CO l" ri,, 11I © I L\( : (:olll l'''i l'r
.-\ pplictl iOIl . ... 10 Pr()n',. .... (:OIlI r()1. \' ic.:IlIl ;1.
. -\ ria. I
OPTIMAL CONTROL ALGORITHM FOR
HYDROPOWER PLANTS CHAIN SHORT-TERM
OPERATION
A. F. Sakr and H. T. Dorrah
Abstract. This paper presents a new optimization algorithm for solving the scheduling
problem of hydroelectric power plant chain . Hydropower plants within the same ri ver
basin tend to be highly interdependant in the short-term. These interdependancies are
accommodated within the mode l incorporating some hydraulic properties in the flow model
of the river between the plants . A continuous dynamic model is developed that , inheren-
tly, considers the time lags nature of the problem . The hydropower plants operation
problem is t hen formulated us ing the optimal control t heory in order to maximize the
benefit of a large power system comprising hydrostorage and thermal power plants us ing
marginal price approach. It is shown that the optimization problem is reduced to the
Prntryagin ' s Maximum Principle which can be efficiently solved by a suggested methodo -
logy. Fina ll y, it is demonstrated , with an application to the Egyptian power system
having five hydrostorage plants , that the proposed technique is pragmatical for real
case studies , developing optimal operation policy matchable with any specified elec -
tric load demand and providing a considerable improvement in the short -t erm energy ge -
neration.
Keywords. Digital computer applications ; Hydraulic systems.; Maximum principle; Modeling;
Opt i mal con t rol; Power generation; Water resources.
INTRODUCTION
The problem of optimal hydropower production can
be class ified int o three categories . These are
(Vide Fig . l):
i) Long-term operat i on schedule:
This is the problem of maxim izin g the hydroenergy
over an extended period of time and obta i ning an
operat ing schedule for intervals of time much
longer than the time lags between the storages in
series. The problem with such a long duration is
cha ra cter ize d by its stochastic nature arising
from the undeterministic inflow from the river
(Turgeon (1980».
ii) Medium and short -te rm schedule:
This problem is seek ing for the optimal solution
over a time duration that is not too small as in
the dispatch problem to assume system stat i onarity
and not too long to ignore ti me lags between sto -
rages. A dynamic model for the system is essen -
tial to cons ide r the effect of the time lag in the
de pendence of each storage on others . Most of the
work in this d ir ect ion is based on employing a d is-
crete model of the form
(1 )
where
Xi(k) is the water content of reservoir i at
start of inter v al k,
Ui (k) denotes the re le ase from reservoir i dur -
ing period k,
is the losses from th e reservoir i dur -
ing period k, and
li designates the time lag between reser-
vo irs i-l and i.
The above model together with the system constr -
aints enables the use of the dynamic programing or
one of its successive approximate methods (Bellman
and Dreyfus (1962), Murray and Yakowitz (1979».
Hi. "
iii) Load dispatch:
Load dispatch problem is concerned with the econo -
mi cal operation of power systems compr ising con -
vential hydrostorage and thermal power plants . This
is carried out for the purpose of load d ispat ch at
a g i ven loading condition during a small interval
of time to the extend that the dynamic behav i or of
the system can be ignored . The formulation of the
problem is static in nature (Ghoneim and Colleagues
(1981». Power dispatch problems are simple mathe -
matical exercise and today many power sys tem s are
operating under economic dispatch.
This paper considers the short - term hydropower
plant operating strategy problem which is highly
affected by the dynamic behavior of the hydro re -
servoir chain system. The short - term strategy is
considered as one level of the infinite horizon op -
timal strategy elucidated in Fig.l . The intera c-
tion with other levels is the final and initial
system conditions as well as the fixed amount of
water during the time interval obtained in the
higher levels of optimization. From th is view
point , the algorithm is an open - loop control s cheme .
The goal of this control scheme is to obtain t he
most economical util iz at i on of the who le power
generating system including hydro and thermal units.
Thermal power generating plants are taken i nto con -
sideration by means of marginal price curve , as if
the thermal stations are purchasing hydropower with
di fferent prices at d iff eren t l oading cond iti ons to
encourage hydropower production dur ing peak per -
i ods. This leads thermal stations to operate uni-
formly , wh i ch certainly reduces their electrical
energy uni t price.
System Description
The system under consideration, Fig. 2, consists of
n cascaded reserv oi rs wit h power plants, locate d
on one rive r basin . For this chain, let us intro-
duce t he following notations :
Hi The head upstream the i th reservoir measured