Stochastic power flow using
cumulants and Von Mises
functions
L A Sanabria* and T S Dillon**
* Dept. of Electrical Engineering, Monash University,
Clayton, Victoria 3168, Australia
** Dept. of Computer Science, Latrobe University,
Bundoora, Victoria 3083, Australia
A method of solution of the stochastic load flow problem
based on the method of cumulants and Von Mises function
is developed. The method has significant computational
advantages when compared to presently existing methods.
In addition, it can be used for the solution of problems,
where arbitrary distribution functions are used to specify
special loads existing in the system.
Keywords: distribution systems, load flow, mathematical
programming
I. Introduction
In recent years there has been an increased interest in find-
ing ways to solve the probabilistic load flow problem when
the random variables involved are not normally distributed.
Researchers 1'9 have shown that, generally, the output
random variables of the probabilistic load flow problem are
not normally distributed, and that the Central Limit
Theorem cannot be used as a basis for asserting that they
are. Even if the input variables are normally distributed,
the output variables will not be normals due to the non-
linearity of the power flow equations. On the other hand,
for a more realistic representation of the problem, input
variables other than normals must be used.
The starting point for the probabilistic load flow problem
is to find approximations for the power equations. Various
first-order and second-order transformations have been
proposed. The second-order ones are based on the expan-
sion of the equations around a solution point, using the
Taylor series2'3. The first-order ones use a linearization of
the equations around the deterministic solution 1. The next
step is the mathematical manipulation of the input random
variables to obtain the output random variables. In the
second-order transformation, Sauer-Heydt 2 and Brucoli
et al. 3 have represented the random variables by their
statistical moments. For the first-order transformations,
Received: 21 November 1984, Revised: 20 February 1985
the classical convolutions of random variables using Laplace
Transform is normally used.
In this paper we solve the probabilistic load flow problem
by using cumulants. We will show that the method of
cumulants gives more accurate results, and works faster
than other recent methods to convolve random variables,
such as the Fast Fourier Transform, and the convolution
by Laplace Transform. An additional advantage is that our
method uses the same subroutines employed in the deter-
ministic solution of the problem, so the computer imple-
mentation of the method is very simple.
It is useful to discuss the method for solving the general
probabilistic load flow problem, using a second-order
transformation and convolution by using statistical moments 2.
From now on we will write RVs for random variables. They
will be represented by capital letters.
In this method a power equation of the form
Y = f(V) (1)
is expanded using Taylor series around an operating point
(V o, yo). A second-order approximation for Y can be
obtained by consitlering V as an RV. The random nature
of the variable V is dne to random changes around V°, i.e.
V=V°+AV (2)
hence,
.l.r AVT"I ~V ]
iv] = [Wl + [av] +: i_X.~)~.ii~.S~j
(3)
where:
J'J= .=¢o
H~k- ~2fi I
~V~ v=¢'
Vol 8 No 1 January 1986 0142-0615/86/010047-14 © 1986 Bultterworth & Co (Publishers) Ltd 47