Вы находитесь на странице: 1из 4

Newton-Raphson method

The Newton-Raphson method of solving the power-flow problem is an iterative algorithm for solving a set of simultaneous non-linear equations in an equal number of unknowns based on Taylors series expansion for a function of two or more variables. All partial derivatives of order greater than 1 are neglected. The partial derivatives are written in a form of a square matrix called the Jacobian. The Newton-Raphson algorithm will converge quadratically if the functions have continuous first derivates in the neighbourhood of the solution, the Jacobian is non-singular, and the initial approximations are close to the actual solution. The linear system of mismatch equations for the power flow problem is shown in (1.a) and (1.b) for polar and rectangular coordinates where P is the real power, Q is imaginary power, V is the line voltage, is the voltage angle, e is the line voltage real component, and f is the line voltage imaginary component.

P P P V Q = Q Q V V
P P e P f e Q = Q e Q f f Decoupled Power-flow Method

(1.a)

(1.b)

The decoupled Power-flow method is a variation of the Newton-Raphson procedure based on the fact that a change in the voltage angle at a bus primarily affects the flow of real power in the transmission lines and leaves the flow of reactive power relatively unchanged, while the change in the voltage magnitude at a bus primarily affects the flow of reactive power and leaves the flow of real power relatively unchanged.

0 P P = Q Q V V 0

(2.a)

For test purposes, an analogue modification can be made in the rectangular co-ordinates.

P f e P 0 Q = Q e 0 f
The equations are solved as two independent systems (thus the name decoupled):

(2.b)

[P] = [P ][] [Q] = [Q V ][ V ]


And analogously:

(3.a)

[P] = [P f ][f ] [Q] = [Q e][e]

(3.b)

Another simplification of the Newton-Raphson procedure that greatly speeds up the process is accomplished by re-evaluating the Jacobian in only the first few iterations. This procedure is called the fast decoupled power-flow method. Even though this simplification may greatly decrease the calculation time it may increase the number of iteration necessary to reach the desired accuracy.

LINEAR PROBLEM
Solving the system of linear equations can be numerically very intensive when the number of equations is large. Both equations in the systems (3) can be written as:

Ax=b

(4)

To solve the power-flow problem equation (4) must be solved several times. This linear algebraic equation (4) represents a frequent power-engineering problem where A is characterized as large, sparse, symmetric or incidence symmetric. However in power flow calculations A is among the sparsest matrices in all engineering problems. Conventional approaches to solving the linear system are often too slow for the power flow problem. We are considering the solution of the linear system (4) where A is a sparse matrix.

Triangular factorisation
The problem (4) can be solved by triangular factorisation.

A = LDU

(5)

where L, D, and U are lower triangular, diagonal and upper triangular matrices respectively. In power flow calculations, the matrix A is symmetric or incidence symmetric, depending on the method applied. As a result the sparsity structure of matrices L and U are identical. Equation (5) can be written as: where

A = L U' U' = D U

(6) (7)

After the triangular factorisation (6) the solution of (4) can be obtained by successively solving two canonical equations:

L z = b U' x = z

(8.a) (8.b)

The factorisation of (6) or the substitutions of (8.a) and (8.b) are performed by row so that the precedence relationships between the computations are preserved. One of the advantages of such factorisation is that both triangular matrices can be stored in memory as single matrix and the triangular factorisation can be conceived as a simple recalculation (operation) on a single matrix. The convention code for such a triangular factorisation is shown in CodeBox 1.
k := 1 .. Size(A) | | j := (k+1) .. Size(A) | | | | A(j,k) := A(j,k)/A(k,k); | | | i := (k+1) .. Size(A) | | | | j := (k+1) .. Size(A) | | | | | | A(i,j) := A(i,j) - A(i,k)*A(k,j); | | |

CODEBOX 1: Basic triangular factorisation a matrix

Pi = Vi Yij Vjcos( i ij j )
j=1 n

Q i = Vi Yij Vjsin( i ij j )
j=1

i = 1, 2, n; i r

2D Primjer 1 2D Primjer 2

Вам также может понравиться