Abstract: Optimization in random walking algorithm with compulsive evolution (RWCE) for heat exchanger network synthesis (HENS) turns to slow down in late evolution as heat load of all heat exchangers are optimized simultaneously. A random walk algorithm with compulsive evolution combined with restrictive-evolution strategy for heat unit (RS-RWCE) is proposed, in which number of evolved heat units in HEN is restricted in each random walk evolution to keep fast convergence of total annual cost in early evolution process and fine search in late evolution process. Specific examples are applied to verify high computational efficiency and accuracy of the strategy. It gives consideration to integer variable and continuous variable. The results are encouraging.