Control of Low-Inertia Power Systems: Naive & Foundational - - PowerPoint PPT Presentation

control of low inertia power systems naive foundational
SMART_READER_LITE
LIVE PREVIEW

Control of Low-Inertia Power Systems: Naive & Foundational - - PowerPoint PPT Presentation

Acknowledgements Control of Low-Inertia Power Systems: Naive & Foundational Approaches ! ! ! ! INCITE Seminar @ Universitat Polit` ecnica de Catalunya Florian D orfler B.K. Poolla C. Arghir T. Jouini P. L utolf D. Gro S.


slide-1
SLIDE 1

Control of Low-Inertia Power Systems: Naive & Foundational Approaches

INCITE Seminar @ Universitat Polit` ecnica de Catalunya

Florian D¨

  • rfler

Acknowledgements

B.K. Poolla

  • C. Arghir
  • T. Jouini
  • P. L¨

utolf

  • D. Groß
  • S. Bolognani
  • S. Curi
  • M. Colombino

! ! !

! 2 / 56

What do we see here?

Hz *10 sec BEWAG UCTE

3 / 56

Frequency of West Berlin when re-connecting to Europe

Source: Energie-Museum Berlin

Hz *10 sec BEWAG UCTE

December 7, 1994

before re-connection: islanded operation based on batteries & single boiler afterwards connected to European grid based on synchronous generation

4 / 56

slide-2
SLIDE 2

Essentially, the pre/post West Berlin curves date back to. . .

Fact: all of AC power systems built around synchronous machines ! At the heart of it is the generator swing equation: M d dt ω(t) = Pgeneration(t) − Pdemand(t) change of kinetic energy = instantaneous power balance τm θ, ω τ M

demand generation

5 / 56

Operation centered around bulk synchronous generation

49.88 49.89 49.90 49.91 49.92 49.93 49.94 49.95 49.96 49.97 49.98 49.99 50.00 50.01 50.02 16:45:00 16:50:00 16:55:00 17:00:00 17:05:00 17:10:00 17:15:00

  • 8. Dezember 2004

f [Hz] 49.88 49.89 49.90 49.91 49.92 49.93 49.94 49.95 49.96 49.97 49.98 49.99 50.00 50.01 50.02 16:45:00 16:50:00 16:55:00 17:00:00 17:05:00 17:10:00 17:15:00

  • 8. Dezember 2004

f [Hz]

Frequency Athens f - Setpoint Frequency Mettlen, Switzerland PP - Outage PS Oscillation

Source: W. Sattinger, Swissgrid Primary Control Secondary Control Tertiary Control Oscillation/Control Mechanical Inertia

6 / 56

Renewable/distributed/non-rotational generation on the rise

synchronous generator new workhorse scaling

new primary sources

location & distributed implementation focus today on non-rotational generation

7 / 56

The foundation of today’s power system

Synchronous machines with rotational inertia M d dt ω ≈ Pgeneration − Pdemand Today’s grid operation heavily relies on

1 robust stabilization of frequency and voltage by generator controls 2 self-synchronization of machines through the grid 3 kinetic energy 1

2Mω2 as safeguard against disturbances

We are replacing this solid foundation with . . .

8 / 56

slide-3
SLIDE 3

Tomorrow’s clean and sustainable power system

Non-synchronous generation connected via power electronics As of today, power electronic converters

1 lack robust control for voltage and frequency 2 do not inherently synchronize through the grid 3 provide almost no energy storage

What could possibly go wrong ?

9 / 56

The concerns are not hypothetical: South Australia event

UPDATE REPORT ! BLACK SYSTEM EVENT IN SOUTH AUSTRALIA ON 28 SEPTEMBER 2016

AN UPDATE TO THE PRELIMINARY OPERATING INCIDENT REPORT FOR THE NATIONAL ELECTRICITY MARKET. DATA ANALYSIS AS AT 5.00 PM TUESDAY 11 OCTOBER 2016.

my conclusion from official report: blue low-inertia area 5 was not resilient; conventional system would have survived

10 / 56

Black System Event in South Australia (Sep 2016)

Key events1

1 intermittent voltage disturbances due to line faults 2 loss of synchronism between SA and remainder of the grid 3 SA islanded: frequency collapse in a quarter of a second

“Nine of the 13 wind farms online did not ride through the six voltage disturbances experienced during the event.”

1AEMO: Update Report - Black System Event in South Australia on 28 September 2016 11 / 56

Low inertia issues have been broadly recognized

by TSOs, device manufacturers, academia, funding agencies, etc.

MIGRATE project: Massive InteGRATion of power Electronic devices

ERCOT is recommending the transition to the following five AS products plus one additional AS that would be used during some transition period:

  • 1. Synchronous Inertial Response Service (SIR),
  • 2. Fast Frequency Response Service (FFR),
  • 3. Primary Frequency Response Service (PFR),
  • 4. Up and Down Regulating Reserve Service (RR), and
  • 5. Contingency Reserve Service (CR).
  • 6. Supplemental Reserve Service (SR) (during transition period)

ERCOT CONCEPT PAPER Future Ancillary Services in ERCOT

PUBLIC

Frequency Stability Evaluation Criteria for the Synchronous Zone

  • f Continental Europe

– Requirements and impacting factors – RG-CE System Protection & Dynamics Sub Group

However, as these sources are fully controllable, a regulation can be added to the inverter to provide “synthetic inertia”. This can also be seen as a short term frequency support. On the other hand, these sources might be quite restricted with respect to the available capacity and possible activation time. The inverters have a very low

  • verload capability compared to synchronous machines.

The relevance of inertia in power systems

Pieter Tielens n, Dirk Van Hertem ELECTA, Department of Electrical Engineering (ESAT), University of Leuven (KU Leuven), Leuven, Belgium and EnergyVille, Genk, Belgium Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/rser

Renewable and Sustainable Energy Reviews

Renewable and Sustainable Energy Reviews 55 (2016) 999–1009

Impact of Low Rotational Inertia on Power System Stability and Operation

Andreas Ulbig, Theodor S. Borsche, Göran Andersson ETH Zurich, Power Systems Laboratory Physikstrasse 3, 8092 Zurich, Switzerland ulbig | borsche | andersson @ eeh.ee.ethz.ch

MIGRATE consortium: green-field approach to control of zero-inertia grids

12 / 56

slide-4
SLIDE 4

Low-inertia issues close to home

# frequency violations in Nordic grid (source: ENTSO-E)

15

Number * 10 5000 10000 15000 20000 25000 30000 Duration [s] Events [-] Months of the year Number * 10 Duration 2001 2002 2003 2004 2006 2005 2007 2008 2009 2010

t eal

same in Switzerland (source: Swissgrid) a day in Ireland (source: F. Emiliano) a year in France (source: RTE)

13 / 56

Obvious insight: loss of inertia & frequency stability

We loose our giant electromechanical low-pass filter: M d dt ω(t) = Pgeneration(t) − Pdemand(t) change of kinetic energy = instantaneous power balance

τm θ, ω τ M

demand generation

5 10 15 20 25 30 35 49 49.2 49.4 49.6 49.8 50

J

Time t [s] f [Hz]

M

14 / 56

Berlin curves before and after re-connecting to Europe

Source: Energie-Museum Berlin islanded Berlin grid

loss of 146 MW loss of 2500 MW

Berlin re-connected to Europe

loss of 1200 MW

15 / 56

  • bvious insights lead to
  • bvious (naive) answers
slide-5
SLIDE 5

Baseline solution: virtual inertia emulation

!""" #$%&'%(#!)&' )& *)+"$ ','#"-'. /)01 23. &)1 2. -%, 2456 5676

!89:;8;<=><? />@=AB: !<;@=>B >< CD!EFGBH;I +><I *JK;@ E;<;@B=>J<

  • JLB88BI@;MB DBNLB@> -J?LBIIB8 %@B<>! "#$%&'# (&)*&+! ,---. B<I "LBO D1 ":F'BBIB<P! "&'./+ (&)*&+! ,---

!"#$%&'"'() %* +$,(-.'() /'-#%(-' .( 0.1$%2$.3- 4-.(2 5.$)6,7 !('$).,

8.".-9 :%(.! "#$%&'# (&)*&+! ,---; :6$<,(,$,<,(, =%%77,! (&)*&+! ,---; ,(3 06>67 ?@ ?9,(3%$>,$! (&)*&+! ,---

Virtual synchronous generators: A survey and new perspectives

Hassan Bevrani a,b,⇑, Toshifumi Ise b, Yushi Miura b

a Dept. of Electrical and Computer Eng., University of Kurdistan, PO Box 416, Sanandaj, Iran b Dept. of Electrical, Electronic and Information Eng., Osaka University, Osaka, Japan

!"#$%&'()*+,-+#'"(./#0*/1(2-33/*04($(5&*0-$1(( 6#+*0&$(7*/8&9+9(:"(!&;0*&:-0+9(<#+*="(20/*$=+( 0/(6;/1$0+9(7/>+*(2";0+%;((

?$-0@&+*(!+1&11+A(!"#$"%&'()))A(B*-#/()*$#C/&;A(*"+,-%'!"#$"%&'()))A($#9(?&11+;(D$1$*$#=+(

!""" #$%&'%(#!)&' )& *)+"$ ','#"-'. /)01 23. &)1 2. -%, 2456

!789:;< "=>?<:;@7 (@7:9@? ':9<:8AB C@9 /'(DE/F( #9<7G=;GG;@7 'BG:8=G

H;8I8; JK>. (<=LI8?? F1 M@@:K. N9<;7 *1 %O<=. %7O98P H1 $@GQ@8. <7O (K9;G N1 M9;AK:

!"#$%&#'$%()*+'",'"%-#,.%/#",012%3#*',#4% 5,)"16'%

7898:%+1*%;'<'*=''4>?%@%58;8A8%$'%A11*?@%!"#$%&'("()"&*'+,,,@%98%/1"'21B%1*$%38%/#<<4.'"C@% !"#$%&'("()"&'+,,,%

M d dt ω(t) = Pgeneration(t)−Pdemand(t) ≈ derivative control on ω(t) ⇒ focus today: where to do it? how to implement it properly? . . . we are not just loosing inertia . . .what else to do ?

16 / 56

Outline

Introduction System Level: Optimal Placement of Virtual Inertia network, disturbances, & performance metrics matter Device Level: Proper Virtual Inertia Emulation Strategy maybe we should not think about frequency and inertia A Foundational Control Approach restart from scratch for low-inertia systems Conclusions

Virtual inertia is becoming a technology and a product

so let’s see how we can make use of it

17 / 56

  • ptimal placement
  • f virtual inertia
slide-6
SLIDE 6

General power system & inertia emulation model

˜ Mis + ˜ Di Tis + 1 . . . . . .

power system model

ω

τm τe iαβ if

Lg Lg Lg iP V Lg

virtual inertia & damping

synchronous machines, governors, loads, transmission, batteries, PLL, …

disturbance inputs performance outputs

(implemented as causal PD)

controlled injections measured frequencies

ω u

(e.g., generator frequencies) (e.g., loss of load/generation) (e.g., at PV, batteries, etc.) (e.g., at AC voltage bus via a PLL) (detailed & linearized)

18 / 56

which metric(s) should

  • ur controller optimize ?

Conventional metrics: spectrum, RoCoF, & total inertia

!"#$%&'$()*(+,-&./+%$/0& 1/$&%2(&'*%*$(&34&5/6($&!-7%("&

5(%($&8#00& &9(:#$&!2#"7;&& <0#=>">$&?($@>A#& ?2(&B+>C($7>%-&/1&& D#+,2(7%($& D#+,2(7%($;&BE& <#+=#=&F#">=>& .2#$0/%%(&3$#+%& 9#%>/+#0&3$>=& 8#$6>,G;&BE& H/*:0#7&8>07/+;&& !(I+&9/$$>7& E-$>#G>&D#0(G#& J07%/"&3$>=& K=>+L*$:2;&BE& .#"ML(00&4//%2;&& N>%(+:&F/+:;&& J+=$(6&O/7,/(& ?2(&B+>C($7>%-&/1& !%$#%2,0-=(& 30#7:/6;&BE& &

RoCoF frequency nadir

source: http://www.think-grid.org

damping ratio

Need for synthetic inertia (SI) for frequency regulation

ENTSO-E guidance document for national implementation for network codes on grid connection

19 / 56

are these suitable metrics ? let’s look at some simulations

slide-7
SLIDE 7

Running example: modified Kundur three-area case study

25 km 10 km 25 km 10 km 25 km 110 km 110 km 110 km 1 2 3 4 5 6 7 8 9 10 11 12 1570 MW 1000 MW 100 Mvar 567 MW 100 Mvar 400 MW 490 MW 611 MW 164 Mvar 1050 MW 284 Mvar 719 MW 133 Mvar 350 MW 69 Mvar 700 MW 208 Mvar 700 MW 293 Mvar 200 Mvar 350 Mvar

added third area to standard case PLLs at all buses for inertia emulation (overall device response time ∼100ms) transformer reactance 0.15 p.u, line impedance (0.0001+0.001i) p.u./km

  • riginal inertia 40s: removed
  • f rotational 28s which can be

re-allocated as virtual inertia added governors & droop control at all generators

20 / 56

Fact: RoCoF, spectrum, & total inertia are poor metrics

  • alloc. 2
  • alloc. 1
  • riginal

Node mi + ˜ mi [s]

1 2 3 4 5 6 7 8 9 10 12 5 10

allocation 2 allocation 1

Real Axis Imaginary Axis

−100 −80 −60 −40 −20 −50 50 100

metrics allocation 1 allocation 2 total inertia 40.85 s 40.85 s damping ratio 0.1190 0.1206 RoCoF 0.8149 Hz/s 0.8135 Hz/s ω nadir

  • 84.8 mHz
  • 65.1 mHz

peak injection 118.38 MW 7.0446 MW control effort 15.581 2.699 comparison for 100 MW load step at bus 7

allocation 2 allocation 1

t [s] ω [mHz]

1 2 3 −80 −60 −40 −20

21 / 56

Re-visiting performance metrics for low-inertia systems

f nominal frequency ROCOF (max rate of change of frequency) frequency nadir energy unbalance restoration time secondary control

System norm quantifying signal amplifications disturbances: impulse (fault), step (loss of unit), white noise (renewables) system

η

performance outputs: integral, peak, ROCOF, restoration time, . . .

22 / 56

Integral-quadratic coherency performance metric

  • ther metrics are poor, hard-to-optimize, & characterize a high (not low) inertia system

∞ x(t)TQ x(t) dt

f nominal frequency

H2 system norm interpretation: system

η y

1 performance output: y = Q1/2x 2 impulsive η (faults) −

→ output energy ∞

0 y(t)T y(t) dt

3 white noise η (renewables) −

→ output variance lim

t→∞ E

  • y(t)T y(t)
  • 23 / 56
slide-8
SLIDE 8

Constraints on control inputs

1 energy constraint:

0 uTR u dt directly captured in H2 framework

2 power constraint: ui = ˜

Mi ˙ ωi + ˜ Di ωi must satisfy ui(t)ℓ∞ ≤ ui

−0.2−0.15−0.1−0.05 0.05 0.1 0.15 −0.01 0.01 Frequency deviation [Hz] RoCoF [Hz/s] 100 101 102 103 104 105

European frequency data (source: RTE)

D 0.5 1 1.5 0.5 1 1.5 2 0.2 Hz ˜ 0.01 Hz s−1 ˜ M

corresponding bounds on gains

⇒ (ωi(t), ˙ ωi(t))1, ( ˜ Di, ˜ Mi)∞ bounded ⇒ ui(t)ℓ∞ bounded

3 budget constraint for finitely many devices:

i ui = const.

24 / 56

(sub)optimize performance and see what we learn

Modified Kundur case study: 3 areas & 12 buses

added governors (droop) at generators & PLLs to obtain frequency for inertia emulation

10 9 5 1 11 12 7 6 3 4 2 8

25 / 56

Test case

inertia emulation control via PLL & batteries:

ui = ˜ Mi ˜ Di

  • xPLL,i

˙ x = Ax + Bu + Gd ui = ˜ Mi ˜ Di

  • xPLL,i

d yperf u xPLL

dynamics: swing equation, droop via governor & turbine, and PLL state: x =

  • generator states , frequencies , governor control , PLL
  • cost penalizes

frequencies, droop control, & inertia emulation effort:

  ω ugov u  

yperf

=   I Kgov  

  • =Q1/2

x +   I  

  • =R1/2

u

26 / 56

slide-9
SLIDE 9

Algorithmic approach to desperate & non-convex problem

structured state-feedback with constraints on gains computation H2 norm, gradient, & projections:

˙ x = Ax + Bu + Gd ui = ˜ Mi ˜ Di

  • xPLL,i

d yperf u xPLL

1 observability and controllability Gramians via Lyapunov equations

(A − BK)⊤P + P(A − BK) + Q + K ⊤RK = 0 (A − BK)L + L(A − BK)⊤ + GG ⊤ = 0

2 H2 norm J =Trace(G ⊤PG) and gradient ∇KJ = 2(RK − B⊤P)L 3 projection on structural & ∞-norm constraint: Π ˜

M, ˜ D[∇KJ ]

⇒ ˜ M and ˜ D can be optimized by first-order methods, IPM, SQP, etc.

27 / 56

Results & insights for the three-area case study

Optimal allocation:

◮ location of inertia &

damping matters

◮ outperforms heuristic

uniform allocation

◮ need penalty on

droop control effort

◮ power constraint

results in ˜ D ≈ 2 ˜ M Fault at bus #4:

◮ strong reduction of

frequency deviation

◮ much less control

effort than heuristic

50 60 50

  • 150

28 / 56

can we make this control design strategy useful ?

Recall: South Australia event

UPDATE REPORT ! BLACK SYSTEM EVENT IN SOUTH AUSTRALIA ON 28 SEPTEMBER 2016

AN UPDATE TO THE PRELIMINARY OPERATING INCIDENT REPORT FOR THE NATIONAL ELECTRICITY MARKET. DATA ANALYSIS AS AT 5.00 PM TUESDAY 11 OCTOBER 2016.

29 / 56

slide-10
SLIDE 10

Control & optimization design scale up to large systems

low-inertia Eastern-Australian grid: removed rotational generation at buses 101, 402, 403 and 502 added controllable power sources with PLLs at 15 buses tractable model for design: linearization of nonlinear model balanced reduction to 140 states

30 / 56

H2-optimal virtual inertia allocation with ℓ∞ constraints

allocation at core area 2 and critical areas 4 & 5 improves performance

  • f low-inertia

& original case post-fault frequencies & control input well-behaved

31 / 56

placement & metrics matter! can we get analytic insights ?

Inertia placement in swing equations

simplified network swing equation model: mi ¨ θi + di ˙ θi = pgen,i − pdem,i generator swing equations pdem,i ≈

j bij (θi − θj)

linearized DC power flow

τm θ, ω τ M

demand generation

η

likelihood of disturbance at #i: ηi ≥ 0 (available from TSO data) H2 performance metric: ∞

  • i,j aij(θi − θj)2 +
  • i si ˙

θ2

i dt

decision variable is inertia: mi ∈ [mi , mi] (additional nonlinearity: enters as m−1

i

in constraints & objective)

32 / 56

slide-11
SLIDE 11

Closed-form results for cost of primary control

recall: primary control di ˙ θi effort was crucial ∞ ˙ θ(t)TD ˙ θ(t) dt

(computations show that insights roughly generalize to other costs)

allocation: the primary control effort H2 optimization reads equivalently as minimize

mi

  • i

ηi mi subject to

  • imi ≤ mbdg

mi ≤ mi ≤ mi key take-away is disturbance matching:

◮ optimal allocation m⋆ i ∝ √ηi or m⋆ i = min{mbdg, mi}

⇒ disturbance profile known from historic data, but rare events are crucial

◮ suggests robust minm maxη allocation to prepare for worst case

⇒ valley-filling solution: η⋆

i /m⋆ i = const. (up to constraints)

33 / 56

Robust min - max allocation for three-area case study

Scenario: fault (impulse) can

  • ccur at any single node

◮ disturbance set

η ∈ {e1 ∪ · · · ∪ e12} ⇒ min/max over convex hull

◮ inertia capacity constraints ◮ robust inertia allocation

  • utperforms heuristic

max-capacity allocation

◮ results become intuitive:

valley-filling property

◮ same for uniform allocation

Cost Original, Robust, and Capacity allocations

1 2 4 5 6 8 9 10 12 node 50 100 150 inertia m m∗ rob m 0.05 0.1 0.15 0.2 0.25 0.3 cost

allocation subject to capacity constraints

Cost Original, Robust, and Uniform allocations

1 2 4 5 6 8 9 10 12 node 10 20 30 40 50 60 70 80 inertia 0.05 0.1 0.15 0.2 0.25 0.3 cost m m∗ rob muni

allocation subject to the budget constraint

34 / 56

Outline

Introduction System Level: Optimal Placement of Virtual Inertia network, disturbances, & performance metrics matter Device Level: Proper Virtual Inertia Emulation Strategy maybe we should not think about frequency and inertia A Foundational Control Approach restart from scratch for low-inertia systems Conclusions

Grid-following inverters

PLL v ˆ ω

ω

P ≈ P

35 / 56

slide-12
SLIDE 12

A stiff grid with grid-following sources . . .

36 / 56

If everyone follows...

36 / 56

we are not just loosing inertia —————————— interestingly, many so-called “virtual inertia” controllers are grid-following design of robust grid-forming mechanisms

slide-13
SLIDE 13

Modeling: signal space in three-phase AC power systems

three-phase AC xa(t)

xb(t) xc(t)

  • =

xa(t + T)

xb(t + T) xc(t + T)

  • periodic with 0 average

1 T

T

0 xi(t)dt = 0

  • π

π 2π −1 1

δ xabc

balanced (nearly true) = A(t)

  • sin(δ(t))

sin(δ(t) − 2π

3 )

sin(δ(t) + 2π

3 )

  • so that

xa(t)+xb(t)+xc(t)=0

  • π

π 2π −1 1

δ xabc

synchronous (desired) =A

  • sin(δ0 + ω0t)

sin(δ0 + ω0t − 2π

3 )

sin(δ0 + ω0t + 2π

3 )

  • const. freq & amp

⇒ const. in rot. frame

  • π

π 2π −1 1

δ xabc

assumption : signals are balanced ⇒ 2d-coordinates x(t) = [xα(t) xβ(t)]

(equivalent representation: complex-valued polar/phasor coordinates)

37 / 56

Averaged power converter model

iload

+ vx iαβ R L ic C

+

− vαβ idc gdc Cdc ix

+

− vdc

DC cap & AC filter equations:

Cdc ˙ vdc = −Gdcvdc + idc − 1 2m⊤iαβ L ˙ iαβ = −Riαβ + 1 2mvdc − vαβ C ˙ vαβ = −iload + iαβ

modulation: vx = 1

2mvdc , ix = 1 2m⊤iαβ

control/dist. inputs: (idc, iload) synchronous generator: mechanical + stator flux + AC cap

˙ θ = ω M ˙ ω = −Dω + τm + i⊤

αβLmif

− sin(θ) cos(θ)

  • Ls ˙

iαβ = −Riαβ − vαβ − ωLmif

  • − sin(θ)

cos(θ)

  • C ˙

vαβ = −iload + iαβ

if θ

38 / 56

Standard power electronics control would continue by

iload − + vx iαβ R L ic C + − vαβ idc gdc Cdc ix + − vdc

reference synthesis (virtual sync gen, droop/inertia, etc.) tracking control (cascaded PIs)

  • 1

3 2 4 4

1 acquiring & processing

  • f AC measurements

2 synthesis of references

(voltage/current/power)

3 track error signals at

converter terminals

4 actuation via modulation

(inner loop) and/or via DC source (outer loop) I guess you can see the problems building up . . .

39 / 56

Challenges in power converter implementations

Virtual synchronous generators: A survey and new perspectives

Hassan Bevrani a,b,⇑, Toshifumi Ise b, Yushi Miura b

a Dept. of Electrical and Computer Eng., University of Kurdistan, PO Box 416, Sanandaj, Iran b Dept. of Electrical, Electronic and Information Eng., Osaka University, Osaka, Japan Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Abstract- The method to investigate the interaction between a Virtual Synchronous Generator (VSG) and a power system is presented here. A VSG is a power-electronics based device that To better study and witness the effects of virtual inertia, the hardware of a real VSG should be tested within a power

  • system. Investigating the interaction between a real VSG and

a power system is not easy as a power system cannot be

Real Time Simulation of a Power System with VSG Hardware in the Loop

Vasileios Karapanos, Sjoerd de Haan, Member, IEEE, Kasper Zwetsloot Faculty of Electrical Engineering, Mathematics and Computer Science Delft University of Technology Delft, the Netherlands E-mails: vkarapanos@gmail.com, v.karapanos@tudelft.nl, s.w.h.dehaan@tudelft.nl

  • 1 delays in measurement acquisition,

signal processing, & actuation

2 accuracy in AC measurements

(averaging over multiple cycles)

3 constraints on currents,

voltages, power, etc.

4 certificates on stability,

robustness, & performance

Frequency Stability Evaluation Criteria for the Synchronous Zone

  • f Continental Europe

– Requirements and impacting factors – RG-CE System Protection & Dynamics Sub Group

However, as these sources are fully controllable, a regulation can be added to the inverter to provide “synthetic inertia”. This can also be seen as a short term frequency support. On the other hand, these sources might be quite restricted with respect to the available capacity and possible activation time. The inverters have a very low

  • verload capability compared to synchronous machines.

let’s do something smarter . . .

40 / 56

slide-14
SLIDE 14

See the similarities & the differences ?

iload

+ vx iαβ R L ic C

+

− vαβ idc Gdc Cdc ix

+

− vdc

DC cap & AC filter equations:

Cdc ˙ vdc = −Gdcvdc + idc − 1 2m⊤iαβ L ˙ iαβ = −Riαβ + 1 2mvdc − vαβ C ˙ vαβ = −iload + iαβ

modulation: vx = 1

2mvdc , ix = 1 2m⊤iαβ

passive: (idc, iload)→(vdc, vαβ) synchronous generator: mechanical + stator flux + AC cap

˙ θ = ω M ˙ ω = −Dω + τm + i⊤

αβLmif

− sin(θ) cos(θ)

  • Ls ˙

iαβ = −Riαβ − vαβ − ωLmif − sin(θ) cos(θ)

  • C ˙

vαβ = −iload + iαβ

if θ

41 / 56

Model matching (= emulation) as inner control loop

iload

+ vx iαβ R L ic C

+

− vαβ idc Gdc Cdc ix

+

− vdc

DC cap & AC filter equations:

Cdc ˙ vdc = −Gdcvdc + idc − 1 2m⊤iαβ L ˙ iαβ = −Riαβ + 1 2mvdc − vαβ C ˙ vαβ = −iload + iαβ

matching control: ˙

θ = Km·vdc , m = ˆ m·

  • − sin(θ)

cos(θ)

  • with Km, ˆ

m > 0 ⇒ equivalent inertia M = Cdc

K 2

m , imbalance signal ω = Km · vdc, etc.

⇒ pros: uses physical storage, uses DC measurements, & remains passive

42 / 56

Further properties of machine matching control

1 base for outer loops

⇒ idc = PD(vdc) gives virtual inertia & damping

2 reformulation of

m = ˆ m · − sin(θ) cos(θ)

  • as adaptive oscillator:

˙ m = Km vdc · 1 −1

  • m

iload

+ vx iαβ R L C

+

− vαβ idc Gdc + Kp,ic Cdc + Kd,ic ix

+

− vdc

Cdc ˙ vdc = −Gdcvdc + i∗

dc − 1

2m⊤iαβ C ˙ vαβ = −iload + iαβ L ˙ iαβ = −Riαβ + 1 2mvdc − vαβ ˙ ξ = ω ·

  • 1

−1

  • ξ

Km ˆ m m vdc (idc, iload) (vdc, vαβ) inverter modulation ω ξ

43 / 56

Summary: bottlenecks to inertia emulation

power system model on grid level: M d dt ω = Pgeneration − Pdemand

τm θ, ω τ M

demand generation

inertia emulation on device level:

iload − + vx iαβ R L ic C + − vαβ idc gdc Cdc ix + − vdc

reference synthesis (virtual sync gen, droop/inertia, etc.) tracking control (cascaded PIs)

  • I/O mismatch: none of the converter inputs or outputs are present in

the swing-equation, e.g., frequency is not a state in the converter inertia emulation ` a la PD problematic both in theory & practice ⇒ maybe matching control ˙ m = Km vdc · 1 −1

  • m was quite clever ?

44 / 56

slide-15
SLIDE 15

Outline

Introduction System Level: Optimal Placement of Virtual Inertia network, disturbances, & performance metrics matter Device Level: Proper Virtual Inertia Emulation Strategy maybe we should not think about frequency and inertia A Foundational Control Approach restart from scratch for low-inertia systems Conclusions

Low-inertia power system model from first principles

vdc idc m iI v LI τm θ, ω vf v if τe is Lθ v iT LT C G Gq C v Cdc RT iI M rf rs rs Gdc RI

◮ balanced three-phase system

(α, β) coordinates

◮ synchronous machines

first principle, 5th order

◮ DC/AC inverters

averaged-switched

◮ nonlinear loads G(v) ◮ voltage bus charge dynamics ◮ dynamic transmission lines: Π-model

Port-Hamiltonian model

˙ x =

  • J(x, u)−R(x)
  • ∇H(x)+g(x)u

nonlinear & large, but insightful

45 / 56

Desired steady-state locus & control specifications

vdc idc m iI v LI τm θ, ω vf v if τe is Lθ v iT LT C G Gq C v Cdc RT iI M rf rs rs Gdc RI

steady-state specifications for nonlinear system: synchronous frequency constant amplitude three-phase balanced

L

˙

zαβ

y a constant: R C

L

n n

  • AC quantities v, is, iI, iT:

˙ zαβ = ω0 · −1 1

  • zαβ

rotor angles: ˙ θ = ω0 DC quantities vdc, vf , ω: ˙ z = 0 desired dynamics: ˙ x = fdes(x, ω0) controls idc, m, τm, if to be found

46 / 56

Proving the obvious (?)

steady-state locus: physics & desired closed-loop vector field coincide (point-wise in time) on set S :=

  • (x, u, ω0) : fphys(x, u) = fdes(x, ω0)
  • control-invariance: steady-state operation

(x, u, ω0) ∈ S for all time if and only if

L

˙

zαβ

y a constant: R C

L

n n

  • 1 synchronous frequency ω0 is constant

2 network satisfies power flow equations with impedances R + ω0JL 3 at each generator: constant torque τm & excitation if 4 at each inverter: constant DC current idc & inverter duty cycle with

constant amplitude & synchronous frequency: ˙ m = ω0 · −1 1

  • m

⇒ internal models & feedforward input-to-steady-state map

47 / 56

slide-16
SLIDE 16

Reduction to a tractable model for synthesis

internal oscillator model for inverter duty cycle with inputs ωm, ˆ m ˙ θI = ωm, m = ˆ m − sin(θ) cos(θ)

  • model reduction steps

1

rotating coordinate frame with synchronous frequency ω0 ⇒ time scales of AC quantities scaled by 1/ω0

2

DC/AC time-scale separation via singular perturbation (ǫ → 0) slow DC variables: xr = (θ, ω, if , θI, vdc), ˙ xr = fz(xr, zα,β, u) fast AC variables: zα,β = (is, iI, v, iT), ǫ ˙ zα,β = fα,β(xr, zα,β, u)

3

reformulation via relative angles δ with respect to synchronous motion

48 / 56

Insights from reduced model: vdc ∝ power imbalance

nonlinear reduced order model in rotating frame: ˙ θ = ω M ˙ ω = −Dω + τm − τe(xr, u) Lf ˙ if = −Rf if + vf − vEMF(xr, u) ˙ θI = ωm Cdc ˙ vdc = −Gdcvdc + idc − isw(xr, u) interconnection via τe, isw, vEMF analogies: suggest matching control: ωm ∼ vdc generator inverter interpretation

1 2Mω2 1 2Cdcv2 dc

energy stored in device τm idc energy supply τe isw energy flow to grid ω vdc power imbalance

49 / 56

Completing the control design

Thus far:

1 desired steady-state locus requires internal oscillator model

˙ θI = ωm, m = ˆ m − sin(θ) cos(θ)

  • 2 converter/generator analogies suggest model matching control

ωm = Km · (vdc − v∗

dc)

Remaining steps:

3 performance requires design of structured & optimal MIMO control 50 / 56

Decentralized MIMO control architecture

˙ x = Ax + Bu + Gd τm vf

  • =

Kdroop KPSS KAVR ω v

 ωm idc ˆ m   =   Km KI,1 Kdroop KI,2 KPSS KAVR  

  • vdc

v

  • d

y ugen xgen uinv xinv

states x = (δ, ω, if , vdc, v) & output y = (ω, vdc, v) included measurement devices for AC voltage magnitude v H2-optimal tuning of decentralized MIMO converter controller

51 / 56

slide-17
SLIDE 17

Illustrative conceptual example

test case: generator & inverter impedance load 10% load increase at t =0 no inverter control: ωm and idc constant power imbalance: ωG, vdc governor stabilizes ωG controlled inverter: reduced peak in ωG vdc stabilized via idc ωm and ωG synchronize

10 MW 12 MW 3 MW

Generator Inverter Generator Inverter

t [sec] iDC [kA] t [sec] iDC [kA] t [sec] vDC [kV] t [sec] vDC [kV] t [sec] ω [Hz] t [sec] ω [Hz] inverter control active inverter control inactive

5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 8.5 9 9.5 10 10.5 −0.02 −0.01 8.5 9 9.5 10 10.5 −0.02 −0.01 0.6 0.8 1 1.2 0.6 0.8 1 1.2 52 / 56

Modified Kundur two-area case study

G G AC/DC AC/DC 25 km 10 km

P =967 MW Q=100 MVAr Q=−200 MVAr

110 km 110 km 10 km 25 km

P =1767 MW Q=100 MVAr Q=−350 MVAr

2 4 1 5 6 7 8 9 10 11 3

P =700 MW V =1.03 p.u. P =700 MW V =1.01 p.u. P =719 MW V =1.03 p.u. P =700 MW V =1.01 p.u.

400 MW Area 1 Area 2

standard line parameters and power flows synchronous machines with droop control and voltage regulator two synchronous machines replaced by DC/AC inverters all dirt effects modeled: saturation, nonlinearities, etc. simulation scenarios: load step (×2) & outage of synchronous machine

53 / 56

Scenario: load step & different converter controllers

2 4 6 8 10

  • 2
  • 1.5
  • 1
  • 0.5

2 4 6 8 10 0.2 0.4 0.6 0.8 1 1.2

feedforward control (power point tracking)

2 4 6 8 10

  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 2 4 6 8 10 0.5 0.6 0.7 0.8 0.9 1 1.1

matching control & un-tuned MIMO gains

2 4 6 8 10

  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 2 4 6 8 10 0.5 0.6 0.7 0.8 0.9 1 1.1

H2-optimal control (all gains tuned)

54 / 56

Scenario: outage of a synchronous machine

2 4 6 8 10

  • 0.6
  • 0.4
  • 0.2

2 4 6 8 10 0.8 0.85 0.9 0.95 1 1.05 1.1

feedforward control (power point tracking)

2 4 6 8 10

  • 0.6
  • 0.4
  • 0.2

2 4 6 8 10 0.8 0.85 0.9 0.95 1 1.05 1.1

matching control & un-tuned MIMO gains

2 4 6 8 10

  • 0.6
  • 0.4
  • 0.2

2 4 6 8 10 0.8 0.85 0.9 0.95 1 1.05 1.1

H2-optimal control (all gains tuned)

55 / 56

slide-18
SLIDE 18

conclusions

Conclusions on virtual inertia emulation

Where to do it?

1 H2-optimal (non-convex) allocation 2 numerical approach via gradient computation 3 closed-form results for cost of primary control

How to do it?

1 down-sides of naive inertia emulation 2 machine matching reveals power imbalance in DC voltage

What else to do?

1 first-principle low-inertia system model 2 nonlinear steady-state control specifications 3 reduction to tractable model for synthesis 4 first promising controller synthesis:

internal model + matching + H2 performance loops

56 / 56