Load distribution is a key technology in strip hot rolling process, which influences the coil′s microstructure and performance. Currently, Newton-Raphson algorithm is applied to load distribution of hot tandem mills in many hot rolling plants and has some serious defects such as having a strict restriction on initial iterative calculation value and requiring coefficient matrix of nonlinear equations to be nonsingular. To eliminate these defects and improve the online performance of the process control computer, Newton descendent numeric algorithm is introduced to this field to widen the initial value range and a new model named error conversion algorithm is put forth to deal with special conditions when the coefficient matrix is singular. Furthermore, considering the characteristics of load distribution, a condition of strip thickness distribution abnormality and corresponding solutions are provided which ensure that rolling parameters can be calculated normally. Simulation results show that the improved algorithm has overcome the defects of the Newton-Raphson algorithm and is suitable for online application. Currently, Newton-Raphson algorithm is applied to load distribution of hot tandem mills in many hot rolling plants and has some serious defects such as having a strict restriction on initial iterative calculation value and require coefficient matrix of nonlinear equations to be nonsingular. To eliminate these online and improve the online performance of the process control computer, Newton descendent numeric algorithm is introduced to this field to widen the initial value range and a new model named error conversion algorithm is put forth to deal with special conditions when the coefficient of load distribution, a condition of strip thickness distribution abnormality and corresponding solutions are provided which ensure that rolling parameters can be calculated normally. Simulation results s how that the improved algorithm has overcome the defects of the Newton-Raphson algorithm and is suitable for online application.