Professional Documents
Culture Documents
01008371discussion of "Z-Bus Loss Allocation
01008371discussion of "Z-Bus Loss Allocation
Discussion of “Z-Bus Loss Allocation” Note, first of all, that bus 3 is correctly allocated more losses than
bus 1 in the base case, because it feeds 55% of the 800 MW load from
Antonio Gómez Expósito and Jesús Riquelme Santos a larger distance. Shifting any of the taps to its limit significantly in-
creases the losses, as shown in the above table, because of adverse re-
active power flows. However, the way those extra losses are allocated
The authors should be commended upon such a motivating and
is not reasonable. For instance, when t1 = 1:1, the extra 13.6 MW are
timely paper.1
roughly split between buses 2 and 3, which benefits bus 1. The opposite
Indeed, as the paper properly states, there is no way to associate
happens when t2 is shifted.
losses with individual buses based solely on physical or electrical con-
More sensitivity is expected to the inclusion of large shunt elements.
siderations. Therefore, all allocation proposals result from mathemat-
The following table shows the allocated losses with and without 100
ical manipulations/simplifications and, consequently, are arguable. For
capacitive MVAR added at bus 1 (t1 = t2 = 0:9).
completeness, we would like to add the following desirable features for
an allocation scheme to be acceptable.
g) Losses allocated to an individual bus should be a fraction of total
losses. That is, jLK j < Ploss .
h) The allocation should be stable enough against minor network
changes whose influence on power flows is negligible.
The idea of decomposing the net active power injections into two
components
P = L+D While the only practical effect of the capacitor is a reduction of the
reactive power generated at bus 1, the consumer at bus 2 would surely
in such a way that complain about the extra losses allocated.
Lk = Pk = Ploss The last example suggests also that there is a perverse incentive to
k k connect large shunt elements in order to reduce the share of losses.
kLk k kDk k This can be proved theoretically as follows. Assume all line shunt sus-
ceptances are very small or negligible compared to a shunt reactor or
is appealing, and may be the origin of future proposals.
capacitor ysh connected at bus k . Then it can be shown that all the
However, we would like to call the attention to the fact that the values
elements of the k th column (row) of Zbus are equal to the inverse of
of Zbus are very sensitive to small changes on network parameters,
particularly at the subtransmission levels where shunt capacitance is
ysh and, hence, purely imaginary. Consequently, the losses allocated
to bus k are null, irrespective of any other consideration (bus powers
small, as a consequence of Ybus being an ill-conditioned matrix.
or branch series impedances).
Consider for instance the 5-bus network shown below, whose data
Other cases tested reveal that the property (g ) above is not always
correspond to Example 7.9 of [1], pp. 253–255 (transformers are as-
satisfied, particularly when line shunt susceptances are very small. A
sumed to have tap changers).
careful analysis of those cases confirms that the values of Real(Zbus )
are significantly influenced by the inclusion of transformer resistances,
which appear in the shunt branches of models for off-nominal taps.
There are more stable ways of building an impedance matrix (see for
instance [6]). Let “s” denote a particular bus, e.g., the slack bus. Then
the nodal equations can be partitioned as
I Y YS
IS
=
YSt YSS
1 VVS
and the losses can be expressed as follows:
Ploss = (I 3 )t 1 Z 1 I + YS YSS
01 Y t V :
S (14)