Research On Artificial Immune Algorithm And Its Applications In Power System Planning

The biologic immune system can defend the invading antigens effectively and keep various antigens coexist. As a new branch of computational intelligence, the Artificial Immune System (AIS) has characters of diversity, distributivity, dynamics, self-adaptability, robustness and so on. It has strong ability to process information and solve complex problems, and as a result, it has been paid significant attention recently.Based on the introduction to the immune system, immune mechanism, AIS and its application status, a new algorithm inspired by the immune system is developed for multimodal problems, and its basic theory is discussed. Moreover, some optimization problems of power systems are solved by the immune algorithm in this dissertation. Contributions are summarized as follows,(1) The disadvantage of the genetic algorithm which can not solve multimodal problems effectively for premature convergence is discussed. Based on the immune diversity principle and the immune network theory, a multimodal immune algorithm (MIA) is proposed which adopts the mutation with the high-bit remained and the multi-population parallel evolution method. The key parameters are adjusted self-adaptively in order to control the search scope and the precision degree. The search ability of the mutation operator and the global convergence are analyzed theoretically, and the mechanism of MIA to prevent premature convergence is also presented. By comparing with the algorithms of sharing, deterministic crowding and restricted competition selection, the results show that MIA has strong ability of exploitation the best solutions and exploration in whole search-space.(2) The transition probability of MIA's immune operators is obtained by using stochastic process theory. Since the multimodal algorithm requires to converge to multi extreme points, it is proved that MIA is completely convergent based on Markov chains, that is to say, the algorithm can converges to all the peaks in the definition scope with probability of 1, as t→∞. The typical multi-modal optimization functions are used to analyze and to test the algorithm from the respects of global convergence, complete convergence, convergence rate and stability. Comparative analysis between MIA and other immune algorithm shows that MIA can search for multiple quasi-optimal solutions simultaneously, including all local optimum and global optimum, with diverse population and steady convergence. The validity of the theoretical analysis is also verified by the simulation.(3) Integrating the three dimension geographic information system (3DGIS) with the artificial immune algorithm, an exploring study on intelligent optimization of power transmission line planning is carried on. A new modal based on 3DGIS is proposed to estimate the power transmission line cost considering both direct cost of line construction and indirect cost of environment impact. An immune algorithm of this model is also presented. Integer dynamic coding strategy is adopted. A new way to produce initial population is designed to avoid invalid route. Global and self-adaptive vaccinations are introduced for the importance of experience in engineering, which increases the convergence speed. Furthermore, self-adaptive mutation and immune regulation are used to maintain the diversity of population and to enhance local search ability. The proposed modal and algorithm are applied to a practical system and the results prove its feasibility and validity.(4) An immune fuzzy algorithm is presented and used in transmission network planning. The concentration of antibodies was adjusted based on the conception of informative entropy of the gene. Considering both the probability of affinity and the probability of concentration suppression, the algorithm can keep the specificity of the antibodies and maintain the diversity of the population at the same time. The mutation rate of the somatic hypermutation can be adaptively set according to the changes of input values of fuzzy control, which can increase the diversity of the population, and can search the local optima simultaneously. The simulation demonstrated that the proposed algorithm can escape the premature convergence with better performance than that of other methods. The feasibility and effectiveness of this algorithm for transmission network planning are also proved.