In 1995 Li Shoufu and Su Kai constructed a class of Parallel Hybrid Methods (PHMs) for solving stiff differential equations. The computational speed of these methods are almost the same as that of Backward Differentiation Formulas (BDFs), whereas PHMs are superior to BDFs in numerical stability properties. However stage order of PHMs is one less than convergence order in the classical sense, this disadvantageous condition leads to the reduction of stiff computational accuracy. In order to overcome the defect, in the present paper we modify PHMs approperiately and construct a class of new parallel multivalue methods. On the basis of basically keeping the various advantages of PHMs, the new methods increase stage order and B-convergence order one. Numerical experiment show that the computational speed of the new methods performed in a parallel environment are almost the same as that of BDFs of same order, wherease the stability properties and the computational accuracy is superior to BDFs of same order.
展开▼