Researdh And Application Of Differential Evolution In Dynamic Environments

Differential evolution (DE) is a population-based global optimization algorithm. Because of its simple structure, superior performance and easy implementation, the DE algorithm is widely used in lots of fields such as traffic optimization, industrial design and wireless sensor networks. Now, it has become a hot spot in the field of computational intelligence research.Although the DE algorithm has gotten a lot of attention and been successfully applied to a number of static optimization problems, many real-world optimization problems are complex and uncertain, the objective functions or constrains might change over time, and thus, the optimum of the problem might change as well. This brings new challenges to DE even to evolutionary computation as a whole.In this dissertation, theory, method and application of the DE algorithm for dynamic environments are studied. The dissertation mainly includes the following parts:1. The research of single-objective DE for dynamic environments is made. The dynamic optimization problem is defined and described mathematically. Then, the test functions and the performance criteria are analyzed in detail. And the predict space based multi-strategy DE is proposed. Three schemes are introduced into DE: population core based multi-population strategy, predict space scheme and the new local research strategy. The simulation results show that these strategies enhance the performance of the proposed algorithm for dynamic environments. And the proposed algorithm can track the optima better than the compared algorithms.2. The research of multi-objective DE for dynamic environments is made. The definition of the dynamic multi-objective optimization problem is given, and the test functions and performance criteria is analyzed. Then an adaptive hybrid immigration based DE is proposed. Adaptive hybrid immigration strategy, Regression forecast and gauss disturbance based forecast strategy are introduced into the proposed algorithm. The simulation results verify the effectiveness of the proposed algorithm for dynamic multi-objective optimization problems, and the results also demonstrate that the proposed algorithm is able to track the POF and POS effectively.3. The research of high dimensional DE for dynamic environments is made. As the dimension of the search space rises, almost all the evolutionary algorithm suffers "the curse of dimensionality". The analysis of reasons is made, and then the statistical correlation based cooperative DE is proposed. The simulation results verify the excellent performance of the proposed algorithm.4. The application of improved DE in the evacuation routing optimization. The analysis of the evacuation routing problem in large public place is made. In the process of the evacuation, the congestion and the speed of the pedestrians are changed over time, thus, the evacuation routing optimization problem is the dynamic optimization problem. An improved DE is proposed for the single-objective evacuation routing problem in which the evacuation time is set to be the single object and the congestion degree is set to be the constraint. The simulation shows that the proposed algorithm is competent enough to solve this problem. Finally, a multi-objective DE is introduced for solving the multi-objective evacuation routing problem, which has three objects:minimum time, shortest path and minimum congestion degree. The simulation results verify that using the proposed algorithm to solve the evacuation routing problem is feasible and effective.