RESULTS OF THE CONTRIBUTIONS TO THE COMPETITION ON WIND TURBINE FAULT DETECTION AND ISOLATION Presented by Peter Fogh Odgaard* At Wind Turbine Control.

RESULTS OF THE CONTRIBUTIONS TO THE COMPETITION ON WIND TURBINE FAULT DETECTION AND ISOLATIONPresented by Peter Fogh Odgaard*At Wind Turbine Control Symposium at Aalborg University 28th-29th November 2011*kk-electronic a/s, Denmark, [email protected] from: Stoustrup, J., Kinnaert, M., Laouti, N., Sheibat-Othman, N., Othman, S., Zhang, X., Zhang, Q., Zhao, S., Ferrari, R. M., Polycarpou, M. M., Parisini, T., Ozdemir, A., Seiler, P., Balas, G., Chen, W., Ding, S., Sari, A., Naik, A., Khan, A., Yin, S., Svard, C. & Nyberg, M.

mailto:[email protected]

Outline

• Motivation• FDI/FTC Benchmark Model and Competition• Description of Selected Contributions• Results of the Selected Contributions• Planned Continuations• Is the Objectives meet?

Motivation

• Increased reliability is of high important in order to minimize cost of energy of wind turbines.

• Fault Detection and isolation (FDI) and Fault Tolerant Control (FTC) are some of the important solutions in obtaining this.

Objective

• The benchmark model1 and competition should:– To attract attention from Academia to the FDI & FTC

problem on wind turbines.– Provide a platform some how relating to wind turbines

which all can use, and which can be used for comparisons.

– A part of showing the potential of FDI and FTC in Wind Turbines.

1 Odgaard, P.; Stoustrup, J. & Kinnaert, M. Fault Tolerant Control of Wind Turbines – a benchmark model Proceedings of the 7th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, 2009, 155-160

FDI/FTC Benchmark

• A generic 4.8MW wind turbine is used.

Blade & Pitch System

Drive TrainGenerator & Converter

Controller

Model Details

• Pitch actuators Second order transfer function with constraints for each blade. Close loop system.

• Converter First order transfer function with constraints. Close loop system.

• Drive train modeled with a 3 state model. Including inertias from generator and rotor.

• Simple Cp curve based aerodynamic model• Sensors modeled by band limited random noise blocks.

Wind Speed Input

Faults

• Sensor Faults– 1m1 fixed to 5 deg. 2000s-2100s– 2m2 scaled with a factor of 1.2. 2300s-2400s– 3m1 fixed to 10 deg. 2600s-2700s– r,m1 fixed to 1.4 rad/s. 1500s-1600s– r,m2 scaled with 1.1 and g,m1 scaled with 0.9. 1000s-1100s

Faults (II)

• Actuator Faults– Hydraulic pressure drop in pitch actuator 2. Abrupt

changed actuator dynamics. 2900s-3000s.– Increased air content in hydraulic oil in pitch actuator 3.

Slowly changing actuator dynamics. 3500s-3600s.– Offset on with 100 Nm. 3800s- 3900s.

• System Faults– Changed dynamics of drive train 4100s-4300s

Faults III

• Seven additional tests were performed with time shifted fault occurrences, resulting in other point of operations for the faults.

FDI Requirements

• Requirements to detection times– Sensors 10Ts– Converter 3Ts– Hydraulic oil leakage 8Ts– Air in oil 100Ts

• Requirement to interval between false positive detections – 100000 samples, and three successive detections are accepted.

• All faults should be detected.

Gausian Kernel Support Vector Machine solution2

• This scheme is based on a Support Vector Machine build on a Gaussian kernel.

• In this design a vector of features is defined for each fault containing 2-4 relevant measurements, filtered measurements or combinations of these.

• Data with and without faults were used for learning the model for FDI of the specific faults, based on this the vectors, kernel were found.

2 Laouti, N., Sheibat-Othman, N. & Othman, S., Support Vector Machines for Fault Detection in Wind Turbines Proceedings of IFAC World Congress 2011, 2011, 7067-7072

Estimation Based solution3

• A fault detection estimator is designed to detect faults, and an additional bank of N isolation estimators are designed to isolate the faults.

• The estimators used for fault detection and isolation are designed based on the provided models including model parameters.

• Each isolation estimator is designed based on a particular fault scenario under consideration.

3 Zhang, X., Zhang, Q., Zhao, S., Ferrari, R. M., Polycarpou, M. M. & Parisini, T.Fault Detection and Isolation of the Wind Turbine Benchmark: An Estimation-Based Approach. Proceedings of IFAC World Congress 2011, 2011, 8295-8300

Up-Down Counter solution4

• Up-down counters are used in this solution for decision of fault detection and isolation based on residuals for each of the faults.

• The fault detection and isolation residuals are based on residuals obtained by physical redundancy, parity equations and different filters.

• Up-down counters based decisions depends on discrete-time dynamics and amplitude of the residuals.

4 Ozdemir, A., Seiler, P. & Balas, G. Wind Turbine Fault Detection Using Counter-Based Residual Thresholding Proceedings of IFAC World Congress 2011, 2011, 8289-8294

Combined Observer and Kalman Filter solution5

• A diagnostic observer based residual generator is used for the faults in the Drive Train, in which the wind speed also is considered as a disturbance. It is decoupled from the disturbance and optimal.

• A Kalman filter based scheme is designed for the other two subsystems.

• GLR test and cumulative variance index are used for fault decision.

• Filter banks are used for fault isolation.

5 Chen, W., Ding, S., Sari, A., Naik, A., Khan, A. & Yin, S. Observer-based FDI Schemes for Wind Turbine Benchmark Proceedings of IFAC World Congress 2011, 2011, 7073-7078

General Fault Model solution6

• An automatic generated solution for FDI. • Main steps in the design are:

– Generate a set of potential residual generators.– Select the most suitable residual – Design the diagnostic tests for the selected set of

residual generators are designed. • A comparison between the estimated probability

distributions of residuals is used for diagnostic tests and evaluated with current and no-fault data.

6 Svard, C. & Nyberg, M. Automated Design of an FDI-System for the Wind Turbine Benchmark Proceedings of IFAC World Congress 2011, 2011, 8307-8315

Results Simulation – Fault 1

Fault # GKSV EB UDC COK GFM

1 Td:Mean 0.02s, Min 0.02s, Max 0.02sFd:Mean 0, Min 0, Max 0

Td:Mean 0.02s, Min 0.01s, Max 0.02sFd:Mean 0, Min 0, Max 0 MD:Mean 3%, Min 0% Max 20%

Td:Mean 0.03s, Min 0.02s, Max 0.03sFd:Mean 0, Min 0, Max 0

Td:Mean 10.32s, Min 10.23s, Max 10.33sFd:Mean 0.89, Min 0, Max 1




2 Td:Mean 47.24s, Min 3.23s, Max 95.09sFd:Mean 0, Min 0, Max 0MD:Mean 56%, Min 0% Max 100%


Td:Mean 69.12s, Min 7.60s, Max 95.72sFd:Mean 0, Min 0, Max 0MD:Mean 67%, Min 0% Max 100%







Td:Mean 0.04s, Min 0.03s, Max 0.10sFd:Mean 0, Min 0, Max 0MD:Mean 3%, Min 0% Max 20%












5 Td:Mean 25.90s, Min 1.24s, Max 87.49sFd:Mean 0, Min 0, Max 0MD:Mean 3%, Min 0% Max 20%



Td:Mean 31.32s, Min 1.54s, Max 91.13sFd:Mean 0.26, Min 0, Max 2MD:Mean 14%, Min 0% Max 40%




6 MD:Mean 100%, Min 100% Max 100%







7 MD:Mean 100%, Min 100% Max 100%







8 Td:Mean 0.01s, Min 0.01s, Max 0.01sFd:Mean 0, Min 0, Max 0 MD:Mean 97%, Min 0% Max 100%





Planned Continuations

• Competition Part II – FTC – 2 invited sessions proposals submitted to IFAC Safeprocess 2012

• An extended version of this benchmark model by merging it with FAST. Planning a invited session on this for ACC 2013. Details and model available in primo 2012. With Kathryn Johnson

• Competition Part III (2013) & Part IV (2014) on a simple wind farm model with faults. Details and model available in primo 2012. With Jakob Stoustrup

Is the Objective Meet?

• Yes!– Higher than expected interest in the FDI and FTC parts

of the competition.– General interest in the problem and benchmark model.

• We hope to continue the momentum of this interest into the new initiatives.

RESULTS OF THE CONTRIBUTIONS TO THE COMPETITION ON WIND TURBINE FAULT DETECTION AND ISOLATION Presented by Peter Fogh Odgaard* At Wind Turbine Control.

Documents

wind turbine fault detection

fault detection estimator

wind speed input slide

fault tolerant control

mw wind turbine

wind turbine control

isolation fdi

benchmark model proceedings