2.10 NRLF in Rectangular co-ordinates [NRLF (Rect.

)]
In rectangular co-ordinates, every complex quantity is expressed in terms of its real and imaginary
parts. Hence, let V̄k = Vk ejθk = ek + jfk ; I¯i = ai + jci and Ȳik = gik + jbik .
n n
Thus, from equation (2.25), I¯i = ∑ Ȳik V̄k . Or, ai + jci = ∑(gik + jbik )(ek + jfk ).
k=1 k=1
Or,
n ⎫
⎪
ai = ∑ (gik ek − bik fk ) ⎪
⎪
⎪
⎪
⎪
⎬
k=1
(2.56)
n
⎪
ci = ∑ (bik ek + gik fk ) ⎪
⎪
⎪
⎪
k=1
⎪
⎭
Complex power injected at bus ‘i’ is, S̄i = Pi + jQi = V̄i I¯i∗ . Or, Pi + jQi = (ei + jfi )(ai − jci ).
Or,
n
Pi = ai ei + ci fi = ∑[ei (gik ek − bik fk ) + fi (bik ek + gik fk )] (2.57)
k=1

and n
Qi = fi ai − ei ci = ∑[fi (gik ek − bik fk ) − ei (bik ek + gik fk )] (2.58)
k=1

Equations (2.57) and (2.58) can be re-written as;

n
Pi = gii (e2i + fi2 ) + ∑ [ei (gik ek − bik fk ) + fi (bik ek + gik fk )] (2.59)
k=1
≠i

n
Qi = −bii (e2i + fi2 ) + ∑ [fi (gik ek − bik fk ) − ei (bik ek + gik fk )] (2.60)
k=1
≠i

Now, again let us consider a ‘n’ bus system having ‘m’ generators (bus 1 being the slack bus). For
each of the ‘(n-m)’ PQ buses, both voltage magnitude and angle are unknown. In other words, for
these buses both real and imaginary parts of the voltages are unknown. For each of the ‘(m-1)’ PV
buses, even though the voltage magnitude is known, real and imaginary parts of the voltage are not
known as there can be many combination of ei and fi to give a specified value of Vi . Hence, for
each of the PV buses also, the real and imaginary parts of the voltage are unknown. Therefore, total
number of unknown quantities is = 2(n − m) + 2(m − 1) = (2n − 2). To solve for these (2n − 2)
unknown quantities, (2n − 2) independent equations are also needed.
Now, let us look at the specified quantities. As discussed earlier, for each of the ‘(n-1)’ buses,
the quantity Pi is specified [given in equation (2.59)]. Similarly, for each of the ‘(n-m)’ buses, the
quantity Q1 is also known [given in equation (2.60)]. Thus, total number of specified quantities is
(n − 1 + n − m) = (2n − m − 1). Hence, still 2n − 2 − (2n − m − 1) = (m − 1) specified quantities are
needed to make the NRLF (rectangular) a well posed problem. These additional specified quantities
would be available from the specified voltage magnitudes at each of the ‘(m-1) buses. At each of
these ‘(m-1)’ PV buses, following relation holds good between the specified quantity (Vi ) and the

53
unknown quantities (ei , fi ).
Vi2 = e2i + fi2 (2.61)

Thus, in NRLF rectangular co-ordinates method, the unknown quantities are; (ei , fi ) for i =
2, 3, ⋯⋯ n (total (2n − 2) in number). The specified quantities are; a) Pi ; for i = 2, 3, ⋯⋯ n; b) Qi ;
for i = (m + 1)⋯⋯ n and c) Vi ; for i = 2, 3, ⋯⋯ m. Hence, the total number of specified quantities
is also (2n − 2). The relations connecting the unknown quantities to the specified quantities are
given by equations (2.59) - (2.61), which are solved by the standard Newton-Rabhosn technique to
determine the unknown quantities. As in the case of NRLF (polar) method, in NRLF (rectangular)
technique also, flat voltage profile is assumed at the starting.
The standard, linearized form of Newton-Raphron solution of the equations (2.59) - (2.61) are
given below (following the basic Newton-Raphron solution method discussed previously);

⎡ ∆P ⎤ ⎡J J ⎤
⎢ ⎥ ⎢ 1 2⎥
⎢ ⎥ ⎢ ⎥
⎢ ∆Q ⎥ = ⎢J3 J4 ⎥ [∆e]
⎢ ⎥ ⎢ ⎥ ∆f (2.62)
⎢ ⎥ ⎢ ⎥
⎢∆V2 ⎥ ⎢J5 J6 ⎥
⎣ ⎦ ⎣ ⎦

In equation (2.62), the sizes of different vector are as follows: ∆e → (n − 1) × 1; ∆f → (n −

1) × 1; ∆P → (n − 1) × 1; ∆Q → (n − m) × 1; ∆V2 → (m − 1) × 1. From these sizes
∂P
of the vectors, the sizes of different sub-matrices J1 ⋯⋯ J6 can be deduced as; J1 → →
∂e
∂P ∂Q ∂Q
(n − 1) × (n − 1); J2 → → (n − 1) × (n − 1); J3 → → (n − m) × (n − 1); J4 → →
∂f ∂e ∂f
∂V 2 ∂V 2
(n − m) × (n − 1); J5 → → (m − 1) × (n − 1); J6 → → (m − 1) × (n − 1).
∂e ∂f
Equation (2.62) can be succinctly written as,

[J] [∆X] = [∆M ] (2.63)

In equation (2.63), as before, the quantities ∆M , ∆X and J are known as the mismatch vector,
correction vector and the Jacobian matrix respectively. Note that the vector ∆M and ∆X each
has a size of (2n − 2) × 1 whereas the Jacobian matrix has a size of (2n − 2) × (2n − 2).
Equation (2.63) forms the basis of NRLF (rectangular) algorithm as discussed below. Note that
the algorithm discussed below assumes that there is no violation of generator reactive power limits.

Basic NRLF (Rectangular) algorithm

Step 1: Assume flat start profile and denote the initial real and imaginary parts of the bus
voltages as e(0) and f (0) respectively.
Step 2: Set iteration counter k = 1.
Step 3: Compute the vectors Pcal and Qcal with the vectors e(k−1) and f (k−1) thereby forming
the vector ∆M . Let this vector be represented as ∆M = [∆M1 , ∆M2 , ⋯⋯ ∆M2n−2 ] .
T

54
 Step 4: Compute error = max (∣∆M1 ∣ , ∣∆M2 ∣ , ⋯⋯ ∣∆M2n−2 ∣).
Step 5: If error ≤  (pre - specified tolerance), then the final solution vectors are e(k−1) and
f (k−1) and print the results. Otherwise go to step 6.
Step 6: Evaluate the Jacobian matrix with the vectors e(k−1) and f (k−1) .
Step 7: Compute the correction vector ∆X by solving equation (2.63).
Step 8: Update the solution vectors e(k) = e(k−1) +∆e and f (k) = f (k−1) +∆f . Update k = k +1
and go back to step 3.

As can be seen, in step 6, the Jacobian matrix needs to be evaluated at each iteration. For this
purpose, the analytical expressions of the elements of the Jacobian matrix are needed, which are
derived next.

Derivation of Jacobian matrix elements for NRLF (rectangular) technique

We derive the elements of the Jacobian matrix by utilising equations (2.59), (2.60) and (2.61). From
these three equations, the Jacobian matrix elements can be obtained as follows:

∂P
Matrix J1 (= ) (in this case, i = 2, 3, ⋯⋯ n, j = 2, 3, ⋯⋯ n)
∂e
n
∂Pi
= 2ei gii + ∑ (gik ek − bik fk ); j=i (2.64)
∂ej k=1
≠i

∂Pi
= (ei gij + fi bij ); j≠i (2.65)
∂ej
∂P
Matrix J2 (= ) (in this case, i = 2, 3, ⋯⋯ n, j = 2, 3, ⋯⋯ n)
∂f

n
∂Pi
= 2fi gii + ∑ (bik ek + gik fk ); j=i (2.66)
∂fj k=1
≠i

∂Pi
= (fi gij − ei bij ); j≠i (2.67)
∂fj
∂Q
Matrix J3 (= ) (in this case, i = (m + 1), (m + 2), ⋯⋯ n, j = 2, 3, ⋯⋯ n)
∂e
n
∂Qi
= −2ei bii + ∑ (−bik ek − gik fk ); j=i (2.68)
∂ej k=1
≠i

∂Qi
= (fi gij − ei bij ); j≠i (2.69)
∂ej

55
 ∂Q
Matrix J4 (= ) (in this case, i = (m + 1), (m + 2), ⋯⋯ n, j = 2, 3, ⋯⋯ n)
∂f

n
∂Qi
= −2fi bii + ∑ (gik ek − bik fk ); j=i (2.70)
∂fj k=1
≠i

∂Qi
= −(fi bij + ei gij ); j≠i (2.71)
∂fj
∂V 2
Matrix J5 (= ) (in this case, i = 2, 3, ⋯⋯ m, j = 2, 3, ⋯⋯ n)
∂e

∂Vi2 ∂Vi2
= 2ei for j = i; and = 0 for j ≠ i; (2.72)
∂ej ∂ej

∂V 2
Matrix J6 (= ) (in this case, i = 2, 3, ⋯⋯ m, j = 2, 3, ⋯⋯ n)
∂f

∂Vi2 ∂Vi2
= 2fi for j = i; and = 0 for j ≠ i; (2.73)
∂fj ∂fj

With these expressions of Jacobian elements, the matrix J can be evaluated at each iteration as
discussed earlier.

Now, in the above algorithm, the violations of generation reactive power limits have not been
taken into account. As in the case of NRLF (polar), in this case also, if a generator violates its Q-
limits at any particular iteration, it is treated as a PQ bus for that iteration. Otherwise, it continues
to be treated as a PV bus. The detailed step-by-step procedure for taking the generation Q limit
violation into account in given below.

Complete NRLF (rectangular) algorithm

(0) (0)
Step 1: Initialise ej = Vjsp for j = 2, 3, ⋯⋯ m and ej = 1.0 for j = (m + 1), (m + 2), ⋯⋯ n.
(0)
Also initialise fj = 0.0 for j = 2, 3, ⋯⋯ n. Let the vectors of the initial real and imaginary parts
of the voltages be denoted as e(0) and f (0) respectively.
Step 2: Set iteration counter k = 1.
(k)
Step 3: For i = 2, 3, ⋯⋯ m, calculate Qi from equation (2.60) by using e(k−1) and f (k−1) . If
(k)
the calculated Qi is within its respective limit, then this bus would be retained as a PV bus in the
current iteration. If either the lower limit or the upper limit is violated, then this bus is converted to
a PQ bus. In general, if l generator buses (l ≤ (m − 1)) violate their corresponding reactive power
limits at step 3, then the size of ∆Q vector increases from (n − m) to (n − m + l) and the size
of ∆V 2 vector decreases from (m − 1) to (m − l − 1). As a result, the new sizes of the matrices
J3 , J4 , J5 and J6 become (n − m + l) × (n − 1), (n − m + l) × (n − 1), (m − l − 1) × (n − 1) and

56
(m − l − 1) × (n − 1) respectively. The sizes of the matrices J1 and J2 remain same. Also, the total
sizes of the mismatch vector, correction vector and the Jacobian matrix remain same.
Step 4: Compute the calculated vectors P, Q and V2 from equations (2.59) - (2.61) with
the vectors e(k−1) and f (k−1) thereby forming the vector ∆M . Let this vector be represented as
∆M = [∆M1 , ∆M2 , ⋯⋯∆M2n−2 ] .
T

Step 5: Compute error = max (∣∆M1 ∣ , ∣∆M2 ∣ , ⋯⋯ ∣∆M2n−2 ∣).

Step 6: If error ≤  (pre - specified tolerance), then the final rotation vectors are e(k−1) and
f (k−1) and print the results. Otherwise go to step 7.
Step 7: Evaluate the Jacobian matrix with the vectors e(k−1) and f (k−1) .
Step 8: Compute the correction vector ∆X by solving equation (2.63).
Step 9: Update the solution vectors e(k) = e(k−1) +∆e and f (k) = f (k−1) +∆f . Update k = k +1
and go back to step 3.

In the next lecture, we will look at an example of NRLF (rectangular) technique.

57