Real-time Cascading Reservoirs Operation using the Genetic Algorithm and Seasonally Varied Models

The Perak cascading scheme located in the state of Perak, Malaysia, consists of four
reservoirs, namely, Temenggor, Bersia, Kenering and Chenderoh. The reservoirs are
used for hydroelectric power generation and flood control. The hydroelectric power
potential of the cascading scheme is 578 MW, while the annual long-term historical
average (HA) hydroelectric power generation was around 228 MW. It was about
39.46% of the potential capacity. Accordingly, the study aimed to improve the
hydroelectric power generation of the scheme.
The genetic algorithm and the seasonally varied models have been developed to
maximize the annual hydroelectric power generation of the Perak cascading reservoir.
The fitness function of the genetic algorithm model (GAM) was to minimize the
difference between the potential capacity and actual generation of the scheme. GAM
was established with a total of 208 and I 04 equality and inequality constraints,
respectively. The optimal release decisions were found after checking the optimality
of the population size (PS), the crossover probability (CRP) and the generation
number (GN). Whereas, a seasonally varied model (SVM) has been developed after
the analysis of the long-term HA operation data. Consequently, from the annual
variation of the headrace level of Temenggor, the most upstream reservoir in the
cascading scheme, four seasons are identified. The seasons are the refill, upper level,
deplete and lower level that occurs in sequentially and repeatedly per year. The two
seasons that require a ranking order to maximize the energy-storage and to minimize
the spill of water in the scheme are the refill and deplete. Hence, the refill rank order
performed according to the decrease order of the change of power production in the
change of storage volume, while the order of depletion has been conducted with the
increase order of the storage effectiveness ratio.