ECE 5984: Power Distribution System Analysis

Lecture 11: DistFlow and LinDistFlow

Reference: see publications list at the end

Instructor: V. Kekatos

1
Outline

1. Branch flow model (BFM)

2. DistFlow model

3. DistFlow model for power flow

4. DistFlow model for optimal power flow

5. LinDistFlow model for approximate analysis

2
Branch-bus incidence matrix

• Single-phase and radial feeder represented by tree graph

<latexit sha1_base64="1jyZQDaD1A74Dh0g9x0FuwWx6mI=">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</latexit>

G = (N , E) with |N | = N + 1 and |E| = L = N

• Branch-bus incidence matrix

buses
lines <latexit sha1_base64="X/BrbsIiXEZaWfYN/rkl3YxxGL8=">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</latexit>

2 3
1 1 reduced branch-bus
6 1 1 7
6 incidence matrix 7
6 1 1 7
6 7
6 1 1 7
6 7
à = 6
6 1 1 7
7
6 1 1 7
6 7
6 1 1 7
6 7
<latexit sha1_base64="Vser/hRCRC/9GRhWTC6FY9UuMIE=">AAACGHicbVDLSsNAFJ3UV62vqEs3wSK4qoko6qJQEMGFlAr2AU0pk+mkHTqZhJkbsaTtX7jxV9y4UMRtd/6N0weo1QMXDufcy733eBFnCmz700gtLC4tr6RXM2vrG5tb5vZORYWxJLRMQh7KmocV5UzQMjDgtBZJigOP06rXvRz71XsqFQvFHfQi2ghwWzCfEQxaappHfTfA0CGYJ8VBP+/YQxfoAyRYtAbDb+9Kezf5Yv6iaWbtnD2B9Zc4M5JFM5Sa5shthSQOqADCsVJ1x46gkWAJjHA6yLixohEmXdymdU0FDqhqJJPHBtaBVlqWH0pdAqyJ+nMiwYFSvcDTneNL1bw3Fv/z6jH4542EiSgGKsh0kR9zC0JrnJLVYpIS4D1NMJFM32qRDpaYgM4yo0Nw5l/+SyrHOec0Z9+eZAu5WRxptIf20SFy0BkqoGtUQmVE0CN6Rq/ozXgyXox342PamjJmM7voF4zRF8V4oMA=</latexit>

|N | = 10 and |E| = L = N = 9 4 5
1 1
breadth-first 1 1
(vs. depth-first)
numbering
à = [a0 A]

• The line feeding bus n is indexed as line n-th

3
Branch-bus incidence matrix (cont’d)

reduced branch-bus incidence matrix

<latexit sha1_base64="T8qCXtkxOgmDXWoySZNGBeOzDhM=">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</latexit>

2 3
1
6 1 1 7
6 7
6 1 1 7
6 7
6 1 1 7
6 7
A=6
6 1 1 7
7
6 1 1 7
6 7
6 1 1 7
6 7
4 1 1 5
1 1

• Branch-bus incidence matrix à = [a0 A]

1
<latexit sha1_base64="XgpNk0Sp0t35veyclzDnYzIf6hs=">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</latexit>

Ã1N +1 = 0 ) F=A =
a0 + A1N = 0
1
1N = A a0

4
Branch flow model (BFM)
V ⇡n Vn

Sn = Pn + jQn

zn = rn +jxn

s⇡n = p⇡n + jq⇡n sn = pn + jqn

Line n feeding bus n from its parent bus ⇡n

• Branch flow equations on x(s) = (S, I, V, s0 )
<latexit sha1_base64="jCpAulyxKm1R/XUpwzW+DAD45Ik=">AAACJnicbZBLS8NAEMc3Pmt9RT16CRahhVISUfRSKHjRW0X7gDaEzXbTLt1swu5GLCGfxotfxYuHiog3P4qbNvXROrDw2//MMDN/N6RESNP80JaWV1bX1nMb+c2t7Z1dfW+/KYKII9xAAQ1424UCU8JwQxJJcTvkGPouxS13eJnmW/eYCxKwOzkKse3DPiMeQVAqydGrXR/KgevFD0lxhiIpVb8/t0l5htc/2EzKwjFLjl4wK+YkjEWwMiiALOqOPu72AhT5mElEoRAdywylHUMuCaI4yXcjgUOIhrCPOwoZ9LGw48mZiXGslJ7hBVw9Jo2J+rsjhr4QI99VlemaYj6Xiv/lOpH0LuyYsDCSmKHpIC+ihgyM1DOjRzhGko4UQMSJ2tVAA8ghksrZvDLBmj95EZonFeusYt6cFmpmZkcOHIIjUAQWOAc1cAXqoAEQeATPYAxetSftRXvT3qelS1rWcwD+hPb5BTLaph8=</latexit>

V ⇡n V n = zn I n
<latexit sha1_base64="0fVFQ6+0wNNgsuhzv8anCvYDCJQ=">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</latexit>

Sn = V⇡n In⇤
X
2
Sn zn |In | + sn = Sk
k: n!k

• Boundary conditions?

• Given s, solve 2L+N+1 equations in 2L+N+1 complex unknowns [3]

• Equivalent with typical bus injection model (BIM); a.k.a. power flow equations
5
Branch flow model squared

• Introduce squared voltage and current magnitudes

vn = |Vn |2 and `n = |In |2

<latexit sha1_base64="GCQut6cNU9xDG77b/gAVYVp1FWM=">AAACEnicbVA9SwNBEN2LXzF+RS1tDoOgTbgLijZCwEa7COYDknjsbSbJkr29Y3cuGC7JX7Dxr9hYKGJrZee/8S5JoYkPBh7vzTAzzw0E12hZ30ZqaXlldS29ntnY3Nreye7uVbQfKgZl5gtf1VyqQXAJZeQooBYooJ4roOr2rhK/2geluS/vcBBA06MdyducUYwlJ3vSd+TlsOLI4X1hPG4gPGBEZWs0HmcaIERi3kzMjJPNWXlrAnOR2DOSIzOUnOxXo+Wz0AOJTFCt67YVYDOiCjkTMMo0Qg0BZT3agXpMJfVAN6PJSyPzKFZaZttXcUk0J+rviYh6Wg88N+70KHb1vJeI/3n1ENsXzYjLIESQbLqoHQoTfTPJx2xxBQzFICaUKR7farIuVZRhnGISgj3/8iKpFPL2Wd66Pc0V87M40uSAHJJjYpNzUiTXpETKhJFH8kxeyZvxZLwY78bHtDVlzGb2yR8Ynz+fiZ35</latexit>

• Rearrange power injection equations <latexit sha1_base64="kLmfXSKGq6I8CQZaqYSNj+ighMA=">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</latexit>

X
X
<latexit sha1_base64="aH+CniwBpLBJTXe6vE+qjhCk70U=">AAACF3icbVDLSsNAFJ3UV62vqEs3g0UQxJIURRGEghvdVWof0MYwmU7bIZNJmJkoNdavcOOvuHGhiFvd+TdO2yy09cCFwzn3cu89XsSoVJb1bWRmZufmF7KLuaXlldU1c32jJsNYYFLFIQtFw0OSMMpJVVHFSCMSBAUeI3XPPxv69RsiJA35lepHxAlQl9MOxUhpyTULFZfv37n8/kLXdXFPuvy0JePATfyTB94StNtTSIjwFvqDiuu7Zt4qWCPAaWKnJA9SlF3zq9UOcRwQrjBDUjZtK1JOgoSimJFBrhVLEiHsoy5paspRQKSTjP4awB2ttGEnFLq4giP190SCAin7gac7A6R6ctIbiv95zVh1jp2E8ihWhOPxok7MoArhMCTYpoJgxfqaICyovhXiHhIIKx1lTodgT748TWrFgn1YsC4P8iUrjSMLtsA22AU2OAIlcA7KoAoweATP4BW8GU/Gi/FufIxbM0Y6swn+wPj8AZw+oCk=</latexit>
P k = p n + Pn rn ` n
Sn zn |In |2 + sn = Sk k: n!k
X
k: n!k Q k = qn + Q n xn `n
k: n!k

• Ohm’s law squared (multiply both sides by complex conjugate)

V n = V ⇡n zn I n )
<latexit sha1_base64="ovsY+IC4PMRxFAxl8jZXj6StNjM=">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</latexit>

Vn Vn⇤ = (V⇡n zn In )(V⇡n zn I n ) ⇤ )

V ⇡n V n = zn I n
<latexit sha1_base64="X5/FfQeoN5hSb1o6hk/teVAd/a0=">AAAB/3icbVBNS8NAEJ3Ur1q/qoIXL4tF8WJJRNGLUPCitwo2LbQhbLabdulmE3Y3Qo09+Fe8eFDEq3/Dm//GbZuDVh8MPN6bYWZekHCmtG1/WYW5+YXFpeJyaWV1bX2jvLnlqjiVhDZIzGPZCrCinAna0Exz2kokxVHAaTMYXI795h2VisXiVg8T6kW4J1jICNZG8ss7B66fdRLmi9GR64uLe1+ga98YFbtqT4D+EicnFchR98ufnW5M0ogKTThWqu3YifYyLDUjnI5KnVTRBJMB7tG2oQJHVHnZ5P4R2jdKF4WxNCU0mqg/JzIcKTWMAtMZYd1Xs95Y/M9rpzo89zImklRTQaaLwpQjHaNxGKjLJCWaDw3BRDJzKyJ9LDHRJrKSCcGZffkvcY+rzmnVvjmp1Ow8jiLswh4cggNnUIMrqEMDCDzAE7zAq/VoPVtv1vu0tWDlM9vwC9bHN+7xlVg=</latexit>

v n = v ⇡n 2Re [zn⇤ V⇡n In⇤ ] + |zn |2 `n )

v n = v ⇡n 2rn Pn 2xn Qn + (rn2 + x2n )`n

• Definition of complex power flow squared

Pn2 + Q2n
<latexit sha1_base64="G0Hrzj0qRJNlVnpyrPVzs+w0pZo=">AAACDnicbVDLSsNAFJ3UV62vqEs3g6UgCCUpim6EghuXLdgHNDFMppN26GQSZiaFEvIFbvwVNy4UcevanX/jpM1CWw/cy+Gce5m5x48Zlcqyvo3S2vrG5lZ5u7Kzu7d/YB4edWWUCEw6OGKR6PtIEkY56SiqGOnHgqDQZ6TnT25zvzclQtKI36tZTNwQjTgNKEZKS55ZcwhjHr9xAoFw2vL4Q+O8nfcsnXqpE1OPZ1nFM6tW3ZoDrhK7IFVQoOWZX84wwklIuMIMSTmwrVi5KRKKYkayipNIEiM8QSMy0JSjkEg3nZ+TwZpWhjCIhC6u4Fz9vZGiUMpZ6OvJEKmxXPZy8T9vkKjg2k0pjxNFOF48FCQMqgjm2cAhFQQrNtMEYUH1XyEeIx2M0gnmIdjLJ6+SbqNuX9at9kW1aRVxlMEJOAVnwAZXoAnuQAt0AAaP4Bm8gjfjyXgx3o2PxWjJKHaOwR8Ynz/uHpv5</latexit>

V⇡n In⇤
<latexit sha1_base64="fd30SsquRjdupYd3u7jwsIL2YMU=">AAAB/HicbVBNS8NAEJ3Ur1q/oj16WSyCeCiJKHoRCl70VtF+QBvDZrtpl242YXcjlFD/ihcPinj1h3jz37htc9DWBwOP92aYmRcknCntON9WYWl5ZXWtuF7a2Nza3rF395oqTiWhDRLzWLYDrChngjY005y2E0lxFHDaCoZXE7/1SKVisbjXo4R6Ee4LFjKCtZF8u3zni8umn3UT5osxuvHFw7FvV5yqMwVaJG5OKpCj7ttf3V5M0ogKTThWquM6ifYyLDUjnI5L3VTRBJMh7tOOoQJHVHnZ9PgxOjRKD4WxNCU0mqq/JzIcKTWKAtMZYT1Q895E/M/rpDq88DImklRTQWaLwpQjHaNJEqjHJCWajwzBRDJzKyIDLDHRJq+SCcGdf3mRNE+q7lnVuT2t1Jw8jiLswwEcgQvnUINrqEMDCIzgGV7hzXqyXqx362PWWrDymTL8gfX5A8eClCU=</latexit>

Sn = `n =
v ⇡n

6
Relaxed branch flow model
v ⇡n vn
Sn = Pn + jQn
• Relaxed BFM on y(s) := (S, `, v, p0 , q0 )

X `n
Pk = p n + Pn rn ` n s⇡n sn
k: n!k
X
Q k = qn + Q n xn `n vn = |Vn |2 and `n = |In |2
<latexit sha1_base64="GCQut6cNU9xDG77b/gAVYVp1FWM=">AAACEnicbVA9SwNBEN2LXzF+RS1tDoOgTbgLijZCwEa7COYDknjsbSbJkr29Y3cuGC7JX7Dxr9hYKGJrZee/8S5JoYkPBh7vzTAzzw0E12hZ30ZqaXlldS29ntnY3Nreye7uVbQfKgZl5gtf1VyqQXAJZeQooBYooJ4roOr2rhK/2geluS/vcBBA06MdyducUYwlJ3vSd+TlsOLI4X1hPG4gPGBEZWs0HmcaIERi3kzMjJPNWXlrAnOR2DOSIzOUnOxXo+Wz0AOJTFCt67YVYDOiCjkTMMo0Qg0BZT3agXpMJfVAN6PJSyPzKFZaZttXcUk0J+rviYh6Wg88N+70KHb1vJeI/3n1ENsXzYjLIESQbLqoHQoTfTPJx2xxBQzFICaUKR7farIuVZRhnGISgj3/8iKpFPL2Wd66Pc0V87M40uSAHJJjYpNzUiTXpETKhJFH8kxeyZvxZLwY78bHtDVlzGb2yR8Ynz+fiZ35</latexit>

k: n!k
v n = v ⇡n 2rn Pn 2xn Qn + (rn2 + x2n )`n current and voltage phases
have been dropped!
Pn2 + Q2n
`n =
v ⇡n

• Boundary conditions?

• Current mags. can be eliminated; equations remain nonlinear

• Given s, solve 2(L+N+1) equations in 3L+N+2 real unknowns [1]-[2]

• In radial grids, we get 4N+2 equations in 4N+2 real unknowns

• Unique solution for practical networks with v0 ' 1 and small {(rn , xn )}

7
Recovering phases
v ⇡n vn
Sn

`n
s⇡n sn

• After the relaxed branch flow equations have been solved [3]

• Recover voltage phases

V ⇡n V n = zn I n )
<latexit sha1_base64="v5JO0p9zuHq7STKEDDJicBnrL/o=">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</latexit>

V⇡⇤n Vn⇤ = zn⇤ In⇤ )

V⇡n Vn⇤ = v⇡n zn⇤ Sn )
✓ ⇡n ✓n = \(v⇡n zn⇤ Sn ) linear system can be inverted
only when L=N
<latexit sha1_base64="szTPKyG6NiRcF5jyxnSso9mVlMM=">AAACFHicbZDLSsNAFIYn9VbrLerSTbAIFaEkouhGKLjRXUV7gSaGyXTSDp1MwsyJUEIewo2v4saFIm5duPNtnF4W2vrDwMd/zuHM+YOEMwW2/W0UFhaXlleKq6W19Y3NLXN7p6niVBLaIDGPZTvAinImaAMYcNpOJMVRwGkrGFyO6q0HKhWLxR0ME+pFuCdYyAgGbfnm0bUvLlxOQ6i4ocQku/VFnjX9zE2YptyVrNeHw3sXK/DNsl21x7LmwZlCGU1V980vtxuTNKICCMdKdRw7AS/DEhjhNC+5qaIJJgPcox2NAkdUedn4qNw60E7XCmOpnwBr7P6eyHCk1DAKdGeEoa9mayPzv1onhfDcy5hIUqCCTBaFKbcgtkYJWV0mKQE+1ICJZPqvFuljnQ3oHEs6BGf25HloHled06p9c1Ku2dM4imgP7aMKctAZqqErVEcNRNAjekav6M14Ml6Md+Nj0lowpjO76I+Mzx+op58p</latexit>

✓ ◆⇤
Sn
• Recover current phasors In =
V ⇡n

8
Linearized distribution flow (LinDistFlow)
• Approximate model to overcome the complexity of quadratic equations [1]-[2]
• Derived from forward DistFlow model upon dropping terms related to losses
v ⇡n vn

Sn

s⇡n sn
DistFlow (forward form) LinDistFlow
X X
Pk = p n + Pn rn ` n Pk ' p n + Pn
k: n!k k: n!k
X X
Q k = qn + Q n xn `n Q k ' qn + Q n
k: n!k k: n!k
v n = v ⇡n 2rn Pn 2xn Qn + (rn2 + x2n )`n v n ' v ⇡n 2rn Pn 2xn Qn
Pn2 + Q2n
`n =
v ⇡n

Voltage drop and line power flows are approximately linearly related to power injections
9
Comparison to Lecture 3

• Drop in squared voltage magnitudes from LDF <latexit sha1_base64="WQRNUF7Ham67xiPCMCHuKHyGM9w=">AAACDHicbVDLSgMxFM34rPVVdekmWARBLDNF0GXBjcsW7AM6Q8ikmTY0yYxJpliGfoAbf8WNC0Xc+gHu/BvTdhbaeiBwOOdcbu4JE860cd1vZ2V1bX1js7BV3N7Z3dsvHRy2dJwqQpsk5rHqhFhTziRtGmY47SSKYhFy2g6HN1O/PaJKs1jemXFCA4H7kkWMYGMlVCqPUOYnDMnJxQhJXzNB76sKyTqS59UHJBtIFm3KrbgzwGXi5aQMctRR6cvvxSQVVBrCsdZdz01MkGFlGOF0UvRTTRNMhrhPu5ZKLKgOstkxE3hqlR6MYmWfNHCm/p7IsNB6LEKbFNgM9KI3Ff/zuqmJroOMySQ1VJL5oijl0MRw2gzsMUWJ4WNLMFHM/hWSAVaYGNvftARv8eRl0qpWPMsbl+Wal9dRAMfgBJwBD1yBGrgFddAEBDyCZ/AK3pwn58V5dz7m0RUnnzkCf+B8/gBLkJsK</latexit>

v ⇡n vn ' 2rn Pn + 2xn Qn

• Drop in voltage magnitudes from chapter 3 |V⇡n |

<latexit sha1_base64="ev3QJy1i2CMcaYVHF5OXWHcKll8=">AAACF3icbZC9TsMwFIUd/il/AUYWiwqJhSpBSDBWYoENEC1ITRU57k2xsJ1gO0glzVuw8CosDCDEChtvg1MyQOFIlj6de69874lSzrTxvE9nYnJqemZ2br62sLi0vOKurrV1kikKLZrwRF1GRANnElqGGQ6XqQIiIg4X0fVhWb+4BaVZIs/NIIWuIH3JYkaJsVboNobtMA9SFspiuGNZDgPNBNzgQBBzpUR+BkWQ34XyOJRBUQvdutfwRsJ/wa+gjiqdhO5H0EtoJkAayonWHd9LTTcnyjDKoagFmYaU0GvSh45FSQTobj66q8Bb1unhOFH2SYNH7s+JnAitByKyneW2erxWmv/VOpmJD7o5k2lmQNLvj+KMY5PgMiTcYwqo4QMLhCpmd8X0iihCjY2yDMEfP/kvtHcbvuXTvXrTr+KYQxtoE20jH+2jJjpCJ6iFKLpHj+gZvTgPzpPz6rx9t0441cw6+iXn/QtmmaCb</latexit>
|Vn | ' Re{zn In }

• How are these two approximations related?

• Consider first-order Taylor series expansion around |V0 | = 1 (in per unit wlog)
<latexit sha1_base64="lT3iq5ayc/kLQfBG45/oOKnWALw=">AAAB7nicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0YtQ8OKxgv2ANpTNdtMu3WzC7kQoaX+EFw+KePX3ePPfuG1z0NYHA4/3ZpiZFyRSGHTdb6ewtr6xuVXcLu3s7u0flA+PmiZONeMNFstYtwNquBSKN1Cg5O1EcxoFkreC0d3Mbz1xbUSsHnGccD+iAyVCwShaqTVp9tzJrdcrV9yqOwdZJV5OKpCj3it/dfsxSyOukElqTMdzE/QzqlEwyaelbmp4QtmIDnjHUkUjbvxsfu6UnFmlT8JY21JI5urviYxGxoyjwHZGFIdm2ZuJ/3mdFMMbPxMqSZErtlgUppJgTGa/k77QnKEcW0KZFvZWwoZUU4Y2oZINwVt+eZU0L6reVdV9uKzUqnkcRTiBUzgHD66hBvdQhwYwGMEzvMKbkzgvzrvzsWgtOPnMMfyB8/kDityO+w==</latexit>

vn = |Vn |2 ' |V0 |2 + 2|V0 | (|Vn | |V0 |)

v ⇡n vn ' 2 (|V⇡n | |Vn |)
= 1 + 2(|Vn | 1) = 2|Vn | 1
<latexit sha1_base64="dyEE45E8adPcF5LQ3QqUdsy0BNE=">AAACHXicbZBNSwMxEIazflu/qh69BIuiB2W3FPRY8OJRwdZCtyzZdLYNZrNrMlso2/4RL/4VLx4U8eBF/DemtYJaBwIP7zvDZN4wlcKg6344M7Nz8wuLS8uFldW19Y3i5lbdJJnmUOOJTHQjZAakUFBDgRIaqQYWhxKuw5uzkX/dA21Eoq6wn0IrZh0lIsEZWikoVnpB7qciUMOjXqD2fSNiuKVlX0KEB4P6tzk4sqwGvhadLh4GxZJ77I6LToM3gRKZ1EVQfPPbCc9iUMglM6bpuSm2cqZRcAnDgp8ZSBm/YR1oWlQsBtPKx9cN6Z5V2jRKtH0K6Vj9OZGz2Jh+HNrOmGHX/PVG4n9eM8PotJULlWYIin8tijJJMaGjqGhbaOAo+xYY18L+lfIu04yjDbRgQ/D+njwN9fKxZ/myUqp6kziWyA7ZJQfEIyekSs7JBakRTu7IA3kiz8698+i8OK9frTPOZGab/Crn/RNF9KKV</latexit>

<latexit sha1_base64="L01/d7BdPApQzpqBNo79iubkDJ4=">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</latexit>

rn Pn + xn Qn = Re{zn Sn⇤ } = Re{zn In V⇡⇤n } ' Re{zn In }

<latexit sha1_base64="IyattFfpbprPgJBQeULLZgqt+g8=">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</latexit>

• Equivalent useful approximation |V⇡n | |Vn | ' rn Pn + xn Qn

<latexit sha1_base64="BLmqyxRHMI+Jvpdf7uFgbK37gdI=">AAACD3icbZDLSgMxFIYzXmu9jbp0EyyKIJYZUXRZcOOyBXuBdgiZNNOGZjJjkhHLtG/gxldx40IRt27d+TZm2llo6w+Bj/+cw8n5/ZgzpR3n21pYXFpeWS2sFdc3Nre27Z3dhooSSWidRDySLR8rypmgdc00p61YUhz6nDb9wXVWb95TqVgkbvUwpl6Ie4IFjGBtLGQfjRoo7cQMifHo1LAYdRQL6R2USFSROHlAooZEEdklp+xMBOfBzaEEclWR/dXpRiQJqdCEY6XarhNrL8VSM8LpuNhJFI0xGeAebRsUOKTKSyf3jOGhcbowiKR5QsOJ+3sixaFSw9A3nSHWfTVby8z/au1EB1deykScaCrIdFGQcKgjmIUDu0xSovnQACaSmb9C0scSE20izEJwZ0+eh8ZZ2b0oO7XzUsXJ4yiAfXAAjoELLkEF3IAqqAMCHsEzeAVv1pP1Yr1bH9PWBSuf2QN/ZH3+ABJQnJc=</latexit>

10
LDF in compact form
• Express LDF in matrix-vector notation
X
P̂k ' pn + P̂n Dr = dg(r), Dx = dg(x)
<latexit sha1_base64="LS6vaHdOIHPceWKNLOiLWGnSrcw=">AAACMnicbVDLSsNAFJ3UV62vqEs3wSK0ICURQTdCQRe6q2Af0JQymU7aoZMHMzfSEvJNbvwSwYUuFHHrRzhpI9S2BwYO55zL3HuckDMJpvmm5VZW19Y38puFre2d3T19/6Ahg0gQWicBD0TLwZJy5tM6MOC0FQqKPYfTpjO8Tv3mIxWSBf4DjEPa8XDfZy4jGJTU1e9sD8PAceObpCuubKAjEF7c6yelP0Mk5dOZ0GhpaJSUu3rRrJgTGIvEykgRZah19Re7F5DIoz4QjqVsW2YInRgLYITTpGBHkoaYDHGfthX1sUdlJ56cnBgnSukZbiDU88GYqLMTMfakHHuOSqYrynkvFZd57Qjcy07M/DAC6pPpR27EDQiMtD+jxwQlwMeKYCKY2tUgAywwAdVyQZVgzZ+8SBpnFUvx+/Ni1czqyKMjdIxKyEIXqIpuUQ3VEUFP6BV9oE/tWXvXvrTvaTSnZTOH6B+0n18qqaxz</latexit>

k: n!k
X
Q̂k ' qn + Q̂n p = A> P̂
k: n!k
q = A> Q̂
v̂n ' v̂⇡n 2rn P̂n 2xn Q̂n
Av̂ + v0 a0 = 2Dr P̂ + 2Dx Q̂
<latexit sha1_base64="7wyYaiKRZU6RZX2sn2uUwtseuf4=">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</latexit>

<latexit sha1_base64="Z0bGdrgd+eobRK16UW4DONQNA3s=">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</latexit>

v̂ = v0 1 + 2FDr F> p + 2FDx F> q

<latexit sha1_base64="Bj1axVpRMujwPISqoMIaOVOamPM=">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</latexit>
= v0 1 + 2Rp + 2Xq

• Matrices (R, X) are symmetric positive definite and have positive entries

• Both matrices are almost

equally important

11
IEEE 13-bus feeder
• Assume transposed lines; average diagonal and off-
diagonal entries; take positive-sequence impedance

v̂ = v0 1 + 2Rp + 2Xq
<latexit sha1_base64="MriwiUxcO3n0bGZoVBzXePUg9qE=">AAACM3icbZBNS8NAEIY3ftb6FfXoZbEIglCSIuhFKHgRT1XsB7QhbLabdulmE3c3hRLyn7z4RzwI4kERr/4HN20qtXVg4dl3ZpiZ14sYlcqyXo2l5ZXVtfXCRnFza3tn19zbb8gwFpjUcchC0fKQJIxyUldUMdKKBEGBx0jTG1xl+eaQCElDfq9GEXEC1OPUpxgpLbnmTaePVNIJkOp7fjJM08uha02/dnpamfJdOqVoRm39qg+pa5assjUOuAh2DiWQR801nzvdEMcB4QozJGXbtiLlJEgoihlJi51YkgjhAeqRtkaOAiKdZHxzCo+10oV+KPTjCo7V2Y4EBVKOAk9XZhvK+Vwm/pdrx8q/cBLKo1gRjieD/JhBFcLMQNilgmDFRhoQFlTvCnEfCYSVtrmoTbDnT16ERqVsa749K1Xt3I4COARH4ATY4BxUwTWogTrA4BG8gHfwYTwZb8an8TUpXTLyngPwJ4zvH/svrOE=</latexit>

• To find entry Rmn connect buses n and m to the substation, and add the
resistances of the common lines, e.g., R10,12 = r01 + r15
12
Southern California Edison 47-bus feeder

Matrix X

13
IEEE 123-bus feeder

Matrix X

14
 LDF approximation error
• Express DistFlow in matrix-vector notation

X
Pk = p n + Pn rn ` n
k: n!k
X p = A> P + D r `
Q k = qn + Q n xn `n
k: n!k
q = A> Q + D x `
v n = v ⇡n 2rn Pn 2xn Qn + (rn2 + x2n )`n Av + v0 a0 = 2Dr P + 2Dx Q
<latexit sha1_base64="F9+UCW6J6WxO/KP1BAsfG99Bp20=">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</latexit>
D2r + D2x `

• LDF gives an over-estimator for squared voltage magnitudes

⇥ ⇤ ⇥ ⇤
v = v̂ + FDr I 2F> Dr ` + FDx I 2F> Dx `  v̂
| {z }
<latexit sha1_base64="Ui+nA0ITkDzMMdfHf6QuoLsROFk=">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</latexit>
0

• LDF gives an under-estimator for line flows P = F> p F> D r ` P̂

• Approximation accuracy depends on loading conditions

15
Linearized power flow models
• Recall linearized or so-termed DC power flow model in transmission systems p = B✓

• It has been derived under three approximations:

1. Voltage magnitudes close to unity |Vn | = 1 + ✏n with ✏n ' 0
<latexit sha1_base64="hLD/baKDuGO/KYwvQ5SK9HXr+B8=">AAACHHicbVDJSgNBEO2Je9yiHr00BkEQwowLehECXjwqmETIhNDTqZgmPT1jd40aJsl/ePFXvHhQxIsHwb+xsyBuDwpev1dFV70glsKg6344mYnJqemZ2bns/MLi0nJuZbVsokRzKPFIRvoiYAakUFBCgRIuYg0sDCRUgvbxwK9cgzYiUufYiaEWskslmoIztFI9t9st11X3yNv2ITZCWkn1+z7CLaY3Als9+/gyfCNCuKJuPZd3C+4Q9C/xxiRPxjit5978RsSTEBRyyYypem6MtZRpFFxCL+snBmLG2+wSqpYqFoKppcPjenTTKg3ajLQthXSofp9IWWhMJwxsZ8iwZX57A/E/r5pg87CWChUnCIqPPmomkmJEB0nRhtDAUXYsYVwLuyvlLaYZR5tn1obg/T75LynvFLz9gnu2ly8WxnHMknWyQbaIRw5IkZyQU1IinNyRB/JEnp1759F5cV5HrRlnPLNGfsB5/wQX7qMr</latexit>

2. Voltage angle differences across lines close to zero ✓nm = ✓n ✓m ' 0

<latexit sha1_base64="x/YMH7netCvhCrr7TkS+70yo77o=">AAACDnicbZA9SwNBEIb3/Izx69TSZjEINh53omgjBGwsI5goJEfY20zM4u7euTsnhCO/wMa/YmOhiK21nf/GTbzCrxcWHt6ZYXbeJJPCYhh+eFPTM7Nz85WF6uLS8sqqv7besmluODR5KlNzmTALUmhookAJl5kBphIJF8n1ybh+cQvGilSf4zCDWLErLfqCM3RW19/u4ACQdQutRscl690SVMcKBTc07Pq1MAgnon8hKqFGSjW6/nunl/JcgUYumbXtKMwwLphBwSWMqp3cQsb4NbuCtkPNFNi4mJwzotvO6dF+atzTSCfu94mCKWuHKnGdiuHA/q6Nzf9q7Rz7R3EhdJYjaP61qJ9LiikdZ0N7wgBHOXTAuBHur5QPmGEcXYJVF0L0++S/0NoLooMgPNuv1YMyjgrZJFtkh0TkkNTJKWmQJuHkjjyQJ/Ls3XuP3ov3+tU65ZUzG+SHvLdPod6cYw==</latexit>

3. Ignoring line resistances and shunt elements

• Repeat the same analysis for a meshed grid without the third assumption [6] 2 <latexit sha1_base64="uGhPbU2nkaPNvNEcTIUM6PIXuZc=">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</latexit>

3
✏n
• Consider voltages Vn = (1 + ✏n )ej✓n and Vm = (1 + ✏m )ej✓m x = 4 ✏m 5
<latexit sha1_base64="gSaG+OKWevZRAJdOp47wOUg6+VA=">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</latexit>

✓nm
• Consider power flow from bus n to m: Snm = Vn (Vn⇤ Vm⇤ )ynm
⇤ ⇤
<latexit sha1_base64="6m/DuXJ5WDdc5Y3vrz9CMkVZkBc=">AAACHnicbVDLSsNAFJ3UV62vqks3wSLUgiURi24KBTcuK9q00EeYTCft0JlJmJmIJfRL3PgrblwoIrjSv3GSFtHWAxcO59zLvfd4ISVSWdaXkVlaXlldy67nNja3tnfyu3uODCKBcAMFNBAtD0pMCccNRRTFrVBgyDyKm97oMvGbd1hIEvBbNQ5xl8EBJz5BUGnJzVdu3JizSdVxeVFXr3TiuKxXOh6ncq9U9YsdBtXQ8+P7yY/q5gtW2UphLhJ7Rgpghrqb/+j0AxQxzBWiUMq2bYWqG0OhCKJ4kutEEocQjeAAtzXlkGHZjdP3JuaRVvqmHwhdXJmp+nsihkzKMfN0Z3KqnPcS8T+vHSn/ohsTHkYKczRd5EfUVIGZZGX2icBI0bEmEAmibzXREAqIlE40p0Ow519eJM5p2a6UreuzQs2axZEFB+AQFIENzkENXIE6aAAEHsATeAGvxqPxbLwZ79PWjDGb2Qd/YHx+A2TMof8=</latexit>

= f (x)ynm

f (x) = |Vn |2
<latexit sha1_base64="OcL+MJ7S7WMZoZmJZbMzJDQyt+g=">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</latexit>

|Vn ||Vm |(cos ✓nm + j sin ✓nm )

= (1 + ✏n )2 (1 + ✏n )(1 + ✏m )(cos ✓nm + j sin ✓nm )

• First-order Taylor’s series expansion f (x) ' f (0) + (rx f (0))> (x

<latexit sha1_base64="si/40/AgA4tEMX2/v2rNYyR0li0=">AAACR3icbVDNS8MwHE3n15xfU49egkPYEEcrih4HXjxOcB+wdiXN0i2YpjVJxVH233nx6s1/wYsHRTyaboV9+SDweO/9kl+eFzEqlWm+G7mV1bX1jfxmYWt7Z3evuH/QlGEsMGngkIWi7SFJGOWkoahipB0JggKPkZb3cJP6rSciJA35vRpGxAlQn1OfYqS05Ba7ftkOkBp4fvI8qtiSBuQRTjVzVDkt2xx5DLnT3Jxf6doqjGZuOZsx3WLJrJpjwGViZaQEMtTd4pvdC3EcEK4wQ1J2LDNSToKEopiRUcGOJYkQfkB90tGUo4BIJxn3MIInWulBPxT6cAXH6uxEggIph4Gnk+mKctFLxf+8Tqz8ayehPIoV4XjykB8zqEKYlgp7VBCs2FAThAXVu0I8QAJhpasv6BKsxS8vk+Z51bqsmncXpZqZ1ZEHR+AYlIEFrkAN3II6aAAMXsAH+ALfxqvxafwYv5NozshmDsEccsYfyjqyZg==</latexit>

0)

• Observe that f (0) = 0

<latexit sha1_base64="hWqGWdYBjCEJ0jrN8dMv85v08/I=">AAAB+XicbVDLSgMxFL3js9bXqEs3wSLUTcmIohuh4MZlBfuAdiiZNNOGZjJDkimUoX/ixoUibv0Td/6NmXYW2nogcDjnXu7JCRLBtcH421lb39jc2i7tlHf39g8O3aPjlo5TRVmTxiJWnYBoJrhkTcONYJ1EMRIFgrWD8X3utydMaR7LJzNNmB+RoeQhp8RYqe+6KKz2ImJGQZjh2cUd7rsVXMNzoFXiFaQCBRp996s3iGkaMWmoIFp3PZwYPyPKcCrYrNxLNUsIHZMh61oqScS0n82Tz9C5VQYojJV90qC5+nsjI5HW0yiwk3lIvezl4n9eNzXhrZ9xmaSGSbo4FKYCmRjlNaABV4waMbWEUMVtVkRHRBFqbFllW4K3/OVV0rqsedc1/HhVqeOijhKcwhlUwYMbqMMDNKAJFCbwDK/w5mTOi/PufCxG15xi5wT+wPn8AeKZknc=</latexit>

16
Linearized power flow models (cont’d)
• Compute gradient of f (x) = (1 + ✏n )2 (1 + ✏n )(1 + ✏m )(cos ✓nm + j sin ✓nm )
<latexit sha1_base64="p+UdlTDwxWUuwTr15UVrHYf8nVM=">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</latexit>

2 3 2 3 <latexit sha1_base64="F8de18w9KqxCk71XDxiQZtNyP/k=">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</latexit>

+1
<latexit sha1_base64="z+triqjhFN/f44FdoO12xVvMUQg=">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</latexit>

2(1 + ✏n ) (1 + ✏m )(cos ✓nm + j sin ✓nm )

rx f (x) = 4 (1 + ✏n )(cos ✓nm + j sin ✓nm ) 5 rx f (0) = 4 1 5
(1 + ✏n )(1 + ✏m )( sin ✓nm + j cos ✓nm ) j

• Linearization j][✏n ✏m ✓nm ]>

<latexit sha1_base64="RJKi/FCDEXXiY0PQ0NS8QFujzjI=">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</latexit>

f (x) ' 0 + [+1 1

• Therefore, power flow on line (n,m) can be linearized as

Pnm ' gnm (✏n

<latexit sha1_base64="fc+f4gquPv7mnyu0jtWM5nldPxA=">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</latexit>

✏m ) + bnm (✓n ✓m )
Snm ' [(✏n
<latexit sha1_base64="oArfzE+CMzLpepxcAK/gskqP4Fk=">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</latexit>

✏m ) j(✓n ✓m )] (gnm + jbnm )

Qnm ' bnm (✏n ✏m ) gnm (✓n ✓m )

• Stacking line power flows P = Dg A✏ + Db A✓

Q = Db A✏ Dg A✓

• Converting to power injections p = A> P = G✏ + B✓ G := A> Dg A

compare to ‘DC’ model for q = A> Q = B✏ G✓ B := A> Db A
transmission grids
17
Linearized power flow models (cont’d)

• Solve equations wrt voltage magnitudes and angles

1 1 1 1
✏ = G + BG B p + B + GB G q
1 1 1 1
<latexit sha1_base64="RNG51S8/JP5WjJWnGi9kx2ETNNg=">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</latexit>
✓ = B + GB G p G + BG B q

• Formula is general; holds even for meshed grids

• For radial grids (square and invertible A), equations simplify to

✏ = Rp + Xq
✓ = Xp Rq

• Compare to LDF; linear approximation for voltage angles too

• Linearization conducted at flat voltage profile

• Another reference state can be used; but (R,X,B,G) will depend on that state

18
Power flow via convex relaxation
• Instead of the BF solver, solve the PF problem as a minimization [3]-[4]
N
X
min rn ` n
n=1
over P, Q, v, `, p0 , q0
X
s.t. Pk = p n + Pn rn ` n
k: n!k
X
Qk = qn + Qn xn `n
k: n!k
v n = v ⇡n 2rn Pn 2xn Qn + (rn2 + x2n )`n
Pn2 + Q2n Pn2 + Q2n
`n =  `n
v ⇡n v ⇡n
• Non-convex constraint relaxed to second-order cone constraints (SOC)
2 3
2Pn
4 2Qn 5  ` n + v ⇡n
` n v ⇡n 2

• It can be solved efficiently as a second-order cone program (SOCP)

• Oftentimes, the relaxation is exact: SOC are satisfied with equality
19
Optimal power flow via convex relaxation

• OPF has to be solved to perform any meaningful grid optimization task

1. power loss minimization

2. voltage regulation

3. conservation voltage reduction

4. demand response

5. electric vehicle charging

6. optimal coordination of energy storage

• Power injections s become control variables rather than fixed (inelastic load)

• Optimally control devices while satisfying the PF equations and network constraints

20
Optimal power flow via convex relaxation
• Solving OPF in single-phase radial grids through via an SOCP [3]

N
X N
X N
X
min rn ` n + cn pgn + ↵n vn
n=1 n=1 n=1
over P, Q, v, `, p0 , q0 , s injection
X p = pg pc
s.t. Pk = p n + Pn rn ` n q = qg qc constraints
k: n!k
X pgn  pgn  pgn , 8n
Q k = qn + Q n xn `n
q gn  qng  q gn , 8n
k: n!k
v n = v ⇡n 2rn Pn 2xn Qn + (rn2 + x2n )`n (pgn )2 + (qng )2  sgn , 8n
Pn2 + Q2n v n  vn  v n , 8n
`n network
v ⇡n relaxed BFM equations `n  `n , 8n constraints

• Oftentimes, the relaxation is exact: SOCs are satisfied with equality

21
Exactness under load over-satisfaction
Theorem ([3]): If power injections are unbounded below, the relaxation is exact

• Assume problem has been solved, but SOC for line n is inexact Pn2 + Q2n < `n v⇡n
v⇡0 n = v⇡n vn0 = vn
zn ✏
Sn0 = Sn
2

`0n = `n ✏

zn ✏ zn ✏ injections can be reduced

s0⇡n = s ⇡n s0n = sn without bound
2 2
• Given current solution (S, s, v, `, s0 ) , construct another point (S0 , s0 , v0 , `0 , s00 )
by changing only the quantities related to line n as shown above

• Show that new point is feasible; satisfies SOC with equality; and yields lower cost!
X
Pk0 = p0n + Pn0 rn `0n
k: n!k
X
Q0k = qn0 + Q0n xn `0n
k: n!k
v n = v ⇡n 2rn Pn0 2xn Q0n + (rn2 + x2n )`0n
(Pn0 )2 + (Q0n )2
`0n =
v⇡n 22
Exactness of SOCP convex relaxation

• Exactness of SOCP relaxation for OPF in radial grids has been studied extensively [6]

• Different sets of sufficient conditions have been derived:

§ no reverse power flows
§ identical r/x ratios for all lines
§ r/x increase downstream and there are no reverse active power flows
§ r/x decrease downstream and there are no reverse reactive power flows

• If the SOCP is exact, the minimizer is unique

• To make BFM exact for meshed grids, add phase shifters to implement angle
differences [3]
• Otherwise, one can use a semidefinite program relaxation based on the bus injection
model (BIM) [4]

• How do these schemes extend to multiphase grids? [7]

23
Multiphase branch flow model
v ⇡n vn

n from scalars to vectors

and matrices
Zn
s ⇡n sn

V ⇡n V n = zn I n v ⇡n v n = Z n in
Sn = V⇡n In⇤ S n = v ⇡n i H
n matrix variable?
X X
Sn zn |In |2 + sn = Sk dg Sn H
Z n in in + s n = dg (Sk )
k: n!k k: n!k

⇥ ⇤
• Power received at node n dg v n iH
n = dg (v⇡n Zn in )iH
n

• Actual power sent from parent bus n = dg(Sn )

24
Relaxed multiphase BFM
v ⇡n vn

Zn
s ⇡n sn

• ‘Square’ (multiply by conjugate transpose) the voltage drop equation

v n = v ⇡n Z n in vn vnH = v⇡n v⇡Hn + Zn in iH H

n Zn v⇡n iH H
n Zn Zn in v⇡Hn

• Define ‘squared’ voltages and currents Vn = vn vnH L n = in iH

<latexit sha1_base64="H34ccZpZ4xpw0NpcIGkQgUKU5Mc=">AAACSHicbVBNS8MwGE7n15xfU49egkPwNFpR9CIMvOzgYYL7gLWWNE23sDStSToYZT/Pi0dv/gYvHhTxZtZVqJsvJDzv87wPefN4MaNSmearUVpZXVvfKG9WtrZ3dveq+wcdGSUCkzaOWCR6HpKEUU7aiipGerEgKPQY6Xqjm5neHRMhacTv1SQmTogGnAYUI6Upt+raIVJDL0g7U5df/zZj3RTxQ9N+TJCfXTC/c/m26KMFH818brVm1s2s4DKwclADebXc6ovtRzgJCVeYISn7lhkrJ0VCUczItGInksQIj9CA9DXkKCTSSbMgpvBEMz4MIqEPVzBji44UhVJOQk9PzraUi9qM/E/rJyq4clLK40QRjucPBQmDKoKzVKFPBcGKTTRAWFC9K8RDJBBWOvuKDsFa/PIy6JzVrYu6eXdea5h5HGVwBI7BKbDAJWiAJmiBNsDgCbyBD/BpPBvvxpfxPR8tGbnnEPypUukH1dG1QQ==</latexit>

• Express ‘squared’ voltage drop as V n = V ⇡n + Z n L n Z H

n Sn ZH
n Zn SH
n

• Linear equation; but complexity is hidden under ‘squared’ variables (Vn , Ln , Sn )

25
Relaxed multiphase BFM (cont’d)
v ⇡n vn

Zn
s ⇡n sn

‘square’ relax
• In single-phase grids Sn = V⇡n In⇤ |Sn |2 = v⇡n `n |Sn |2  v⇡n `n

   H
v ⇡n Sn V ⇡n V⇡n
• Relaxation can be also written = ⌫ 0 and rank 1
Sn⇤ `n In In

• In multi-phase grids, the relaxation becomes

   H
V⇡n Sn v ⇡n v ⇡n
= ⌫ 0 and rank-1
S⇤n Ln in in

• Semidefinite program (SDP) constraint captures all quadratic relationships

26
OPF with multiphase BFM

min losses and/or CVR and/or generation cost

over {Sn , sn , Vn , Ln }n
X p = pg pc injection
s.t. dg (Sn Zn Ln ) + sn = dg (Sk )
q = qg qc
constraints
k: n!k
pgn,  pgn,  pgn, , 8n,
V n = V ⇡n + Zn Ln ZH
n Sn ZH
n Zn SH
n g
 q gn,  qn,  q gn, , 8n,
V ⇡n S n
⌫0 (pgn, )2 + (qn,
g
)2  sgn, , 8n,
S⇤n Ln
relaxed BFM equations
v  dg(Vn )  v, 8n
network
dg(Ln )  `, 8n constraints

• Relaxation is exact (constraint satisfied with equality) under practical conditions

27
Linear approximation for multiphase grids

• Ignore losses to get approximate power conservation

X X
dg (Sn Zn Ln ) + sn = dg (Sk ) n + sn = k
k: n!k k: n!k

• Voltage drop requires approximating the full matrix Sn

V n = V ⇡n + Z n L n Z H
n Sn ZH
n Zn SH
n

• Assuming approximately balanced voltages (and currents)

2 3
1
v⇡n ' V⇡n ↵, in ' In ↵ ↵ = 4 ↵⇤ 5 , ↵ = ej2⇡/3
↵
H
• Power flow matrix can be approximated as Sn = ↵↵ dg( n)

⇤
dg(Vn ) = dg(V⇡n ) dg(↵↵H dg( H
n )Zn ) dg(Zn dg( n) ↵↵H )
28
Inter-phase coupling
• Simplify approximate voltage drop using the property

dg (Adg(x)B) = (A
<latexit sha1_base64="wBC44a9ltCeKwXlOuB+i8iFWzOc=">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</latexit>
B> )x, : entry-wise (Hadamard) product

• Approximate voltage drop

⇤
v ⇡n
<latexit sha1_base64="fvjnrCJ7o6lcGtxOTXfDzPPjSDg=">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</latexit>
vn ' 2Re Z̄n n , where Z̄n = Zn ↵⇤ ↵>

2 3
0.530 + 1.112i 0.127 + 0.404i 0.126 + 0.423i
Zn = 4 0.127 + 0.404i 0.545 + 1.043i 0.133 + 0.374i 5 HW2-Exercise 1
0.126 + 0.423i
<latexit sha1_base64="DWHSTyeywaMLRPxMsLnvWjwa5uQ=">AAACvnicbVFdi9QwFE3r11q/Rn30JTi4CEJJv3YEEQZ88XEFZ3dxWoY0czsTJ01LkopD6Z8UfPDfmGmrrrt7IXA499yT5Ny8FlwbQn457q3bd+7eO7rvPXj46PGTydNnZ7pqFIMFq0SlLnKqQXAJC8ONgItaAS1zAef57sOhf/4NlOaV/Gz2NWQl3UhecEaNpVaTn2lJzTYv2i/dSr5PBRRmmeaw4bKlStF91zLGOo/4SUTwGxz4QRByfIwx8YNwZhnixyT+y5wMTBjxNPVukiRx0tuQOOIY42OriaJeE83iP0P/XEbfS5JhKInDwSY54V4Kcj2+NlV8szXZajIlPukLXwfBCKZorNPV5Ee6rlhTgjRMUK2XAalNZk0NZwI6L2001JTt6AaWFkpags7aPv4Ov7LMGheVskca3LOXJ1paar0vc6s8hK2v9g7kTb1lY4q3Wctl3RiQbLioaAQ2FT7sEq+5AmbE3gLKFLdvxWxLFWXGbtyzIQRXv3wdnIV2pT75FE/nZIzjCL1AL9FrFKAZmqOP6BQtEHPeOdT56uzcuVu4pVsNUtcZZ56j/8r9/hu3IsAy</latexit>
0.133 + 0.374i 0.542 + 1.056i
2 3
0.530 + 1.112i 0.286 0.312i 0.430 0.103i
Z̄n = 4 0.413 0.092i 0.545 + 1.0429i 0.258 0.303i 5
0.304 0.321i
<latexit sha1_base64="nfL/IPKa5rE09dF00Hg5hkq6bcQ=">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</latexit>
0.391 0.072i 0.542 + 1.056i

• How do complex injections affect voltage drops?

2 3 2 3
+ + ⇥ ⇤ +
⇥ ⇤
sign Re Z̄n =4 + + 5 sign Im Z̄n = 4 + 5
+ + <latexit sha1_base64="wS1Fm8UQmisEFiqgBujmj5mnZxo=">AAACgnicbVFNb9QwEHXSAiV8beHIxWJVhKi6TaASHIpUiQvcisS2Feto5XgnWau2E9kTpJWVH9K/1Ru/BrzZVIKWkcZ+fvNh+03RKOkwTX9F8db2vfsPdh4mjx4/efpstPv8zNWtFTAVtartRcEdKGlgihIVXDQWuC4UnBeXn9fx859gnazNd1w1kGteGVlKwTFQ89EV0xyXVnsnK9MxBSXObqivekMwzwpufU8Xpf/RdXPDrKyWyLrNnn8aKguopPHcWr7qvBCiS/bpa3oQnB4wlvSA9tTNqV/2EwZmMdQNLeejcTpJe6N3QTaAMRnsdD66ZotatBoMCsWdm2Vpg3loilIo6BLWOmi4uOQVzAI0XIPLfS9hR/cCs6BlbYMbpD37d4Xn2rmVLkLmWgZ3O7Ym/xebtVh+zL00TYtgxOaislUUa7qeB11ICwLVKgAurAxvpWLJLRcYppYEEbLbX74Lzt5NsoC/HY1P0kGOHfKSvCJvSEY+kBPyhZySKRHkd7QXTaLDeDt+G2fx+01qHA01L8g/Fh//AVVav4A=</latexit>
+
<latexit sha1_base64="4kpP6UZMQr+T1SteyZDuQdPSJ0M=">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</latexit>

See Section IV of [8] for an analysis of these patterns 29

References
[1] M. Baran and F. Wu, ‘Optimal sizing of capacitors on a radial distribution system,’ IEEE Trans.
on Power Delivery, Vol. 4, No. 1, Jan. 1989.

[2] M. Baran and F. Wu, ‘Network reconfiguration in distribution systems for loss reduction and
load balancing,’ IEEE Trans. on Power Delivery, Vol. 4, No. 2, Apr. 1989.

[3] M. Farivar and S. Low, ‘Branch flow model: Relaxations and convexification – Part I,’ IEEE
Trans. on Power Systems, Vol. 28, No. 3, Aug. 2013.

[4] S. Low, ‘Convex relaxation of optimal power flow—Part I: Formulations and equivalence,’ IEEE
Trans. on Control of Network Systems, Vol. 1, No. 1, March 2014.

[5] S. Low, ‘Convex relaxation of optimal power flow—Part II: Exactness,’ IEEE Trans. on Control of
Network Systems, Vol. 1, No. 2, June 2014.

[6] D. Deka, S. Backhaus, and M. Chertkov, ‘Structure learning in power distribution networks,’
IEEE Trans. on Control of Network Systems, early access, 2018.

[7] L. Gan and S. Low, ‘Convex relaxations and linear approximation for optimal power flow in
multiphase radial networks,’ in proc. Power System Computation Conf., Feb. 2015, Wroclaw, Poland.

[8] V. Kekatos, L. Zhang, G. B. Giannakis, and R. Baldick, "Voltage Regulation Algorithms for
Multiphase Power Distribution Grids," IEEE Trans. on Power Systems, Vol. 31, No. 5, Sep. 2016.
30