In this paper, time-varying nonlinear optimization (TNO) problems are studied, and a new discrete time-varying zeroing neural dynamics (DZND) model is proposed to solve the discrete time-varying equality constrained optimization (DTECO) problems. Firstly, a continuous time-varying zeroing neural dynamics (CZND) model of equality constraints is established using the Lagrange method. Thereafter, the four-step Zhang discretization formula (FZDF) is employed to discretize the CZND model, resulting in the four-step DZND (FDZND) model. Notably, the truncation error of the FDZND model is raised to the fourth-order, thereby improving calculative accuracy. Theoretical analysis indicates that the FDZND model is convergent and stable. Besides, the FDZND model is compared with the Taylor-type DZND (TDZND) model and Euler-type DZND (EDZND) model to further verify its effectiveness and superiority. Among them, the residual errors of TDZND and EDZND are \(\boldsymbol{O}(g^3)\)  and \(\boldsymbol{O}(g^2)\) , respectively, where g denotes the sampling gap. During the experiment, the impact of different values of parameter on the residual errors and the effective interval of step size s is also assessed. The experimental results indicate that the truncation error of the FDZND model is smaller.