impact of dc voltage droop control structures on dc...

1
5. Results: GSCs use different CS 4. Results: Both GSCs use same CS Impact of DC Voltage Droop Control Structures on DC Oscillations in a MTDC F. Thams Technical University of Denmark (DTU) Center for Electric Power and Energy Email: [email protected] CS3 (! "# −% "# ) achieves significantly lower singular values at costs of higher control action for & > 10 *+" , CS1 (% "# −! "# ) fulfils the maximum gain requirement in steady state, but not for the whole frequency range – but same control action as remaining CSs CS2,4-8 do not even fulfil maximum gain requirement Ø If power spectrum of wind farm includes frequency of resonance peaks: Δ% "# > 10%% "#,0 or ! #102 > 110%! 3 Introduction: Offshore wind farms raised the interest in HVDC transmission based on voltage source converter (VSC-HVDC). It is the most appropriate technology for a multi-terminal HVDC (MTDC) grid facilitating the change of power direction and enabling parallel connections. The control of a MTDC grid is still an open research topic. For all CSs except CS3 (! "# −% "# ) : CS at GSC closer to WFC dominates SV representation (Fig. 7a) CS3 (! "# −% "# ) leads to significant lower SV in steady state, but does not improve damping of resonance peaks (Fig. 7b) Better damping of CS3 (! "# −% "# ) comes at the cost of increased control action (Fig. 8) Combination of any CS with CS3 increases control action already in steady state first loop second loop PI + - 1 k droop,i I dc V dc V dc I dc + + - i q Control structure 1 first loop second loop - + 1 k droop,i V dc V dc + + i q i c,q Control structure 2 first loop second loop PI + - k droop,i I dc V dc V dc I dc + + - i q Control structure 3 first loop second loop PI - - + i c,q V dc V dc i c,q + + i q k droop,i Control structure 4 first loop second loop + - i q V dc V dc + P dc - PI P dc + 1 k droop,p Control structure 5 first loop second loop - + i q V dc V dc + P ac - PI P ac + 1 k droop,p Control structure 6 first loop second loop - + i q V dc V dc + P dc - PI P dc + k droop,p Control structure 7 i q V dc V dc + PI - k droop,p P ac - + + P ac first loop second loop Control structure 8 2. Model: 3 terminal grid with two grid side converters (GSCs) GSCs connected by LC filters to two different equivalent ac grids modeled as Thévenin equivalent. Fig. 2.: MTDC grid Fig. 3.: Control structure of multi-terminal HVDC grid 3. Methodology: Performance of the CSs is analyzed by the singular value representation (SV) of the system transfer function matrix 4 5 (7 8 9&) = < 5 (7 8 = 9& 7 8 9& ) The maximum SV indicates the maximum amplification of the corresponding inputs (> 8? ) by the system seen from a specific output (Δ% "# ) . The max. gain of SV for max. voltage error of 10% of the nominal value at each terminal is calculated as [2]: The research leading to these results was co-funded by the European Union’s Programme for Research, Technological Development and Demonstration under the grant agreement no. 612748 Fig. 4.: Singular value representation of 7 8 (9&) (dc voltage error - wind power input) for the case where both GSCs use the same CS a) Control structure 1 ( % "# −! "# ) and control structure 2 ( % "# −! +# ) b) Control structure 3 ( % "# −! "# ) and control structure 2 ( % "# −! +# ) Fig. 7.: Singular value representation of 7 8 (9&) (dc voltage error - wind power input) for the case where both GSCs use different CS Fig. 6.: Participation factor analysis of the eigenvalues of 7 8 (9&) with damping frequency corresponding to the resonance peak for the case where both GSCs use CS1 ( % "# −! "# ) Fig. 8.: Singular value representation of A 8 B5C (D) (current loop references - wind power input) for a combination of CS2 and CS3 DC grid Δu Δw = ΔP gs,n GSVSCs Droop Controller Current Controller AC Voltage Controller Δz = ΔV dc,n ΔP wf,n ΔV dc,n Δy n Δy dc ΔV dc,n Δy n Δy gsvsc = Δi ac,n ΔP ac,n Δr + - ΔU n ΔU n + - Δu iq Δe i q c,n i dc,n Vdc Vdc3 Idc13 Idc23 States 0 20 40 Participation factor (%) Damp. ratio:2.56%, Damp.F.:269.7 Fig. 5.: Singular value representation of A 8 B5C (9&) (current controller references - wind power input) for the case where both GSCs use the same CS – results of CS1 comparable to CS2,4-8 Conventional current controller in SRF tuned by Internal Model Control (IMC) technique with a settling time of 10ms Saturation limits included DC lines are modeled using a single π-equivalent model Closed loop transfer function matrices considering ac and dc dynamics: ΔE D = ΔF D − ΔG D = 7 8 D7 * D ΔH D ΔG D = ΔI 5C = (A 8 B5C (D)A * B5C (D)) ΔH D ΔG D = Origin of resonance peaks determined by participation factor analysis: related to dc values (voltage / current) at the wind farm 4J 7 8 0 E 0 L H 0 L = 20 log QR E STQ +E STL > V = −81.85 Z[ Max. allowed current flowing through the converter = 110% of the nominal current value (represents the control action) [2]: 4 A 8 B5C 0 ≤ −109.74 Z[ 5. Conclusion All CSs need damping of dc oscillations / control actions in order to comply with boundaries over whole frequency range Closeness of GSCs to uncontrolled node (wind farm) needs to be considered in design of MTDC Current damping solutions focus on damping at GSC Ø In this work, we show the need for development of appropriate damping of dc oscillations at wind farm converter [1] T. K. Vrana, “System Design and Balancing Control of the North Sea Super Grid,” Ph.D. dissertation, NTNU, Trondheim, 2013. [2] E. Prieto-Araujo, A. Egea-Alvarez, S. F. Fekriasl, and O. Gomis-Bellmunt, “DC voltage droop control design for multi-terminal HVDC systems considering AC and DC grid dynamics,” IEEE Transactions on Power Delivery, vol. 31, no. 2, pp. 575 – 585, 2015. Fig. 1.: Analysed dc voltage droop control structures E. Prieto-Araujo Technical University of Catalonia CITCEA, Barcelona, Spain Email: [email protected] R. Eriksson Swedish National Grid Markets and System Development Email: [email protected] S. Chatzivasileiadis Technical University of Denmark (DTU) Center for Electric Power and Energy Email: [email protected] 1. Control of MTDC grid: Preferable to have a decentralized control structure with multiple units actively participate in the dc voltage control One of the preferred control structures (CSs) is dc voltage droop control. However, in technical literature several alternative droop control schemes have been discussed. They can be categorized by eight different types of structures Comparable tuning for all control structures with a settling time of 100ms achieved by using robust control techniques Comparable power / current based droop gains [1]: _ "*11`,5 = % "# 1 _ "*11`,` −! "#/C Power based Current based

Upload: others

Post on 05-Oct-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Impact of DC Voltage Droop Control Structures on DC ...chatziva.com/presentations/Thams_PESGM2016_poster.pdf · F. Thams Technical University of Denmark (DTU) Center for Electric

5. Results: GSCs use different CS4. Results: Both GSCs use same CS

Impact of DC Voltage Droop Control

Structures on DC Oscillations in a MTDCF. Thams Technical University of Denmark (DTU)Center for Electric Power and Energy Email: [email protected]

• CS3 (!"# − %"#) achieves significantly lower singular values at costs of higher control action for & > 10

*+"

,

• CS1 (%"# − !"#) fulfils the maximum gain requirement in steady state, but not for the whole frequency range – but same control action as remaining CSs

• CS2,4-8 do not even fulfil maximum gain requirement Ø If power spectrum of wind farm includes frequency of

resonance peaks: Δ%"# > 10%%"#,0 or !#102 > 110%!3

Introduction: Offshore wind farms raised the interest in HVDC transmission based on voltage source

converter (VSC-HVDC). It is the most appropriate technology for a multi-terminal HVDC (MTDC) grid

facilitating the change of power direction and enabling parallel connections. The control of a MTDC grid

is still an open research topic.

• For all CSs except CS3 (!"# − %"#) : CS at GSC closer to WFC dominates SV representation (Fig. 7a)

• CS3 (!"# − %"#) leads to significant lower SV in steady state, but does not improve damping of resonance peaks (Fig. 7b)

• Better damping of CS3 (!"# − %"#) comes at the cost of increased control action (Fig. 8)

• Combination of any CS with CS3 increases control action already in steady state

first loop second loop

PI+

-

1

kdroop,i

Idc

V⇤dc

Vdc

I⇤dc

++- i⇤q

Control structure 1

first loop second loop

- + 1

kdroop,i

Vdc

V⇤dc

++ i⇤q

i⇤c,q

Control structure 2

first loop second loop

PI+

-kdroop,iIdc

Vdc

V⇤dcI⇤dc

++- i⇤q

Control structure 3

first loop second loop

PI-

- +ic,q

Vdc

V⇤dci⇤c,q

++ i⇤qkdroop,i

Control structure 4

first loop second loop

+- i⇤qVdc

V⇤dc

+

P⇤dc

-PI

Pdc

+1

kdroop,p

Control structure 5

first loop second loop

- + i⇤qVdc

V⇤dc

+

P⇤ac

-PI

Pac

+1

kdroop,p

Control structure 6

first loop second loop

- + i⇤q

V⇤dc

Vdc

+Pdc -

PI

P⇤dc

+kdroop,p

Control structure 7

i⇤q

V⇤dc

Vdc

+PI

-kdroop,pPac

-+

+

P⇤ac

first loop second loop

Control structure 8

2. Model: • 3 terminal grid with two grid side converters (GSCs) • GSCs connected by LC filters to two different equivalent

ac grids modeled as Thévenin equivalent.

Fig. 2.: MTDC grid

Fig. 3.: Control structure of multi-terminal HVDC grid

3. Methodology:• Performance of the CSs is analyzed by the singular value

representation (SV) of the system transfer function matrix45(78 9&) = <5(78

= 9& 78 9& )• The maximum SV indicates the maximum amplification of

the corresponding inputs (>8?)by the system seen from a specific output (Δ%"#).

• The max. gain of SV for max. voltage error of 10% of the nominal value at each terminal is calculated as [2]:

The research leading to these results was co-funded by the European Union’s Programme for Research, Technological Development and Demonstration under the grant agreement no. 612748

Fig. 4.: Singular value representation of 78(9&) (dc voltage error - wind power input) for the case where both GSCs use the same CS a) Control structure 1 (%"# − !"#) and control structure 2 (%"# − !+#)

b) Control structure 3 (%"# − !"#) and control structure 2 (%"# − !+#)

Fig. 7.: Singular value representation of 78(9&) (dc voltage error - wind power input) for the case where both GSCs use different CS

Fig. 6.: Participation factor analysis of the eigenvalues of 78(9&) with damping frequency corresponding to the resonance peak for the case where both GSCs use CS1 (%"# − !"#)

Fig. 8.: Singular value representation of A8B5C(D) (current loop

references - wind power input) for a combination of CS2 and CS3

DC grid �u�w = �Pgs,n

GSVSCs Droop Controller

Current Controller

AC Voltage Controller

�z = �Vdc,n

�Pwf,n

�Vdc,n

�yn

�ydc

�V ⇤

dc,n

�y⇤n

�ygsvsc =

�iac,n�Pac,n

��r

+�

�Un

�U⇤n+

� �uiq

�e

iq⇤c,n

id⇤c,n

Vdc Vdc3 Idc13 Idc23

States

0

20

40

Part

icip

atio

n fact

or

(%) Damp. ratio:2.56%, Damp.F.:269.7

Fig. 5.: Singular value representation of A8B5C(9&) (current controller

references - wind power input) for the case where both GSCs use the same CS – results of CS1 comparable to CS2,4-8

• Conventional current controller in SRF tuned by Internal Model Control (IMC) technique with a settling time of 10ms

• Saturation limits included• DC lines are modeled using a single π-equivalent model• Closed loop transfer function matrices considering ac and dc

dynamics: ΔE D = ΔF D − ΔG D= 78 D 7* D ΔH D ΔG D

=

ΔI5C = (A8B5C(D)A*

B5C(D)) ΔH D ΔG D

=

• Origin of resonance peaks determined by participation factor analysis: related to dc values (voltage / current) at the wind farm

4J 78 0 ≤E 0 L

H 0 L= 20 logQR

ESTQ + ESTL

>V= −81.85Z[

• Max. allowed current flowing through the converter = 110% of the nominal current value (represents the control action) [2]:

4 A8B5C

0 ≤ −109.74Z[

5. Conclusion• All CSs need damping of dc oscillations / control actions in

order to comply with boundaries over whole frequency range• Closeness of GSCs to uncontrolled node (wind farm) needs to

be considered in design of MTDC• Current damping solutions focus on damping at GSCØ In this work, we show the need for development of appropriate

damping of dc oscillations at wind farm converter

[1] T. K. Vrana, “System Design and Balancing Control of the North Sea Super Grid,” Ph.D. dissertation, NTNU, Trondheim, 2013. [2] E. Prieto-Araujo, A. Egea-Alvarez, S. F. Fekriasl, and O. Gomis-Bellmunt, “DC voltagedroop control design for multi-terminal HVDC systems considering AC and DC griddynamics,” IEEE Transactions on Power Delivery, vol. 31, no. 2, pp. 575 – 585, 2015.

Fig. 1.: Analysed dc voltage droop control structures

E. Prieto-AraujoTechnical University of Catalonia CITCEA, Barcelona, SpainEmail: [email protected]

R. ErikssonSwedish National GridMarkets and System DevelopmentEmail: [email protected]

S. Chatzivasileiadis Technical University of Denmark (DTU)Center for Electric Power and Energy Email: [email protected]

1. Control of MTDC grid: • Preferable to have a decentralized control structure with multiple units actively

participate in the dc voltage control • One of the preferred control structures (CSs) is dc voltage droop control.• However, in technical literature several alternative droop control schemes have

been discussed. • They can be categorized by eight different types of structures• Comparable tuning for all control structures with a settling time of 100ms

achieved by using robust control techniques• Comparable power / current based droop gains [1]:_"*11`,5 =

%"#∗

1_"*11`,`

− !"#/C∗

Pow

er b

ased

C

urre

nt b

ased