资源环境科技发展态势分析平台

This paper presents a simulated annealing multiobjective algorithm developed to include novel features for promoting the convergence toward the best Pareto front that shows diversity and uniformity in the distribution of solutions. The new algorithm, MultiObjective Simulated Annealing with new Generation and Reannealing procedures (MOSA-GR), is a trajectory-based algorithm, where, in a first phase, diversified generation strategies are used to define candidate solutions at different stages of the search procedure. In a second phase a reannealing process starting at low temperature is implemented to intensify the search based on the last solutions of the first phase. An extensive comparison with solutions obtained by various multiobjective evolutionary algorithms (MOEAs) described in the literature is provided. Using a limited budget for the number of function evaluations, the results show that MOSA-GR has a very good performance and gives rise to better fronts, even compared with those obtained by merging the results from different MOEAs implemented in similar conditions.