Artificial Intelligence in Power System Optimization

CHAPTER

4

HYDROTHERMAL SCHEDULING

*

4.1 INTRODUCTION

A modern power system comprises a large number of thermal and hydro power

plants of various types. The integrated operation of hydro and thermal generation

in a hydrothermal system to minimize the cost of generation has been referred to

as hydrothermal coordination scheduling. The hydrothermal scheduling problem

is even more complex than a purely thermal system optimization problem since

hydroelectric plants are coupled both electrically and hydraulically. In addition,

all energy resources should be fully utilized in the most economic manner. The

operating cost of thermal plants is very high, though their capacity cost is low. In

general, the startup of thermal plants is slow and their response speed is also low.

Hence, once they are started, they have to be operated continuously for many hours/

days. On the contrary, the operating cost of hydroelectric plants is low but their

capital cost is high. Hydroelectric plants can be started quickly and their response is

so fast that they can easily cope with fl uctuating loads. Therefore, it is economical

and convenient to have both thermal and hydro power plant in a generation

system. Thermal power plants are generally scheduled as base load plants whereas

hydroelectric plants are used during peak load periods. Hydrothermal optimal

scheduling offers a substantial savings on fuel cost and enables the utilization of

the full potential of limited water resources.

Hydroelectric plants are constrained by limited storage reservoirs and discharge

rates. Available water resources are limited by the capacities of reservoirs and

infl ows, pre-specifi ed amounts of water withdrawals from the reservoirs for meeting

agricultural or fl ood control demands, or amounts to be preserved for navigational

purposes or ecological requirements (e.g., for fi sh ladders).

Due to the usually cyclic nature of reservoir water infl ows, hydro scheduling

is classifi ed into long range and short range problems. The long range problem

is normally confi ned to a year, a season or a month, the short range problem may

cover a day or a week.

*This chapter has been written with assistance from Nit Petcharaks