The utility of optimal‐control theory for the deterministic
operation of very large multireservoir systems is demonstrated for a
real situation, the complex multireservoir Mahaweli system in Sri Lanka.
This system includes 19 nodes, or reservoirs, and 35 release links. The
model developed for the Mahaweli is designed to minimize a
hydroelectric energy shortage objective and satisfy prespecified
irrigation demand constraints. Two alternative approaches are explored
for optimal operation of the Mahaweli system. The first involves monthly
application of the optimal‐control algorithm to find an optimal policy
for the next year, based on current storage and forecasted or historical
inflows and demands. The second alternative is an implicit stochastic
approach, in which linear operating rules are derived using
deterministic optimal control and historical data. Both of the
alternatives give reasonable and comparable results. The implicit
stochastic optimization alternative has a great advantage regarding
computer time and storage requirements. This makes it usable on a small
personal computer, providing the system operator with decisions in a few
seconds, even with very large systems.