Simultaneous One-Step Approximations of Real and Reactive Power Flow ()
1. Introduction
For AC network power flow calculations there are two widely used models: the exact AC model and the DC linear approximation. Being linear, the DC model solves in one step; it is used when this advantage outweighs the disadvantage of being approximate. See [2] -[5] .
The DC approximation is based on several assumptions. However, one of these assumptions is not needed to obtain a linear system: the assumption of negligible line resistances. A linear system without this assumption is more accurate, in particular of course if this assumption is far from satisfied. Also, the DC approximation models only real power.
The rest of the paper is organized as follows. We start with the exact AC model in Section 2. In Section 3 the linear Taylor approximation is applied. This results in two quite similar models. Then we add the unneeded assumption, arriving at a cruder model, of which the DC model is a part. The estimation of the line losses is discussed shortly in Section 4. In Section 5 the accuracies of the linear models are computed using a simple example. Finally some conclusions are given and in the Appendix a mathematical detail is discussed.
2. The Exact Model
Consider a power network to which applies the standard Short Line model; i.e., without shunt capacitances. We have for the voltages at the ends of a link between node and node:
(1)
with and hence. Complex variables are bold upright (except in the Appendix).
Then the conjugate complex power flow from node in the direction of node is given by:
. (2)
Substituting this into the equations stating that the net power flow from any network node is zero, gives the well-known simultaneous nonlinear system of equations, with as many equations as unknown node voltages. (At a node with a generator, we have a given voltage and a unknown power supply from the generator.)
After solving this system for the voltages, the currents can be computed using (1).
3. The Linear Taylor Models
The standard method for the linearization of a function is the first-order Taylor approximation. This will be applied here with complex differentiation, using a lemma about the derivative of a conjugate. The same result can be obtained without this lemma, differentiating with respect to the real and imaginary parts of the voltages. Finally, a slightly different result is obtained when differentiating with respect to the polar coordinates of the voltages.
3.1. Complex Differentiation
Rewrite (2) as
. (3)
We shall alternatively consider as a function of the complex and themselves, or as a function of the rectangular or the polar coordinates of the complex voltages.
Consider the first order Taylor approximation of around the point defined by
(4)
where is a given voltage, such as the nominal voltage, usually chosen to be real. It is assumed that all voltages in the network are close to this point. At this point we have.
The two derivatives at (4) are:
(5)
For the second equality sign in the first derivative, see the Appendix. Then we have the following Taylor expansion around (4):
. (6)
The total system is now linear in all nodal voltages. This boils down to replacing the in (2) by the given.
3.2. Scalar Differentiation: Cartesian Coordinates
The above result can also be obtained without the Appendix, by differentiating with respect to the real and imaginary parts of the voltages, as follows.
Let and similarly and. Then we have
. (7)
The four derivatives at (4) are:
(8)
Then we have the following Taylor expansion around (4):
(9)
This is the same result as (6) above.
3.3. Scalar Differentiation: Polar Coordinates
Instead of the cartesian coordinates, we might express the complex voltages in their magnitude and their argument (phase angle). Then, instead of point (4), we now have:
(10)
with given and with denoting the voltage phase angle difference: node minus node. At this point we have, as above.
Rewrite in terms of the unknown magnitudes and:
. (11)
The three derivatives at (10) are:
(12)
Then we have the following Taylor expansion around (10):
(13)
with
(14)
the voltage magnitude drop. Hence we have:
(15)
This result differs from the above linearization.
We have here a generalization of Equations (48) and (49) of [1] . The generalization is our complex given voltage, against a real voltage in [1] , indicated by. The two equations of [1] give the real and reactive power respectively:
(16)
. (17)
Using we readily find:
(18)
which is equal to (15) with replaced by.
3.4. The DC Approximation
Adding the assumption of negligible line resistances changes into. Then (16) becomes:
. (19)
This is the well known DC approximation, modelling only real power and ignoring line resistance. There is little to recommend this, compared with the other linear approximations above.
Probably these two reductions (no and no) stem from the original physical analog direct-current system, half a century ago. See for instance [4] , first paragraph of Section III.
4. Line Losses
With each of the above linear approximations, the power flow from in the direction of is the negation of the power from in the direction of. Hence the lines losses must be computed as, using Equation (1).
5. Example
One would expect the two Taylor approximations to be more accurate than the DC approximation. This is indeed the case in the example presented in Table 1, modelling one single-phase low-voltage network link at the end of a line in a residential area. The parameters, shown at the foot of the table, are taken from Appendix F.6 of [1] with a small change of the sending end voltage. The magnitude of this voltage is below the nominal voltage, due to the voltage drop from the generator down to this link.
The errors with polar Taylor are easy to explain from (15): the is 4% larger than the voltages in this line segment and hence is 4% too small and is 8% too small.
With the DC, the voltage lag error is more than ten times larger than with the two Taylor models. A similar result is found in the forthcoming [6] , where the polar Taylor is investigated.
6. Conclusions
Two new non-iterative approximations of the power flow over a short transmission line have been introduced, based on the general method for linear approximations: the first-order Taylor approximation. The equations of
Table 1. The voltage change over a link in a low-voltage network.
Nominal voltage 230 V. Sending end complex voltage 221-j V.
these approximations are complex, modelling in this way both real and reactive power.
They result from the same assumptions as the DC approximation, except the needless assumption of small ratios. In a way, this paper is not based on an assumption, but on the removal of an assumption.
A simple example shows what one might expect from this model: much more accurate voltages than the DC approximation. There seems to be little motive to use the DC approximation.
Further work is needed to test this with larger applications (as in [6] ) and to extend this approach to the Medium-Length Line model (i.e., including shunt capacitances).
Appendix. The Magnitude of
A complex-valued function is said to be differentiable (in the complex sense) at if the following differential quotient exists and is finite:
. (20)
The complex derivative of the conjugate is
(21)
with. This limit does not exist. Loosely speaking, this can be shown as follows. The ratio equals 1 if is real, and if is purely imaginary. (Formally: the Cauchy-Riemann equa- tions are not satisfied.)
However, we have:
(22)
Hence, although one might say that itself has a different value for each direction in the complex plane along which approaches zero, the value of its magnitude is always unity. Hence for practical purposes we can consider as a complex number with unit magnitude but without argument, which might be called the “Complex Unity”:
(23)
Compare with Complex Infinity (loosely speaking, a complex number with infinite magnitude but without argument) and with Complex Zero (a complex number with zero magnitude and without argument).
A practical use of this result is as follows: for any real or complex we have:
(24)
and in particular:
(25)
and hence also:
(26)
This last result is used in section 3.1, Equation (5).
NOTES
*Rob Aalbers (CPB) suggested the use of Taylor while I was working on [1] .