Abstract: To solve large-scale optimization problems(OP),the algorithm with global convergence was constructed for solving OP based on the characteristics of memory principles(MP) and cellular automata(CA)．In the algorithm,the theoretical search space of OP is divided into the discrete space,and the discrete space is defined as celullar space where each cell is an alternative solution of OP；the memorizing and forgeting rules of MP are used to control transition of states of each cell；a cellular state consists of position,increment of position and residual memory which is divided into three kinds of memory state such as instantaneous,short and long-term memory,each of which is strengthed or wea-kened by accepted stimulus strength．A cell is forgotted and then discarded when its residual memory is lower than a threshould．During evoluation process,a cell’s transferring from one state to another realizes the search of cellular space on the theoretical search space．The stability condition of a reducible stochastic matrix was applied to prove the global convergence of the algorithm．The case study shows that the algorithm is efficient.

Key words: Optimization,Cellular automata,Memory principles,Global convergence