#SAT is used widely in the field of artificial intelligence, and many real-world problems can be reduced to #SAT to get the number of models of a propositional theory. By an in-depth study of #SAT solvers using extension rule, this paper finds that the order of reduction clause has great impact on the size of maxterm space. Thus, in this paper two heuristic strategies, MW and LC&MW are proposed to enhance the efficiency of solving #SAT. MW chooses the clause with maximum weight as reduction clause in every calling procedure; LC&MW chooses the longest clause as reduction clause in every calling procedure and takes the clause with maximum weight as tie breaker. The algorithm CER_MW is designed by using MW, and the algorithm CER_LC&MW is designed by using LC&MW. According to the experimental results, CER_MW and CER_LC&MW show a significant improvement over previous #SAT algorithms. Comparing the new #SAT solvers with other state-of-the-art #SAT solvers using extension rule also shows that CER_MW and CER_LC&MW are significantly improved over the previous #SAT algorithms in solving efficiency and solving capacity. In terms of efficiency, the speed of CER_MW and CER_LC&MW is 1.4~100 times that of other algorithms. In terms of capacity, CER_MW and CER_LC&MW can solve more instances in a time limit.