EEE 305: Power System I

(Part-04)

Instructor:
Dr. Md. Nasim Ahmed Dewan
Professor, EEE

Class Routine:
Sat-9:00, Sun-9:00, Tue-10:00
 The Gauss-Seidel Method
• The complexity of obtaining a formal solution for power-flow in a
power system arises because of the differences in the type of data
specified for the different types of buses

• Formulation of sufficient equations to match the number of

unknown state variables is difficult

• The closed form of solution is not practical

𝑎1 𝑥 + 𝑏1 𝑦 = 𝐶1 , 𝑎2 𝑥 + 𝑏2 𝑦 = 𝐶2 has a closed form solution

• Digital solutions of the power-flow problems follow an iterative

process
– by assigning estimated values to the unknown bus voltages
– by calculating a new value for each bus voltage from the estimated values at the
other buses and the real and reactive power specified
 The Gauss-Seidel Method
• A new set of values for the voltage at each bus is thus obtained
and used to calculate still another set of bus voltages

• Each calculation of new set of voltages is called an iteration.

• The iterative process is repeated until the changes at each bus are
less than a specified minimum value

• With the slack bus designated as bus number (1), computation starts
with bus (2).

If P2,sch and Q2,sch are the scheduled real and reactive power,
entering the network at bus (2), we can write
𝑃2, 𝑠𝑐ℎ − 𝑗𝑄2, 𝑠𝑐ℎ = 𝑉2∗ 𝑌21 𝑉1 + 𝑌22 𝑉2 + 𝑌23 𝑉3 + 𝑌24 𝑉4 (14)
 The Gauss-Seidel Method
1 𝑃2, 𝑠𝑐ℎ −𝑗𝑄2, 𝑠𝑐ℎ
𝑉2 = − 𝑌21 𝑉1 + 𝑌23 𝑉3 + 𝑌24 𝑉4 (15)
𝑌22 𝑉2∗

For now let us assume, buses (3) and (4) are also load buses with real
and reactive power specified. So at bus (3) we have
1 𝑃3, 𝑠𝑐ℎ −𝑗𝑄3, 𝑠𝑐ℎ
𝑉3 = − 𝑌31 𝑉1 + 𝑌32 𝑉2 + 𝑌34 𝑉4 (16)
𝑌33 𝑉3∗

• If we were to equate real and imaginary parts of Eqs. (15) and (16),
and the equation of bus (4), we would obtain six equations in the six
state variables 𝛿2 , 𝛿3 , 𝛿4 , and |𝑉2 |, |𝑉3 | and |𝑉4 |
– However, we solve for the complex voltages directly from equations as they
appear

• The solution proceeds by iteration based on–

– The scheduled real and reactive power at buses (2), (3) and (4)
– The scheduled slack bus voltage 𝑉1 = 𝑉1 ∠𝛿1
(0) (0) (0)
– The initial estimates 𝑉2 , 𝑉3 , and 𝑉4 at the other buses
 The Gauss-Seidel Method
(1)
Solution of Eq. (15) gives the corrected voltage 𝑉2 as,
(1) 1 𝑃2, 𝑠𝑐ℎ −𝑗𝑄2, 𝑠𝑐ℎ (0) (0)
𝑉2 = (0)∗ − 𝑌21 𝑉1 + 𝑌23 𝑉3 + 𝑌24 𝑉4 (17)
𝑌22 𝑉 2
• 𝑃2,𝑠𝑐ℎ , 𝑄2,𝑠𝑐ℎ , 𝑉1 → Fixed specifications
0 ∗ 0 0
• 𝑉2 , 𝑉3 , 𝑉4 → Initial estimates

1 0
• 𝑉2 and 𝑉2 will not agree

• Agreement would be reached to a good degree of accuracy after

several iterations
– Would be the correct value for V2 with estimated voltages but without regard to
power at other buses
– This value would not be the solution for V2 for the specific power-flow conditions.
Because the voltage on which this calculation for V2 depends, are the estimated
0 0 0
values 𝑉2 , 𝑉3 and 𝑉4 -- actual values are not yet known
 The Gauss-Seidel Method
• As the corrected voltage is found at each bus, it is used to calculate the
corrected value at the next bus

1
• Substituting 𝑉2 in Eq. (16), we obtain the first calculated value at bus
(3),
(1) 1 𝑃3, 𝑠𝑐ℎ −𝑗𝑄3, 𝑠𝑐ℎ (1) (0)
𝑉3 =𝑌 (0)∗ − 𝑌31 𝑉1 + 𝑌32 𝑉2 + 𝑌34 𝑉4 (18)
33 𝑉3

• The process is repeated at bus (4) and at each bus consequently

throughout the network (except at the slack bus)
– To complete the first iteration
– Calculated values are found for each state variable

• Then entire process is carried out again and again until the amount of
correction in voltage at every bus is less than some pre-determined
precision index

• This process is known as the Gauss-Seidel iterative method

 The Gauss-Seidel Method
• Convergence upon an erroneous solution is usually avoided if the
initial values are of reasonable magnitude and do not differ too
widely in phase

• It is common practice to set the initial estimates of the unknown

voltages at all nodes equal to 1.0∠0° per unit
– Such initialization is called a “Flat start” because of the uniform voltage profile
assumed

For a system of N buses the general equation for the calculated voltage
at any bus (i) where P and Q are scheduled is,
(𝑘) 1 𝑃𝑖, 𝑠𝑐ℎ −𝑗𝑄𝑖, 𝑠𝑐ℎ (𝑘) (𝑘−1)
𝑉𝑖 = (𝑘−1)∗ − σ𝑖−1
𝑗=1 𝑌𝑖𝑗 𝑉𝑗 − σ𝑁
𝑗=𝑖+1 𝑌𝑖𝑗 𝑉𝑗 (19)
𝑌𝑖𝑖 𝑉𝑖
k = The number of iteration in which the voltage is currently being calculated
k-1 = the number of preceding iteration
 The Gauss-Seidel Method
• The values of the voltages used (right hand side) in the above
equation are the most recently calculated values
– Or the estimated voltages if k = 1, i.e., no iteration has yet been made for that
particular bus

• The Eq. (19) applies only at load buses where real and reactive power
are specified

• An additional step is necessary at voltage-controlled buses

– Voltage magnitude is constant in these buses
 The Gauss-Seidel Method
Example 9.2:
Generators: bus (1), (4)
Loads: all buses
Base values of transmission system: 100 MVA, 230 kV

Fig. 9.2
 The Gauss-Seidel Method
Bus- i Bus- j
Series 𝑌 = 𝑍 −1
Shunt Y/2: From bus to ground
 The Gauss-Seidel Method
Net 𝑷𝒊, 𝒔𝒄𝒉 , 𝑸𝒊, 𝒔𝒄𝒉
-ve for bus 2, 3
 The Gauss-Seidel Method
Assuming the iteration should start at bus 2, find the value of V2 for the first iteration.

Solution:
• We will do computation for an accuracy of 6 decimal places
• From Table 9.2 we construct Ybus of Table 9.4.

For bus 2,
𝑌21 = − 3.815629 − 𝑗19.078144 = −3.815629 + 𝑗19.078144

𝑌24 = − 5.169561 − 𝑗25.847809 = −5.169561 + 𝑗25.847809

𝑌22 = −𝑌21 + 𝑗0.05125 − 𝑌24 + 𝑗0.03875 = 8.985190 − 𝑗44.835953

From Eq. (17)

(1) 1 𝑃2, 𝑠𝑐ℎ − 𝑗𝑄2, 𝑠𝑐ℎ (0)
𝑉2 = (0)∗
− 𝑌21 𝑉1 + 𝑌24 𝑉4
𝑌22 𝑉 2
1 −1.7 + 𝑗1.0535
= ൤ − −3.815629 + 𝑗19.078144
8.985190 − 𝑗44.835953 1.0 − 𝑗0.0
× 1.0 + 𝑗0.0 − 5.169561 + 𝑗25.847809 × 1.02 + 𝑗0.0 ൨
 The Gauss-Seidel Method
(1)
⟹ 𝑉2 = 0.983564 − 𝑗0.032316 pu

• Experience with Gauss-Seidel method of solution of power-flow

problems has shown that the number of iterations required may be
reduced considerably
– if the correction in voltage at each bus is multiplied by some constant
– that increase the amount of correction to bring the voltage closer to the value it
approaching

• The multiplier that accomplishes this improved convergence is called

an acceleration factor

• The difference between the newly calculated voltage and the best
previous voltage at the bus is multiplied by the appropriate
acceleration factor to obtain a better correction to be added to the
previous value
 The Gauss-Seidel Method
For example, at bus 2 in the first iteration we have the accelerated value
defined by straight line formula,
(1) (0) (1) (0) 1 0
𝑉2,𝑎𝑐𝑐 = 1 − 𝛼 𝑉2 + 𝛼𝑉2 = 𝑉2 + 𝛼 𝑉2 − 𝑉2 (20)
where, 𝛼 = acceleration factor

More generally, for bus i during iteration k,

(𝑘) (𝑘−1) (𝑘) (𝑘−1) 𝑘 𝑘−1
𝑉𝑖,𝑎𝑐𝑐 = 1 − 𝛼 𝑉𝑖,𝑎𝑐𝑐 + 𝛼𝑉𝑖 = 𝑉𝑖,𝑎𝑐𝑐 + 𝛼 𝑉𝑖 − 𝑉𝑖,𝑎𝑐𝑐 (21)

• If 𝛼 = 1, then the Gauss-Seidel computed value of Vi is stored as the current

value

• If 0 < 𝛼 < 1, then the value to be stored is a weighted average of the Gauss-
Seidel value and the value store from the previous iteration

• If 1 < 𝛼 < 2, then the value to be stored is essentially an extrapolated value

 The Gauss-Seidel Method
• In power-flow studies, 𝛼 is generally set at about 1.6 and cannot
exceed 2 if convergence is to occur

Substituting the results of Ex. 9.1 and an acceleration factor of 1.6 in

Eq.(20), we find-

(1) (0) 1 0
𝑉2,𝑎𝑐𝑐 = 𝑉2 + 𝛼 𝑉2 − 𝑉2
= 1.0 + 𝑗0.0 + 1.6 0.983564 − 𝑗0.032316 − 1.0 + 𝑗0.0
= 0.973703 − 𝑗0.051706 pu

Using in similar calculation for bus 3 gives first iterative value,

(1)
𝑉3,𝑎𝑐𝑐 = 0.953949 − 𝑗0.066708 pu

• Bus 4 is voltage controlled-- we must treat it differently

 The Gauss-Seidel Method
• The acceleration factor for the real component of the correction may
differ from that for the imaginary component

• For any system optimum values for acceleration factor exists

– Poor choice of factors may result in less rapid convergence or make convergence
impossible

• An acceleration factor of 1.6 for both the real and imaginary

components is usually a good choice but studies may be made to
determine the best choice for a particular system

Voltage-controlled buses:
• When voltage magnitude rather than reactive power is specified at
bus (i), the real and imaginary components of voltage for each
iteration are found by first computing a value for the reactive power
 The Gauss-Seidel Method
From equation (4) we have,
𝑄𝑖 = −𝐼𝑚 𝑉𝑖∗ σ𝑁
𝑗=1 𝑌𝑖𝑗 𝑉𝑗 (22)

The equivalent algorithmic expression,

(𝑘) (𝑘−1)∗ (𝑘) (𝑘−1)
𝑄𝑖 = −𝐼𝑚 𝑉𝑖 σ𝑖−1
𝑗=1 𝑌𝑖𝑗 𝑉
𝑗 + σ𝑁
𝑗=𝑖 𝑌𝑖𝑗 𝑉
𝑗 (23)
(𝑘)
Here, 𝑄𝑖 is evaluated for the best previous voltage values

(𝒌) (𝑘)
Using this value of 𝑸𝒊 in Eq. (19) we have, a new value of 𝑉𝑖 ,
(𝑘)
(𝑘) 1 𝑃𝑖, 𝑠𝑐ℎ −𝑗𝑄𝑖 (𝑘) (𝑘−1)
𝑉𝑖 = (𝑘−1)∗ − σ𝑖−1
𝑗=1 𝑌𝑖𝑗 𝑉𝑗 − σ𝑁
𝑗=𝑖+1 𝑌𝑖𝑗 𝑉𝑗 (19)
𝑌𝑖𝑖 𝑉𝑖

(𝑘) 𝑘
The components of the new 𝑉𝑖 are then multiplied by 𝑉𝑖 / 𝑉𝑖
where, |𝑉𝑖 |= the specified constant magnitude of the voltage-controlled bus
 The Gauss-Seidel Method
• The result is the corrected complex voltage of the specified magnitude

Using Eq. (23) for bus (4),

(1) (0)∗ (1) (1) (0)
𝑄4 = −𝐼𝑚 𝑉4 𝑌41 𝑉1 + 𝑌42 𝑉2,𝑎𝑐𝑐 +𝑌43 𝑉3,𝑎𝑐𝑐 + 𝑌44 𝑉4 (24)

(𝟏)
Using 𝑸𝟒 for 𝑸𝟒, 𝒔𝒄𝒉 in Eq. (19) we have,
(1)
(1) 1 𝑃4, 𝑠𝑐ℎ −𝑗𝑄4 1 1
𝑉4 = (0)∗ − 𝑌41 𝑉1 + 𝑌42 𝑉2, 𝑎𝑐𝑐 + 𝑌43 𝑉3, 𝑎𝑐𝑐 (25)
𝑌44 𝑉4
• All quantities on the right-hand side of Eq. (25) are now known

(𝟏) (𝟏)
Since, 𝑽𝟒 is specified, we correct the magnitude of 𝑽𝟒 as,
1
(1) 𝑉4
𝑉4,𝑐𝑜𝑟𝑟 = 𝑉4 1 (26)
𝑉4
(1)
Proceed to the next step with stored value of 𝑉4,𝑐𝑜𝑟𝑟 having the specified magnitude in the
remaining calculations of iteration
 The Gauss-Seidel Method
• As discussed before, either voltage magnitude or reactive power
must be specified at every bus, except the slack bus, where voltage is
specified by both magnitude and angle

• At buses with generation the voltage magnitude is specified as well

as the real power Pg supplied by the generator

• The reactive power Qg entering the network from the generator is

then determined by the computer

• From a practical viewpoint the Qg must be within definite limits

given by the inequality,
𝑄𝑚𝑖𝑛 ≤ 𝑄𝑔 ≤ 𝑄𝑚𝑎𝑥

where, 𝑄𝑚𝑖𝑛 = Minimum limit of the Qg

𝑄𝑚𝑎𝑥 = Maximum limit of the Qg
 The Gauss-Seidel Method
• If calculate value of Qg is outside either limit, then Qg is set equal to
the limit violated
– The originally specified voltage magnitude at the bus is relaxed
– The bus is then treated as a P-Q bus

• In subsequent iteration the computer program tries to sustain the

originally specified voltage at the bus while ensuring that Qg is
within the permitted range of values

• This could well be possible since other changes may occur elsewhere
in the system to support the local action of the generator excitation
as it adjusts to satisfy the specified terminal voltage
 The Gauss-Seidel Method
Now using Eq. (24) we have,
(1) 0 ∗ 1 1 0
𝑄4 = −𝐼𝑚 𝑉4 𝑌42 𝑉2,𝑎𝑐𝑐 +𝑌43 𝑉3,𝑎𝑐𝑐 + 𝑌44 𝑉4

−5.169561 + 𝐽25.847809 0.973703 − 𝐽0.0511706

= −𝐼𝑚 1.02 + −3.023705 + 𝐽15.118528 0.953949 − 𝐽0.066708
+ 8.193267 − 𝐽40.863838 1.02 + 𝐽0.0

= 1.654151 pu

(𝟏)
This value of 𝑸𝟒 is now substituted into Eq. (25) to yield,
 The Gauss-Seidel Method
(𝟏)
This value of 𝑸𝟒 is now substituted into Eq. (25) to yield,
(1)
(𝟏) 1 𝑃4, 𝑠𝑐ℎ − 𝑗𝑄4 1 1
𝑽𝟒 = (0)∗
− 𝑌42 𝑉2, 𝑎𝑐𝑐 + 𝑌43 𝑉3, 𝑎𝑐𝑐
𝑌44 𝑉 4

1 2.38 − 𝑗1.654151
= ቈ
8.193267 1.02 − 𝑗0.0
− ሼ −5.16956 + 𝑗25.847809 0.973703 − 𝑗0.051706
+ −3.023705 + 𝑗15.118528 0.953949 − 𝑗0.066708 ሽ቉

= 1.017874 − 𝑗0.010604 pu

1
Now, |𝑉4 | = 1.017929, and so we must correct the magnitude to 1.02,
(1) |𝑉4 | 1 1.02
𝑉4, 𝑐𝑜𝑟𝑟 = 𝑉4 = 1.017874 − 𝑗0.010604
𝑉
1 1.017929
4
= 1.019945 − 𝑗0.010625 pu
 The Gauss-Seidel Method
(1)
• In this example 𝑄4 is found to be 1.654151 pu in the first iteration
– If Q generation at bus (4) were limited below 1.654151 pu, the specified limit value
(1)
would be used for 𝑄4
– and bus (4) in that case would be treated as a load bus within the iteration.

• The same strategy is used within any other iteration in which the
generator Q-limit are violated

• The Gauss-Seidel procedure is one method of solving the power-flow

problem

• Today’s industry-based studies generally employ the alternative Newton-

Raphson iterative method,
– reliable in convergence
– computationally faster
– more economical in storage requirements
 The Newton-Raphson Method
• Taylor’s series expansion for a function of two or more variables is
the basis for the Newton-Raphson method of solving power-flow
problem

Let,
ℎ1 𝑥1 , 𝑥2 , 𝑢 = 𝑏1 ; 𝑏1 = Constant

Re-arranging the equation we can write,

𝑔1 𝑥1 , 𝑥2 , 𝑢 = ℎ1 𝑥1 , 𝑥2 , 𝑢 − 𝑏1 = 0 (27)

Consider a second equation involving function h2 such that,

𝑔2 𝑥1 , 𝑥2 , 𝑢 = ℎ2 𝑥1 , 𝑥2 , 𝑢 − 𝑏2 = 0 (28)
where, 𝑏2 = Constant
 The Newton-Raphson Method
• The symbol u represents an independent control, which is
considered constant for this moment

• As in Eqs. (10) and (11), the functions g1 and g2 are introduced for
convenience to allow us to discuss the differences between
calculated values of h1 and h2 and their respective specified values b1
and b2

• For a specified value of ‘u’ let us estimate the solutions of this

(0) (0)
equations to be 𝑥1 and 𝑥2 , where 𝑥1∗ and 𝑥2∗ are the actual
solutions of these equations
 The Newton-Raphson Method
(0) (0) (0) (0)
∴ 𝑔1 𝑥1∗ , 𝑥2∗ , 𝑢 = 𝑔1 𝑥1 + Δ𝑥1 , 𝑥2 + Δ𝑥2 , 𝑢 = 0 (29)
(0) (0) (0) (0)
and, 𝑔2 𝑥1∗ , 𝑥2∗ , 𝑢 = 𝑔2 𝑥1 + Δ𝑥1 , 𝑥2 + Δ𝑥2 , 𝑢 = 0 (30)
(0) (0)
Here Δ𝑥1 and Δ𝑥2 are the corrections needed to reach the solution

(0) (0)
• Now, our problem is to solve Δ𝑥1 and Δ𝑥2

(0) (0)
Expanding Eqs. (29) and (30) about the assumed solution, 𝑥1 and 𝑥2
we have,
(0) (0)
(0) (0) 0 𝜕𝑔1 0 𝜕𝑔1
𝑔1 𝑥1∗ , 𝑥2∗ , 𝑢 = 𝑔1 𝑥1 , 𝑥2 , 𝑢 + Δ𝑥1 𝜕𝑥 ቚ + Δ𝑥2 𝜕𝑥 ቚ +⋯=0 (31)
1 2
(0) (0)
(0) (0) 0 𝜕𝑔2 0 𝜕𝑔2
𝑔2 𝑥1∗ , 𝑥2∗ , 𝑢 = 𝑔2 𝑥1 , 𝑥2 , 𝑢 + Δ𝑥1 𝜕𝑥 ቚ + Δ𝑥2 𝜕𝑥 ቚ +⋯=0 (32)
1 2
Partial derivatives of order greater than 1 have not been shown in the equations
 The Newton-Raphson Method
Here,
𝜕𝑔1 (0) 𝜕𝑔1 (0) 𝜕𝑔2 (0) 𝜕𝑔2 (0)
ቚ , ቚ , ቚ , ቚ → The value of derivatives are evaluated for
𝜕𝑥1 𝜕𝑥2 𝜕𝑥1 𝜕𝑥2
0 0
the estimate 𝑥1 and 𝑥2

Neglecting the partial derivatives of order higher than 1, and re-writing

the above equations in matrix form, we have
𝜕𝑔1 𝜕𝑔1 (0)
(0) 0 0 0 0
𝜕𝑥1 𝜕𝑥2 Δ𝑥1 0 − 𝑔1 (𝑥1 , 𝑥2 , 𝑢) b1 − ℎ1 (𝑥1 , 𝑥2 , 𝑢)
𝜕𝑔2 𝜕𝑔2 (0)
= 0 0
= 0 0
(33)
Δ𝑥2 0− 𝑔2 (𝑥1 , 𝑥2 , 𝑢) b2 − ℎ2 (𝑥1 , 𝑥2 , 𝑢)
𝜕𝑥1 𝜕𝑥2

𝜕𝑔1 𝜕𝑔1 (0)

𝜕𝑥1 𝜕𝑥2
Here, 𝐉 (0) = 𝜕𝑔2 𝜕𝑔2
= jacobian;
𝜕𝑥1 𝜕𝑥2
(0) (0)
𝐉 (0) indicates that the initial estimates 𝑥1 and 𝑥2 are used to compute all the
partial derivatives in the J matrix
 The Newton-Raphson Method
0 0
• 𝑔1 𝑥1 , 𝑥2 , 𝑢 does not have the zero value, as specified by Eq.
0 0
(27), unless the estimates 𝑥1 and 𝑥2 are correct and equal to the
actual solutions

(0)
Let us designate Δ𝑔1 as the difference between the specified and the
(0)
estimated values of g1, and similarly Δ𝑔2 for g2 . Then we have,
(0) (0)
Δ𝑥1 Δ𝑔1
𝐉 (0) (0)
= (0)
(34)
Δ𝑥2 Δ𝑔2

• By solving the mismatch equations, either by triangular factorization

of the jacobian or (for very small problem) by finding inverse, we can
determine Δ𝑥10 and Δ𝑥20
 The Newton-Raphson Method
• Since we truncated the series expansion, these values added to our initial
guesses will not determine the correct solution

1 1
• We must try again by assuming new estimates 𝑥1 and 𝑥2 where,
(1) (0) (0) (1) (0) (0)
𝑥1 = 𝑥1 + Δ𝑥1 ; 𝑥2 = 𝑥2 + Δ𝑥2 (35)

• We repeat the process until correction becomes so small in magnitude that

they satisfy a chosen precision index 𝜖 > 0; that is, until Δ𝑥1 , Δ𝑥2 < 𝜖

Example 9.4:
Using the newton-Raphson method, solve for 𝑥1 and 𝑥2 of the nonlinear equations
𝑔1 𝑥1 , 𝑥2 , 𝑢 = ℎ1 𝑥1 , 𝑥2 , 𝑢 − 𝑏1 = 4𝑢𝑥2 sin 𝑥1 + 0.6 = 0
𝑔2 𝑥1 , 𝑥2 , 𝑢 = ℎ2 𝑥1 , 𝑥2 , 𝑢 − 𝑏2 = 4𝑥22 − 4𝑢𝑥2 cos 𝑥1 + 0.3 = 0

Treat the parameter ‘𝑢’ as a fixed number equal to 1.0, and choose the initial conditions
(0) (0)
𝑥1 = 0 rad and 𝑥2 = 1.0. The precision index 𝜖 is 10-5.
 The Newton-Raphson Method
Solution:
We have,
𝑔1 𝑥1 , 𝑥2 , 𝑢 = 4𝑢𝑥2 sin 𝑥1 + 0.6
𝑔2 𝑥1 , 𝑥2 , 𝑢 = 4𝑥22 − 4𝑢𝑥2 cos 𝑥1 + 0.3
𝜕𝑔1 𝜕𝑔1
𝜕𝑥1 𝜕𝑥2 4𝑢𝑥2 cos 𝑥1 4𝑢𝑠𝑖𝑛 𝑥1
Here, 𝐉 = =
𝜕𝑔2 𝜕𝑔2 4𝑢𝑥2 sin 𝑥1 8𝑥2 − 4𝑢 cos 𝑥1
𝜕𝑥1 𝜕𝑥2
Note: Here the parameter ‘u’ has the fixed value equal to 1.0, but in some
studies it could be treated as s specifiable or control variable.

Putting u = 1.0
𝑔1 𝑥1 , 𝑥2 , 1.0 = 4𝑥2 sin 𝑥1 + 0.6
𝑔2 𝑥1 , 𝑥2 , 1.0 = 4𝑥22 − 4𝑥2 cos 𝑥1 + 0.3
4𝑥2 cos 𝑥1 4𝑠𝑖𝑛 𝑥1
And, 𝐉 =
4𝑥2 sin 𝑥1 8𝑥2 − 4 cos 𝑥1
 The Newton-Raphson Method
(0) (0)
First iteration: 𝑥1 = 0 rad, 𝑥2 = 1.0.
(0)
Δ𝑔1 = 0 − 𝑔1,𝑐𝑎𝑙𝑐 = 0- 0 + 0.6 = −0.6
(0)
Δ𝑔2 = 0 − 𝑔2,𝑐𝑎𝑙𝑐 = 0- 4 − 4 + 0.3 = −0.3
4 0
𝐉 (𝟎) =
0 4
1
0
𝟎 −𝟏 4
⟹𝐉 = 1
0 4

From Eq. (34) we have,

(0) (0)
Δ𝑥1 Δ𝑔1
𝐉 (0) (0)
= (0)
(34)
Δ𝑥2 Δ𝑔2

(0) (0) 1
Δ𝑥1 Δ𝑔1 0 −0.6 −0.150
⟹ =𝐉 0 −1 = 4
=
(0) (0) 1 −0.3 −0.075
Δ𝑥2 Δ𝑔2 0
4
 The Newton-Raphson Method
(1) (0) (0)
𝑥1 = 𝑥1 + Δ𝑥1 = 0 + (−0.150) = −0.150 rad
(1) (0) (0)
𝑥2 = 𝑥2 + Δ𝑥2 = 1.0 + (−0.075) = 0.925
The correction exceeds the specified tolerance, so we can continue.

(1) (1)
Second iteration: 𝑥1 = −0.150 rad, 𝑥2 = 0.925
𝑔1 𝑥1 , 𝑥2 , 1.0 = 4𝑥2 sin 𝑥1 + 0.6
𝑔2 𝑥1 , 𝑥2 , 1.0 = 4𝑥22 − 4𝑥2 cos 𝑥1 + 0.3
4𝑥2 cos 𝑥1 4𝑠𝑖𝑛 𝑥1
And, 𝐉 =
4𝑥2 sin 𝑥1 8𝑥2 − 4 cos 𝑥1
(1)
Δ𝑔1 −4 0.925 sin −0.150 − 0.6 −0.047079
= =
Δ𝑔2
(1) −4 0.925 2 + 4 0.925 cos −0.150 − 0.3 −0.064047

4 0.925 cos(−0.150) 4 sin −0.150

𝐉 (𝟏) =
4 0.925 sin(−0.150) 8 0.925 − 4 cos(−0.150)
3.658453 −0.59775
=
−0.55292 3.444916
 The Newton-Raphson Method
−1
∴𝐉 𝟏 −𝟏
=
3.658453 −0.59775
=
0.280701 0.048706
−0.55292 3.444916 0.045053 0.2981
Now,
(1) (1)
Δ𝑥1 1 −1
Δ𝑔1
(1)
=𝐉 (1)
Δ𝑥2 Δ𝑔2
0.280701 0.048706 −0.047079 −0.01633
= =
0.045053 0.2981 −0.064047 −0.02121

(2) (1) (1)

𝑥1 = 𝑥1 + Δ𝑥1 = −0.150 + (−0.01633) = −0.166335 rad
(2) (1) (1)
𝑥2 = 𝑥2 + Δ𝑥2 = 0.925 + (−0.075) = 0.903786
These corrections also exceeds the precision index, and so we move on to the next
iteration.
(2) (2)
Third iteration: 𝑥1 = − 0.166335 rad, 𝑥2 = 0.903786
(2) (2)
𝚫𝑥1 = − 0.000540644 rad, Δ𝑥2 = −0.000728353

(3) (2) (2)

𝑥1 = 𝑥1 + Δ𝑥1 = −0.166875236 rad
(3) (2) (2)
𝑥2 = 𝑥2 + Δ𝑥2 = 0.903058137
These corrections also exceeds the precision index, and so we move on to the next
iteration.
 The Newton-Raphson Method
(3) (3)
Fourth iteration: 𝑥1 = − 0.166875 rad, 𝑥2 = 0.903058
(3) (3)
𝚫𝑥1 = − 6.17158 × 10−7 rad, Δ𝑥2 = −8.39051 × 10−7

Both of these corrections are smaller in magnitude than stipulated

tolerance of 10-5. Accordingly, we calculate the solution,
(4) (3) (3)
𝑥1 = 𝑥1 + Δ𝑥1 = −0.166876 rad
(4) (3) (3)
𝑥2 = 𝑥2 + Δ𝑥2 = 0.903057

Now, let us calculate the mismatches after 4th iteration:

(4)
Δ𝑔1 = −4 0.903057 sin −0.166876 − 0.6 = −2.15683 × 10−12
(4) 2
Δ𝑔2 = −4 0.903057 + 4 0.903057 cos −0.166875 − 0.3 = −3.15037 × 10−12

These mismatches are very much insignificant and the solution may be well
accepted.
 The Newton-Raphson Method
Now consider the following power system:
Bus-1: slack bus

Writing the mismatch equations at bus (2) we have,

𝑔2′ = 𝑃2 − 𝑃2,𝑠𝑐ℎ = 𝑃2 − 𝑃𝑔2 − 𝑃𝑑2 = 0
𝑔2′′ = 𝑄2 − 𝑄2,𝑠𝑐ℎ = 𝑄2 − 𝑄𝑔2 − 𝑄𝑑2 = 0

From the figure,

𝑃𝑔2 = 𝑄𝑔2 = 0
𝑃𝑑2 = 0.6 and 𝑄𝑑2 = 0.3

Putting these values in the

above mismatch equations,
𝑔2′ = 𝑃2 + 0.6
𝑔2′′ = 𝑃2 + 0.3
Fig. 9.3
 The Newton-Raphson Method
Now the total real and reactive power entering the power system at bus 2,
2

𝑃2 = ෍ 𝑌2𝑛 𝑉2 𝑉𝑛 cos(𝜃2𝑛 + 𝛿𝑛 − 𝛿2 )
𝑛=1
2

𝑄2 = − ෍ 𝑌2𝑛 𝑉2 𝑉𝑛 sin(𝜃2𝑛 +𝛿𝑛 − 𝛿2 )

𝑛=1

Now,
1
𝑌21 = − = 𝑗4 = 4∠90°
𝑗0.25
1
𝑌22 = = −𝑗4 = 4∠ − 90°
𝑗0.25

𝑃2 = 𝑌21 𝑉2 𝑉1 cos(𝜃21 + 𝛿1 − 𝛿2 ) + 𝑌22 𝑉2 𝑉2 cos 𝜃22 + 𝛿2 − 𝛿2

= 4 𝑉2 𝑉1 cos(90° + 0° − 𝛿2 ) + 4 𝑉2 𝑉2 cos −90°
= 4 𝑉1 𝑉2 sin 𝛿2
 The Newton-Raphson Method
𝑄2 = − 𝑌21 𝑉2 𝑉1 sin 𝜃21 + 𝛿1 − 𝛿2 − 𝑌22 𝑉2 𝑉2 sin(𝜃22 + 𝛿2 − 𝛿2 )
= −4 𝑉2 𝑉1 sin 90° + 0° − 𝛿2 − 4 𝑉2 𝑉2 sin(−90°)
= −4 𝑉1 𝑉2 cos 𝛿2 + 4 𝑉2 2

Now,
𝑔2′ = 𝑃2 + 0.6 = 4 𝑉1 𝑉2 sin 𝛿2 + 0.6
𝑔2′′ = 𝑃2 + 0.3 = −4 𝑉1 𝑉2 cos 𝛿2 + 4 𝑉2 2 + 0.3

If we replace 𝑔2′ and 𝑔2′′ by g1 and g2, respectively, we have exactly the same
equations that we had in the above problem.
where,
𝑥1 → 𝛿2
𝑥2 → |𝑉2 |
Also, 𝑢 → |𝑉1 |
 The Newton-Raphson Power-Flow Solution
• To apply N-R method we use polar form of voltages and admittances

• When we separate the term n = i in Eqs. (6) and (7), we have

𝑃𝑖 = σ𝑁
𝑛=1 𝑌𝑖𝑛 𝑉𝑖 𝑉𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) (6)
𝑄𝑖 = − σ𝑁
𝑛=1 𝑌𝑖𝑛 𝑉𝑖 𝑉𝑛 sin 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 (7)

𝑃𝑖 = 𝑌𝑖𝑖 𝑉𝑖 𝑉𝑖 cos(𝜃𝑖𝑖 + 𝛿𝑖 − 𝛿𝑖 ) + ෍ 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 )

𝑛=1
𝑛≠𝑖
= 𝑉𝑖 2 𝐺𝑖𝑖 + σ𝑁
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) (38)
𝑛≠𝑖

𝑄𝑖 = − 𝑌𝑖𝑖 𝑉𝑖 𝑉𝑖 sin 𝜃𝑖𝑖 + 𝛿𝑖 − 𝛿𝑖 − σ𝑁

𝑛=1 𝑌𝑖𝑛 𝑉𝑖 𝑉𝑛 sin 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖
𝑛≠𝑖
= − 𝑉𝑖 2 𝐵𝑖𝑖 − σ𝑁
𝑛=1 𝑌𝑖𝑛 𝑉𝑖 𝑉𝑛 sin 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 (39)
𝑛≠𝑖
 The Newton-Raphson Power-Flow Solution
• The above equations can be readily differentiated with respect
to voltage angles and magnitudes

• 𝑌𝑖𝑖 = 𝑌𝑖𝑖 ∠𝜃𝑖𝑖 = 𝑌𝑖𝑖 cos 𝜃𝑖𝑖 + 𝑗 𝑌𝑖𝑖 sin 𝜃𝑖𝑖 = 𝐺𝑖𝑖 + 𝑗𝐵𝑖𝑖

Let us consider all buses (except slack bus) as load buses with known
𝑷𝒅𝒊 and 𝑸𝒅𝒊
– The slack bus has specified values for 𝛿1 and |𝑉1 |
– Each of other buses has two state variables— 𝛿𝑖 and |𝑉𝑖 | to be
calculated

• The known values of 𝑃𝑑𝑖 and 𝑄𝑑𝑖 correspond to the negative of ‘b’
constants shown in Eqs. (27) and (28)
𝑔1 𝑥1 , 𝑥2 , 𝑢 = ℎ1 𝑥1 , 𝑥2 , 𝑢 − 𝑏1 = 0 (27)
𝑔2 𝑥1 , 𝑥2 , 𝑢 = ℎ2 𝑥1 , 𝑥2 , 𝑢 − 𝑏2 = 0 (28)
 The Newton-Raphson Power-Flow Solution
• At each non-slack bus estimated values of 𝛿𝑖 and |𝑉𝑖 |
(0) (0)
corresponds to the estimates 𝑥1 and 𝑥2 in the preceding section

• Correspondence to the Δ𝑔 mismatches of Eq. (34) follows from Eqs.

(8) and (9) by writing the power mismatches for typical load bus (i),
Δ𝑃𝑖 = 𝑃𝑖,𝑠𝑐ℎ − 𝑃𝑖,𝑐𝑎𝑙𝑐 (40)
Δ𝑄𝑖 = 𝑄𝑖,𝑠𝑐ℎ − 𝑄𝑖,𝑐𝑎𝑙𝑐 (41)

Δ𝑃𝑖 = 𝑃𝑖,𝑠𝑐ℎ − 𝑃𝑖,𝑐𝑎𝑙𝑐 = 𝑃𝑔𝑖 − 𝑃𝑑𝑖 − 𝑃𝑖,𝑐𝑎𝑙𝑐 (8)

Δ𝑄𝑖 = 𝑄𝑖,𝑠𝑐ℎ − 𝑄𝑖,𝑐𝑎𝑙𝑐 = 𝑄𝑔𝑖 − 𝑄𝑑𝑖 − 𝑄𝑖,𝑐𝑎𝑙𝑐 (9)

(0) (0)
Δ𝑥1 Δ𝑔1
𝐉 (0) (0)
= (0)
(34)
Δ𝑥2 Δ𝑔2
 The Newton-Raphson Power-Flow Solution
Writing the mismatch equation for a four bus system,
𝜕𝑃𝑖 𝜕𝑃𝑖 𝜕𝑃𝑖
Δ𝑃𝑖 = Δ𝛿 + Δ𝛿 + Δ𝛿
𝜕𝛿2 2 𝜕𝛿3 3 𝜕𝛿4 4
𝜕𝑃𝑖 𝜕𝑃 𝜕𝑃
+ Δ|𝑉2 | + 𝑖 Δ|𝑉3 | + 𝑖 Δ|𝑉4 | (42)
𝜕|𝑉2 | 𝜕|𝑉3 | 𝜕|𝑉4 |

Multiplying and dividing the last three terms by their respective voltage
magnitudes,
𝜕𝑃𝑖 𝜕𝑃𝑖 𝜕𝑃𝑖
Δ𝑃𝑖 = Δ𝛿 + Δ𝛿 + Δ𝛿
𝜕𝛿2 2 𝜕𝛿3 3 𝜕𝛿4 4
𝜕𝑃𝑖 Δ 𝑉2 𝜕𝑃𝑖 Δ 𝑉3 𝜕𝑃𝑖 Δ 𝑉4
+ 𝑉2 ⋅ + 𝑉3 ⋅ + 𝑉4 ⋅ (43)
𝜕 𝑉2 𝑉2 𝜕 𝑉3 𝑉3 𝜕 𝑉4 𝑉4

A similar mismatch equation for reactive power 𝑸𝒊 ,

𝜕𝑄𝑖 𝜕𝑄𝑖 𝜕𝑄𝑖
Δ𝑄𝑖 = Δ𝛿2 + Δ𝛿3 + Δ𝛿4
𝜕𝛿2 𝜕𝛿3 𝜕𝛿4
𝜕𝑄𝑖 Δ 𝑉2 𝜕𝑄𝑖 Δ 𝑉3 𝜕𝑄𝑖 Δ 𝑉4
+ 𝑉2 ⋅ + 𝑉3 ⋅ + 𝑉4 ⋅ (44)
𝜕 𝑉2 𝑉2 𝜕 𝑉3 𝑉3 𝜕 𝑉4 𝑉4
 The Newton-Raphson Power-Flow Solution
• Each non-slack bus has two equations like Eqs. (43) and (44)

Collecting all the mismatch equations in vector matrix form,

𝜕𝑃2 𝜕𝑃2 𝜕𝑃2 𝜕𝑃2

… |𝑉2 | … |𝑉4 |
𝜕𝛿2 𝜕𝛿4 𝜕|𝑉2 | 𝜕|𝑉4 | Δ𝛿2
⋮ 𝐉11 ⋮ ⋮ 𝐉12 ⋮ ⋮ Δ𝑃2
𝜕𝑃4 𝜕𝑃4 𝜕𝑃4 𝜕𝑃4 Δ𝛿4 ⋮
… |𝑉2 | … |𝑉4 | Δ𝑃4
𝜕𝛿2 𝜕𝛿4 𝜕|𝑉2 | 𝜕|𝑉4 | Δ 𝑉2
𝜕𝑄2 𝜕𝑄2 𝜕𝑄 𝜕𝑄
=
Δ𝑄2
(45)
… |𝑉2 | 2 … |𝑉4 | 2 𝑉2
𝜕𝛿2 𝜕𝛿4 𝜕|𝑉2 | 𝜕|𝑉4 | ⋮ ⋮
⋮ 𝐉21 ⋮ ⋮ 𝐉22 ⋮ Δ 𝑉4 Δ𝑄4
𝜕𝑄4 𝜕𝑄4 𝜕𝑄4 𝜕𝑄4 𝑉4
… |𝑉2 | … |𝑉4 |
𝜕𝛿2 𝜕𝛿4 𝜕|𝑉2 | 𝜕|𝑉4 |
Mismatches
Corrections

jacobian
 The Newton-Raphson Power-Flow Solution
• We cannot include mismatches for the slack bus, since Δ𝑃1 and
Δ𝑄1 are undefined, when 𝑃1 and Q1 are not scheduled

• We also omit all terms involving Δ𝛿1 and Δ|𝑉1 | from the
equations because those corrections are both zero at slack bus

• The partitioned form of Eq. (45) emphasizes the four different

types of partial derivatives

• The elements of 𝐉12 and 𝐉22 have voltage magnitude

multipliers because a simpler and more symmetrical jacobian
results
– In choosing this format we have used the identity:
𝜕𝑃𝑖 Δ 𝑉𝑗 𝜕𝑃𝑖
𝑉𝑗 × = × Δ 𝑉𝑗 (46)
𝜕 𝑉𝑗 𝑉𝑗 𝜕 𝑉𝑗

Element of J12 Correction

 The Newton-Raphson Power-Flow Solution
Δ 𝑉𝑗
• We used corrections rather than Δ 𝑉𝑗
𝑉𝑗

The solution of Eq. (45) is found by iteration as follows:

(0) (0)
• Estimate values 𝛿𝑖 and 𝑉𝑖 for the state variables
• Use the estimates to calculate:
(0) (0)
– 𝑃𝑖,𝑐𝑎𝑙𝑐 and 𝑄𝑖,𝑐𝑎𝑙𝑐 from the Eqs. (38) and (39)
(0) (0)
– Mismatches Δ𝑃𝑖 and Δ𝑄𝑖 from Eqs. (40) and (41)
– The partial derivative elements of the jacobian J

𝑃𝑖 = 𝑉𝑖 2 𝐺𝑖𝑖 + σ𝑁
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) (38)
𝑛≠𝑖
𝑄𝑖 = − 𝑉𝑖 2 𝐵𝑖𝑖 − σ𝑁
𝑛=1 𝑌𝑖𝑛 𝑉𝑖 𝑉𝑛 sin 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 (39)
𝑛≠𝑖
Δ𝑃𝑖 = 𝑃𝑖,𝑠𝑐ℎ − 𝑃𝑖,𝑐𝑎𝑙𝑐 (40)
Δ𝑄𝑖 = 𝑄𝑖,𝑠𝑐ℎ − 𝑄𝑖,𝑐𝑎𝑙𝑐 (41)
 The Newton-Raphson Power-Flow Solution
(0) Δ 𝑉𝑖 0
• Solve Eq. (45) for the initial corrections Δ𝛿𝑖 and 𝑉𝑖 0
• Add the solved corrections to the initial estimates to obtain,
(1) (0) (0)
𝛿𝑖 = 𝛿𝑖 + Δ𝛿𝑖 (47)
𝑉𝑖 (1) = 𝑉𝑖 (0) + Δ 𝑉𝑖 (0) (48)
(1) (1)
• Use the new values 𝛿𝑖 and 𝑉𝑖 as starting values for iteration 2 and
continue

• In more general terms, the updated formulas for the starting values
of the state variables are,
(𝑘+1) (𝑘) (𝑘)
𝛿𝑖 = 𝛿𝑖 + Δ𝛿𝑖 (49)
(𝑘+1) (𝑘) (𝑘) 𝑘 Δ 𝑉𝑖 𝑘
𝑉𝑖 = 𝑉𝑖 + Δ 𝑉𝑖 = 𝑉𝑖 1+ (50)
𝑉𝑖 𝑘
 The Newton-Raphson Power-Flow Solution
• For the four bus system submatrix J11 has the form,
𝜕𝑃2 𝜕𝑃2 𝜕𝑃2
𝜕𝛿2 𝜕𝛿3 𝜕𝛿4
𝜕𝑃3 𝜕𝑃3 𝜕𝑃3
𝐉11 = 𝜕𝛿2 𝜕𝛿3 𝜕𝛿4
(51)
𝜕𝑃4 𝜕𝑃4 𝜕𝑃4
𝜕𝛿2 𝜕𝛿3 𝜕𝛿4
• Expressions for the elements of this equation are easily found by
differentiating the appropriate term in Eq. (38)
𝑃𝑖 = 𝑉𝑖 2 𝐺𝑖𝑖 + σ𝑁
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) (38)
𝑛≠𝑖

When the variable n equals the particular value j, only one of the cosine terms
in the summation of Eq. (38) contains 𝜹𝒋 . So, we obtain the typical off-diagonal
(𝒊 ≠ 𝒋 ) elements of J11,
𝜕𝑃𝑖
= − 𝑉𝑖 𝑉𝑗 𝑉𝑖𝑗 sin 𝜃𝑖𝑗 + 𝛿𝑗 − 𝛿𝑖 (52)
𝜕𝛿𝑗
 The Newton-Raphson Power-Flow Solution
Every term in summation of Eq. (38) contains 𝜹𝒊 . So, the typical
diagonal (𝒊 = 𝒋 ) element of J11,
𝜕𝑃𝑖 𝜕𝑃
= σ𝑁 𝑁
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 sin(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) = − σ𝑛=1 𝜕𝛿
𝑖
(53)
𝜕𝛿𝑖 𝑛
𝑛≠𝑖 𝑛≠𝑖

Now we have,
𝑄𝑖 = − 𝑉𝑖 2 𝐵𝑖𝑖 − σ𝑁
𝑛=1 𝑌𝑖𝑛 𝑉𝑖 𝑉𝑛 sin 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 (39)
𝑛≠𝑖
By comparing the expression in Eq. (53) and Eq. (39) we obtain,
𝜕𝑃𝑖
= −𝑄𝑖 − 𝑉𝑖 2 𝐵𝑖𝑖 (54)
𝜕𝛿𝑖

In a quite similar manner, we can derive formulas for the elements of

sub-matrix J21 as,
Off-diagonal elements:
𝜕𝑄𝑖
= − 𝑉𝑖 𝑉𝑗 𝑉𝑖𝑗 cos 𝜃𝑖𝑗 + 𝛿𝑗 − 𝛿𝑖 (55)
𝜕𝛿𝑗
 The Newton-Raphson Power-Flow Solution
Diagonal elements:
𝜕𝑄𝑖 𝜕𝑄
𝜕𝛿𝑖
= σ𝑁 𝑁 𝑖
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) = − σ𝑛=1 𝜕𝛿 (56)
𝑛≠𝑖 𝑛≠𝑖 𝑛

Now we have,
𝑁
𝑃𝑖 = 𝑉𝑖 2 𝐺𝑖𝑖 + σ𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) (38)
𝑛≠𝑖

By comparing the expression in Eq. (56) and Eq. (38) we can show that,
𝜕𝑄𝑖
𝜕𝛿𝑖
= 𝑃𝑖 − 𝑉𝑖 2 𝐺𝑖𝑖 (57)

The off-diagonal elements of sub-matrix J12 are easily found by first finding
𝝏𝑷𝒊
the expression for derivative and then multiplying by |𝑽𝒋 | as,
𝝏|𝑽𝒋 |
𝜕𝑃𝑖
𝑉𝑗 = 𝑉𝑗 𝑉𝑖 𝑌𝑖𝑗 cos 𝜃𝑖𝑗 + 𝛿𝑗 − 𝛿𝑖 (58)
𝜕 𝑉𝑗
 The Newton-Raphson Power-Flow Solution
We have,
𝜕𝑄𝑖
= − 𝑉𝑖 𝑉𝑗 𝑉𝑖𝑗 cos 𝜃𝑖𝑗 + 𝛿𝑗 − 𝛿𝑖 (55)
𝜕𝛿𝑗

Comparison with Eq. (55) yields,

𝜕𝑃𝑖 𝜕𝑄𝑖
|𝑉𝑗 | =− (59)
𝜕 𝑉𝑗 𝜕𝛿𝑗

• This the most useful result, because it reduces the computation

involved in forming the jacobian, since the off-diagonal elements of
the J12 are now simply the negative of the corresponding elements in
J21
𝜕𝑃𝑖
– This fact would not become apparent if we had not multiplied by the
𝜕 𝑉𝑗
magnitude |𝑉𝑗 | in Eq. (43)

𝜕𝑃 𝜕𝑃 𝜕𝑃𝑖 Δ 𝑉2 𝜕𝑃𝑖 Δ 𝑉4
Δ𝑃𝑖 = 𝜕𝛿 𝑖 Δ𝛿2 + ⋯ + 𝜕𝛿 𝑖 Δ𝛿4 + 𝑉2 𝜕 𝑉2
⋅ 𝑉2
+ ⋯ + 𝑉4 𝜕 𝑉4
⋅ 𝑉4
(43)
2 4
 The Newton-Raphson Power-Flow Solution
In an analogous manner, diagonal elements of J12 can be found. We
have,
𝑃𝑖 = 𝑉𝑖 2 𝐺𝑖𝑖 + σ𝑁
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) (38)
𝑛≠𝑖
𝜕𝑃𝑖
So, |𝑉𝑖 | = 𝑉𝑖 2 𝑉𝑖 𝐺𝑖𝑖 + σ𝑁
𝑛=1 𝑉𝑛 𝑌𝑖𝑛 cos 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 (60)
𝜕 𝑉𝑖
𝑛≠𝑖

Comparing this result with Eqs. (56) and (57), we arrive at the formula,
𝜕𝑃𝑖 𝜕𝑄𝑖
𝑉𝑖 = + 2 𝑉𝑖 2 𝐺𝑖𝑖 = 𝑃𝑖 + 𝑉𝑖 2 𝐺𝑖𝑖 (61)
𝜕 𝑉𝑖 𝜕𝛿𝑖

𝜕𝑄𝑖 𝜕𝑄
= σ𝑁 𝑁
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 cos(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) = − σ𝑛=1 𝜕𝛿
𝑖
(56)
𝜕𝛿𝑖 𝑛
𝑛≠𝑖 𝑛≠𝑖
𝜕𝑄𝑖
𝜕𝛿𝑖
= 𝑃𝑖 − 𝑉𝑖 2 𝐺𝑖𝑖 (57)
 The Newton-Raphson Power-Flow Solution
Finally, the off-diagonal and diagonal elements of sub-matrix J22 are
determined as,
𝑄𝑖 = − 𝑉𝑖 2 𝐵𝑖𝑖 − σ𝑁
𝑛=1 𝑌𝑖𝑛 𝑉𝑖 𝑉𝑛 sin 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 (39)
𝑛≠𝑖
𝜕𝑄𝑖
∴ 𝑉𝑗 = − 𝑉𝑗 𝑉𝑖 𝑌𝑖𝑗 sin 𝜃𝑖𝑗 + 𝛿𝑗 − 𝛿𝑖
𝜕 𝑉𝑗

Comparing with Eq. (52) we have,

𝜕𝑄𝑖 𝜕𝑃𝑖
𝑉𝑗 = (62)
𝜕 𝑉𝑗 𝜕𝛿𝑗
So, off-diagonal elements of the J22 are now simply equal to the corresponding
elements in J11

𝜕𝑃𝑖
= − 𝑉𝑖 𝑉𝑗 𝑉𝑖𝑗 sin 𝜃𝑖𝑗 + 𝛿𝑗 − 𝛿𝑖 (52)
𝜕𝛿𝑗
 The Newton-Raphson Power-Flow Solution
Diagonal elements:
𝑁
𝜕𝑄𝑖
|𝑉𝑖 | = 𝑉𝑖 −2 𝑉𝑖 𝐵𝑖𝑖 − ෍ 𝑉𝑛 𝑌𝑖𝑛 sin 𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖
𝜕 𝑉𝑖
𝑛=1
𝑛≠𝑖

Comparing with Eqs. (53) and (54),

𝜕𝑄𝑖 𝜕𝑃𝑖
𝑉𝑖 =− − 2 𝑉𝑖 2 𝐵𝑖𝑖 = 𝑄𝑖 − 𝑉𝑖 2 𝐵𝑖𝑖 (63)
𝜕 𝑉𝑖 𝜕𝛿𝑖

𝜕𝑃𝑖 𝜕𝑃
𝜕𝛿𝑖
= σ𝑁 𝑁
𝑛=1 𝑉𝑖 𝑉𝑛 𝑌𝑖𝑛 sin(𝜃𝑖𝑛 + 𝛿𝑛 − 𝛿𝑖 ) = − σ𝑛=1 𝜕𝛿
𝑖
(53)
𝑛≠𝑖 𝑛≠𝑖 𝑛

𝜕𝑃𝑖
= −𝑄𝑖 − 𝑉𝑖 2 𝐵𝑖𝑖 (54)
𝜕𝛿𝑖
 The Newton-Raphson Power-Flow Solution
Let us now bring together the results developed above in the following
definitions:
Off-diagonal elements, 𝑖 ≠ 𝑗
𝜕𝑃𝑖 𝜕𝑄𝑖
𝑀𝑖𝑗 ≜ = |𝑉𝑗 | (64)
𝜕𝛿𝑗 𝜕 𝑉𝑗
𝜕𝑄𝑖 𝜕𝑃𝑖
𝑁𝑖𝑗 ≜ = −|𝑉𝑗 | (65)
𝜕𝛿𝑗 𝜕 𝑉𝑗

Diagonal elements, 𝑖 = 𝑗
𝜕𝑃𝑖 𝜕𝑄𝑖
𝑀𝑖𝑖 ≜ 𝑉𝑖 = −𝑀𝑖𝑖 − 2 𝑉𝑖 2 𝐵𝑖𝑖 (66)
𝜕𝛿𝑖 𝜕 𝑉𝑖
𝜕𝑄𝑖 𝜕𝑃𝑖
𝑁𝑖𝑖 ≜ 𝑉𝑖 = 𝑁𝑖𝑖 + 2 𝑉𝑖 2 𝐺𝑖𝑖 (67)
𝜕𝛿𝑖 𝜕 𝑉𝑖

• Interrelationships among the elements in the four sub-matrices of the

jacobian are more clearly seen if we use the above definitions to re-write Eq.
(45) in the following form:
 The Newton-Raphson Power-Flow Solution
Let us now bring
𝑀22 𝑀23 𝑀24 𝑁22 + 2 𝑉2 2 𝐺22 −𝑁23 −𝑁24
𝑀32 𝑀33 𝑀34 −𝑁32 𝑁33 + 2 𝑉3 2 𝐺33 −𝑁34
𝑀42 𝑀43 𝑀44 −𝑁42 −𝑁43 𝑁44 + 2 𝑉4 2 𝐺44
𝑁22 𝑁23 𝑁24 −𝑀22 − 2 𝑉2 2 𝐵22 𝑀23 𝑀24
𝑁32 𝑁33 𝑁34 𝑀32 −𝑀33 − 2 𝑉3 2 𝐵33 𝑀34
𝑁42 𝑁43 𝑁44 𝑀42 𝑀43 −𝑀44 − 2 𝑉4 2 𝐵44

Δ𝛿2
Δ𝛿3 Δ𝑃2
Δ𝛿4 Δ𝑃3
Δ 𝑉2 Δ𝑃4
× 𝑉2 =
Δ𝑄2
(68)
Δ 𝑉3
Δ𝑄3
𝑉3
Δ 𝑉4 Δ𝑄4
𝑉4
 The Newton-Raphson Power-Flow Solution
Voltage-controlled buses:
• In the polar form of power-flow equations voltage controlled buses
are easily taken into account

• If bus (4) is voltage-controlled, then 𝑉4 has a specified constant

Δ 𝑉4
value, i.e., = 0 is always true
𝑉4

• Sixth column of the jacobian of Eq. (68) is always multiplied by 0, and

it may be removed

• Furthermore, since 𝑄4 is not specified, the mismatch Δ𝑄4 cannot be

defined, so we must omit the sixth row of the Eq. (68)
– Off course, 𝑄4 can be calculated after the power-flow solution become available
 The Newton-Raphson Power-Flow Solution
Considering bus (4) as a voltage-controlled bus:

Δ𝛿2
𝑀22 𝑀23 𝑀24 𝑁22 + 2 𝑉2 2 𝐺22 −𝑁23 Δ𝛿3 Δ𝑃2
𝑀32 𝑀33 𝑀34 2
−𝑁32 𝑁33 + 2 𝑉3 𝐺33 Δ𝛿4 Δ𝑃3
𝑀42 𝑀43 𝑀44 −𝑁42 −𝑁43 Δ 𝑉2 = Δ𝑃4
𝑁22 𝑁23 𝑁24 −𝑀22 − 2 𝑉2 2 𝐵22 𝑀23 𝑉2 Δ𝑄2
𝑁32 𝑁33 𝑁34 𝑀32 −𝑀33 − 2 𝑉3 2 𝐵33 Δ 𝑉3 Δ𝑄3
𝑉3

• In general, if there are Ng voltage-controlled buses (besides the slack

bus), a row and column for each such bus is omitted
– Which then has (2N-Ng-2) rows and (2N-Ng-2) columns consistent with Table 9.1