Parallel And Dual-systems Cooperative Co-evolutionary Differential Evolution Algorithms And Their Application

Co-evolutionary algorithm is a new optimization algorithm, which is inspired by biological co-evolutionary phenomenon in natural. In recent years, co-evolutionary algorithm has become an important research field in evolutionary algorithm, which provides an effective way for solving complex optimization problems. Co-evolutionary algorithm is also an efficient algorithm for solving engineering system optimization problem, known as the engineering systems design method. This method has gradually caused extensive concern of engineering fields.The basic idea of co-evolutionary algorithm is that multiple species co-evolve, and individuals are evaluated by their performance relative to other individuals. According to the ways to evaluate individuals, co-evolutionary algorithm is divided into two categories: competitive co-evolutionary algorithm and cooperative co-evolutionary algorithm. The thesis mainly studies the Cooperative Co-evolutionary Algorithm (CCEA). Cooperative co-evolutionary algorithm for solving high dimensional optimization problems shows better computing performance, but there are still insufficient. This paper focuses on two problems. First, how to better implement the parallel computing of CCEA and further improve the calculation accuracy and efficiency of CCEA. Second, how to improve the ability to solve strong coupling problems (especially for non-separable problems) by CCEA. So this paper discusses on improving CCEA for solving strong coupling optimization problems with high computational expense and related fields. This paper presentes a parallel cooperative co-evolutionary differential evolution(PCCDE) and a dual-systems cooperative co-evolutionary differential evolution(DCCDE). The main contributions are as follows:(1) To solve the high computational complexity of cooperative co-evolutionary algorithm for large-scale optimization problems, the paper presentes a Parallel Cooperative Co-evolutionary Differential Evolution (PCCDE) algorithm based on Cooperative Co-evolutionary Differential Evolution (CCDE) and Bulk Synchronous Parallel Computing Model (BSP). In this algorithm, an improved Archive collaboration mechanism replaces the original mechanism in CCDE. Then, the PCCDE uses the BSP model to implement the parallel computation. Finally, the simulation results based on a set of widely used benchmark function have indicated that the algorithm has superior calculation efficiency and quality. (2) To solve strong coupling optimization problems by cooperative co-evolutionary algorithm, this paper proposes a Dual-systems Cooperative Co-evolutionary Differential Evolution (DCCDE) based on dual-systems cooperative co-evolutionary framework. The co-evolution of DCCDE uses two systems A and B. The individual migration between the sub-populations of two system increases diversity of the population and improves the coordination between subsystems. DCCDE improves the coordination mechanism of DVGCCGA, and replaces the genetic algorithm (GA) with efficient global search algorithm DE. Simple cross local search strategy is employed to enhance the DCCDE computing performance. To improve the capability of DCCDE for solving coupled problems, the three main aspects of the variable grouping, coordination mechanism and exploration capabilities are studied. With a test suit 20 benchmark functions, the numerical results indicate that DCCDE has more competitive in solution quality and premature convergence compared with some similar algorithms. It also showes that the algorithms based on CCDE framework for solving complex coupled optimization problems has a great potential.(3) The experimental results from layout of design of satellite module problem show that the PCCDE and DCCDE algorithm can achieve satisfactory optimization solution. PCCDE and DCCDE not only solve high dimensional function optimization problem but also are used to optimize a class of complex layout design problems. This work expects to help CCEA to apply in engineering coupled system and be conducive to the theory study of co-evolutionary algorithm.