁ ,U ₂ ,…,U _q ), which corresponds to the eigenvalues Eigenvector, Y is the main component, sorted according to the order of the principal component variance from large to small, Y _q is the qth principal component, λ _ji is the eigenvalue corresponding to the i-th principal component under working condition j, and δ is the cumulative variance Contribution rate, the principal component analysis results of each type of working condition j select _ωj principal components according to the cumulative variance contribution rate greater than δ, π(t _g ) is the probability distribution of each state of the fan component at the monitoring point t _g , and A is the state transition matrix, Δt is an arbitrary time interval, _t0 is the time step of state transition, and S _k is the vector matrix of each state quantization value.

Assuming that the failure rate λ _k of the system component k is composed of the natural degradation failure rate λ _0k and the component cumulative failure rate increment Δλ _ski (t) caused by multiple typhoon impacts, the life of the unit component is a non-negative continuous random variable, The natural degradation process is described by Weibull distribution, combined with the component failure rate increment model under typhoon impact and the degradation and impact correlation model, the time-varying failure rate model of offshore wind turbine components considering the dependence of degradation and impact is obtained:

为部件退化与冲击相关性函数，表示部件k在运行状态为

时受到冲击对故障率增量的影响；t_si和t_ei分别表示台风对机组冲击影响的首达时刻和解除时刻；β为形状参数，η为尺度参数。Among them, i is the number of typhoon impacts; N(t) is the cumulative number of typhoon impacts up to time t; D _i (t) is the impact amplitude at time t during the ith typhoon period;

is the component degradation and impact correlation function, which means that the component k is in the running state of

t _si and t _ei represent the first arrival time and release time of the typhoon’s impact on the unit, respectively; β is a shape parameter, and η is a scale parameter.

The remaining life prediction model of the unit components considering the dependence of degradation and impact is:

表示考虑随机台风冲击影响的部件故障率期望值函数，n_z表示台风冲击场景的总次数，Z为台风冲击场景，单次台风冲击场景为台风最大风速半径、移动速度、中心气压差、最大风速半径处的风速和台风的移动方向的集合，z为台风冲击场景变量，T为部件发生故障的时刻，运行至t时刻部件剩余寿命为

为剩余寿命期望值，x为剩余寿命时间变量，

表示剩余寿命期望值

的分布函数，

为根据历史台风冲击下部件可靠度期望值函数，

表示考虑随机台风冲击场景下的可靠度函数

为

的导函数，t_s为使取t＝0时剩余寿命期望值概率密度函数

最大值对应的时刻，

为可靠度故障阈值。Among them, the component reliability R _k (t) represents the probability that the normal operation time of the component is greater than t, and P represents the probability of the event,

Indicates the component failure rate expectation function considering the influence of random typhoon impact, n _z indicates the total number of typhoon impact scenarios, Z is the typhoon impact scenario, and a single typhoon impact scenario is the typhoon maximum wind speed radius, moving speed, central air pressure difference, and maximum wind speed radius The set of the wind speed and the moving direction of the typhoon, z is the scene variable of typhoon impact, T is the time when the component fails, and the remaining life of the component at time t is

is the expected value of remaining life, x is the time variable of remaining life,

Remaining life expectancy

distribution function of

is the component reliability expectation function under the impact of historical typhoons,

Represents the reliability function considering random typhoon impact scenarios

for

The derivative function of , t _s is the probability density function of the expected value of the remaining life when t=0

The moment corresponding to the maximum value,

is the reliability failure threshold.

Through the statistics of the life frequency of a certain component of the wind turbine in the offshore wind farm, the maximum likelihood estimation method is used to estimate the parameters of the natural degradation failure rate model, and on the basis of the natural degradation, the component failures of n _w units under the impact of the typhoon are extracted Rate increment model for parameter estimation:

和y_j(t_gi)分别为机组j在监测点t_gi时刻的剩余寿命估计值和剩余寿命实际值，n_g为监测点的个数，n_w为抽取的实验风机样本数量，

为n_w机组的平均均方根误差。Among them, β is the shape parameter, η is the scale parameter, f _0k (t) is the probability density distribution function of Weibull distribution, L(β,η) is the likelihood function, t _z is the sample value of component life, lnL(β, η) is the logarithmic likelihood function, n _c is the sample size, U _j is the root mean square error of unit j,

and y _j (t _gi ) are the estimated value of remaining life and the actual value of remaining life of unit j at the time of monitoring point t _gi respectively, n _g is the number of monitoring points, n _w is the number of experimental fan samples drawn,

is the average root mean square error of n _w units.

Compared with the prior art, the present invention has the following beneficial effects:

(1) The present invention takes into account the impact of typhoon random impact on the prediction results, and establishes a time-varying failure rate model and an incremental component failure rate model for components of offshore wind turbines that are dependent on degradation and impact, which can accurately describe the actual situation of the unit under the harsh operating environment at sea. The law of degenerate evolution.

(2) The present invention establishes a component degradation and impact correlation model, accurately describes the correlation between random impact and degradation state, and uses the actual monitoring data of the component to correct the remaining life prediction model, so that the predicted degradation process of the component is more in line with the actual degradation The process improves the accuracy of life prediction.

Description of drawings

Fig. 1 is method flowchart of the present invention;

Figure 2 is a schematic diagram of the failure rate evolution of offshore wind turbine components under the impact of a typhoon;

Figure 3 is a schematic diagram of the Batts typhoon wind field model and its impact on offshore wind farms;

Fig. 4 is a schematic diagram of component status division;

Fig. 5 is the diagram of the influence of working condition division on the results of principal component clustering analysis;

Figure 6 is a schematic diagram of the remaining life of the unit components under the random typhoon impact scenario;

Fig. 7 is the flow chart that particle swarm optimization algorithm carries out parameter estimation;

Figure 8 shows the generator current change and state evaluation results;

Figure 9 is the variation of the quantized expected value of the random state of the generator with the running time;

Figure 10 is the probability density distribution of the remaining life expectancy of the generator;

Figure 11 shows the reliability change process of the three life prediction methods;

Figure 12 is the reliability change process corresponding to the correction process of method 3;

Figure 13 is a comparison of the remaining life prediction results of the three methods;

Figure 14 is the probability density curve of the remaining life expectancy under different typhoon arrival rates λ;

Figure 15 shows the impact of the number of random typhoon impacts on the forecast error.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is carried out on the premise of the technical solution of the present invention, and detailed implementation and specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.

A method for predicting the remaining life of offshore wind turbines considering the impact of typhoons, as shown in Figure 1, includes the following steps:

1) Based on the impact mechanism of typhoon on unit component degradation, the influence of direct impact and indirect impact on component failure rate is analyzed, and an incremental model of component failure rate under typhoon impact is constructed.

1-1) Obtain typhoon data in typhoon-prone areas;

1-2) Build the Batts typhoon wind field model; the typhoon wind field model and the schematic diagram of the impact on the offshore wind farm are shown in Figure 3;

1-3) Assuming that the number of typhoon shocks encountered by offshore wind turbines during operation obeys the Poission process with parameter λ, and the typhoon wind field model remains unchanged before and after the typhoon passes through the wind farm, the Monte Carlo sampling method is used to obtain each typical typhoon Shock probability, establish a typhoon random shock model;

The typhoon random impact model is:

Among them, P(N(t)=n) represents the probability of n times of typhoon impact within (0, t), λ is the arrival rate of typhoon, f _G (t _si ) is the required time for the first arrival of the i-th impact Probability density at time t _si , Γ(i) is the gamma function value; V _Rmax is the gradient wind speed at the maximum wind radius, K is an empirical parameter, the value range is 6.93-6.97, Δp is the typhoon center pressure difference, R _max is the radius of the maximum wind speed, f is a scientific parameter; v is the instantaneous wind speed of the typhoon, V _s is the moving speed of the typhoon; r is the distance from any position of the typhoon to the center of the typhoon; x is an empirical parameter, and its value ranges from 0.5 to 0.7.

1-4) Based on the random impact model of typhoon, analyze the influence of direct impact and indirect impact on component failure rate, and construct the incremental model of component failure rate under typhoon impact:

为部件退化与冲击相关性函数，表示部件k在运行状态为

时受到冲击对故障率增量的影响，t_si和t_ei分别表示台风对机组冲击影响的首达时刻和解除时刻，α、ε为部件的台风直接或间接冲击修正系数，v为台风瞬时风速。in,

is the component degradation and impact correlation function, which means that the component k is in the running state of

t _si and t _ei represent the first arrival time and release time of the impact of typhoon on the unit, respectively, α and ε are the typhoon direct or indirect impact correction coefficients of components, and v is the typhoon instantaneous wind speed .

Distinguishing the effects of direct shocks and indirect shocks, the component failure rate increment model under direct shocks is expressed as:

The incremental model of component failure rate under indirect shock is expressed as:

Among them, Δλ _sk1i is the failure rate of external components of the unit caused by the direct impact of typhoon, α _k1 and ε _k1 are the correction coefficients of direct typhoon impact respectively, w is the wind load perpendicular to the surface of the fan, ρ is the air density, and C _s is the shape coefficient , S _α is the projected area on the plane perpendicular to the wind direction; Δλ _sk2i represents the failure rate of the internal components of the unit caused by the indirect typhoon impact, α _k2 and ε _k2 are the correction coefficients of typhoon indirect impact respectively, p is the power of the fan, and C _p is the wind energy Conversion efficiency coefficient, λ _a is the tip speed ratio of the fan, β _a is the pitch angle of the fan blade, and r _a is the radius of the fan blade.

In this embodiment, an offshore wind farm in my country is taken as an example. The offshore wind farm includes 36 3MW wind turbines arranged in 4 rows and 9 columns. The distance between the wind turbines is 0.5 km in the north-south direction and 1 km in the east-west direction.

Based on the historical typhoon yearbook and the wind farm's statistics on historical typhoon process monitoring data, the estimated values of typhoon random impact model parameters are shown in Table 1.

Table 1 Estimated values of typhoon model parameters

2) Aiming at the problem of quantifying the correlation between component degradation process and impact, combined with the monitoring data of unit operating status, a correlation model between degradation and impact of component operating status is constructed to quantify the ability to withstand external impacts.

2-1) Obtain historical operation monitoring data of offshore wind turbines;

2-2) Extract the health state characteristics of offshore wind turbine components under different operating conditions through principal component cluster analysis, and divide different component states according to the difference between other states and the healthy state;

2-3) According to the state division results of components under different working conditions and the historical state evolution process, based on the Markov chain state transition process, a stochastic state model is established to describe the historical state transition process of components, and the degradation and impact correlation model is obtained.

在台风天气的随机扰动下，随着机组部件随机失效状态的加剧，部件对外部冲击愈发敏感，承受冲击的能力逐步下降。部件退化与冲击相关性函数

为单调递增函数。Using Markov Chain State Transition Process to Describe the Stochastic State Model of Components

Under the random disturbance of typhoon weather, with the aggravation of the random failure state of unit components, the components become more sensitive to external shocks, and the ability to withstand shocks gradually decreases. Component Degradation and Shock Correlation Function

is a monotonically increasing function.

The degradation-shock correlation model is expressed as:

R=X'X/ _nq

Y=XU=[Y ₁ ,Y ₂ ,…,Y _q ]

为部件k的运行状态，

为部件退化与冲击相关性函数，μ为状态量修正系数，风机的某一部件相关的SCADA数据标准化处理后的输入变量矩阵X包含q维初始参量，每维参量包含n_q个样本，X'表示X的转置，相关系数矩阵R的特征方程有q个特征值，λ₁≥λ₂≥…≥λ_q，U＝(U₁,U₂,…,U_q)，为特征值对应的特征向量，Y为主成分，按照主成分方差从大到小的顺序进行排序，Y_q为第q主成分，Y₁为第一主成分，具有最大方差，能够解释数据的大部分信息，λ_ji为工况j下第i个主成分对应的特征值，δ为累积方差贡献率，每类工况j的主成分分析结果按照累积方差贡献率大于δ选取ω_j个主成分，π(t_g)为监测点t_g时刻风机部件各状态的概率分布，A为状态转移矩阵，Δt为任意时间间隔，t₀为状态转移时间歩长，S_k为各状态量化值向量矩阵。in,

is the operating state of component k,

is the component degradation and impact correlation function, μ is the state quantity correction coefficient, the input variable matrix X after normalization processing of the SCADA data related to a certain component of the fan contains q-dimensional initial parameters, and each dimensional parameter contains n _q samples, X' Represents the transposition of X, the characteristic equation of the correlation coefficient matrix R has q eigenvalues, λ ₁ ≥λ ₂ ≥…≥λ _q , U=(U ₁ ,U ₂ ,…,U _q ), which corresponds to the eigenvalues Eigenvector, Y is the main component, sorted according to the order of the principal component variance from large to small, Y _q is the qth principal component, Y ₁ is the first principal component, has the largest variance, and can explain most of the information of the data, λ _ji is the eigenvalue corresponding to the i-th principal component under working condition j, and δ is the cumulative variance contribution rate. For the principal component analysis results of each type of working condition j, ω _j principal components are selected according to the cumulative variance contribution rate greater than δ, π(t _g ) is the probability distribution of each state of the fan components at the monitoring point _tg , A is the state transition matrix, Δt is an arbitrary time interval, _t0 is the state transition time step, and S _k is the vector matrix of quantized values of each state.

In this embodiment, taking a generator with a relatively high failure rate and serious failure consequences as an example, the states of the generator are divided into four states: healthy, slightly abnormal, abnormal, and faulty. As shown in Figure 4, the initial probability distribution is π(t _g = 0) = [1000].

Combined with the SCADA monitoring parameters related to the generator, according to the initial principal component cluster analysis results, it is divided into three types of operating conditions under three wind speeds: low, medium and high. The principal component cluster analysis is performed on the data of the three types of working conditions, and the principal components are selected according to the cumulative variance contribution rate greater than 90%. The clustering results of generator health characteristics under different working conditions after principal component analysis are shown in Figure 5.

3) Based on the time-varying randomness of the unit performance degradation caused by the random impact of the typhoon, combined with the component failure rate increment model and the degradation-impact correlation model under the typhoon impact, the time-varying failure rate of offshore wind turbine components considering the dependence of degradation and impact is constructed Model.

Assuming that the failure rate λ _k of the system component k is composed of the natural degradation failure rate λ _0k and the component cumulative failure rate increment Δλ _ski (t) caused by multiple typhoon impacts, the life of the unit component is a non-negative continuous random variable, The natural degradation process is described by Weibull distribution, combined with the component failure rate increment model under typhoon impact and the degradation and impact correlation model, the time-varying failure rate model of offshore wind turbine components considering the dependence of degradation and impact is obtained:

为部件退化与冲击相关性函数，表示部件k在运行状态为

时受到冲击对故障率增量的影响；t_si和t_ei分别表示台风对机组冲击影响的首达时刻和解除时刻；β为形状参数，η为尺度参数。Among them, i is the number of typhoon impacts; N(t) is the cumulative number of typhoon impacts up to time t; D _i (t) is the impact amplitude at time t during the ith typhoon period;

is the component degradation and impact correlation function, which means that the component k is in the running state of

t _si and t _ei represent the first arrival time and release time of the typhoon’s impact on the unit, respectively; β is a shape parameter, and η is a scale parameter.

The schematic diagram of the failure rate evolution of offshore wind turbine components under the impact of typhoon is shown in Fig. 2.

4) Based on the time-varying failure rate model of offshore wind turbine components, in the typhoon impact scenario, the remaining life prediction model of the unit components is constructed, and the particle swarm optimization algorithm is used to optimize the component failure rate increment model and the degradation and The parameters of the shock correlation model are used to obtain an optimized remaining life prediction model.

According to the distribution probability of historical typhoons, the Monte Carlo sampling method is used to randomly sample and establish N(t) random typhoon impact scenarios within (0, t). A collection of wind speeds at the wind speed radius and the typhoon's moving direction. According to the historical state monitoring data of the unit, a stochastic state model of the unit components based on the Markov transfer process is constructed to obtain the historical state transfer process of the components. Combined with the actual typhoon weather data before any operating time t _g and the monitoring data of component operating status, the typhoon impact scene and component state transition process are corrected, so as to obtain the remaining life prediction model of unit components considering the dependence of degradation and impact:

表示考虑随机台风冲击影响的部件故障率期望值函数，n_z表示台风冲击场景总次数，Z为台风冲击场景，z为台风冲击场景变量，T为部件发生故障的时刻，运行至t时刻部件剩余寿命为

为剩余寿命期望值，x为剩余寿命时间变量，

表示剩余寿命期望值

的分布函数，

为根据历史台风冲击下部件可靠度期望值函数，

表示考虑随机台风冲击场景下的可靠度函数

为

的导函数，t_s为使取t＝0时剩余寿命期望值概率密度函数

最大值对应的时刻，

为可靠度故障阈值。Among them, the component reliability R _k (t) represents the probability that the normal operation time of the component is greater than t, and P represents the probability of the event,

Represents the expected value function of component failure rate considering the impact of random typhoon, n _z indicates the total number of typhoon impact scenarios, Z is the typhoon impact scenario, z is the variable of the typhoon impact scenario, T is the time when the component fails, and the remaining life of the component runs to time t for

is the expected value of remaining life, x is the time variable of remaining life,

Remaining life expectancy

distribution function of

is the component reliability expectation function under the impact of historical typhoons,

Represents the reliability function considering random typhoon impact scenarios

for

The derivative function of , t _s is the probability density function of the expected value of the remaining life when t=0

The moment corresponding to the maximum value,

is the reliability failure threshold.

Through the statistics of the life frequency of a certain component of the wind turbine in the offshore wind farm, the maximum likelihood estimation method is used to estimate the parameters of the natural degradation failure rate model, and on the basis of the natural degradation, the component failures of n _w units under the impact of the typhoon are extracted Rate increment model for parameter estimation:

和y_j(t_gi)分别为机组j在监测点t_gi时刻的剩余寿命估计值和剩余寿命实际值，n_g为监测点的个数，n_w为抽取的实验风机样本数量，U为n_w机组的平均均方根误差。Among them, β is the shape parameter, η is the scale parameter, f _0k (t) is the probability density distribution function of Weibull distribution, L(β,η) is the likelihood function, t _z is the sample value of component life, lnL(β, η) is the logarithmic likelihood function, n _c is the sample size, U _j is the root mean square error of unit j,

and y _j (t _gi ) are the estimated value of remaining life and the actual value of remaining life of unit j at the time of monitoring point t _gi respectively, n _g is the number of monitoring points, n _w is the number of experimental fan samples drawn, and U is n The average root mean square error of _w unit.

The schematic diagram of the remaining life of the unit components under the random typhoon impact scenario is shown in Fig. 6. Fig. 7 is a flowchart of the particle swarm optimization algorithm. In this embodiment, the initial particle number is set to 40, the learning factor is also 1.494, and the maximum particle velocity is 0.8. Through particle swarm optimization algorithm parameter optimization, the parameter estimation results are shown in Table 2:

Table 2 Estimation results of remaining life prediction model parameters

5) Based on the optimized remaining life prediction model, combined with the typhoon impact monitoring data and unit real-time operating status monitoring data, the optimized remaining life prediction model is revised and the remaining life prediction is realized.

Fig. 8 shows the evaluation results of the generator status for a period of time before the unit stops due to abnormal generator current. Fig. 9 shows the variation of the quantized expected value of the random state of the generator with the running time at any monitoring point under different initial states according to the Markovian state transition process.

Figure 10 shows the distribution diagram of the probability density distribution function of the expected value of the remaining life of the generator considering the random impact of the typhoon.

Taking the remaining life prediction of a unit generator as an example, combined with historical monitoring data, the validity of the model is verified. The three methods are compared respectively, method 1: natural degradation prediction model without considering the impact of typhoon impact; method 2: random typhoon impact scenario, considering the prediction model of degeneration and typhoon impact interdependence; method 3: on the basis of method 2, combined with the actual typhoon Shock-corrected forecasting models. According to the interval of historical typhoons and the duration of each state, the model is corrected every 50 days.

Figure 11 shows that method 1 does not consider the impact of typhoons, but only considers the natural degradation process of unit components, and the prediction results are the most optimistic, which overestimates the reliability and remaining life of components. Compared with method 1, methods 2 and 3 consider the influence of typhoon on the degradation process of unit components, and the reliability evolution process is closer to reality, and the root mean square error (RMSE) of remaining life prediction is reduced by 17.4% and 25.1%, respectively. Compared with method 2, method 3 can further reduce the prediction error by combining the actual typhoon impact correction model, which is 7.7% lower than method 2.

Figure 12 shows that after the unit has experienced the impact of typhoon "Hai Kui" and typhoon "Bulavan" during its operation, it partially corrects the model in Method 3.

Figure 13 shows that with the increase of monitoring data, the prediction result of method 3 is closer to the actual life value, and the prediction accuracy is improved.

The influence of typhoon arrival rate λ on the remaining life prediction and the influence of random typhoon impact times on the prediction error are respectively analyzed. Figure 14 shows that with the increase of typhoon impact frequency, the peak value of the probability density curve moves to the upper left, and the shape of the curve becomes narrower, and the probability distribution is more concentrated, which indicates that the increase of typhoon impact frequency accelerates the reliability reduction of unit components, and the remaining Life prediction expectations are reduced, probability distributions are more focused, and uncertainty in remaining life predictions is reduced. Figure 15 shows that when the number of typhoon impacts is the same as the actual number of occurrences, the remaining life prediction error RMSE is the smallest, and the correction of monitoring data is more beneficial to the reduction of RMSE. When the number of typhoon impacts deviates from the actual number of typhoon occurrences, the remaining life prediction error caused by the deviation of typhoon impact times is reduced.

It can be seen from this embodiment that the method proposed by the present invention is effective and feasible, and can provide a reference for the prediction of the remaining life of offshore wind turbines considering the influence of typhoons.

The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make many modifications and changes according to the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experiments on the basis of the prior art shall be within the scope of protection defined in the claims.

Claims (10)

1. A method for predicting the remaining life of offshore wind turbines considering typhoon influences, characterized in that it comprises the following steps: Based on the impact mechanism of typhoon on unit component degradation, the impact of direct impact and indirect impact on component failure rate is analyzed, and an incremental model of component failure rate under typhoon impact is constructed; Aiming at the problem of quantifying the correlation between component degradation process and impact, combined with the monitoring data of unit operating status, construct a degradation and impact correlation model of component operating status to quantify the ability to withstand external impacts; Based on the time-varying randomness of performance degradation caused by typhoon random impact, combined with the incremental model of component failure rate and the correlation model between degradation and impact under typhoon impact, a time-varying failure rate model of offshore wind turbine components is constructed considering the dependence of degradation and impact; Based on the time-varying failure rate model of offshore wind turbine components, in the typhoon impact scenario, the remaining life prediction model of the unit components is constructed, and the particle swarm optimization algorithm is used to optimize the component failure rate increment model and the correlation between degradation and impact under the typhoon impact with the minimum prediction error The parameters of the performance model are used to obtain an optimized remaining life prediction model; Based on the optimized remaining life prediction model, combined with the typhoon impact monitoring data and unit real-time operating status monitoring data, the optimized remaining life prediction model is corrected and the remaining life prediction is realized. 2. a kind of offshore wind turbine remaining life prediction method considering typhoon influence according to claim 1, is characterized in that, the construction of component failure rate increment model under the impact of described typhoon comprises the following steps: Obtain typhoon data in typhoon-prone areas; Construct the Batts typhoon wind field model; Assuming that the number of typhoon impacts encountered by offshore wind turbines during operation obeys the Possion process with parameter λ, and the typhoon wind field model remains unchanged before and after the typhoon passes through the wind farm, the Monte Carlo sampling method is used to obtain the impact probability of each typical typhoon, and the establishment Typhoon random impact model; Based on the random impact model of typhoon, the influence of direct impact and indirect impact on component failure rate is analyzed, and the incremental model of component failure rate under typhoon impact is constructed. 3. A method for predicting the remaining life of offshore wind turbines considering the influence of typhoons according to claim 2, wherein the typhoon random impact model is:

Among them, P(N(t)=n) represents the probability of n times of typhoon impact within (0, t), λ is the arrival rate of typhoon, f _G (t _si ) is the required time for the first arrival of the i-th impact Probability density at time t _si , Γ(i) is the gamma function value; V _Rmax is the gradient wind speed at the maximum wind radius, K is an empirical parameter, the value range is 6.93-6.97, Δp is the typhoon center pressure difference, R _max is the radius of the maximum wind speed, f is a scientific parameter; v is the instantaneous wind speed of the typhoon, V _s is the moving speed of the typhoon; r is the distance from any position of the typhoon to the center of the typhoon; x is an empirical parameter, and its value ranges from 0.5 to 0.7. 4. according to claim 1 or 2, a kind of offshore wind turbine remaining life prediction method considering typhoon influence, is characterized in that, the component failure rate incremental model under the impact of typhoon is:

in,

is the component degradation and impact correlation function, which means that the component k is in the running state of

t _si and t _ei represent the first arrival time and release time of the impact of typhoon on the unit, respectively, α and ε are the typhoon direct or indirect impact correction coefficients of components, and v is the typhoon instantaneous wind speed . 5. A method for predicting the remaining life of offshore wind turbines considering the influence of typhoon according to claim 4, characterized in that, the influence of direct impact and indirect impact is distinguished, and the component failure rate incremental model under direct impact is expressed as:

The incremental model of component failure rate under indirect shock is expressed as:

Among them, Δλ _sk1i is the failure rate of external components of the unit caused by the direct impact of typhoon, α _k1 and ε _k1 are the correction coefficients of direct typhoon impact respectively, w is the wind load perpendicular to the surface of the fan, ρ is the air density, and C _s is the shape coefficient , S _α is the projected area on the plane perpendicular to the wind direction; Δλ _sk2i represents the failure rate of the internal components of the unit caused by the indirect typhoon impact, α _k2 and ε _k2 are the correction coefficients of typhoon indirect impact respectively, p is the power of the fan, and C _p is the wind energy Conversion efficiency coefficient, λ _a is the tip speed ratio of the fan, β _a is the pitch angle of the fan blade, and r _a is the radius of the fan blade. 6. A method for predicting the remaining life of offshore wind turbines considering the influence of typhoons according to claim 1, wherein the construction of the degradation and impact correlation model based on the operating state of the components comprises the following steps: Obtain historical operation monitoring data of offshore wind turbines; Extract the health state characteristics of offshore wind turbine components under different operating conditions through principal component cluster analysis, and divide different component states according to the difference between other states and health states; According to the component state division results and historical state evolution process under different working conditions, based on the Markov chain state transition process, a stochastic state model is established to describe the historical state transition process of components, and the degradation and impact correlation model is obtained. 7. A method for predicting the remaining life of an offshore wind turbine considering the impact of a typhoon according to claim 6, wherein the degradation and impact correlation model is expressed as:

R=X'X/ _nq Y=XU=[Y ₁ ,Y ₂ ,…,Y _q ]

in,

is the operating state of component k,

is the component degradation and impact correlation function, μ is the state quantity correction coefficient, the input variable matrix X after normalization processing of the SCADA data related to a certain component of the fan contains q-dimensional initial parameters, and each dimensional parameter contains n _q samples, X' Represents the transposition of X, the characteristic equation of the correlation coefficient matrix R has q eigenvalues, λ ₁ ≥λ ₂ ≥…≥λ _q , U=(U ₁ ,U ₂ ,…,U _q ), which corresponds to the eigenvalues Eigenvector, Y is the main component, sorted according to the order of the principal component variance from large to small, Y _q is the qth principal component, λ _ji is the eigenvalue corresponding to the i-th principal component under working condition j, and δ is the cumulative variance Contribution rate, the principal component analysis results of each type of working condition j select _ωj principal components according to the cumulative variance contribution rate greater than δ, π(t _g ) is the probability distribution of each state of the fan component at the monitoring point t _g , and A is the state transition matrix, Δt is an arbitrary time interval, _t0 is the time step of state transition, and S _k is the vector matrix of each state quantization value. 8. A kind of offshore wind turbine remaining life prediction method considering typhoon influence according to claim 1, is characterized in that, suppose the failure rate λ _k of system component k is by the failure rate λ _0k of natural degradation and suffers multiple typhoon impacts The component cumulative failure rate increment Δλ _ski (t) is formed. The life of the unit component is a non-negative continuous random variable, and its natural degradation process is described by Weibull distribution. Combining the component failure rate increment model under typhoon impact and the degradation and The shock correlation model is used to obtain the time-varying failure rate model of offshore wind turbine components considering the dependence of degradation and shock:

Among them, i is the number of typhoon impacts; N(t) is the cumulative number of typhoon impacts up to time t; D _i (t) is the impact amplitude at time t during the ith typhoon period;

is the component degradation and impact correlation function, which means that the component k is in the running state of

t _si and t _ei represent the first arrival time and release time of the typhoon’s impact on the unit, respectively; β is a shape parameter, and η is a scale parameter. 9. A method for predicting the remaining life of an offshore wind turbine considering the impact of a typhoon according to claim 8, wherein the model for predicting the remaining life of a unit component that considers the dependence of degradation and impact is:

Among them, the component reliability R _k (t) represents the probability that the normal operation time of the component is greater than t, and P represents the probability of the event,

Indicates the component failure rate expectation function considering the influence of random typhoon impact, n _z indicates the total number of typhoon impact scenarios, Z is the typhoon impact scenario, and a single typhoon impact scenario is the typhoon maximum wind speed radius, moving speed, central air pressure difference, and maximum wind speed radius The set of the wind speed and the moving direction of the typhoon, z is the typhoon impact scene variable, T is the time when the component fails, and the remaining life of the component is T _t = Tt,

is the expected value of remaining life, x is the time variable of remaining life,

Remaining life expectancy

distribution function of

is the component reliability expectation function under the impact of historical typhoons,

Represents the reliability function considering random typhoon impact scenarios

for

The derivative function of , t _s is the probability density function of the expected value of the remaining life when t=0

The moment corresponding to the maximum value,

is the reliability failure threshold. 10. A method for predicting the remaining life of an offshore wind turbine considering the impact of a typhoon according to claim 1, characterized in that, by counting the life frequency of a certain part of the wind turbine in the offshore wind farm, using the maximum likelihood estimation method, Estimate the parameters of the natural degradation failure rate model, and on the basis of natural degradation, select n _w units to estimate the parameters of the component failure rate incremental model under typhoon impact:

Among them, β is the shape parameter, η is the scale parameter, f _0k (t) is the probability density distribution function of Weibull distribution, L(β,η) is the likelihood function, t _z is the sample value of component life, lnL(β, η) is the logarithmic likelihood function, n _c is the sample size, U _j is the root mean square error of unit j,

and y _j (t _gi ) are the estimated value of remaining life and the actual value of remaining life of unit j at the time of monitoring point t _gi respectively, n _g is the number of monitoring points, n _w is the number of experimental fan samples drawn,

is the average root mean square error of n _w units.

Images