FAO “3414” design plus no fertilizer treatment called “3415” design was used in field plot experiment to investigate the effect of different application of N, P and K fertilizer on flue-cured tobacco under local rice-tabacco rotation system. The results showed that in the region applying rice-tabacco rotation, soil contributed little to the yield of tobacco crops, yield in no fertilizer area only accounted for 39.4% of that in area with balanced NPK fertilization; with the increasing of N, P and K, the yield increased first, and then declined. The fertilizer treatment with the balanced NPK showed the highest tobacco yield (3 315.9 kg/hm2), the best economic yield (34 371.4 yuan/hm2) and the best yield increasing effect. The lack of N influenced the yield of tobacco most, followed by P and K. The yield in treatments with deficiency of N, P and K accounted respectively for 78.3%, 79.4% and 89.7% of that in treatment with balanced NPK fertilization. The data from “3415” experiment did not fit the ternary quadratic fertilizer response function, but fit the one-place quadratic fertilizer response function. By one-place quadratic fertilizer response function, the best application of N, P and K were 159.8 kg/hm2, 86.1 kg/hm2 and 152.2 kg/hm2 respectively and the optimum yield was 3 182.8 kg/hm2, approaching the best application amount of N, P, K and the highest tobacco yield in field plot experiment.