The invention relates to a submerged arc furnace working condition prediction method based on confidence rule base reasoning, which comprises the following steps of: analyzing and mining a characteristic vector related to working condition prediction by utilizing expert experience knowledge, industrial field data and a mechanism, establishing a belief rule base, and reflecting a complex nonlinearmapping relation between an input parameter variable and working condition output by using the belief rule base; adopting a real-time decision filtering algorithm to carry out real-time filtering on industrial data, so that the characteristic vector more truly reflects the change of the internal working condition of the submerged arc furnace; And establishing a confidence rule base nonlinear optimization model, so that the problem of inaccurate parameter initial values given by experts is solved. The method not only can improve the intelligent level of production operation, but also can improve the production efficiency, and is very important for later power demand prediction to realize energy conservation and emission reduction in the production process and energy efficiency optimizationscheduling.