In order to precisely forecast the power of Tower combined diamond circular saw blade during material sawing, a model for sawing power was established using the average thickness of an undeformed chip from a single grinding grain of a solitary saw blade within the combination saw as a medium, and was thereafter refined. A fast but accurate prediction model, requiring only a small number of samples, was presented. Sawing power was measured through a range of parameter combinations via sawing experiments, and model coefficients were obtained through fitting the data using multivariate linear regression techniques. An optimization model was then established, with sawing parameters as optimization variables. The objectives of this model were to minimize the sawing specific energy and to reduce the sawing time. An optimized particle swarm algorithm was adopted to solve the model. The experimental results reveal that the parameter model can completely elucidate the influence of various sawing parameters on sawing power, with the model accurately forecasting the sawing power under different saw blade combinations. The improved particle swarm algorithm displays strong optimization performance, with optimized parameters contributing to significant reductions in sawing power.