To identify the transient and distributed internal surface heat flux of the slab mold in continuous casting process, a fuzzy inference method is proposed in this work. For temporal and spatial distribution characteristics of the internal surface heat flux of continuous casting mold, a decentralized fuzzy inference (DFI) identification scheme possessed of a decoupling characteristic in time and space is established. For each temperature measurement point, the fuzzy inference processes are, respectively, executed from the correspondingly observed temperature sequence through corresponding DFI units. In the time domain, according to sensitivity coefficients, the weighing and synthesizing processes for the decentralized inference results are performed to get the temporal compensation vector for the internal surface heat flux of mold. Then, in the space domain, according to the normal distribution function, the weighing and synthesizing processes for the temporal compensation vectors are performed to get the spatial compensation vector for the internal surface heat flux of mold. Numerical tests are carried out to research the influence of the number of thermocouples and measurement errors on the identification results, which prove the effectiveness of proposed scheme in this work.