Research Of Dynamic Optimization Method Based On Evolutionary Algorithms

Evolutionary Algorithms (EA) are widely applied to static optimization problems and a considerable number of valuable achievements have been reported. However, many real-world optimization problems are dynamic, time-varying, they will change when the fitness function, constraints, or environmental parameters change from time to time, frequent changes in the solution space will make the optimal solution also change over time, and this is the dynamic optimization problems. As a result of traditional evolutionary algorithms will lose the ability to adapt to environmental changes in the late period of the evolution. So improving EA to solve dynamic optimization problems, tracking changes in a dynamic environment has become a new research field.First of all, on the basis of making an overview of dynamic optimization problems and evolutionary algorithms, memory enhanced genetic algorithms (MEEA) are introduced, which is one of typical memory-based evolutionary algorithms to solve dynamic optimization problems. The simulation results show the method is superior to the traditional EA in the cyclical changesThen, according to the limitations of detection method of environmental change in the traditional EA, an improved detection method is proposed. A detection method -based EA is also raised up. The algorithm not only applies the new detection method, but also adds a pretreatment operator to improve performance of the algorithm. The algorithm is tested in a simulated dynamic environment, and experimental results show that the tracking performance of the algorithm is better than MEEA.Finally, the fixed memory size is the reason that the traditional memory-based EA can't show the expected performance. A variable size memory-based EA is presented, in which the size of memory and the population change in some way. In order to promote and maintain diversity, this algorithm also applies hyper-mutation. Simulation results show that the tracking error of the algorithm is less than MEEA, and thus show the algorithm is effective.