Genetic algorithms in Stochastic optimisation

Genetic Algorithms (GA) have been successfully used in a variety of optimisation problems. They are especially strong in the solution of difficult problems, which cannot be solved or are hard to solve using conventional linear or non-linear optimisation. One of those problems is the Constrained Stochastic Optimisation (CSO) problem. The central characteristic of these kinds of problems is that some or all variables of the problem are given in the form of random variables. Random variables capture the uncertainties associated with system behaviour. These kinds of variables must be used whenever the problem parameters fluctuate within very large range of values and/or it is difficult to assess their expected values. Problems of this type arise in a variety of engineering fields, in power systems, transport engineering, internet access, communication networks, etc. In these and many other areas, the system has to be designed for mid to long-term optimum operation forcing the design engineer to use CSO models. Solution of the CSO problem using conventional methods is very complicated. Genetic Algorithms offer simple yet accurate solutions using computer efficient techniques. To illustrate the method, the problem of finding the optimum design of an Intranet server is solved.