1

Using Grid-Forming Inverters to Improve Transient Stability of Dynamic Distribution Systems

Wei DuSenior Research EngineerElectricity & Infrastructure GroupPacific Northwest National LaboratoryMay 30, 2019, Salt Lake City

This work is funded by the Microgrid R&D program, which is funded by the U.S. Department of Energy’s (DOE) Office of Electricity. The Microgrid R&D Program is managed by Mr. Dan Ton.

2

OUTLINE

Grid-Forming Inverter & Grid-Following Inverter

Inverter Modeling Work at PNNL & Model Validation against CERTS/AEP Microgrid Field Test Results

Transient Stability Study of Large-Scale Distribution Systems with High Penetration of PV Grid-Forming Inverters

3

Grid-Following vs. Grid-Forming

E XL

ω

E∠δ

P,Q

I θ+φ V∠θ

P,Qgrid

I∠(θ+φ)

+ Current Control (PLL+ Current loop)+ Control P & Q

- Do not control voltage & frequency- ‘Negative Load’ to Power System- Reduce System Inertia

+ Voltage & Frequency Control+ Autonomous Load Tracking

- No Direct Control of Current- Overload Issues

Grid-Following (Current Source) Grid-Forming (Voltage Source)

Most grid-connected sources (PV, fuel cells) use grid-following control

Grid-forming source is crucial for islanded microgrid operation

Inverter Control

4

• Challenges in simulation: Neither the electromagnetic simulation nor the positive sequence-based transient stability simulation works for large-scale, three-phase, unbalanced distribution systems simulation

• Solutions: PNNL developed an open-source software—GridLAB-D, which can conduct three-phased transient stability simulations for distribution systems. Dynamic models include:• Synchronous generators• Motors• Grid-Forming/Grid-Following inverters• …

How do thousands of grid-forming/grid-following inverters impact the transient stability of large-scale

distribution systems?

5

XLE∠δa Eω

XL

E ω

XLE∠δc E ω

E∠δb Distribution

System NetworkSolution

Grid-Forming Controller

Pi,Qi,Vgi

E,ω

Inverter model and the interface to the distribution system

Modeling of Grid-Forming Inverters in GridLAB-D• Grid-Forming Inverter: Three-phase balanced voltage source behind XL &

Controller• Network Solution: Extension of Current Injection Method, which allows three-

phase power flow calculation

P-f Droop and Overload Mitigation

Q-V Droop

6

-50

0

50

100

150

A1 Power kWB1 Power kWESS Power kWLB4 Power kW

-200

0

200

A1 Current phase a AA1 Current phase b AA1 Current phase c A

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

-200

0

200

Time Seconds

B1 Current phase a AB1 Current phase b AB1 Current phase c A

AEP Test Data 3/18/2016 10:53:09.5829

-50

0

50

100

150

A1 Power kWB1 Power kWESS Power kWLB4 Power kW

-200

0

200

A1 Current phase a AA1 Current phase b AA1 Current phase c A

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

-200

0

200

Time Seconds

B1 Current phase a AB1 Current phase b AB1 Current phase c A

Simulation

Field Test Results

PSCAD Simulation, Time Step = 50 μs

Under-Frequency Load Shedding: Generator vs. Inverter

Feeder A

Feeder B

Inverter-based Source Inverter-based Source

Energy Storage Synchronous Generator

Load Bank 3 Load Bank 4

Load Bank 5

Feeder C

Load Bank 6Grid

Static Switch

A1

ESS B1

A2

• In most microgrids, load shedding is achieved through fast communication. The CERTS microgrid can do under-frequency load shedding.

• By ignoring electromagnetic transients, GridLAB-D is able to conduct transient analysis of large-scale, three-phase, unbalanced distribution systems

CERTS/AEP Microgrid

Time Step = 5 ms

Validation against CERTS/AEP Microgrid Test Results

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

-50

0

50

100

150

P [k

W]

PA1

PB1

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

-50

0

50

Q [k

var]

QA1

QB1

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

0.9

0.95

1

1.05

1.1

Vol

tage

[pu]

Va

Vb

Vc

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

58.5

59

59.5

60

60.5Fr

eque

ncy

[Hz]

fA1

fB1

fLB4

7

An Islanded IEEE 123-Node Test Feeder with High Penetration of PVs

• Synchronous Generators: 3*600 kW• 6 Distributed PV Inverters: 1400 kW• Peak Load: 2150 kW + 490 kvar• PV Penetration: 65% (compared to

peak load)

• Substation voltage is lost due to extreme weather events, three microgrids work as an islandednetworked microgrid

• Contingency: Loss of Generator 3

Simulation on a Modified IEEE 123-Node Test Feeder

Gen PV

8

PV: Grid-Following controlled at MPP

• PV operates at Maximum Power Point• Contingency: Loss of Generator 3• Frequency drops fast due to the low inertia, 12%

Loads are tripped

GridLAB-D Simulation

0 200 400 600 800 1000 1200

Voltage[V]

0

280

560

840

1120

1400

1680

Pow

er[k

W]

dP/dV>0

stableunstable

dP/dV<0

Maximum Power Point

T=25 ° C

G=600W/m 2

Simulation on a Modified IEEE 123-Node Test Feeder

4 6 8 10 12 14

57

58

59

60

61

Freq

uenc

y [H

z]

4 6 8 10 12 14

Time [s]

0.9

0.95

1

1.05

1.1

Vol

tage

[V]

Va-97

Vb-97

Vc-97

4 6 8 10 12 140

500

1000

1500

Act

ive

Pow

er [k

W]

Diesel:1+2

Diesel:3

Total PV

Load shedding

9

• As the penetration of PVs continues increasing, dynamic reserve of PVs has the potential to improve the transient stability

• How do we use this reserve?

33% Reserve

Simulation on a Modified IEEE 123-Node Test Feeder

0 200 400 600 800 1000 1200

Voltage[V]

0

280

560

840

1120

1400

1680

Pow

er[k

W]

dP/dV>0

stableunstable

dP/dV<0

Maximum Power Point

T=25 ° C

G=600W/m 2

10

Simulation on a Modified IEEE 123-Node Test Feeder

33% Reserve

0 200 400 600 800 1000 1200

Voltage[V]

0

280

560

840

1120

1400

1680

Pow

er[k

W]

dP/dV>0

stableunstable

dP/dV<0

Maximum Power Point

T=25 ° C

G=600W/m 2

• PV: Grid-Following with Frequency-Watt• Not only reduces the system inertia, but also results in

oscillations between generators and inverters• Loads are still tripped

4 6 8 10 12 1458

59

60

61

Freq

uenc

y [H

z]

4 6 8 10 12 14

Time [s]

0.9

0.95

1

1.05

1.1

Vol

tage

[V]

Va-97

Vb-97

Vc-97

4 6 8 10 12 140

500

1000

1500

Act

ive

Pow

er [k

W]

Diesel:1+2

Diesel:3

Total PVLoad shedding

Grid-Following with Frequency-Watt

0.05s 4% droop 0.25s

Frequency-Watt Control

11

4 6 8 10 12 1458

59

60

61

Freq

uenc

y [H

z]

4 6 8 10 12 14

Time [s]

0.9

0.95

1

1.05

1.1

Vol

tage

[V]

Va-97

Vb-97

Vc-97

4 6 8 10 12 140

500

1000

1500

Act

ive

Pow

er [k

W]

Diesel:1+2

Diesel:3

Total PVLoad shedding

Grid-Following with Frequency-Watt

4 6 8 10 12 1458

59

60

61

Freq

uenc

y [H

z]

4 6 8 10 12 14

Time [s]

0.9

0.95

1

1.05

1.1

Vol

tage

[V]

Va-97

Vb-97

Vc-97

4 6 8 10 12 140

500

1000

1500

Act

ive

Pow

er [k

W]

Diesel:1+2

Diesel:3

Total PV

Improve frequency control

Simulation on a Modified IEEE 123-Node Test Feeder• Grid-Forming PV with dynamic reserve can improve

frequency stability, load shedding is avoided• Grid-forming inverters response to load changes

instantaneously0 200 400 600 800 1000 1200

Voltage[V]

0

280

560

840

1120

1400

1680

Pow

er[k

W]

dP/dV>0

stableunstable

dP/dV<0

Maximum Power Point

T=25 ° C

G=600W/m 2

Grid-Forming

12

4 6 8 10 12 1458

59

60

61

Freq

uenc

y [H

z]

4 6 8 10 12 14

Time [s]

0.9

0.95

1

1.05

1.1

Vol

tage

[V]

Va-97

Vb-97

Vc-97

4 6 8 10 12 140

500

1000

1500

Act

ive

Pow

er [k

W]

Diesel:1+2

Diesel:3

Total PVLoad shedding

Grid-Following with Frequency-Watt

4 6 8 10 12 1458

59

60

61

Freq

uenc

y [H

z]

4 6 8 10 12 14

Time [s]

0.9

0.95

1

1.05

1.1

Vol

tage

[V]

Va-97

Vb-97

Vc-97

4 6 8 10 12 140

500

1000

1500

Act

ive

Pow

er [k

W]

Diesel:1+2

Diesel:3

Total PV

Mitigate voltage unbalance

Grid-Forming

Simulation on a Modified IEEE 123-Node Test Feeder

• Grid-Forming inverters can improve voltage control, voltage unbalance is mitigated

• All PV inverters become three-phase, balanced voltage source behind XL, system is stiffer

13

Conclusion

• A three-phase, electromechanical modelling approach is proposed to model Grid-Forming/Grid-Following inverters for large-scale distribution system study

• The developed Grid-Forming inverter model in GridLAB-D is validated against CERTS Microgrid test results

• PV Grid-Forming inverters with dynamic reserve can improve Frequency & Voltage stability of distribution systems

Thank you

14

15

Backup Slides

16

Simulation on a Modified IEEE 123-Node Test Feeder

• The voltage profile of the entire distribution feeder is significantly improved

• The three-phase unbalanced characteristic cannot be easily reflected by positive-sequence-based simulation

Grid-FormingGrid-Following

4 4.5 5 5.5 6 6.5 7 7.5 8

Time (s)

0.9

0.92

0.94

0.96

0.98

1

1.02

1.04

1.06

1.08

1.1

Vol

tage

[pu]

4 4.5 5 5.5 6 6.5 7 7.5 8

Time (s)

0.9

0.92

0.94

0.96

0.98

1

1.02

1.04

1.06

1.08

1.1

Vol

tage

[pu]

17

Grid-Following with Frequency-Watt Grid-Forming

Currents Currents

• Grid-Forming inverters provide larger currents than grid-following inverters, because they need to provide reactive power to support the voltage

4 5 6 7 8 9 10 11 12 13 14 15

Time (sec)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Tota

l Cur

rent

[pu]

Ia

Ib

Ic

4 5 6 7 8 9 10 11 12 13 14 15

Time (sec)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Tota

l Cur

rent

[pu]

Ia

Ib

Ic

18

-50

0

50

100

150

A1 Power kWA2 Power kWESS Power kWLB4 Power kW

-200

0

200

A1 Current phase a AA1 Current phase b AA1 Current phase c A

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

-200

0

200

Time Seconds

A2 Current phase a AA2 Current phase b AA2 Current phase c A

AEP Test Data 5/12/2016 09:18:51.8316

-50

0

50

100

150

A1 Power kWA2 Power kWESS Power kWLB4 Power kW

-200

0

200

A1 Current phase a AA1 Current phase b AA1 Current phase c A

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

-200

0

200

Time Seconds

A2 Current phase a AA2 Current phase b AA2 Current phase c A

Simulation

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

-50

0

50

100

150

P [k

W]

PA1

PA2

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

-50

0

50

Q [k

var]

QA1

QA2

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

0.9

0.95

1

1.05

1.1

Vol

tage

[pu]

Va

Vb

Vc

3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8

Time (sec)

59

59.5

60

Freq

uenc

y [H

z]

fA1

fA2

fLB4

Under-Frequency Load Shedding: Inverter vs. Inverter

Feeder A

Feeder B

Inverter-based Source Inverter-based Source

Energy Storage Synchronous Generator

Load Bank 3 Load Bank 4

Load Bank 5

Feeder C

Load Bank 6Grid

Static Switch

A1

ESS B1

A2

Validation against CERTS/AEP Microgrid Test Results

CERTS/AEP Microgrid

Time Step = 5 ms

By ignoring electromagnetic transients, GridLAB-D is able to conduct transient analysis of large-scale, three-phase, unbalanced distribution systems

PSCAD Simulation, Time Step = 50 μs

Field Test Results

19

Tecogen’s response to fault

20

2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45 2.51.01

1.015

1.02

1.025

volta

ge [p

u]

ugd-pu-A

2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45 2.50.4

0.6

0.8

1

curr

ent [

pu]

igd-pu-A

igd-ref-A

2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45 2.5

Time (sec)

-0.4

-0.2

0

curr

ent [

pu]

igq-pu-A

igq-ref-A

Current Loop PI gains: kp=0.05, ki=5, but with feedforward term F=0.3

CoordinateTransformation

ugqi

δPLLi

kpPLL

∆ωi1/s

δPLLi Ugi∠δgi

ω0

fPLLi

2π

+

+

kiPLL/s

++

+-

+-

kpc+

-

+

++

+

igdi_ref

igdi

igqi_ref

igqi

ωL

ωL

ugdi

ugqi

edi

eqi

kic/s+

+

kpc

kic/s+

+

Pref changes from 0.5 pu to 1 pu at 2.1 s

Grid-Following Inverters

21

Grid-Forming Inverter & Droop Control

• Grid-Forming inverter: voltage source behind a coupling reactance XL

A well designed XL is important for controller design: P∝δ, Q∝E

• Droop Control: make multiple inverters work together in a microgrid P vs. f droop ensure sources are synchronized Q vs. V droop avoids large circulating reactive power between sources

P vs. f droop Q vs. V droop

Inverter model and the interface to the distribution system

Phase-Lock Loop

Current Loop

22

Modeling of Grid-Following Inverters in GridLAB-D• Grid-Following inverters are usually modeled as controllable PQ node in

traditional transient stability analysis software, PLL and current loops are ignored• In GridLAB-D, they are modeled as controllable voltage sources, and detailed

PLL and current loop are modeled

23

Synchronous Generator• Why synchronous generators need inertia?

• A synchronous generator behaves as a voltage source behind Xd’’ in transient • Pe responses to load changes instantaneously• Pm from prime mover changes slowly• The unbalance between Pe and Pm results in the change of speed

Engine

Governor

ValveGenerator

Pm Pe

Load PL

Fuel

Speed ω

E∠δ Xd

’’

2 m edH P Pdtω= −

Pe responses fastPm responses slow