The nonlinearity of the power flow equation is the significant cause of the non-convexity of optimization problems in the power system. The existing linear power flow model is derived based on special mathematical approximation and numerical analysis. In this paper, a method to quantify the linearization error of the power flow equation is proposed. The truncation error of the Taylor series expansion of the power flow equation is studied. The linear power flow models with the independent variable $(v,\theta)$ and $(\pmb{v}^{\mathbf{2}},\theta)$ are compared. The proportion of power loss in high-order components is analyzed. The advantage of $(\pmb{v}^{\mathbf{2}},\theta)$ is demonstrated from a new perspective.