Inverter Applications •Motor Drives • Power back-up systems •Others: Example HVDC Transmission systems 1 Single Phase Inverter Square-wave “Modulation” (1) + q1(t) Vdc 1-q1(t) - Q1 1-q2(t) Q’2 +v out q2(t) Q’1 Vdc vout(t) Q2 -Vdc sin(n / 2) vout (t ) Vdc cos(nt n0 ) n n 1 4 Vout 1 Vdc 1.27Vdc 4 2 t Single Phase Inverter Square-wave “Modulation” (2) THD = 0.48 Characteristics: - High harmonic content. - Low switching frequency. - Difficult filtering. - Little control flexibility. 3 Single Phase Inverter Square-wave “Modulation” (3) + q1(t) Q1 1-q2(t) +v out q2(t) Vdc 1-q1(t) Q’1 - Q’2 Vdc vout(t) Q2 t -Vdc sin(n / 2) cosn S t cosn S t n vout (t ) n1 n 4 Vout 1 Vdc cos 2 2Vin 4 Single Phase Inverter Square-wave “Modulation” (4) Example with Vout-1=1.21Vdc THD = 0.3 Characteristics: - High harmonic content. - Low switching frequency. - Difficult filtering. - More control flexibility. 5 Single Phase Inverter Pulse Width Modulation (1) vout DVdc (1 D)( Vdc ) Vdc (2 D 1) D is the duty cycle of switch Q1. D is the portion of the switching cycle during which Q1 will remain closed. TON Q11 D TS In PWM D is made a function of time D=D(t) 6 Single Phase Inverter Pulse Width Modulation (2) •Let’s 1 1 D m(t ) 2 2 m(t ) m[cos(0t )] where Modulation index Fundamental Signal Vout 1 m Vdc 7 Modulation function Single Phase Inverter Pulse Width Modulation (3) 1 D [1 m cos(0t )] 2 n m(t ) sin 4Vdc 2 2 cos(n t ) vout (t ) m(t )Vdc S n1 n Moving average vout Vdc (2 D 1) m(t )Vdc 8 t Single Phase Inverter Pulse Width Modulation (4) Implementation issue: time variable “t” needs to be sampled. Two basic sampling methods: UPWM NSPWM 9 Single Phase Inverter Pulse Width Modulation (5) t 10 Single Phase Inverter Pulse Width Modulation (6) Considering that m(t ) m cos( 0t ) Fundamental (n=1) and its sidebands NSPWM mVdc no sidebands UPWM nm J 4Vdc k n 2k é n ù sin ê n ú i 1 n 1 2û ë k 4Vdc 1 im i J 0 sin i 1 i 2 2 Carrier 4Vdc im i J sin 0 component i 1 2 2 (n = k) Sidebands of im J the carrier 4Vdc n 2 é ù 4Vdc sin n i i êë i 1 2 úû 11 nm J k n 2k é n ù sin ê n ú 1 2û ë k i 1 n Three Phase Inverter Pulse Width Modulation (1) Active States: Zero States: State Qa Qb Qc S0 0 0 0 S7 1 1 1 12 State Qa Qb Qc S1 0 0 1 S2 0 1 0 S3 0 1 1 S4 1 0 0 S5 1 0 1 S6 1 1 0 Three Phase Inverter Pulse Width Modulation (2) ma (t ) m[cos(0t ) ea 0 (t )] Modulation m (t ) m[cos( t 2 / 3) e (t )] b 0 b0 Functions mc (t ) m[cos(0t 2 / 3) ec 0 (t )] ea 0 (t ) e03 (t ) ea 0H (t ) “Zero-Sequence” e (t ) e (t ) e (t ) b0 03 b 0 H Signals ec 0 (t ) e03 (t ) ec 0H (t ) 13 Three Phase Inverter Pulse Width Modulation (3) Triplen Harmonics (3, 9, 15, …) ma (t ) m[cos(0t ) e03 (t ) ea 0H (t )] Modulation Fundamental Signal index V ph peak m Vdc / 2 14 Other Harmonics (5, 7, 11, 13, 17, ….) Three Phase Inverter Pulse Width Modulation (4) m cos(0t ) t e03 (t ) m(1/ 6) cos(30t ) ma (t ) m[cos( 0t ) e03 (t )] m[cos( 0t ) (1 / 6) cos( 30t )] 15 Classic Approaches to PWM (1) Time Domain Use of modulation signal: Duty cycle computation: D 1 1 m (kT ) a 2 2 a s 1 1 Db mb (kTs ) 2 2 t D 1 1 m (kT ) c 2 2 c s 16 Sector limits Classic Approaches to PWM (2) Classic SVM - Application Park’s Transformation 3 Park’s Transformation ia* (t ) ib* (t ) 2 1 1 ia 1 M 2 2 2 * ib M 3 3 3 0 q i 2 2 c ic* (t ) M (kTs ) M d* (kTs ) * q * d M d (kTs ) - - M de d-PI Controller + q-PI Domain Transformer (Modulator) Space Time Vector M qe Controller M q (kTs ) + M qref (kTs ) M dref ( kTs ) vref (From Vector Controller) M d* cos(t ) Switch Control DC Source M q* sin(t ) Inverter Motor ia cos(t ) ib cos(t 2 / 3) ic cos(t 2 / 3) Disadvantages: Can’t see evolution in time Loss of information about e0-3(t) 17 Classic Approaches to PWM (3) Classic SVM S2 S3 v̂q S6 Space vector domain SECTOR v̂2 v̂6 II SECTOR Bdq vˆd vˆq vref I SECTOR v̂4 2 Bases t v̂3 III v̂0 BSVM vˆi vˆ j v̂7 O v̂ d SECTOR SECTOR S 4 IV v̂1 S1 S In O: 7 S0 18 VI SECTOR V v̂5 S5 BSVM changes in each sector BSVM in sector I vˆi vˆ4 vˆ j vˆ6 Classic Approaches to PWM (4) Classic SVM Transitions within sector I S6 S2 T7=T0 SVM Computation S4 S7 S3 -Track the sector in which vref is in and based on it select the appropriate set of basis Bij S0 S1 S5 -Calculate the coordinates of vref in the basis Bij - Change the coordinates of the reference voltage vector from basis Bdq to basis Bij. The sector dependant transformation yields the period of time Ti that the machine remains in each state in a given sampling period. - When the time Ti is finished move to the next state following the sequence given by the SVM state machine. 19 Mathematical Framework (1) Complete representation involves a 3-D space Then, a 3-D vector can be introduced: v R Space Vector Domain W 20 Control Time Domain Output Time Domain S Mathematical Framework (2) Control Time Domain Functions of time are used as basis tˆa cos(t ) ea 0 (t ) ˆ ˆ ˆ ˆ Bt ta , tb , tc tb cos(t 2 / 3) eb 0 (t ) tˆ cos(t 2 / 3) e (t ) c0 c v m(tˆa tˆb tˆc ) 21 Mathematical Framework (3) Space Vector Domain Bdq 0 dˆ , qˆ , nˆ0 t dˆ 1 0 0T T qˆ 0 1 0 T n 0 0 e ( t ) ˆ 0 3 0t v vd (t )dˆ vq (t )qˆ v0 (t )nˆ0t When e0-H(t)=0, v describes a circumference with radius equal to m. 22 Mathematical Framework (4) Output Time Domain vˆa 1 0 0 T vˆb 0 1 0 T vˆc 0 0 1 T BS vˆa , vˆb , vˆc v va (t )vˆa vb (t )vˆb vc (t )vˆc 23 Mathematical Framework (5) Output Time Domain Length of sides equal to 2 Each corner represents one state State sequence is obtained naturally Bdq va vb vc 0 Balanced system plane Sectors: six pyramid-shaped volumes bounded by sides of the cube and |vi|=|vj| planes (i, j = a,b,c; i j). 24 Mathematical Framework (6) Matrix R When e0-H(t)=0, v describes a circumference with radius equal to m. 1st Idea: Use Park’s transformation to a synchronous rotating reference frame: m(cos(t ) e03 (t )) vd cos(t ) cos(t 2 / 3) cos(t 2 / 3) 2 v sin( t ) sin( t 2 / 3 ) cos( t 2 / 3 ) m (cos( t 2 / 3 ) e ( t )) q 0 3 3 v 1 m(cos(t 2 / 3) e (t )) 1 1 0 0 3 Instantaneous values of the voltages 25 Mathematical Framework (7) Matrix R 1 vd v 1 q v 2me (t ) 0 0 3 Problem: components in d̂ and q̂ are constant values I am interested in having a constant value in nˆ0 t 2nd Idea: Freeze the rotational reference frame at t=0 1 1 1 2 2 v m(cos(t ) e03 (t )) d m cos(t ) 2 3 3 vq 0 m(cos(t 2 / 3) e03 (t )) m sin(t ) 2 2 v 3 2me (t ) m (cos( t 2 / 3 ) e ( t )) 1 1 1 0 0 3 0 3 26 Mathematical Framework (8) Matrix R 3rd Idea: Rearrange the product. vB dq 0 1 1 1 2 2 cos(t ) e03 (t ) vd 2 3 3 v 0 cos( t 2 / 3 ) e ( t ) m q 0 3 2 2 v 3 1 1 1 cos(t 2 / 3) e03 (t ) 0 cos(t ) (1 / 2) cos(t 2 / 3) (1 / 2) cos(t 2 / 3) m 2 0 ( 3 / 2) cos(t 2 / 3) ( 3 / 2) cos(t 2 / 3) m 3 m e ( t ) e ( t ) e ( t ) 0 3 0 3 0 3 vB t 27 Mathematical Framework (9) Matrix R 4th Idea: Eliminate the dependency on e0-3(t) in order to have a constant coordinate in nˆ0 t cos(t ) (1 / 2) cos(t 2 / 3) (1 / 2) cos(t 2 / 3) 2 R 0 ( 3 / 2) cos(t 2 / 3) ( 3 / 2) cos(t 2 / 3) 3 1 1 1 vd m cos(t ) vq R m m sin(t ) v m 2 0 28 Mathematical Framework (10) Matrix R In order to include e0-H(t) we need to follow the same steps and apply superposition. cos(t ) (1 / 2) cos(t 2 / 3) (1 / 2) cos(t 2 / 3) 2 R 0 ( 3 / 2) cos(t 2 / 3) ( 3 / 2) cos(t 2 / 3) 3 1 1 1 ea 0 H (t ) (1 / 2)eb 0 H (t ) (1 / 2)ec 0 H (t ) 2 0 ( 3 / 2)eb 0 H (t ) ( 3 / 2)ec 0 H (t ) 3 0 0 0 cos(t ) ea 0 H (t ) (1/ 2)(cos(t 2 / 3) eb 0 H (t )) (1 / 2)(cos(t 2 / 3) ec 0 H (t )) 2 R 0 ( 3 / 2)(cos(t 2 / 3) eb 0 H (t )) ( 3 / 2)(cos(t 2 / 3) ec 0 H (t )) 3 1 1 1 29 Mathematical Framework (11) Matrix W 1 1 ˆ d ' vˆa vˆb vˆc 2 2 qˆ ' vˆb vˆc nˆ '0 qˆ dˆ vˆa vˆb vˆc dˆ ' 1 1 / 2 1 / 2 vˆa 1 vˆb qˆ ' 0 1 nˆ ' 1 vˆ 1 1 c 0 W’ 30 Mathematical Framework (12) Matrix W 1 1/ 2 1/ 2 W ' 0 1 1 1 1 1 0 1 2 1 W ' ' 1 3 1 2 3 1 1 0 e03 (t ) 2 1 W 1 3 e03 (t ) 2 3 e03 (t ) 1 31 1) Take the transpose and apply scaling factor 2) Include e0-3(t) in order to have it as a component Mathematical Framework (13) Matrix S S=WR 2 cos(t ) e03 (t ) cost 2 / 3 e03 (t ) cos(t 2 / 3) e03 (t ) 1 S cos(t ) e03 (t ) 2 cos(t 2 / 3) e03 (t ) cos(t 2 / 3) e03 (t ) 3 cos(t ) e03 (t ) cos(t 2 / 3) e03 (t ) 2 cos(t 2 / 3) e03 (t ) 2ea 0 H (t ) eb 0H (t ) ec 0 H (t ) 1 ea 0 H (t ) 2eb 0 H (t ) ec 0 H (t ) 3 ea 0 H (t ) eb 0H (t ) 2ec 0H (t ) 2 cos(t ) e03 (t ) 2ea 0 H (t ) cost 2 / 3 e03 (t ) eb 0 H (t ) cos(t 2 / 3) e03 (t ) ec 0 H (t ) 1 S cos(t ) e03 (t ) ea 0 H (t ) 2 cos(t 2 / 3) e03 (t ) 2eb 0 H (t ) cos(t 2 / 3) e03 (t ) ec 0 H (t ) 3 cos(t ) e03 (t ) ea 0 H (t ) cos(t 2 / 3) e03 (t ) eb 0 H (t ) 2 cos(t 2 / 3) e03 (t ) 2ec 0 H (t ) 32 3D Analysis and Representation v B (W )v B S dq 0 vB dq 0 ( R )v B t v B ( S )v B S t cos(t S ) ea 0 (t S ) va vb m cos(t S 2 / 3) eb 0 (t S ) v cos(t 2 / 3) e (t ) c S c0 S 33 3D Analysis and Representation 1 ei 0 (t ) e03 (t ) cos(3t ) 4 UPWM 34 3D Analysis and Representation SVM UPWM 35 3D Analysis and Representation 3D Representation: Plot evolution of v in output time domain during a complete fundamental period. The resulting curve always lays within the cube defined by the switching states. 1 ei 0 (t ) e03 (t ) cos(3t ) 6 36 3D Analysis and Representation Triplen harmonic distortion: Evolution away from the plane va+vb+vc=0 Other harmonic distortion: non circular projection of the curve over the plane va+vb+vc=0 Sharp corners indicate the presence of higher order harmonics. 37 Commonly used schemes 1 ei 0 (t ) e03 (t ) cos(3t ) 6 ei 0 (t ) 0 SVM Square wave 38 3D Analysis and Representation ei 0 (t ) 0 39 1 ei 0 (t ) e03 (t ) cos(3t ) 6 Square wave 3D Analysis and Representation Phase and Line Voltages (1) 1 vab (t ) Vdc (v .vab ) 2 1 voa (t ) Vdc (v .voa ) 2 vab va vb vˆa vˆb 1 1 voa (vab vca ) (2vˆa vˆb vˆc ) 3 3 Phase Leg Phase voltages Line Voltages 40 13 2 1 a 1 1 T 1 0 T 13 1 0 b 2 1 T 1 1 T 13 1 1 c 1 2 T 0 1 T 3D Analysis and Representation Phase and Line Voltages (2) voa (t) va (t) t 41 Direction of vab Direction of voa 3D Analysis and Representation Phase and Line Voltages (3) e03 (t ) (1/ 4) cos(30t ) e0 H (t ) (1/ 10) cos(50t ) voa (t) va (t) t 42 3D Analysis and Representation Maximum non distorting range Radius of circle is m if it is measured in the Space Vector Domain. There is a scaling factor in W vd vd 1 ei 0 (t ) e03 (t ) cos(3t ) 6 43 Bdq 0 BS vd vˆd vˆa 0.5vˆb 0.5vˆc mmax ND Bdq 0 vd 1 BS 2 2 2 1.15 3 3 3 2 3D Analysis and Representation vi Di (1 Di ) 1 2 Di 1 If fswitch>>ffund then 44 , i a, b, c vi vi 3D Analysis and Representation t 45 3D Analysis and Representation Sector II Sector I T7 T6 T4 T0 T7 T6 Sector III T 2 T0 T7 T 3 Qa Qb Qb Qb Qa Qc Qc Qc Qa T2 T0 T7 DminTs T (1 D )T 0 max s Ti ( Dmax Dmed )Ts , (i 1, 2, 4) T j ( Dmed Dmin )Ts , (j 3, 5, 6) 46 Analysis of SVM (1) T7=T0 Dmin=1-Dmax vmax +vmin=0 S is sector e (t ) e (t ) 1 v (t ) 0 0 3 med 2 dependent (considering e0-H(t)=0) SVM tends to approximate the trajectory of a square wave, but adds 3rd harmonic and higher order triplen harmonics No difference in sequence compared to other schemes 47 Analysis of SVM (2) t Fundamental and zero-sequence signal 48 t Modulating signals Analysis of SVM (3) Control Time Domain Space Vector Domain Output Time Domain Digital implementation related to sampling method selected, not to the modulation function used. 1 ei 0 (t ) cos(3t ) vs. SVM 6 mmax 49 2 1.15 3