Numerical Modelling for Optimization of Wind Farm Turbine ... · & J.Sutton . Curtin University,...
Transcript of Numerical Modelling for Optimization of Wind Farm Turbine ... · & J.Sutton . Curtin University,...
COOPERATIVE RESEARCH CENTRE FOR CONTAMINATION ASSESSMENT AND REMEDIATION OF THE ENVIRONMENT
Numerical Modelling for Optimization of Wind Farm Turbine Performance
M. O. Mughal, M.Lynch, F.Yu, B. McGann, F. Jeanneret & J.Sutton Curtin University, Perth, Western Australia
19/05/2015
OVERVIEW OF PRESENTATION • Acknowledgements
• Introduction
• Objective
• Methodology
• Weather Research Forecasting (WRF) Sensitivity Analysis
• Coherent Doppler LIDAR (CDL) versus WRF Comparison
• Coupling WRF to OpenFOAM
• Conclusions
• Future work
• Q & As
SCHEME OF PRESENTATION
ACKNOWLEDGEMENTS • Cooperative Research Centre for Contamination
Assessment and Remediation of Environment (CRC CARE), Australian Government
• “This work was supported by resources provided by the Pawsey Supercomputing Centre with funding from the Australian Government and the Government of Western Australia."
• Department of Environment Regulation (DER), Government of Western Australia
• Dr Peter Rye, DER, Government of Western Australia • Associate Professor Diandong Ren, Curtin University
ACKNOWLEDGEMENTS
INTRODUCTION
• Significance of short term numerical forecasting (STF)
• Significance and role of CDL in STF
• Mesoscale model shortcomings at wind farm scale
• Significance of meso and microscale model coupling
• Proposed technique
• Wind energy grid integration challenges
INTRODUCTION
THE LAKE TURKANA WIND FARM • Lake Turkana wind farm characteristics
– 325 MW – Located in Marsabit district, near Lake Turkana Kenya, East Africa – 140 to 700 km (width) – 610 to 1524 m (elevation above sea level)
• CRC CARE & DER joint campaign (2009) to map the wind field using CDL
• Three masts [~40 m] and CDL used for numerical model validation
INTRODUCTION
Mast B (38 m) Mast C (46 m) Mast A (39 m)
COHERENT DOPPLER LIDAR • Lockheed
Martin WindTracer 1.6 µm Doppler-Lidar – employed in
determining wind fields at Lake Turkana
– State of the art eye safe technology with a range of 8-12 km
– 250 km2 coverage for winds
COHERENT DOPPLER LIDAR
LAKE TURKANA TOPOGRAPHY
INTRODUCTION
Topographic Height (m) Topographic Height (m)
OBJECTIVES
• Develop a short term forecasting technique
• Investigate on the use of an optimised configuration of (WRF) software
• Evaluate model performance using CDL and mast observations
• Improve wind farm forecasting by applying a coupled WRF and CFD (OpenFOAM) model
• Improve initialization data for WRF using CDL observations
OBJECTIVE
Achieved In Progress
METHODOLOGY
• WRF sensitivity analysis and validation via in situ and CDL observations
• Coupling optimised WRF model with OpenFOAM for prediction of micro-scale wind for improving turbine energy output estimation
• Evaluating impact using comparisons with in situ meteorological measurements
• WRF initialization data tuning incorporating CDL observations
METHODOLOGY
Achieved In Progress
WRF SENSITIVITY ANALYSIS
• WRF performance validation in Western Australia • Sensitivity analysis conducted at Lake Turkana, Kenya site • Sensitivity analysis includes testing
– initialization fields, physical & parametrization schemes, grid configurations, terrain complexity and satellite data
• Validation criteria based on – Root Mean Square Error (RMSE) and Correlation Coefficient between in situ and
predicted winds – Considering Mast as true measurements
• Best results obtained through changing initialization fields
WRF SENSITIVITY ANALYSIS
SENSITIVITY ANALYSIS RESULTS • Results of sensitivity analysis at Lake
Turkana • European Centre for Medium-Range
Weather Forecasts (ECMWF) initialization field selected with – 70 km horizontal resolutions – 60 model levels – 6 hourly temporal resolution
WRF SENSITIVITY ANALYSIS
Topographic Height (m)
Configuration 4 domains 36 model levels
Domain Size km x km
Geographic Resolution
Grid Resolution km
Time Step s
Domain 1 1593 x 1593 10` 27 30
Domain 2 918 x837 5` 9 10
Domain 3 459 x 297 2` 3 3.33
Domain 4 126 x 82 30” 1 1.11
WRF SENSITIVITY ANALYSIS
SENSITIVITY ANALYSIS RESULTS
Mast A vs WRF Wind Speed
ms -1 Mast mean 11.03 WRF mean 11.38 RMSE 1.66 Correlation Coefficient 0.69
CDL vs WRF RMSE 2.24 Correlation coefficient 0.60 LIDAR Mean 10.5 Mast A VS WRF Wind Direction Mast mean 113.305° WRF mean 104.37° RMSE 12.019° Correlation Coefficient 0.44
Comparison of wind speed comparison between mast A (10 mins sampling), WRF predicted wind(10 mins sampling) and CDL at 39 m height (Time UTC)
Comparison of wind speed comparison between mast A (10 mins sampling) and WRF predicted wind (10 mins sampling) at 39 m height (Time UTC)
CDL TERRAIN FOLLOWING WIND MAP COMPARISON WITH WRF
CDL WRF COMPARISON
WRF Wind Speed (m/sec)
Zoomed in data at CDL location
CDL Terrain Following Map
LAKE TURKANA CDL-WRF COMPARISON AT PROPOSED TURBINE LOCATIONS
CDL WRF COMPARISON
RMSE 1.23 ms-1
Correlation Coefficient 0.84
RMSE 1.14 ms-1 Correlation Coefficient 0.81
• WRF-CDL at proposed turbine locations suggests – WRF has captured the wind speeds
well spatially and even at locations other than masts
• Turbine location away from the mountain suggests – terrain complexity is reduced and
better RMSE values compared with mast-WRF comparison are achieved.
WRF TO OPENFOAM COUPLING • Reliable micro-siting and cost energy estimation demands
meso –micro scale model coupling • In STF coupled model can
– ingest inputs from WRF forecast running in real mode and use them to predict turbulence structures affecting wind speed ensuring reliable forecast.
• Spatial and temporal grids are, in general, non-matching – WRF grid moves in the vertical with time-dependent pressure variation.
• Ambiguous mechanism for transferring turbulent energy from one code to the other
• The validation of these models is also difficult as the measurement data available is limited
WRF TO OPENFOAM COUPLING
WRF TO OPENFOAM COUPLING • Proposed technique surpasses other techniques having
real time data from CDL. • The solver can handle complex terrain features
– e.g. topography, temperature & pressure variations etc.
• Capability to capture a complete wind profile from WRF • CDL data can be further ingested
– to improve initial conditions from WRF – to bring fluxes up to right values at solver boundaries
• The software performance is tested in Lake Turkana and the results are validated with CDL
• The results are not an artefact of nudging in WRF • Coupled WRF-CFD predicts
– turbulence structures due to topography catastrophic for turbines
WRF TO OPENFOAM COUPLING
WRF
Coupled CFD
• Terrain extracted from SRTM data • 4X4X2 km mesh (wind direction
SE) • Neumann and Dirichlet conditions
applied to boundaries • Simulation conducted on a Cray
XC40 (24 nodes each with 64GB RAM)
• Results agree well with CDL data as compared to WRF
WRF TO OPENFOAM COUPLING
WRF TO OPENFOAM COUPLING
CONCLUSIONS
• WRF modelling (without coupling) at Lake Turkana in a data sparse region of complex terrain typically achieves – RMSE in speed of 1.6 ms-1 and direction of 12° and a correlation coefficient
of 0.69 when validated against mast observations
• Comparison of WRF predictions with CDL demonstrates – an improved performance of the model with RMSE in wind speed of 1.2 ms-1,
and correlation coefficient 0.84 respectively when validated against meteorological observations; a 25% improvement in wind prediction and a 22% improvement in the correlation coefficient (against in situ mast data).
• The retrieval of vertical winds – is showing significant skill with CFD / OpenFOAM having a low standard
deviation of 3.35 ms-1 compared to WRF’s of 4.71 ms-1 when validated against CDL winds
CONCLUSIONS
FUTURE WORK • Promote the advantages of CDL in wind farm research
particularly in their potential operational role in enhancing the prediction of winds for improved infrastructure protection (e.g. turbulence, severe wind gusts) and for improved lead time energy generation management.
• It is intended to integrate local observations from CDL to improve the initialization fields for WRF
• Analyse the impact of integrating CDL data and assessing its impact on the knowledge of wind speed and direction information
• Determining the best criteria for judging a numerical model output in terms of wind speed and direction
FUTURE WORK
Q & A
?
Questions