A gradient-guided niching method in genetic algorithm for solving continuous optimisation problems

A hybrid genetic algorithm, which embeds a gradient-based local search route into a niching genetic algorithm, is proposed for solving continuous optimisation problems. The optimisation algorithm is applied to three nonlinear functions each having up to 100 variables and multi-minima. The test results show that relative to a standard niching algorithm the combination of a gradient-based search and niching improves the searching precision by several orders and the capability for locating the global optimum is significantly improved.