A novel online final product quality prediction scheme is proposed in this paper for the improvement of quality prediction in multi-phase batch processes. Phase cumulative product quality (PCPQ), which is quality cumulated from the beginning of the phase to the end, is introduced for quality prediction, and final product quality prediction offline is achieved by cumulating all the predicted PCPQ based on the corresponding PCPQ model. In this way, the quality prediction approach proposed not only explores the different effects of process variables in different phases on final product quality, but also takes the common effects of process variables in different phases into account. The PCPQ model and remained phase cumulative product quality (RPCPQ) model are combined to improve online prediction precision, without the missing observations estimation. The proposed approach is applied to a simulated penicillin fermentation process and the results of simulation demonstrate effectiveness and superiority.