Abstract – As the world population

is growing, its water demand is becoming more and more diversified; the

difference between the demand and supply is continuously increasing. The only

solution of this problem is the systematic utilisation of the available water

resources, or by harvesting additional water potential from new water resources

projects and/or by formulating strategies for proper utilization of the

available resources. Considering the fact that most of the River and their

tributaries in India are Seasonal, the Major Steps taken for the remedy of

above mentioned problem is the “Reservoir Projects”.

Development

of monthly release policy for a reservoir is a multistage decision making

process. A model is required to be developed for each individual system of

reservoirs. The essential need for formulation of optimal policy is to determine

of the schedule of releases from a reservoir system, which can maximizes or

minimizes the utilities associated with the release of water.

Various algorithms have

been applied to optimize the Reservoir operation and maximize the net benefit,

but they have their own limitations, the selection of appropriate model for

deriving reservoir operating policies is difficult and more often there is a

scope for further improvement as the model selection depends on the

characteristics of reservoir considered, on the availability of data, on the

specified objectives and constraints. In the present paper, various

optimization Algorithm will be compared along with their Advantage and

Limitation to find the best suited Algorithm for reservoir operation.

Keywords- Reservoir Operation, Optimization

Algorithm

Introduction

– The essential need for formulation of optimal policy is to determine

of the schedule of releases from a reservoir system, which can maximizes or

minimizes the utilities associated with the release of water.

Applying optimization techniques for reservoir operation is not a new idea.

Various techniques have been applied in an attempt to improve the efficiency of

reservoir(s) operation. These techniques include Linear Programming, Nonlinear

Programming and Dynamic Programming. Recent and more advanced techniques are

Artificial Neural Network, Fuzzy logic, Genetic Algorithm, Evolutionary

Algorithm, Ant Colony Optimization, Particle Swarm Optimization and Differential Evolution etc.

A

large number of studies are reported by various investigators on optimal

operation of reservoirs, a few of the recent studies are discussed in

chronological order as below.

CRITIQUE

On the basis of study of the literature reviewed

above, it’s clear that optimization in operation of reservoir is one of the

important activities in the field of reservoir operation that aim at an

effective and efficient utilization of water with maximum benefits. Various

algorithms have been applied to optimize the Reservoir operation and maximize

the net benefit, but they have their own Advantages and Disadvantages.

The conventional Linear

Programming has been used mostly for the planning and design problems of single

reservoir systems. Natural processes are rarely linear and solving the problem

by Linear Programming forces approximations, which may

lead to either,

approximate or sometimes

even to unrealistic solutions. In addition, Linear

Programming yields only point solutions in the policy space and hence it is

unsuitable for the operational problems of reservoirs where decisions are

required to be made successively with the changing state of the system.

Incorporation of inflow stochasticity

further increases the

complexity. All the Linear Programming techniques incorporating stochasticity

viz., Stochastic Linear Programming, Chance Constrained Linear Programming,

Reliability Programming have so far been limited to the design problems of

single reservoirs.

Dynamic Programming is

suitable for sequential decision-making process of reservoir operation

problems. Its use is practically restricted to single reservoirs due to the

“curse of dimensionality”. In some cases, Deterministic Dynamic

Programming has been applied to a system of three to four reservoirs but was

found computationally inefficient. The use of Stochastic Dynamic Programming is

restricted to single reservoirs only because of its requirement of discretization

of state and space, the computer storage requirement increases exponentially

with the increase in the number of states (reservoirs).

In Artificial Neural

Network, the neural networks need training to operate. The modeling results

converge to a local minimum. Generalization and over fitting renders inaccuracy

in some cases. The model provides results, which are hard to interpret

occasionally.

Ant Colony

Optimization too has

some disadvantages like

the theoretical analysis is

difficult, sequences of random decisions are present which are not independent,

probability distribution changes with iteration, research is experimental

rather than theoretical, and time for

convergence is uncertain although convergence is guaranteed.

The simplicity in the

application of Honey Bee Mating optimization (HBMO) is quite an advantage but a

basic disadvantage of the original (HBMO) algorithm is the fact that it may

miss the optimum and provide a near optimum solution in a limited runtime

period.

Particle Swarm Optimization

(PSO) is inspired from the foraging behaviour of birds, because of fast

convergence, fewer parameters setting, and the easiness to implement. Even

though PSO is efficient, it also has some critical problems such as premature

convergence and easily drops into regional optimum. It shares many common

points with Genetic Algorithm (GA). Both algorithms start with a group of a

randomly generated population, both have fitness values to evaluate the

population. Both update the population and search for the optimum with random

techniques. Both systems do not guarantee success.

Differential Evolution

is an improved version of Genetic Algorithms. It relies on mutation rather than

Crossover. Differential Evolution has

several advantages, it can search randomly, requires only fewer parameters

setting, high performance and applicable to high-dimensional complex

optimization problems. But similar to PSO, DE has several drawbacks including

unstable convergence in the last period and easy to drop into regional optimum.

The results show the Fuzzy

inference system based reservoir operating rules (FIS-ORs) FIS-ORs to perform

considerably better than the other operating rules when they are recalibrated

every 10 year. The results also suggest that the comparative performances of

the operating rules to be influenced by weighting of the three problem

objectives. Where the weighting is such that the problem is relatively easy, it

is found the FIS-ORs to have no significant advantage.

Chaos Particle Swarm

Optimization- Differential Evolution CPSO-DE algorithm is formed by making a

few alterations in standard PSO. By sample analysis, and comparison with other

algorithms. The calculation results show that CPSO-DE improved the convergence

accuracy of PSO, the ability of global optimization, and increased the

convergence and stability at a certain degree.

Conclusion -As

every optimization model has its own limitations, the selection of appropriate

model for derivation of reservoir operating policies is difficult and more often

there is a scope for further improvement as the model selection depends on the

characteristics of the reservoir considered, on the availability of data, on

the specified objectives and constraints. In the present Study, Differential

Evolution algorithm, Partial Swarm Optimization and its Hybrids seems to be

better than the previous methods for the optimal operation of a reservoir. From

the optimization results, general operating policy can be derived.