For the short-term scheduling of multipurpose batch process, unit-specific event-driven formulation needs many additional variables to subscribe storage, orelse it would lose some good results. To resolve this problem, a novel event-driven formulation is presented. In this model, the storage state is formulated asa time function using the existent variables, which can exactly reflect the storage state, and the dimension of the model would not be enlarged. The new model has a typical decomposition structure, which consists of two levels of optimization problems. Then a composition algorithm is proposed. Because the sub-problem mainly consists of time consequence constraints and storage constraints by relaxing, a simple algorithm based on simple deduction is given to obtain the optimalor sub-optimal solution. Using the returned solution, the master problem couldbe resolved by a genetic algorithm. So the searching space is greatly reduced. The application of the model is illustrated through two example problems, which also reflects the validity of the algorithm.