Test Title by N08rRQQF

VIEWS: 9 PAGES: 36

									Optimal Rotors for
Distributed Wind Turbines

28th ASME Wind Energy Symposium
47th AIAA Aerospace Sciences Meeting and Exhibition
January 8, 2009


Dr. Curran Crawford & Luke Stack
Department of Mechanical Engineering
University of Victoria




                                                      1
Overview
•Why small wind turbines?
 –Why are they interesting academically?
•Design approach
•Study results




                                           2
There is wind out there if you’re careful
where to look
•Wind resource
 –Studies of wind in urban areas
   • Generally low mean wind speeds
   • Turbulence
   • Some viable locations
     – Purposeful building integration
     – Windy locales
 –Remote locations
   • Open terrain
   • Higher energy costs
 –Tower mounted preferable


                                            3
The top of a 100m pole is always going
have better wind, but...
•MW scale machines are hard to beat
  –Economies of scale
  –Wind shear
•kW scale machines left in the past?
  –MW lessons learnt not applied to small machines
  –Consumer demand
    • Engagement with energy source
  –Distributed energy source
    • Transmission infrastructure
    • Smart grids


                                                     4
There is unique academic value at the
small scale
•Validation data & trying new ideas are easy
  –Design-build-test for low $
•Challenging design space
  –Strive for multi-objective, multi-disciplinary optimum
    • Energy yield, cost, noise, etc.
    • Shape, structure, etc.
    • Matching generator to rotor
  –Tools & methods applicable at all scales




                                                            5
And of course someone has pay the bills!
•Industrial sponsor
  –Ampair in UK
  –Redesigning & up-scaling product range
    • 100 W – 8 kW
  –Some units left off results for commercial sensitivities
•UVic efforts
  –Focusing on rotor blade design
  –HAWTs not VAWTs




                                                              6
So, what’s different at the small scale?
•Reynolds number
  –Operating around transition
•Rotor control
  –Controlled pitching too expensive
  –Failsafe pitching possible (PTS or PTF)
•Generator
  –Cogging torque
    • Startup wind speed
•Power electronics
  –Full power conversion
  –Passive or active generator torque control
                                                7
Our approach to rotor optimization
•A stepwise approach




                                     8
BEM-based analysis tool ExcelBEM was
used for analysis
•Implementation
 –Excel interface
 –Underlying C/Matlab code
•Enhanced BEM method
 –Coned rotors, including blade section sweep
 –Steady and unsteady aero
   • Rigid body blade flapping
 –Implicit optimal control algorithms
   • PTF, PTS, VSS
 –Generator characteristics
 –Blade section structural properties & stresses
 –Low frequency noise                              9
The blade shape is parameterized with
DVs control points of Bezier curves
•Incorporate implicit shape constraints
  –E.g. parabolic tip
•Convex hull property




                                          10
Blade section choice treated as a
parameter
•Want to be confident in aerodynamic
 properties
  –Not much experimental data available!
  –Eppler & NACA 44XX
  –Flat plate extensions for full AOA range
•Structure
  –Glass/carbon thermoplastic skins (Twintex)
  –Foam core




                                                11
The generator characteristic is a
parameter
•Passive PE
  –Single torque-speed-power relationship
    • Torque matching problem
  –Over speed protection from centrifugal PTS
  –E.g. battery charging
•Active PE
  –Variable speed by varying generator load on rotor
  –Torque-speed-power surface
  –Need to find optimum operating profile on surface
•Variable efficiency
  –Not sufficient to find optimal tip-speed ratio
                                                       12
Generator and aerodynamic torque must
be equal for passive PE
•Example of a bad match!




                                        13
Two approaches to active PE optimization
•Implicit
  –Use built-in optimal operating routines in ExcelBEM
  –No extra DVs
  –Computationally intensive
    • Require multiple BEM solutions for each design iterate
•Explicit
  –Add DVs for Ω at 15-20 wind speeds
  –Directly control rotor speeds
  –Extra finite differencing, but better overall



                                                               14
The overall optimization procedure was
developed during the study
•Full process only applied to final example
  –Incremental steps for other design cases
•Final design/optimization process
  –Matlab driven interactive with designer
  –Navigate towards energy yield optimum
    • Pick airfoil set
    • Analytic chord/twist optimum for CP at given λopt
    • Fit shape DVs to analytic optimum (lsqnonlin)
    • Direct numerical optimization for Eann over entire wind speed
    range (fmincon)
  –Careful numerics
    • Finite differencing (round off & truncation errors)
    • Limit step sizes                                            15
Additional constraints can be applied to
the optimization
•Startup wind speed
  –Cogging torque must be overcome
    • Either:
      – Specify Vcut-in (no extra DVs)
      – Add Vcut-in as a DV and avoid over constraining
•Maximum electrical power
•Geometric constraints for manufacture
  –Max. chord and twist
•Others possible, but not used here
  –E.g. tip speed (noise), tower thrust

                                                          16
Additional step for structure definition
•Manual iteration on ply count (1-4) for one of
 two cases:
  –Simulated loading: extreme storm & rated operation
  –IEC 61400-2: root bending moments at various
   conditions
    • Scaled loading for other sections
    • Found that maximum yaw rate tends to limit
•Eventually like to couple aero-structure
  –Not as critical at small scale



                                                    17
Design studies
•Predetermined diameters/powers




                                  18
Ampair 600 rotor redesign
•First design
  –No generator characteristic available
    • Only guesses at rotor speeds at cut-in and
    rated power
    • Assumption that pitch mechanism activates
    just after rated power
  –Could not perform Eann optimization
   directly
  –Numerical optimization for CP
    • Range of λopt tried
    • 6 found as optimum structurally and
    aerodynamically

                                                   19
Field tests were carried out on new rotor
yielding a few insights
 –Clearly require good generator spec.
   • Pitching mechanism performance also related
 –Should be considering dynamic operation




                                                   20
Ampair 300 rotor redesign
•Well characterized generator
  –Passive PE
•Full Eann optimizer not yet available
  –Could now redo and impose torque matching
   constraint
  –Numerical CP optimization
    • Range of λopt tried
  –Post-process
    • Implicit torque matching
    routines
    • Identify rotor with highest
    energy yield
                                               21
Fairly good rotor match to generator
•Generator still a bit aggressive
  –Better if generator curve shifted left




                                            22
Imposition of cogging constraint changes
rotor somewhat
                           •Effects
                             –Larger chords
                             –Startup now
                              at 3m/s, vs.
                              5m/s for
                              unconstrained
                              design
                             –Generator
                              more
                              aggressive
                              relative to
                              rotor
                                           23
A unique test apparatus was available…
•Equipment
 –Modified VW golf
 –12V load
 –Anemometer,
  tachometer, etc.
 –Quiet UK country lane
 –Windless day
 –Repeated/averaged
  runs



                                         24
Constant speed testing atop vehicle
performed for verification
•Quite good agreement!
 –Not just aerodynamics being tested
   • Results include generator matching algorithm




                                                    25
6000 W clean sheet design
•Full Eann optimizer available
•Active PE
  –Variable speed
    • Optimal power capture
    • VSS control above rated
  –Generator characteristic contrived for these results
    • Data from dynamometer is being gathered
    • Expect efficiency degradation with partial power
      – Assume 70%-95% performance




                                                          26
Generator characteristic defined from
“experimental” data points
                            4
                         x 10

                    3




                   2.5




                    2




                   1.5
       Powerelec




                    1




                   0.5




                    0



                                                                                            300
                -0.5
               1200                                                                   200
                           1000   800   600                              100
                                                     400   200   0   0

                                              gen                             Nrot gen           27
Begin with simple analytic CP optimum
and fit blade profiles with chord constraint
•Test range of λopt
  –Implicit optimal operation algorithm




  –Optimum around λ = 5-6
  –Performance hit with simple fitting procedure

                                                   28
Perform full Eann optimization with
increasing cogging torque constraint
•Blade profiles




                                       29
Operating curves




                   30
Performance over generator map

                     Note high cogging torque for
                     higher solidity rotors




                                                    31
Cogging torque impacts energy yield




•Key results
  –Complete optimization process is superior to simple
   chord cap of a CP optimized blade
  –Shape changes predominantly in the chord profiles
    • Some twist to maintain optimal angles of attack
  –Increasing cogging torque constraint leads to:
    • Higher solidity rotors
    • Attendant structural and weight repercussions

                                                         32
Acoustic concerns over downwind rotor
led to study of coning angle/offset
                        •Design point
                          –Rated conditions
                           (11m/s)




                                          33
A compromise of 7° was reached
between acoustics and stress
•Centrifugal moments build past 7°




                                     34
Future directions
•More integration in MDO loop
 –Concurrent structures optimization
 –Generator/PE design
•Add DVs and constraints
 –Optimal blade lengths?
 –Tower thrust tradeoffs
•Include dynamic behavior
 –Stochastic winds
 –Effect on power curve
 –Fatigue loads?

                                       35
Thanks are in order…




                       UVic Engineering Design Office




                                                        36

								
To top