Research On Bacterial Foraging Optimization Algorithm

Traditional optimization methods have been unable to deal with mathematical modeling of engineering problems which tend to be high-dimensional, non-linear and multiple extremisms with the advancement of human science and technology. The demand for more efficient optimization algorithm is becoming increasingly urgent. Swarm intelligence optimization algorithms based on bionic arise at the historic moment and are used in practical optimization problems widely because of its advantages like parallel search and global search. Bacteria Foraging Optimization algorithm(BFO) is a kind of heuristic swarm intelligence algorithm proposed in recent years, its theoretical basis is the foraging behavior mechanism of bacterial and it has many advantages such as speediness, parallelism, and robustness. But as a new kind of optimization algorithm, some defects still exist in BFO, such as premature convergence or late convergence, efficiency decreased with the dimension increased and difficult to find the global optimal solution when solve multimodal optimization problems.Several aspects of BFO such as basic theory, realization, test function, performance measuring indicators and optimization process are systematically discussed in this thesis. According to the theoretical analysis and research result of advantages and disadvantages, an Improved Bacteria Foraging Optimization algorithm(IBFO) is proposed. First, the algorithm adopts a multi-field self-adaptive chemotaxis strategy to improve the optimization precision of BFO. IBFO selects the favorable chemotaxis direction by Metropolis criterion and adjusts the chemotaxis step by using the self-adaptive shift factor. In this way, bacteria always move fast in the fitness direction and change to meticulous search when to close to the optimal point in chemotaxis operation. Second, a premature convergence avoided way based on group diversity is proposed. The algorithm adjusts the group diversity of bacteria to make it in the appropriate threshold by leading disturbance and migrating operation, thus the premature convergence is avoided and the overall performance of algorithm is increased. Third, inspired by the Harmony Search optimization algorithm(HS), the regional dimension migration way is designed to judge and handle the bacteria out of the boundary, retain the elite bacteria, and update the bacterial to be migration according to fitness, thus the search efficiency is increased.Ten sets of experiments indicate that the improved algorithm outperforms the classic algorithm not only in terms of the solution accuracy, but also convergence speed. The improved algorithm can improve the solution precision about five orders of magnitude, increase the convergence speed about 50%, and solve the problem of premature convergence to some extent.It needs to study the effect of parameter settings on the algorithm further, and improve the BFO from different aspects on the basic of summarizing the practical engineering problems in the subsequent research work.