Scaled physical model tests for steam breakthrough were conducted based on the analysis of mechanisms and influence factors of steam breakthrough. Physical simulation results showed that at the initial steam breakthrough, preferential flow channels were formed in narrow sand packs and most residual oil left in these channels was immobile. This shortened the steam breakthrough time of follow-up steam flooding and decreased the increment of oil recovery efficiency. Steam breakthrough occurred easily for a smaller producer-injector spacing, and a bigger difference in physical properties between fluids and rock. Steam breakthrough is more likely to occur at a larger formation permeability (k), greater steam displacement velocity (u) and smaller producer-injector spacing (L). Steam breakthrough time is a function of the parameter group (uk/L), i.e. tb=3.2151 (uk/L) -0.5142. A non piston-like displacement model was built based on steam breakthrough observation for a steam stimulated well in the Jinglou Oilfield, Henan Oilfield Company. The steam volume swept in different directions could be obtained from inter-well permeability capacity and breakthrough angle, and the steam swept pore volume (SSPV) was also determined. Numerical simulation showed that steam sweep efficiency reached its peak value when a slug of profile control agent (slug size 10%-15% SSPV) was set at one half of the inter-well spacing. Field test with 12.5% SSPV of profile control agents in the Jinglou Oilfield achieved success in sealing breakthrough channels and good production performance of adjacent producers.