The cement clinker manufacturing process is a complex process of the conversion and consumption of thermal energy. An energy balance analytical model that can be used for on-line identification is established, aiming at solving the problem that the thermal efficiency can’t be on-line monitored. Considering the input matrix is close to a singular matrix under steady-state condition, a recursive singular value decomposition (RSVD) identification algorithm that can overcome the illconditioned problem in linear regression algorithm is adopted for the on-line estimation of model parameters. In order to avoid the parameter vectors entering complex field in the recursive process, the initial values of the recursive algorithm are improved. The simulation results based on historical data show that the established model and the improved identification algorithm can effectively calculate the thermal efficiency of the cement clinker manufacturing process.