CN106097146A - A kind of meter and the Wind turbines short term reliability Forecasting Methodology of running status - Google Patents

Abstract

The present invention relates to a kind of meter and the Wind turbines short term reliability Forecasting Methodology of running status, comprise the following steps: obtain Wind turbines state parameter by status monitoring with data collecting system, status monitoring parameter is divided into two classes；For device temperature parameter, set up state parameter forecast model based on reverse transmittance nerve network, calculate protection act probability based on prediction residual distribution character, for remaining parameter, calculate protection act probability according to meter and threshold crossing time；By wind energy turbine set operation maintenance data and SCADA data, stoppage in transit number of times and the moment wind speed of stopping transport of Wind turbines are added up, set up the Wind turbines statistics outage model that wind speed is interdependent；Get over limit information in conjunction with stoppage in transit statistical information and state parameter, calculate Wind turbines short-term stoppage in transit probability.This method can evaluate Wind turbines stoppage in transit risk in a short time exactly, increases substantially the accuracy of Wind turbines short-term outage model, and the short term reliability for whole wind energy turbine set is assessed and safety and economic operation provides Technical Reference.

Description

A kind of meter and the Wind turbines short term reliability Forecasting Methodology of running status

Technical field

The invention belongs to new forms of energy power equipment safety assessment technology field, relate to a kind of meter and the wind turbine of running status Group short term reliability Forecasting Methodology.

Background technology

Wind turbines is the key equipment of wind energy turbine set, and its reliability has closely connection with the safe and economical operation of wind energy turbine set System.Owing to long-term work is in severe natural environment, the outage rate of Wind turbines is far above tradition power transmission and transforming equipment.Effectively Prediction Wind turbines reliability within the following short time, exerts oneself prediction to wind energy turbine set and control has important value.

Utilizing statistical method to obtain equipment dependability parameter is the primary hand of current Wind turbines reliability assessment and prediction Section.The reliability data of Wind turbines is added up by external existing multiple mechanisms, utilizes statistical data to establish based on can By property block-scheme method, failure mode and effect analysis method, FTA, appraisal procedure based on markoff process, pattra leaves The Wind turbines long-term reliability model of this network technique and Monte Carlo method etc..But this class model is difficult to apply and Wind turbines Short term reliability prediction, main cause includes: 1) reliability of some mechanical and electric information assembly depends not only on product matter Amount and maintenance condition, also have substantial connection, existing statistical data not to consider these factors with natural environment and operating condition； 2) due to natural environment and operating condition real-time change, static constant reliability statistics index is difficult to reflect that Wind turbines is current Stoppage in transit risk.

Summary of the invention

In view of this, it is an object of the invention to provide the Wind turbines short term reliability prediction of a kind of meter and running status Method, the Wind turbines state parameter that the method is obtained with data acquisition (SCADA) system by status monitoring, according to natural ring Status monitoring parameter is divided into two classes by the dependency in border；For device temperature parameter, set up based on reverse transmittance nerve network (BPNN) state parameter forecast model, proposes protection act method for calculating probability based on prediction residual distribution character；For Remaining parameter, proposes meter and the protection act method for calculating probability of threshold crossing time；By wind energy turbine set operation maintenance data and SCADA Stoppage in transit number of times and the moment wind speed of stopping transport of Wind turbines are added up by data, and the Wind turbines statistics proposing wind speed interdependent is stopped transport Model.Finally, get over limit information in conjunction with stoppage in transit statistical information and state parameter, propose the calculating side of Wind turbines short-term stoppage in transit probability Method, improves the short-term stoppage in transit pre-alerting ability of Wind turbines.

For reaching above-mentioned purpose, the present invention provides following technical scheme:

A kind of meter and the Wind turbines short term reliability Forecasting Methodology of running status, the method comprises the following steps:

S1: obtain Wind turbines state parameter with data acquisition (SCADA) system by status monitoring, according to natural environment Dependency status monitoring parameter is divided into two classes: device temperature parameter and remaining state parameter；

S2: for device temperature parameter, sets up state parameter forecast model based on reverse transmittance nerve network (BPNN), Protection act probability is calculated based on prediction residual distribution character；For remaining parameter, calculate protection according to meter and threshold crossing time dynamic Make probability；

S3: by wind energy turbine set operation maintenance data and SCADA data to the stoppage in transit number of times of Wind turbines and moment wind of stopping transport Speed is added up, and sets up the Wind turbines statistics outage model that wind speed is interdependent；

S4: combining stoppage in transit statistical information and state parameter gets over limit information, calculate Wind turbines short-term stoppage in transit probability, carrying out can By property prediction.

Further, in the method, unit short-term stoppage in transit probability P is obtained based on state parameter out-of-limit protection act probability_z For:

$P_z=1-\underset{i=1}{\overset{N}{\Π}}(1-P_{zi})$

In formula, N is the sum of state parameter, P_ziOut-of-limit protection act probability for state parameter i；

The mode of operation that input is Wind turbines of statistics outage model and wind speed probabilistic forecasting information, when state parameter is got over When limit protection act probability is less, stoppage in transit probability based on statistical data can reflect to a certain extent owing to other reasons causes The probability of compressor emergency shutdown, and when parameter out-of-limit protection act probability is bigger, the reference price of stoppage in transit probability based on statistical data Being worth less, therefore the computational methods of unit short-term stoppage in transit probability are:

P=max (P_t,P_z)

Wherein, P_tFor unit short-term stoppage in transit probability based on statistical data, P_zStop transport for the short-term out-of-limit based on state parameter Probability.

Further, in step s 2, for the two class state parameters out-of-limit protection act probability of calculating:

Described device temperature parameter out-of-limit protection act method for calculating probability specifically includes: 1) set up wind speed short-term forecast mould Type, it was predicted that the wind speed in following 15min；2) temperature parameter forecast model is set up, it was predicted that temperature parameter during each forecasting wind speed value； Due to wind speed too low time, the impact of output-power fluctuation is negligible by compressor emergency shutdown, therefore do not consider low wind speed and The shutdown situation that low temperature causes；3) residual distribution characteristic based on forecasting wind speed model and temperature prediction model, calculates temperature The out-of-limit probability of parameter, owing to predicted time yardstick is relatively big, out-of-limit protection act probability can be approximately considered out-of-limit equal to parameter generally Rate；

Remaining state parameter described, owing to such state parameter is affected more weak by environmental factors, the out-of-limit guarantor of this type of parameter Protecting action probability approximation uses threshold crossing time linearly to express, and is shown below:

$F_i=\left\{\begin{array}{cc}1& t<t_{\lim }\\ t/t_{\lim }& 0<t<t_{\lim }\\ 0& t\&le;0\end{array}\right.$

In formula, t is parameter threshold crossing time, t_limFor setting time；Owing to setting time is shorter, such parameter out-of-limit generally Rate evaluation time yardstick is less, generally within 1min.

Further, in the out-of-limit protection act probability calculation of device temperature parameter:

Described 1) wind speed Short-term Forecasting Model is set up, it was predicted that the wind speed in following 15min includes: use BP neutral net mould Type prediction of wind speed, obtains Parameters of Normal Distribution according to the error statistics data estimation of prediction；Out-of-limit for simplifying calculating temperature parameter Probability, carries out sliding-model control to forecast error, obtains the probability of each error amount according to Parameters of Normal Distribution；

Described 2) temperature parameter forecast model is set up, it was predicted that temperature parameter during each forecasting wind speed value includes: based on BP god The forecast model of each temperature parameter is set up through network model；In conjunction with the operation characteristic of Wind turbines, use wind speed, ambient temperature with And the temperature parameter in upper moment is as input parameter；Consider the seasonality of weather conditions, the individual variation of equipment and fault feelings Condition, choosing state parameter data when unit is properly functioning under each season is that the parametric prediction model of each unit is entered by training sample Row training modeling；

Described 3) residual distribution characteristic based on forecasting wind speed model and temperature prediction model, calculates the out-of-limit of temperature parameter Probability includes: when Wind turbines is properly functioning, and the prediction residual symbol of its temperature parameter is the normal distribution of zero from average, and in short-term The change of interior (in 15min) ambient temperature is inconspicuous, and therefore the out-of-limit probability of the short-term of temperature may largely be determined by wind speed The temperature parameter value in predictive value and a upper moment；Normal distribution characteristic according to temperature prediction residual error, can get in prediction The out-of-limit probability of temperature parameter under wind speed:

O_i(v_j)=P (T_e＞ T_lim-T_vj)=1-F_N(T_lim-T_vj)

In formula, T_vjFor prediction of wind speed v_jThe predictive value of lower temperature parameter i, T_limFor the higher limit of temperature parameter, F_N() is The distribution function of temperature prediction error；

When prediction residual falls outside normal distribution 99% confidence interval, it is calculated as once predicting inefficacy；If at one hour The most at least occur that 3 predictions were lost efficacy, then it is assumed that temperature parameter is abnormal；Data conduct during owing to using Wind turbines properly functioning The state parameter that the forecast model that sample training obtains is difficult to being in abnormal condition carries out Accurate Prediction, therefore this method root Predict the outcome according to the intensity of anomaly correction of parameter, the average of former residual error normal distribution is set to the forecast error at upper a moment, and then Obtain unit when being in abnormal conditions, it was predicted that the out-of-limit probability of the temperature parameter under wind speed:

O_i(v_j)=1-F_N(T_lim-T_vj-ε_t-1)

In formula, ε_t-1Forecast error for upper a moment；

The out-of-limit probabilistic forecasting value of temperature parameter is as follows:

$F_i=\underset{j=1}{\overset{N}{\&Sigma;}}P\left(v_j\right)O_i\left(v_j\right)$

In formula, N is that the centrifugal pump of forecasting wind speed error is counted, P (v_j) represent that wind speed is v in a short time_jProbability；Prediction wind Speed value v_jWith P (v_j) can be obtained by forecasting wind speed result and forecast error.

Further, in step s3, by wind energy turbine set SCADA data and the stoppage in transit of operation maintenance data statistics Wind turbines Number of times and the wind speed in moment of stopping transport, carry out subregion to wind speed with the interval of 1m/s, and under each wind speed interval, Wind turbines outage rate can It is calculated by following formula:

$\mathrm{\&lambda;}\left(v_i\right)=\frac{N_{vi}}{T_{vi}}$

In formula, N_viFor wind speed v_iIn the case of the number of times of always stopping transport of all Wind turbines, T in wind energy turbine set_viFor all units Cumulative operation time；

The Wind turbines short-term stoppage in transit probability of meter and wind speed is shown below:

P_t(t, v)=1-e^λ(v)t。

The beneficial effects of the present invention is: Wind turbines short term reliability Forecasting Methodology of the present invention is easily programmed reality Existing, it is possible to the fault stoppage in transit accident evaluating Wind turbines stoppage in transit risk in a short time, the most only Wind turbines exactly carries For accurate early warning information, increase substantially the accuracy of Wind turbines short-term outage model, be also that the short-term of whole wind energy turbine set can Technical Reference is provided by property assessment and safety and economic operation.

Accompanying drawing explanation

In order to make the purpose of the present invention, technical scheme and beneficial effect clearer, the present invention provides drawings described below to carry out Illustrate:

The Wind turbines short term reliability estimation flow that Fig. 1 provides for the present invention；

Fig. 2 is the out-of-limit method for calculating probability of device temperature parameter；

Fig. 3 is the outage rate under each wind speed；

Fig. 4 is the unit outage probability in embodiment.

Detailed description of the invention

Below in conjunction with accompanying drawing, the preferred embodiments of the present invention are described in detail.

Fig. 1 is Wind turbines short term reliability estimation flow figure.Wind turbines state parameter abnormal information, unit Working mould Formula and the input parameter that wind speed probabilistic forecasting information is Wind turbines short term reliability assessment models.Owing to state parameter occurs more Limit situation may directly result in compressor emergency shutdown, it is also possible to provides alarm signal then to be determined whether to shut down by staff, State parameter generation is out-of-limit will result directly in unit outage.Obviously any state parameter is out-of-limit, and unit all will be caused to stop Fortune, therefore, obtains unit short-term stoppage in transit probability P based on state parameter out-of-limit protection act model_zFor:

$P_z=1-\underset{i=1}{\overset{N}{\&Pi;}}(1-P_{zi})$

In formula, N is the sum of state parameter, P_ziOut-of-limit protection act probability for state parameter i.

The mode of operation that input is Wind turbines of statistics outage model and wind speed probabilistic forecasting information.When state parameter is got over When limit protection act probability is less, stoppage in transit probability based on statistical data can reflect to a certain extent owing to other reasons causes The probability of compressor emergency shutdown, and when parameter out-of-limit protection act probability is bigger, the reference price of stoppage in transit probability based on statistical data It is worth less.Therefore the computational methods of unit short-term stoppage in transit probability are:

P=max (P_t,P_z)

Wherein, P_tFor unit short-term stoppage in transit probability based on statistical data, P_zStop transport for the short-term out-of-limit based on state parameter Probability.

1, state parameter out-of-limit protection act probability

Wind turbines SCADA state parameter is as shown in table 1, can be according to the dependency of each state parameter and natural environment by this A little parameters are divided into two classes, and first kind status monitoring parameter is notable by wind speed and ambient temperature effect, predominantly device temperature parameter； Equations of The Second Kind status monitoring parameter is more weak with the dependency relation of natural environment.Next will be respectively directed to this two classes state parameter calculate Out-of-limit protection act probability.

Table 1 status monitoring parameter

1) first kind parameter out-of-limit protection act method for calculating probability:

The computational methods of first kind state parameter out-of-limit protection act probability are as in figure 2 it is shown, calculating process can be divided into three steps: (1) wind speed in the following 15min of wind speed short-term forecast prediction is set up.(2) temperature parameter forecast model is set up, it was predicted that each wind speed is pre- Temperature parameter during measured value.Due to wind speed too low time, the impact of output-power fluctuation is negligible, therefore by compressor emergency shutdown Do not consider low wind speed and shutdown situation that low temperature causes.(3) based on forecasting wind speed model and the residual error of temperature prediction model Distribution character, calculates the out-of-limit probability of temperature parameter, and owing to predicted time yardstick is relatively big, out-of-limit protection act probability can approximate to be recognized For equal to the out-of-limit probability of parameter.

Short-term wind speed probabilistic forecasting:

Use BP Neural Network model predictive wind speed, obtain normal distribution ginseng according to the error statistics data estimation of prediction Number.Calculate the out-of-limit probability of temperature parameter for simplifying, forecast error is carried out sliding-model control, obtains respectively according to Parameters of Normal Distribution The probability of error amount.Table 2 show the forecast error centrifugal pump of Single-step Prediction in 15min and corresponding probability.

The table 2 forecasting wind speed probability of error

Error/m/s	Probability
		0	0.234
±0.5	0.197
		±1	0.118
±1.5	0.05
		±2	0.018

Temperature parameter probabilistic forecasting:

Forecast model based on each temperature parameter of BP Establishment of Neural Model.In conjunction with the operation characteristic of Wind turbines, adopt With the temperature parameter in wind speed, ambient temperature and a upper moment as input parameter.Consider the seasonality of weather conditions, equipment Individual variation and failure condition, choosing state parameter data when unit is properly functioning under each season is that training sample is to each unit Parametric prediction model be trained modeling.

The out-of-limit probability of temperature parameter:

When Wind turbines is properly functioning, the prediction residual of its temperature parameter generally accord with from average be the normal distribution of zero.And In short time, the change of (in 15min) ambient temperature is inconspicuous, and therefore the out-of-limit probability of the short-term of temperature may largely be determined by The predictive value of wind speed and the temperature parameter value in a upper moment.Normal distribution characteristic according to temperature prediction residual error, can get in advance The out-of-limit probability of temperature parameter under the wind speed surveyed:

O_i(v_j)=P (T_e＞ T_lim-T_vj)=1-F_N(T_lim-T_vj)

In formula, T_vjFor prediction of wind speed v_jThe predictive value of lower temperature parameter i, T_limFor the higher limit of temperature parameter, F_N() is The distribution function of temperature prediction error.

When prediction residual falls outside normal distribution 99% confidence interval, it is calculated as once predicting inefficacy.If at one hour The most at least occur that 3 predictions were lost efficacy, then it is assumed that temperature parameter is abnormal.Data conduct during owing to using Wind turbines properly functioning The state parameter that the forecast model that sample training obtains is difficult to being in abnormal condition carries out Accurate Prediction, therefore this patent root Predict the outcome according to the intensity of anomaly correction of parameter, the average of former residual error normal distribution is set to the forecast error at upper a moment, and then Obtain unit when being in abnormal conditions, it was predicted that the out-of-limit probability of the temperature parameter under wind speed:

O_i(v_j)=1-F_N(T_lim-T_vj-ε_t-1)

In formula, ε_t-1Forecast error for upper a moment.

The out-of-limit probabilistic forecasting value of temperature parameter is as follows:

$F_i=\underset{j=1}{\overset{N}{\&Sigma;}}P\left(v_j\right)O_i\left(v_j\right)$

In formula, N is that the centrifugal pump of forecasting wind speed error is counted, P (v_j) represent that wind speed is v in a short time_jProbability.Prediction wind Speed value v_jWith P (v_j) can be obtained by forecasting wind speed result and forecast error.

2) Equations of The Second Kind parameter out-of-limit protection act method for calculating probability:

Owing to Equations of The Second Kind state parameter is affected more weak by environmental factors, this type of parameter out-of-limit protection act probability can approximate Use threshold crossing time linearly to express, be shown below:

$F_i=\left\{\begin{array}{cc}1& t<t_{\lim }\\ t/t_{\lim }& 0<t<t_{\lim }\\ 0& t\&le;0\end{array}\right.$

In formula, t is parameter threshold crossing time, t_limFor setting time.Owing to setting time is shorter, Equations of The Second Kind parameter out-of-limit Probability assessment time scale is less, generally within 1min.

2, statistics outage model

Fig. 3 is the outage rate under each wind speed, by wind energy turbine set SCADA data and operation maintenance data statistics Wind turbines Stoppage in transit number of times and the wind speed in moment of stopping transport, carry out subregion to wind speed with the interval of 1m/s, and under each wind speed interval, Wind turbines is stopped transport Rate can be calculated by following formula::

$\mathrm{\&lambda;}\left(v_i\right)=\frac{N_{vi}}{T_{vi}}$

In formula, N_viFor wind speed v_iIn the case of the number of times of always stopping transport of all Wind turbines, T in wind energy turbine set_viFor all units Cumulative operation time.The Wind turbines short-term stoppage in transit probability of meter and wind speed is shown below:

P_t(t, v)=1-e^λ(v)t

Embodiment:

The present embodiment causes compressor emergency shutdown accident with one of certain wind energy turbine set domestic because first kind state parameter is out-of-limit As a example by Wind turbines, the short-term stoppage in transit probability of unit is studied, by reliable to meter and the short-term of operating states of the units Property Forecasting Methodology and the relative analysis of short term reliability Forecasting Methodology result based on statistical data, demonstrate the carried side of the present invention The effectiveness of method and accuracy.

This unit on June 17th, 2014 about 17:00 stop transport because of dynamo bearing B temperature over-range.Stopping transport Front in several hours, mean wind speed fluctuates near rated value 11.5m/s, and ambient temperature is in the state being stepped up.Machine Group dynamo bearing B temperature starts occur extremely in early June, and actual value is mostly higher than predictive value.Therefore, unit exposes for a long time In high temperature, high wind speed natural environment under, the lasting rising of bearing B temperature has ultimately resulted in the generation of compressor emergency shutdown accident.

The probabilistic forecasting result of table 3 wind speed

Out-of-limit probability under each wind speed of table 4

With the data of the moment 17:00 that stops transport, compressor emergency shutdown probability is carried out computational analysis.The mean wind speed prediction of this period Really definite value is 11.2m/s.According to the forecasting wind speed probability of error, the probabilistic forecasting value of wind speed can be obtained, as shown in table 3.Each wind speed Under temperature prediction determine that value and corresponding out-of-limit probability are as shown in table 4.Respectively by meter proposed by the invention and running status Stoppage in transit probabilistic model and be based only upon the outage model of statistical data and calculate the stoppage in transit probability in each moment of this unit, such as Fig. 4 Shown in, visible by relative analysis, unit outage model based on running status can the stoppage in transit in a short time of significant reaction unit Risk, although and the outage model based on the statistical data high wind speed period before stoppage in transit has shown that of a relatively high stoppage in transit is general Rate, but still far below this method.Therefore, the outage model that the present invention proposes can effectively reflect unit stoppage in transit wind in a short time Danger.

Finally illustrate, preferred embodiment above only in order to technical scheme to be described and unrestricted, although logical Cross above preferred embodiment the present invention to be described in detail, it is to be understood by those skilled in the art that can be In form and it is made various change, without departing from claims of the present invention limited range in details.

Claims (5)

1. a meter and the Wind turbines short term reliability Forecasting Methodology of running status, it is characterised in that: the method includes following Step:

S1: obtain Wind turbines state parameter with data acquisition (SCADA) system by status monitoring, according to the phase of natural environment Status monitoring parameter is divided into two classes by closing property: device temperature parameter and remaining state parameter；

S2: for device temperature parameter, sets up state parameter forecast model based on reverse transmittance nerve network (BPNN), based on Prediction residual distribution character calculates protection act probability；For remaining parameter, calculate protection act according to meter and threshold crossing time general Rate；

S3: stoppage in transit number of times and the moment wind speed of stopping transport of Wind turbines are entered by wind energy turbine set operation maintenance data and SCADA data Row statistics, sets up the Wind turbines statistics outage model that wind speed is interdependent；

S4: combine stoppage in transit statistical information and state parameter gets over limit information, calculate Wind turbines short-term stoppage in transit probability, carry out reliability Prediction.

A kind of meter the most according to claim 1 and the Wind turbines short term reliability Forecasting Methodology of running status, its feature It is: in the method, obtains unit short-term stoppage in transit probability P based on state parameter out-of-limit protection act probability_zFor:

$P_z=1-\underset{i=1}{\overset{N}{\&Pi;}}(1-P_{zi})$

In formula, N is the sum of state parameter, P_ziOut-of-limit protection act probability for state parameter i；

The mode of operation that input is Wind turbines of statistics outage model and wind speed probabilistic forecasting information, as the out-of-limit guarantor of state parameter Protect action probability less time, stoppage in transit probability based on statistical data can reflect to a certain extent owing to other reasons causes unit The probability shut down, and when parameter out-of-limit protection act probability is bigger, the reference value of stoppage in transit probability based on statistical data is relatively Little, therefore the computational methods of unit short-term stoppage in transit probability are:

P=max (P_t,P_z)

Wherein, P_tFor unit short-term stoppage in transit probability based on statistical data, P_zStop transport for the short-term out-of-limit based on state parameter general Rate.

A kind of meter the most according to claim 2 and the Wind turbines short term reliability Forecasting Methodology of running status, its feature It is: in step s 2, for the two class state parameters out-of-limit protection act probability of calculating:

Described device temperature parameter out-of-limit protection act method for calculating probability specifically includes: 1) set up wind speed Short-term Forecasting Model, Wind speed in the following 15min of prediction；2) temperature parameter forecast model is set up, it was predicted that temperature parameter during each forecasting wind speed value；By When wind speed is too low, the impact of output-power fluctuation is negligible by compressor emergency shutdown, does not therefore consider low wind speed and low The shutdown situation that temperature causes；3) residual distribution characteristic based on forecasting wind speed model and temperature prediction model, calculates temperature ginseng The out-of-limit probability of number, owing to predicted time yardstick is relatively big, out-of-limit protection act probability can be approximately considered equal to the out-of-limit probability of parameter；

Remaining state parameter described, owing to such state parameter is affected more weak by environmental factors, the out-of-limit protection of this type of parameter is moved Making probability approximation uses threshold crossing time linearly to express, and is shown below:

$F_i=\left\{\begin{array}{cc}1& t<t_{\lim }\\ t/t_{\lim }& 0<t<t_{\lim }\\ 0& t\&le;0\end{array}\right.$

In formula, t is parameter threshold crossing time, t_limFor setting time；Owing to setting time is shorter, the out-of-limit probability of such parameter is commented Estimate time scale less, generally within 1min.

A kind of meter the most according to claim 3 and the Wind turbines short term reliability Forecasting Methodology of running status, its feature It is: in the out-of-limit protection act probability calculation of device temperature parameter:

Described 1) wind speed Short-term Forecasting Model is set up, it was predicted that the wind speed in following 15min includes: use BP neural network model pre- Survey wind speed, obtain Parameters of Normal Distribution according to the error statistics data estimation of prediction；The out-of-limit probability of temperature parameter is calculated for simplifying, Forecast error is carried out sliding-model control, obtains the probability of each error amount according to Parameters of Normal Distribution；

Described 2) temperature parameter forecast model is set up, it was predicted that temperature parameter during each forecasting wind speed value includes: based on BP nerve net The forecast model of each temperature parameter set up by network model；In conjunction with the operation characteristic of Wind turbines, use wind speed, ambient temperature and on The temperature parameter in one moment is as input parameter；Consider the seasonality of weather conditions, the individual variation of equipment and failure condition, choosing Taking state parameter data when unit is properly functioning under each season is that the parametric prediction model of each unit is instructed by training sample Practice modeling；

Described 3) residual distribution characteristic based on forecasting wind speed model and temperature prediction model, calculates the out-of-limit probability of temperature parameter Including: when Wind turbines is properly functioning, the prediction residual symbol of its temperature parameter is the normal distribution of zero from average, and in the short time The change of (in 15min) ambient temperature is inconspicuous, and therefore the out-of-limit probability of the short-term of temperature may largely be determined by the pre-of wind speed Measured value and the temperature parameter value in a upper moment；Normal distribution characteristic according to temperature prediction residual error, can get the wind speed in prediction The out-of-limit probability of lower temperature parameter:

O_i(v_j)=P (T_e＞ T_lim-T_vj)=1-F_N(T_lim-T_vj)

In formula, T_vjFor prediction of wind speed v_jThe predictive value of lower temperature parameter i, T_limFor the higher limit of temperature parameter, F_N() is temperature The distribution function of forecast error；

When prediction residual falls outside normal distribution 99% confidence interval, it is calculated as once predicting inefficacy；If in one hour extremely Occur that 3 predictions were lost efficacy less, then it is assumed that temperature parameter is abnormal；Data during owing to using Wind turbines properly functioning are as sample The state parameter that the forecast model that training obtains is difficult to being in abnormal condition carries out Accurate Prediction, and therefore this method is according to ginseng The intensity of anomaly correction of number predicts the outcome, and the average of former residual error normal distribution is set to the forecast error at upper a moment, and then obtains When unit is in abnormal conditions, it was predicted that the out-of-limit probability of the temperature parameter under wind speed:

O_i(v_j)=1-F_N(T_lim-T_vj-ε_t-1)

In formula, ε_t-1Forecast error for upper a moment；

The out-of-limit probabilistic forecasting value of temperature parameter is as follows:

$F_i=\underset{j=1}{\overset{N}{\&Sigma;}}P\left(v_j\right)O_i\left(v_j\right)$

In formula, N is that the centrifugal pump of forecasting wind speed error is counted, P (v_j) represent that wind speed is v in a short time_jProbability；Prediction of wind speed value v_jWith P (v_j) can be obtained by forecasting wind speed result and forecast error.

A kind of meter the most according to claim 4 and the Wind turbines short term reliability Forecasting Methodology of running status, its feature It is: in step s3, by wind energy turbine set SCADA data and the stoppage in transit number of times of operation maintenance data statistics Wind turbines and stoppage in transit The wind speed in moment, carries out subregion to wind speed with the interval of 1m/s, and under each wind speed interval, Wind turbines outage rate can be calculated by following formula Obtain:

$\mathrm{\&lambda;}\left(v_i\right)=\frac{N_{vi}}{T_{vi}}$

In formula, N_viFor wind speed v_iIn the case of the number of times of always stopping transport of all Wind turbines, T in wind energy turbine set_viAdding up for all units Working time；

The Wind turbines short-term stoppage in transit probability of meter and wind speed is shown below:

P_t(t, v)=1-e^λ(v)t。