WP5 status April 20, 2015

1. WP5 statusApril 20, 2015 Tatiana Madsen

2. Table of contents • Introduction • The EDGE project • System description • Evaluation Approach • Scenario description • Results • Conclusion

3. Powertechpaper • On the Impact of information access delays on remote control of a wind turbine

4. System description • Windfarm controller • Requires fatigue estimation input and WT state • Reduce damage to the WT • Follow power reference • 2 Communication networks • IP-network: Fiber optics • Sensor network: Bus topology • 3 Sensors for each WT

5. System description • Wind farm controller • Aims to reduce damage to WT by reducing fatigue • Fatigue estimator requires measurements from the WT • The controller uses the fatigue estimators output to minimize damage • The WT reacts to setpoints set by the controller • The wind acts as a disturbance to the WT

6. System description • 2 Information Access Strategies • Periodic sampling • 1 sample per control period • Offset from controller • Optimal offset based on delay • Event driven sampling • Send sensor information when sensor data change • Different event triggers based on messages per control period • Different event triggers for each sensor

7. Evaluation Approach • We consider 2 metrics for performance evaluation • Accumulated damage: • WT always suffers damage, aim to minimize this • We assume delays in the sensor information will reduce performance • Mismatch probability (mmpr) • Probability that the given information is different from the actual information • We assume high mmpr equals poor performance

8. Scenario description • Communication parameters • 3 sensors on the WT • 1 WT • 150 ms ctrl loop • 50 ms computation time • Exponentially distributed delay • Following scenarios are simulated: • Periodic sampling with increasing mean of exponentially distributed delay • Periodic sampling with varying offset • Event driven sampling with varying event triggers

9. Results – periodic sampling

10. Results – periodic sampling

11. Results – event-driven sampling

12. Results • We have found that increasing the delay of the network decreases the performance of the controller • We have also found that for a given delay we can find an offset that increases the performance over other offsets • This seems to correlate to the mmpr of bothe the fatigue estimator output and the sensor output, however more simulation runs are needed • We have further found that for different event triggers we can increase or decrease the performance of the controller • Here we see a correlation between the mmpr of sensor 2, and the performance of the controller

13. 3-months work plan from now • Analytical modelling for the wind farm use case. Comparison with simulation results. Focus on finding optimal offset. • Enhancing simulation model by implementing technologies for wind farm use case taking onto account cross traffic. • Demand response in residential areas.

14. System description • 2 Communication networks • IP-network: Fiber optics • Sensor network: Bus topology • 3 Sensors for each WT

15. Prelimenary results

16. Prelimenary results