CN109583075B - Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction - Google Patents

Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction Download PDF

Info

Publication number
CN109583075B
CN109583075B CN201811413999.7A CN201811413999A CN109583075B CN 109583075 B CN109583075 B CN 109583075B CN 201811413999 A CN201811413999 A CN 201811413999A CN 109583075 B CN109583075 B CN 109583075B
Authority
CN
China
Prior art keywords
wind turbine
temperature
prediction
service quality
temperature parameter
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
CN201811413999.7A
Other languages
Chinese (zh)
Other versions
CN109583075A (en
Inventor
王宪
赵前程
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.)
Hunan University of Science and Technology
Original Assignee
Hunan 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 Hunan University of Science and Technology filed Critical Hunan University of Science and Technology
Priority to CN201811413999.7A priority Critical patent/CN109583075B/en
Publication of CN109583075A publication Critical patent/CN109583075A/en
Application granted granted Critical
Publication of CN109583075B publication Critical patent/CN109583075B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • 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
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • 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
    • G06Q10/06395—Quality analysis or management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • 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
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/80—Management or planning
    • Y02P90/82—Energy audits or management systems therefor

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • Educational Administration (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Geometry (AREA)
  • Evolutionary Computation (AREA)
  • Computer Hardware Design (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Wind Motors (AREA)

Abstract

The invention discloses a service quality evaluation method of a permanent magnet direct-drive wind turbine based on temperature parameter prediction, which comprises the steps of establishing a temperature parameter time sequence prediction model with the rotating speed of a hub of the wind turbine, the external wind speed, the ambient temperature, the output active power and the blade pitch angle as external input variables and the temperature of a main bearing, the temperature of an engine room and the temperature of the hub as autoregressive prediction variables; obtaining 5 training sample subsets through replaced uniform random sampling, and independently training 5 temperature parameter time sequence prediction models; integrating the 5 models by adopting a prediction result averaging mode, and establishing a temperature parameter integration prediction model; and calculating the service quality index of the temperature sign of the wind turbine according to the temperature parameter prediction error of the integrated prediction model and evaluating the service quality of the wind turbine in real time. The invention can provide key technical support for scientific maintenance and efficient operation of the permanent magnet direct-drive wind turbine in severe working environment.

Description

Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction
Technical Field
The invention relates to the technical field of state monitoring and evaluation of complex electromechanical systems, in particular to a method for evaluating service quality of a permanent magnet direct-drive wind turbine based on temperature parameter prediction.
Background
The wind power generation technology is the green energy utilization technology which is the most mature technology and has the most large-scale development condition at present, and has wide development prospect. The wind turbine, a key device of the wind power generation technology, is a complex electromechanical system, is usually located in remote suburban areas with inconvenient traffic and severe environment and coastal or offshore areas, and is difficult to detect the daily operation state and expensive in maintenance cost due to severe natural environments such as freezing, low air pressure, sand dust, lightning stroke and the like. The method has the advantages of developing a wind turbine state monitoring and evaluating technology, mastering the health state and the development trend of the wind turbine, and having important significance for optimizing a unit maintenance plan, saving operation and maintenance cost, avoiding malignant safety and production accidents and improving the competitiveness of the wind power industry.
The installation of a data acquisition and monitoring control (SCAD) system is a measure generally adopted by the existing wind power plant for monitoring the real-time operation state of a wind turbine, and the aim of improving the operation safety and the economical efficiency of the wind power plant is hopefully achieved. The system has a plurality of monitoring parameters including temperature, wind speed, vibration, voltage, current, yaw angle, motor control and the like, and a wind turbine manufacturer, a wind power plant owner and the like want to evaluate the health state of the wind turbine by analyzing the number acquired by the SCADA system because an additional data acquisition system is not required to be installed in the wind turbine.
In recent years, experts at home and abroad develop a lot of work in the field of wind turbine fault prediction and state evaluation by using SCAD system state data. Guo Peng and the like of the North China Power university adopt a nonlinear state estimation technology as a modeling method, and a tower vibration model is established on the basis of carrying out detailed analysis on the vibration characteristics and the influence factors of a wind turbine generator tower, so that a good foundation is laid for the follow-up wind turbine generator vibration state monitoring and early fault diagnosis to be carried out; by analyzing SCADA data, kusiak et al at Iowa university excavate information related to faults of key parts such as bearings and motors of the wind turbine generator, and construct a data model for revealing correlation characteristics between vibration of a main shaft and a tower and operating parameters of the wind turbine generator. A wind turbine variable pitch system parameter regression model based on SCAD system data is established in the Wangwei of Beijing university of transportation, and the like, and accordingly, an online identification method for the degradation state of a wind turbine variable pitch system based on data mining is provided. Compared with signals such as electricity and vibration, the temperature signal is a more stable state signal, and researchers try to explore a wind turbine short-term reliability evaluation method based on an SCAD system temperature signal prediction model, but the researches do not consider the influence of a long-time-scale historical state of the wind turbine on the temperature parameter change, and the precision of the temperature parameter prediction model and the accuracy of wind turbine state evaluation cannot be guaranteed.
The successful work pushes forward the development of the state monitoring and evaluation technical level of the wind turbine to a certain extent, but most of the wind turbine pays attention to state evaluation and fault early warning in a certain part or a short time period under the operation state of the wind turbine, and the technical requirements of paying attention to the actual evaluation of the whole state of the wind turbine in the full service period (including the operation state and the shutdown state) of the wind turbine cannot be met; the specific objects of the researches are generally double-fed wind turbines, and the number of permanent magnet direct-drive wind turbines which are widely applied is less; still basically staying in the theoretical research stage, there is a distance from practical application. The prior art can not provide solid technical support for scientific maintenance and efficient operation of a wind turbine, particularly a permanent magnet direct-drive wind turbine, in a severe working environment.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a method for evaluating the service quality of a permanent magnet direct-drive wind turbine based on temperature parameter prediction aiming at the defects of the prior art, realize the real-time evaluation of the dynamic health state of the whole working condition of the wind turbine in the service period, and provide key technical guarantee for scientific maintenance and efficient operation of the permanent magnet direct-drive wind turbine in a severe working environment.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows: the technical scheme adopted by the invention is as follows: a permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction comprises the following steps:
1) Establishing a temperature parameter time series prediction model: and establishing a temperature parameter time sequence prediction model which takes the rotating speed of a hub of the wind turbine, the external wind speed, the environment temperature, the output active power and the blade pitch angle as external input variables and takes the temperature of a main bearing, the temperature of an engine room and the temperature of the hub as autoregressive prediction variables.
2) Training and integrating multiple prediction models: reading state data of a wind turbine to be evaluated in the 1 st year of stable service from a data acquisition and monitoring control (SCADA) system of the wind turbine as a training sample set, obtaining 5 sample subsets through replaced uniform random sampling, wherein the number of samples of each sample subset is 3/10 of that of the sample set, and independently training 5 temperature parameter time sequence prediction models by applying the sample subsets respectively; and integrating the trained 5 models by adopting a prediction result averaging mode, and establishing a temperature parameter integrated prediction model.
3) The temperature parameter prediction and the service quality index of the wind turbine temperature sign are as follows: reading current state parameters of the wind turbine from the SCADA system, and predicting a current moment prediction variable on line by applying a temperature parameter integrated prediction model; and calculating the service quality index of the temperature sign of the wind turbine according to the prediction absolute error.
4) Evaluating the service quality of the wind turbine: and evaluating the service quality of the wind turbine in real time according to the temperature sign service quality index value of the wind turbine.
In step 1, the temperature parameter time series prediction model is configured as an autoregressive neural network prediction model with an external input nonlinear time series, and an external input vector X (t) of the model is as follows:
X(t)＝[x(t) 1 x(t) 2 x(t) 3 x(t) 4 x(t) 5 x(t) 6 x(t) 7 ]
wherein t is the current time. x (t) 1 、x(t) 2 、x(t) 3 、x(t) 4 、x(t) 5 、x(t) 6 、x(t) 7 The rotating speed of a hub of the wind turbine at the moment t, the external wind speed, the ambient temperature, the output active power, the pitch angle of the blade 1, the pitch angle of the blade 2 and the pitch angle of the blade 3 are respectively.
The autoregressive prediction vector Y (t) is:
Y(t)＝[y(t) 1 y(t) 2 y(t) 3 ]
wherein, y (t) 1 、y(t) 2 、y(t) 3 The main bearing temperature, the cabin temperature and the hub temperature of the wind turbine at the moment t are respectively.
The temperature parameter time series prediction model PM [. Cndot. ] can be expressed as:
Figure BDA0001879022020000031
wherein a is the order of the model considering the influence of the historical state data on the current temperature parameter, and the value range of a is [2,10 ]]。z(t) 1 ——z(t) s1 As input parameters for the model, z (t) 1 ＝x(t) 1 ,z(t) 2 ＝x(t) 2 ,…，z(t) 7 ＝x(t) 7 ，z(t) 8 ＝x(t-1) 1 ，…，z(t) 7(a+1) ＝x(t-a) 7 ，z(t) 7(a+1)+1 ＝y(t-1) 1 ，…，z(t) s1 ＝y(t-a) 3 (ii) a Wherein s is 1 =10 × a +7. The output of the model is the predicted value of the autoregressive prediction vector
Figure BDA0001879022020000033
Temperature parameter time series prediction model PM [ ·]From 1 input layer L 1 1 output layer L 3 And 1 hidden layer L with delay 2 Constitution, L 1 The number of layer nodes is the number s of input parameters 1 ，L 3 The number of layer nodes is 3,L of output parameters 2 Number of layer nodes s 2 The calculation formula of (c) is:
Figure BDA0001879022020000032
wherein INT (·) is an integer function, c is a constant, and the value range of c is [3,6].
Hidden layer L of temperature parameter time series prediction model 2 Is weighted input L to the ith node 2 b i Comprises the following steps:
Figure BDA0001879022020000041
wherein v is ij Is the connection weight between the jth input layer node and the ith hidden layer node, theta i Is the offset of the ith hidden layer node.
Hidden layer L 2 Weighted output OL of the ith node of 2 b i Comprises the following steps:
Figure BDA0001879022020000042
wherein e is a natural constant.
Output layer L 3 Is weighted input L to the kth node 3 b k And output
Figure BDA0001879022020000043
Comprises the following steps:
Figure BDA0001879022020000044
wherein, w kh Is the connection weight between the h hidden layer node and the k output layer node, gamma k Is the offset of the kth output layer node;
Figure BDA0001879022020000045
auto-regressive prediction vector predictor for prediction model output
Figure BDA0001879022020000046
The (k) th element of (a),
Figure BDA0001879022020000047
namely predicted values of the main bearing temperature, the cabin temperature and the hub temperature of the wind turbine at the moment t.
In step 2, the training method of the temperature parameter time sequence prediction model is a Levenberg-Marquardt iterative algorithm, and the iteration termination condition is as follows:
Figure BDA0001879022020000048
wherein the content of the first and second substances,
Figure BDA0001879022020000049
is the actual output value of the output node k after the r-th training, y (t) kr Reading an expected output value of the output node k after the r training, namely a predicted variable value actually measured by a data acquisition and monitoring control system from a training sample; er is the maximum allowable mean square error, er is in the interval [0.03,0.07]Taking a middle value; ger is effective error reduction speed, and Ger is in the interval of [5 × 10% -8 ，1.5×10 -7 ]A medium value.
In the step 3, the service quality index of the wind turbine temperature sign is composed of two secondary indexes, namely a temperature sign statistical health index and a temperature sign track health index. The service quality evaluation angle of the temperature sign statistical health index is as follows: the statistical conformity degree of the wind turbine temperature change and the prediction model within a period of time; the service quality evaluation angle of the temperature sign track health index is as follows: the deviation degree of the development track of the wind turbine temperature parameter predicted by the wind turbine temperature parameter and the temperature prediction model at the current moment; the two are complementary to each other.
The reference standard for calculating the temperature sign statistical health index is a prediction error standard percentile, and the method comprises the following steps of: predicting all the training sample set autoregressive prediction vectors in the step 2 by using the temperature parameter integrated prediction model obtained in the step 2, and calculating a prediction absolute error to obtain a standard prediction error sample set; percentile set pc of statistical calculation standard prediction error sample set 1 ,pc 2 ,…,pc 99 This is taken as the standard percentile of prediction error. temperature sign statistical health index IS at time t t The calculation method comprises the following steps:
Figure BDA0001879022020000051
CS t is a set formed by the time when the data acquisition and monitoring control system has status parameter records within hb hours before the current time t, and hb is in an interval [4, 72]And taking a middle value, and complementing the missing parameter records by using the state parameters at the time t when the state parameter records before the time t are less than hb hours. aq nt Predicting the model absolute error e for integration with the temperature parameter at nt time nt Closest standard percentile of prediction error pc pi The serial number of (2). nq (n q) nt The calculating method comprises the following steps:
Figure BDA0001879022020000052
aqn is the dispersion nonsingular threshold, and the value of aqn is taken in the interval [67, 95 ].
temperature at time tDegree physical sign track health index IT t The calculating method comprises the following steps:
Figure BDA0001879022020000053
wherein e is t And integrating the absolute prediction error of the prediction model for the temperature parameter at the time t, wherein arctan (-) is an arc tangent function.
Prediction absolute error e of temperature parameter integrated prediction model at time t t The calculation method comprises the following steps:
Figure BDA0001879022020000061
wherein, y (t) k Is the measured value of the kth element of the regression prediction vector Y (t), Y (t) 1 、y(t) 2 、y(t) 3 The temperature of a main bearing of the wind turbine, the temperature of a cabin and the temperature of a hub at the moment t are recorded by a data acquisition and monitoring control system respectively.
In step 4, the service quality IS evaluated according to the temperature sign quality index value of the wind turbine by IS t The index is dominant, IT t Index assisted IS t The service quality is judged by the indexes, and the evaluation criterion is as follows:
1)IS t ∈(0.95，1]the service quality of the wind turbine is excellent; IS t ∈(0.85，0.95]In time, the service quality of the wind turbine is good; IS t ∈(0.7，0.85]In time, the service quality of the wind turbine is slightly degraded, and attention needs to be paid; IS t ∈(0.5，0.7]In time, the service quality of the wind turbine is degraded, and close attention is needed; IS t ∈[0，0.5]In the process, the service quality of the wind turbine is severely degraded, and the wind turbine needs to pay close attention or plan maintenance.
2)IT t ∈(0.8，1]When the wind turbine is used, special attention does not need to be paid to the wind turbine; IT (information technology) device t ∈(0.6，0.8]In time, wind turbine conditions need to be concerned; IT (information technology) device t ∈[0,0.6]When the wind turbine needs to pay close attention to the wind turbine or plan maintenance.
Compared with the prior art, the invention has the beneficial effects that: the influence of a long-time-scale historical state of the wind turbine on the temperature parameter change is fully considered, the change rule of the wind turbine parameters under the healthy state is accurately mastered by establishing a nonlinear time series autoregressive neural network prediction model with the external input of the temperature parameter band of the permanent magnet direct-driven wind turbine, the accurate evaluation of the service quality of the wind turbine is realized according to the difference between the actually measured temperature parameter and the estimated value of the wind turbine parameters under the healthy state, and a key technical guarantee can be provided for scientific maintenance and efficient operation of the permanent magnet direct-driven wind turbine under the severe working environment; the method can realize the online evaluation of the full-working-condition dynamic health state including the shutdown state in the service period of the wind turbine, is not limited to the running state, and can master the health state of the wind turbine more completely. The training sample source of the prediction model is state data of the wind turbine to be evaluated 1 year after the wind turbine formally starts to be in service, further screening is not needed, a sample subset is constructed by adopting uniform random sampling with playback, the influence of uncertainty of the training sample on prediction precision is inhibited in a mode of respectively training integration of a plurality of prediction models, and the effectiveness of temperature parameter prediction can be guaranteed while the model training operation difficulty is simplified. According to the method, the service quality indexes of the temperature signs of the wind turbine are constructed, the service quality information of the wind turbine which is contained in the predicted absolute error data and is not clear is presented accurately, clearly and clearly, and the working intensity and difficulty of a user are reduced. The method can also be popularized and applied to the dynamic health state on-line evaluation of complex electromechanical systems of other types of wind turbines, high-speed trains and the like.
Drawings
FIG. 1 is a flow chart of a method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a temperature parameter time series prediction model according to an embodiment of the present invention; wherein, 1, an input layer; 2. a hidden layer; 3. an output layer;
FIG. 3 is a prediction error standard percentile scatter plot according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating a calculation result of a temperature sign service quality index and abnormal shutdown data of a second year of service of a wind turbine according to an embodiment of the invention; fig. 4 a is the predicted absolute error; b of fig. 4 is the temperature sign statistical health indicator; c of fig. 4 is the temperature sign track health indicator; FIG. 4 d is wind turbine abnormal shutdown data;
FIG. 5 is a schematic diagram illustrating a calculation result of a service quality index of a temperature sign of a third year of service of a wind turbine and abnormal shutdown data according to an embodiment of the invention; fig. 5 a is the predicted absolute error; b of fig. 5 is a temperature sign statistical health indicator; c of fig. 5 is a temperature sign track health indicator; FIG. 5 d is data of abnormal shutdown of the wind turbine.
Detailed Description
As shown in fig. 1, the detection method according to an embodiment of the present invention comprises the following steps:
firstly, establishing a temperature parameter time series prediction model which takes the rotating speed of a hub of a wind turbine, the external wind speed, the ambient temperature, the output active power and the blade pitch angle as external input variables and takes the temperature of a main bearing, the temperature of an engine room and the temperature of the hub as autoregressive prediction variables. Model structure referring to fig. 2, the external input vector X (t) is:
X(t)＝[x(t) 1 x(t) 2 x(t) 3 x(t) 4 x(t) 5 x(t) 6 x(t) 7 ]
wherein t is the current time. x (t) 1 、x(t) 2 、x(t) 3 、x(t) 4 、x(t) 5 、x(t) 6 、x(t) 7 The rotating speed of a hub of the wind turbine at the moment t, the external wind speed, the ambient temperature, the output active power, the pitch angle of the blade 1, the pitch angle of the blade 2 and the pitch angle of the blade 3 are respectively.
The autoregressive prediction vector Y (t) is:
Y(t)＝[y(t) 1 y(t) 2 y(t) 3 ]
wherein, y (t) 1 、y(t) 2 、y(t) 3 The main bearing temperature, the cabin temperature and the hub temperature of the wind turbine at the moment t are respectively.
In this embodiment, the sampling period of the SCADA system of the wind turbine is 10 minutes, and the influence of the historical state data of the previous 40 minutes on the current temperature parameter is considered, so that the order a is 3, and the number s of the input parameters is 1 =10 × 3+7=37, temperatureParameter time series prediction model PM [ ·]Can be expressed as:
Figure BDA0001879022020000081
wherein, z (t) 1 ＝x(t) 1 ,z(t) 2 ＝x(t) 2 ,…，z(t) 7 ＝x(t) 7 ，z(t) 8 ＝x(t-1) 1 ，…，z(t) 28 ＝x(t-3) 7 ，z(t) 29 ＝y(t-1) 1 ，…，z(t) 37 。
Model PM [ ·]Hidden layer L 2 Number of nodes s 2 The calculation formula of (2) is as follows:
Figure BDA0001879022020000082
in this embodiment, the constant c takes the value 4, and the hidden layer L 2 The number of nodes is 10. The number of the input layer nodes and the number of the output layer nodes are respectively 37 and 3.
After a temperature parameter time series prediction model is established, reading the state data of the wind turbine to be evaluated in the 1 st year of stable service from the SCADA system of the wind turbine as a training sample set. The evaluation object of the embodiment is a certain permanent magnet direct-drive wind turbine of a wind field in certain mountainous region in south China, the wind turbine is installed on site in 2012 and starts to be in service stably in 2013. The training sample set is state data recorded by the SCADA system of the wind turbine from 1/2013 to 31/12/2013.
In the embodiment, 5 sample subsets are obtained by uniformly and randomly sampling the training sample set, the number of samples of each sample subset is 3/10 of that of the sample set, and the sample subsets are applied to respectively and independently train 5 temperature parameter time sequence prediction models. The maximum allowable mean square error er at the end of the training iteration is 0.05, and the effective error reduction speed Ger is 1 multiplied by 10 -7 . And integrating the trained 5 models by adopting a prediction result averaging mode, and establishing a temperature parameter integrated prediction model.
Acquired temperature parameter integration predictionAfter the model is measured, predicting the autoregressive prediction vector of a wind turbine training sample set by using the model and calculating a prediction absolute error to obtain a standard prediction error sample set; percentile set pc of statistical calculation standard prediction error sample set 1 ,pc 2 ,…,pc 99 This is taken as the standard percentile of prediction error. The obtained standard percentile of the prediction error is shown in figure 3, and it can be known from figure 3 that in a standard prediction error sample set, 80% of the element values are within 0.12 ℃, and the prediction precision of the temperature parameter integrated prediction model is extremely high.
In the embodiment, when the service quality index of the temperature sign of the wind turbine is calculated, the value of the parameter hb is 48 hours, and the deviation nonsingular threshold aqn is 82. The calculation results of the service quality indexes of the temperature signs in 2014 and 2015 of the wind turbine in the embodiment and abnormal shutdown data refer to fig. 4 and fig. 5. Wherein, a in fig. 4 and a in fig. 5 are the predicted absolute errors in 2014 and 2015, respectively; b of fig. 4 and b of fig. 5 are the calculation results of the temperature sign statistical health indicator in 2014 and 2015, respectively; (ii) a C of fig. 4 and c of fig. 5 are the temperature sign trajectory health indicator results of 2014 and 2015, respectively; d in FIG. 4 and d in FIG. 5 are data of abnormal shutdown of the wind turbine in 2014 and 2015 respectively. As can be seen from the figure, the parameter sequence 4000-7000 in 2014 and the parameter sequence 45000-50000 in 2015 recorded by the SCADA system cause long-time abnormal shutdown of the wind turbine, and the temperature sign statistical health index IS IS generated when the abnormal shutdown IS about to occur t The numerical value is obviously reduced, and most of the time is positioned in the service quality moderate degradation and severe degradation interval of the service quality evaluation criterion of the invention; temperature sign track health index IT t At the moment, the wind turbine state is mostly in the interval between the wind turbine state needing attention and the wind turbine needing close attention or scheduled maintenance. At normal time of wind turbine state (no abnormal shutdown or imminent), IS t And IT t The numerical values are all high, and most parts are located in the interval with good service quality and slightly degraded service quality or without paying special attention to the wind turbine. The method can effectively evaluate the service quality of the wind turbine. Comparing fig. 4 b, fig. 5 b, fig. 4 c, fig. 5 c with fig. 4 a, fig. 5 a, the present invention is shownThe service quality index of the temperature sign of the built wind turbine, particularly the statistical health index of the temperature sign can accurately, clearly and clearly present the service quality information of the wind turbine which is contained in the prediction absolute error data and is not clear; the temperature sign track health index can be used as a conscious supplementary auxiliary temperature sign counting health index to evaluate the service quality of the wind turbine.
According to the method, the influence of a long-time-scale historical state of the wind turbine on the temperature parameter change is fully considered, the change rule of the wind turbine parameter under the health state is accurately mastered by establishing a nonlinear time series autoregressive neural network prediction model with the external input of the temperature parameter of the permanent magnetic direct-drive wind turbine, the accurate evaluation of the service quality of the wind turbine is realized according to the difference between the actually measured temperature parameter and the wind turbine parameter estimated value under the health state, and a key technical guarantee can be provided for scientific maintenance and efficient operation of the permanent magnetic direct-drive wind turbine under the severe working environment; the method can realize the online evaluation of the full-working-condition dynamic health state including the shutdown state in the service period of the wind turbine, is not limited to the running state, and can master the health state of the wind turbine more completely. The training sample source of the prediction model is state data of the wind turbine to be evaluated 1 year after the wind turbine formally starts to be in service, further screening is not needed, a sample subset is constructed by adopting uniform random sampling with playback, the influence of uncertainty of the training sample on prediction precision is inhibited in a mode of respectively training integration of a plurality of prediction models, and the effectiveness of temperature parameter prediction can be guaranteed while the model training operation difficulty is simplified. According to the method, the service quality indexes of the temperature signs of the wind turbine are constructed, the service quality information of the wind turbine which is contained in the predicted absolute error data and is not clear is presented accurately, clearly and clearly, and the working intensity and difficulty of a user are reduced. The method can also be popularized and applied to the dynamic health state on-line evaluation of complex electromechanical systems of other types of wind turbines, high-speed trains and the like.

Claims (10)

1. A permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction is characterized by comprising the following steps:
1) Establishing a temperature parameter time sequence prediction model which takes the rotating speed of a hub of a wind turbine, the external wind speed, the environment temperature, the output active power and the blade pitch angle as external input variables and takes the temperature of a main bearing, the temperature of an engine room and the temperature of the hub as autoregressive prediction variables;
2) Reading state data of a wind turbine to be evaluated in the 1 st year of stable service as a training sample set, obtaining 5 sample subsets through replaced uniform random sampling, wherein the number of samples of each sample subset is 3/10 of that of the sample set, and applying the sample subsets to respectively train 5 temperature parameter time sequence prediction models independently; integrating the trained 5 prediction models by adopting a prediction result averaging mode, and establishing a temperature parameter integrated prediction model;
3) Reading current state parameters of the wind turbine, and predicting a prediction variable at the current moment on line by applying a temperature parameter integrated prediction model; calculating the service quality index of the temperature sign of the wind turbine according to the predicted absolute error;
4) And evaluating the service quality of the wind turbine in real time according to the temperature sign service quality index value of the wind turbine.
2. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction according to claim 1, wherein an external input vector X (t) of the temperature parameter time series prediction model is as follows: x (t) = [ X (t) 1 x(t) 2 x(t) 3 x(t) 4 x(t) 5 x(t) 6 x(t) 7 ](ii) a Wherein t is the current moment; x (t) 1 、x(t) 2 、x(t) 3 、x(t) 4 、x(t) 5 、x(t) 6 、x(t) 7 The rotating speed of a hub of the wind turbine at the time t, the external wind speed, the ambient temperature, the output active power, the variable pitch angle of the blade 1, the variable pitch angle of the blade 2 and the variable pitch angle of the blade 3 are respectively;
the autoregressive predictor variable Y (t) is: y (t) = [ Y (t) 1 y(t) 2 y(t) 3 ](ii) a Wherein, y (t) 1 、y(t) 2 、y(t) 3 The main bearing temperature, the cabin temperature and the hub temperature of the wind turbine at the moment t are respectively.
3. The method for evaluating the service quality of the permanent magnet direct-drive wind turbine based on temperature parameter prediction according to claim 2, wherein the temperature parameter time series prediction model PM [ · ] is expressed as follows:
Figure RE-FDA0001982576230000021
wherein a is the order of the model considering the influence of the historical state data on the current temperature parameter; z (t) 1 ——
Figure RE-FDA0001982576230000022
As input parameters for the model, z (t) 1 ＝x(t) 1 ,z(t) 2 ＝x(t) 2 ,…，z(t) 7 ＝x(t) 7 ，z(t) 8 ＝x(t-1) 1 ，…，z(t) 7(a+1) ＝x(t-a) 7 ，z(t) 7(a+1)+1 ＝y(t-1) 1 ，…，
Figure RE-FDA0001982576230000023
Wherein s is 1 =10 × a +7; the output of the model is the predicted value of the autoregressive prediction vector
Figure RE-FDA0001982576230000024
a has a value range of [2,10 ]]。
4. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction according to claim 3, characterized in that the temperature parameter time series prediction model PM [ · is used]Comprising 1 input layer L 1 1 output layer L 3 And 1 hidden layer L with time delay 2 ，L 1 The number of layer nodes is the number s of input parameters 1 ，L 3 The number of layer nodes is 3,L of output parameters 2 Number of layer nodes s 2 The calculation formula of (2) is as follows:
Figure FDA0001879022010000023
wherein INT (-) is an integer function, c is a constant, and c has a value range of [3,6]](ii) a The hidden layer L 2 Is weighted input L to the ith node 2 b i Comprises the following steps:
Figure FDA0001879022010000024
wherein v is ij Is the connection weight between the jth input layer node and the ith hidden layer node, theta i An offset for the ith hidden layer node; hidden layer L 2 Weighted output OL of the ith node of 2 b i Comprises the following steps:
Figure FDA0001879022010000025
wherein e is a natural constant; output layer L 3 Is weighted input L to the kth node 3 b k And output
Figure FDA0001879022010000026
Comprises the following steps:
Figure FDA0001879022010000027
wherein w kh Is the connection weight between the h hidden layer node and the k output layer node, gamma k An offset for the kth output layer node;
Figure FDA0001879022010000031
auto-regressive prediction vector predictor for prediction model output
Figure FDA0001879022010000032
The (k) th element of (a),
Figure FDA0001879022010000033
namely predicted values of the main bearing temperature, the cabin temperature and the hub temperature of the wind turbine at the moment t.
5. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction according to claim 4, wherein the training method of the temperature parameter time sequence prediction model is a Levenberg-Marquardt iterative algorithm, and the iteration termination condition is as follows:
Figure FDA0001879022010000034
wherein the content of the first and second substances,
Figure FDA0001879022010000035
is the actual output value of the output node k after the r-th training, y (t) kr Outputting an expected output value of the node k after the nth training; er is the maximum allowed mean square error, er is in the interval [0.03,0.07]Taking a middle value; ger is effective error reduction speed, and Ger is in the interval of [5 × 10 -8 ，1.5×10 -7 ]Taking the value in the step (1).
6. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction according to claim 1, wherein the service quality indexes of the temperature signs of the wind turbine comprise temperature sign statistical health indexes and temperature sign track health indexes.
7. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction as claimed in claim 1, wherein the statistical health index IS of the temperature signs at the time t t The calculating method comprises the following steps:
Figure FDA0001879022010000036
CS t is a set formed by the time when the data acquisition and monitoring control system has status parameter records within hb hours before the current time t, and hb is in an interval [4, 72]Taking a middle value, and complementing the missing parameter records by using the state parameters at the moment t when the state parameter records before the moment t are less than hb hours; aq nt Predicting the absolute error e of the model for integrating with the temperature parameter at the time nt nt Nearest prediction error standard percentile pc pi The serial number of (a) is included,
Figure FDA0001879022010000041
aqn is the dispersion nonsingular threshold, aqn is in the interval [67, 95]]Taking the value in the step (1).
8. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction as claimed in claim 1, wherein the temperature sign track health index IT at the time t t The calculation method comprises the following steps:
Figure FDA0001879022010000042
wherein e is t And integrating the absolute prediction error of the prediction model for the temperature parameter at the time t, wherein arctan (·) is an arctangent function.
9. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction as claimed in claim 1, wherein the prediction absolute error e of the temperature parameter integrated prediction model at the time t t The calculation method comprises the following steps:
Figure FDA0001879022010000043
wherein, y (t) k Is the measured value of the kth element of the regression prediction vector Y (t), Y (t) 1 、y(t) 2 、y(t) 3 The temperature of the main bearing of the wind turbine, the temperature of the cabin and the temperature of the hub at the time t are recorded by the data acquisition and monitoring control system respectively.
10. The service quality evaluation method of the permanent magnet direct-drive wind turbine based on temperature parameter prediction according to claim 1, wherein the evaluation criterion for evaluating the service quality of the wind turbine in real time according to the index value of the service quality of the temperature sign of the wind turbine is as follows:
1)IS t ∈(0.95，1]the service quality of the wind turbine is excellent; IS t ∈(0.85，0.95]In time, the service quality of the wind turbine is good; IS t ∈(0.7，0.85]In time, the service quality of the wind turbine is slightly degraded, and attention needs to be paid;IS t ∈(0.5，0.7]in time, the service quality of the wind turbine is degraded, and close attention needs to be paid; IS t ∈[0，0.5]In the process, the service quality of the wind turbine is severely degraded, and the wind turbine needs to be closely concerned or planned to overhaul and maintain;
2)IT t ∈(0.8，1]in time, the wind turbine does not need to be particularly concerned; IT (information technology) device t ∈(0.6，0.8]When the wind turbine needs to be concerned about the state of the wind turbine; IT (information technology) device t ∈[0,0.6]When the wind turbine needs to pay close attention or the maintenance is planned.
CN201811413999.7A 2018-11-26 2018-11-26 Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction Active CN109583075B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811413999.7A CN109583075B (en) 2018-11-26 2018-11-26 Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811413999.7A CN109583075B (en) 2018-11-26 2018-11-26 Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction

Publications (2)

Publication Number Publication Date
CN109583075A CN109583075A (en) 2019-04-05
CN109583075B true CN109583075B (en) 2022-12-02

Family

ID=65924430

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811413999.7A Active CN109583075B (en) 2018-11-26 2018-11-26 Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction

Country Status (1)

Country Link
CN (1) CN109583075B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110544003B (en) * 2019-07-18 2023-09-15 湖南科技大学 Wind turbine generator system frequency converter state evaluation method based on temperature prediction
CN111080039B (en) * 2020-03-17 2023-10-20 浙江上风高科专风实业有限公司 Fan cluster fault prediction method and system
CN111553032B (en) * 2020-04-27 2023-09-15 湖北文理学院 Blisk milling temperature prediction method, device, equipment and storage medium
CN116611741A (en) * 2023-07-14 2023-08-18 湖南省计量检测研究院 Construction method and system of service quality index system based on wind power equipment
CN117111661B (en) * 2023-08-31 2024-05-24 杭州泰龙净化设备工程有限公司 Centralized control system and method for production workshops

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106845826A (en) * 2017-01-18 2017-06-13 西安交通大学 A kind of cold continuous rolling production line military service quality state appraisal procedure based on PCA Cpk
JP2018036196A (en) * 2016-09-01 2018-03-08 富士電機株式会社 Wind speed prediction device, wind speed prediction system, wind speed predication method and program
CN108680358A (en) * 2018-03-23 2018-10-19 河海大学 A kind of Wind turbines failure prediction method based on bearing temperature model

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2018036196A (en) * 2016-09-01 2018-03-08 富士電機株式会社 Wind speed prediction device, wind speed prediction system, wind speed predication method and program
CN106845826A (en) * 2017-01-18 2017-06-13 西安交通大学 A kind of cold continuous rolling production line military service quality state appraisal procedure based on PCA Cpk
CN108680358A (en) * 2018-03-23 2018-10-19 河海大学 A kind of Wind turbines failure prediction method based on bearing temperature model

Also Published As

Publication number Publication date
CN109583075A (en) 2019-04-05

Similar Documents

Publication Publication Date Title
CN109583075B (en) Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction
CN105205569B (en) State of fan gear box online evaluation method for establishing model and online evaluation method
CN115013261B (en) State monitoring method and system for offshore wind farm
CN112733283A (en) Wind turbine generator component fault prediction method
CN102141808A (en) Embedded type fault pre-diagnosis system and method for steam turbine generator unit
CN103711645A (en) Wind generating set state evaluation method based on modeling parameter feature analysis
CN111415070A (en) Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data
CN103940611A (en) Self-adaptive anomaly detection method for rolling bearing of wind generator set under variable working conditions
CN114580666A (en) Multi-mode fusion motor intelligent maintenance system
Pinar Pérez et al. Economic viability analytics for wind energy maintenance management
Wei et al. Comprehensive aging assessment of pitch systems combining SCADA and failure data
CN111396266A (en) GBRT-based wind turbine generator bearing fault early warning method
Yongjie et al. Research on early fault diagnostic method of wind turbines
Wilson et al. Modeling the relationship between wind turbine failure modes and the environment
CN111814849B (en) DA-RNN-based wind turbine generator set key component fault early warning method
CN113111314B (en) Multi-unit cooperation based health state evaluation method for grid-connected wind turbine generator
Wang et al. Reliability assessment of wind farm active power based on sequential monte-carlo method
CN114320773A (en) Wind turbine generator fault early warning method based on power curve analysis and neural network
Guo et al. Early Fault Detection of Wind Turbine Gearbox Based on Adam-trained LSTM
Zhang Comparison of data-driven and model-based methodologies of wind turbine fault detection with SCADA data
CN114048767A (en) Fault monitoring and early warning method for wind power master control system
Joshi et al. Soft computing methods and its applications in condition monitoring of dgs—a review
Skrimpas et al. Detection of wind turbine power performance abnormalities using eigenvalue analysis
CN113915078B (en) Intelligent recognition method and system for working conditions of blades of fiber bragg grating type wind turbine generator
Ravet et al. Wireless Sensor Network Powered with Data Analytics for Small Hydro Reliability Enhancement

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