Modélisation et simulation des turbines éoliennes en vue du
Transcription
Modélisation et simulation des turbines éoliennes en vue du
Modélisation et simulation des turbines éoliennes Masterand Florin Sebastian TUDOR Professeur Dan STEFANOIU Universite ”Politehnica” de Bucharest Abstract • The problem approached within this presentation is to provide a useful and easy to use set of analytical equations regarding the HAWTs dynamics, after being expressed in input-output form, on automatic control purposes. • The equations are based here upon the causal series of events that lead to the wind energy conversion. • The HAWT model introduced next is a medium complexity one, which tries to reach for a good tradeoff between operability and accuracy. Contents 1. Introduction 2. Analytic and I/O modeling of a HAWT plant a. b. c. d. The wind source model The aerodynamics model The mechanics model The generator model 3. Simulation results 4. Conclusions Introduction • The research concerning modeling and control of renewable energy production systems based on wind activity has known an impressive development in the last years. • The stimulus for the development of wind energy was the price of oil and concern over limited fossil-fuel resources. • Now, of course, the main driver for use of wind turbines to generate electrical power is the very low CO2 emissions. • Although there are exciting new developments, particularly in very large wind turbines, and many challenges remain, there is a considerable body of established knowledge concerning the science and technology of wind turbines. Introduction General HAWT Model Break Propeller Gear box Slow Fast axis axis Electrical generator Hub Coupling Blade Connection to electrical network Pillar Nacelle Gyration mechanism Introduction A functional schemata The wind source modeling • Stochastic time-frequency descriptions of wind generator and filtering applied by the propeller. • Wind components: – the average wind speed is a stepwise profile, with durations and heights generated random, with normal distribution; – stochastic disturbances. • Two phenomena are considered when generating the disturbances: – Colored noise: von Karman adaptive filter H N ( s) = KvK (1 + bTvK s ) (1 + TvK s )(1 + aTvK s ) The wind source modeling • the impact between the propeller blades and the air flow is modeled with two harmonic filters – – the periodical effect of wind cutting is expressed by H1; H2 is used to estimate the part of the air flow that changes direction along the propeller plane. H h ,1 ( s) = • b1,0 + b1,1φs 1 + a1,1φs + a1,2 φ2 s 2 Thus the stochastic wind speed is H h ,2 ( s ) = b2,0 + b2,1φs 1 + a2,1φs + a2,2 φ2 s 2 The wind source modeling Matlab / Simulink Diagram The wind source modeling • Simulation results Aerodynamics • One of the most interesting and realistic models is based on Blade Element Momentum (BEM) Theory. • Two main hypotheses: – blades are far enough to each other, so that the turbulent interferences can be neglected; – each blade can be decomposed into a number of elements (10 - 20) with small length and approximately constant chord size. • After computing the local forces (lift and drag) applied by the wind on each element, the overall forces are obtained by integration (summation) along each blade. Aerodynamics BEM Theory G v G vr ∆r φ/ 2 lc G tract Fss Hub ⎛ v ⎞ ϕ ( r ) = arctan ⎜ ⎟ ⎝ ωss r ⎠ G FL inflow lift FD sin ϕ JJJJG ϕ ωss r pitch β π −ϕ−β 2 attack α= G FD drag FL sin ϕ G Ft thrust HUB FL cos ϕ FD cos ϕ wind ⎧ ρavr2 (r) CL ( α(r)) lc (r) ⎪⎪FL (r) = 2 ⎨ 2 ⎪F (r) = ρavr (r) C ( α(r)) l (r) D c ⎪⎩ D 2 ⎧Ft ≡ FL sin ϕ − FD cos ϕ ⎨ ⎩Fss ≡ FL cos ϕ + FD sin ϕ Tss ( r ) = Fss ( r ) ⋅ r ⎧ ∆Ft ,1 + ∆Ft , Ne ⎤ ⎡ Ne = ∆ ∆ − F r F ⎪ t ⎢∑ t , k ⎥ 2 ⎪ ⎣ k =1 ⎦ ⎨ Ne ∆Ft ,1 + Ne ∆Ft , Ne ⎤ ⎪ 2⎡ ⎥ ⎪Tss = (∆r) ⎢∑k ∆Fss,k − 2 k = 1 ⎣ ⎦ ⎩ Aerodynamics BEM Nonlinear Model 2 V wind speed 4 dz/dt Dyb u V2 (r) rel 2 Q 5 d(zeta)/dt u Product1 -C- 2 -K- Q*S Lift Force F L l_c(r) (vector) rho/2 F * sin(φ) L 1 om_ss atan2 t -K- blade speed r Blade Length Elements (vector) Subtract1 Vectorial Gain [dr/2 ; dr ; ... ; dr ; dr/2] dr = R/Ne Trigonometric Function Sum of Elements1 2 Ft F * cos(φ) Product L C 3 beta(r)(rad) -K- F * cos(φ) D -C- Bend Force F (r) φ(r) β(r) α(r) L Drag Force F D T (r) ss pi/2 const1 mod lift polar Add C 2*pi const Math Function F * sin(φ) D D cos drag polar sin r -KVectorial Gain [dr/2 ; dr ; ... ; dr ; dr/2]. dr = R/Ne Sum of Elements2 Spin Moment 1 Tss Mechanical Subsystem • Euler ‐ Lagrange equations: d ⎛ ∂Ec ⎜ dt ⎜⎝ ∂q j ⎞ ∂Ec ∂E p ∂Ed + + = Qj ⎟⎟ − ∂ ∂ ∂ q q q j j j ⎠ q = [y t ζ θ r θ g ] where generalized coordinates T Q = [NFT NFT rb Tr − Tg ] generalized forces T • After manipulations – two linear independent systems x1 ≡ ⎣⎡ z ζ ⎧x i (t ) = Ai xi (t ) + Bi ui (t ) ⎨ ⎩yi (t ) = Ci xi (t ) • T z ζ ⎦⎤ x 2 ≡ [ θs T u1 ≡ Ft u 2 ≡ ⎡⎣Tss ω fs ⎤⎦ T y1 ≡ ⎡⎣ z ζ ⎤⎦ y 2 ≡ ⎡⎣T fs ωss ⎤⎦ Pitch actuator nonlinear model: H p (s) = ωss ] ω2p β ∈ ⎡⎣ −2D ,30D ⎤⎦ s 2 + 2ζ p ω p s + ω2p [ −10,10] β∈ T Electrical subsystem • Two types of generators have to be accounted: – directly connected to the grid (with ON/OFF controller); – indirectly connected to the grid through synchronizers and frequency controllers (SCIG / DFIG). • In this paper: DFIG • Highly nonlinear model ! (induction generators…) DFIG nonlinear model • Input: {u • State: {i • Output: electro‐magnetic torque s ,d s ,d , u s , q , ur , d , u r , q } , is , q , ir , d , ir ,q } fs ≡ T fs − TG JGω TG ≡ 3 N p Lsr ( is , q ir , d − ir , q is , d ) / 2 • DFIG is a 3‐phase machine i 3ϕ ≡ [i0 • Park Transform: conversion to/from (d,q) coordinates ⎡⎣id • iq T 0 ⎤⎦ ≡ i d ,q ≡ Pi 3ϕ DFIG dynamics: iθ i−θ ] T θ = 2π / 3 ⎡cos ( ω fs t ) cos ( ω fs t − θ ) cos ( ω fs t + θ ) ⎤ ⎥ 2 ⎢⎢ P (t ) = sin ( ω fs t ) sin ( ω fs t − θ ) sin ( ω fs t + θ ) ⎥ 3⎢ ⎥ 0.5 0.5 ⎢ 0.5 ⎥ ⎣ ⎦ ⎧⎪x ≡ A ( ω fs , ωc ) x + Bu ⎨ ⎪⎩ y ≡ 3 N p Lsr ( x2 x3 − x1 x4 ) / 2 Electrical subsystem Simulink nonlinear model 2 3 om_fs ' 1 -K- Tfs ωfs 1/Jg 1 s Integrator T ωc Interpreted MATLAB Fcn P Park's matrix Interpreted MATLAB Fcn 2 u_c Saturation u_c A *x G 3 Saturation om_c 1 s Product1 Matrix Multiply Integrator1 BG* u sin Gain1 DFIG_3U Add Park's Transform om_c G Matrix Multiply x' Clock PG ωfs AG Interpreted MATLAB Fcn Product x -K- , Gain . Add1 1 Saturation TG WINTUS at a glance 1 -30/pi V f(u) Wind power rot/min 3 8 om_ss Tfs 2 Pw 1 1 u_c z Unit Delay 4 2 Tss om_c om_ss V Tss V T_f s om_ss Vm Ft beta(r)(rad) Wind Generator om_f s dz/dt 3 beta_c beta_c Tf s 10 TG TG dz/dt d(zeta)/dt u_c PG Mechanics Ft beta Pitch actuator om_ss Tss -1 p 11 PG d(zeta)/dt om_c Aerodynamics om_f s DFIG Model 5 6 7 d(zeta)/dt dz/dt Ft -30/pi rot/min. 9 om_fs Simulation results Wind speed Wind spectrum 23 160 Measured wind speed Mean wind speed 22 Wind speed [m/s] SNR = 78.3527 dB 140 Spectral power of wind speed [dB] 22.5 21.5 21 20.5 120 100 80 60 40 20 0 20 -20 19.5 0 500 1000 1500 2000 2500 Time [s] (Ts = 0.2 s) 3000 3500 -40 4000 0 0.05 0.1 0.15 4 Slow shaft angular speed 14 22.4 22.38 0.2 0.25 0.3 0.35 Normalized Frequency 0.4 0.45 0.5 Slow shaft torque x 10 12 22.36 10 22.32 Tss [Nm] ωss [rot/min] 22.34 22.3 22.28 22.26 8 6 4 22.24 2 22.22 22.2 0 500 1000 1500 2000 Time [sec] 2500 3000 3500 0 0 500 1000 1500 2000 2500 Time [sec] 3000 3500 4000 -4 1 Fast shaft torque Axial displacement speed of the nacelle x 10 4000 3500 3000 T [Nm] 0 2000 fs dz/dt [m/s] 2500 1500 1000 500 -1 0 500 1000 1500 2000 Time [sec] 2500 3000 3500 0 4000 0 500 1000 Thrust force 1500 2000 Time [sec] 2500 3000 3500 4000 3000 3500 4000 Generator torque 4000 5000 4000 3000 3000 2000 2000 T [Nm] F [N] 1000 0 G t 1000 -1000 0 -2000 -3000 -1000 -4000 -2000 -5000 0 500 1000 1500 2000 Time [sec] 2500 3000 3500 4000 0 500 1000 5 Fast shaft angular speed 5 900 1500 2000 Time [sec] 2500 Generated power x 10 4 3 fs G P [W] ω [rot/min] 895 2 1 890 0 885 0 500 1000 1500 2000 Time [sec] 2500 3000 3500 4000 -1 0 500 1000 1500 2000 Time [sec] 2500 3000 3500 4000 Conclusions • We introduced WINTUS – a flexible and yet realistic simulator of usual HAWTs. • The design is modular and the analytical equations flow is based on the causality principle. The BEM Theory was successfully employed. • Although some functional features were not modeled (the azimuth orientation, the braking system or the gearbox), the simulator is sufficiently complex to allow connections to other systems, such as automatic controllers. Thank you ! ☺ Q & A ? [email protected]