Nature Inspired Real-Valued Optimization And Applications

Real-valued optimization has widespread applications in the real world. Many scientific and engineering problems can be solved as real-valued optimization prob-lems. Due to their ease of implementation and capability of global optimization, nature inspired optimization algorithms, which mostly are a class of population-based meta-heuristic methods, have received intensive attention in the area of real-valued optimiza-tion. However, these methods often suffer from premature convergence, difficulty of tuning control parameters, and unsatisfactory scalability. Such limitations restrict their practicability in real-world applications. Based on the current development of nature inspired optimization research, this dissertation has the following three main goals:1. Researching and designing efficient and easy-to-use nature inspired algorithms for real-valued optimization problems.2. Developing nature inspired algorithms that are able to tackle large-scale opti-mization problems.3. Validating the efficacy of the proposed algorithms by applying them to real-world applications.With these targets, in this dissertation we carried out a series of fundamental al-gorithm design, analyses and applications based on the research of population-based nature inspired computation paradigms, i.e. differential evolution and cooperative co-evolution. Specifically, the main contents and contributions of this dissertation are as follows.1. Based on the analysis of the search characteristics in differential evolution, a more efficient algorithm, i.e. differential evolution with neighborhood search, is proposed by adopting Gaussian and Cauchy operators to control the search step size.2. Several parameter self-adaptation schemes are proposed through the discussions on existing strategies. An efficient and easy-to-use self-adaptive differential evo-lution is proposed on the basis of these parameter adaptation mechanisms.3. The idea of decision variables grouping for problem decomposition is introduced to design a new cooperative coevolution framework for large-scale optimization. Based on the framework, a new algorithm with the ability to tackle real-valued optimization problems with up to 1000 variables is proposed.4. Group size is hard to determine in the grouping-based cooperative coevolution framework. To deal with the problem, a multilevel grouping strategy is proposed to pursue the self-adaptive control of the group size parameter.5. By formulating the parameter identification in building thermal models as a real-valued optimization problem, an improved self-adaptive differential evolution al-gorithm is applied to identify the focused system parameters.6. A cooperative coevolution algorithm is proposed for a shape design optimiza-tion application. The efficacy of the proposed method is evaluated using two 2-dimensional target shape design optimization problems.In summary, this dissertation first proposes several efficient and easy-to-use dif-ferential evolution algorithms. Then, decision variables grouping and multilevel group-ing methods for large-scale problem decomposition are introduced into the cooperative coevolution framework, which results in two new algorithms that are able to tackle real-valued optimization problems with up to 1000 decision variables. Moreover, the designed algorithms are applied to two practical applications to further verify their practicability. The work in this dissertation is of significance not only to the funda-mental research of nature inspired optimization, but also to the development of related application domains.