Although voltage control of an active distribution network (ADN) is regulated by voltage regulation devices which can be classified according to the speed of voltage regulation, devices, slow devices lack intra-day regulation flexibility, whereas fast devices have a narrow voltage regulation range. To enable intra-day cooperative regulation of fast and slow devices in an ADN, we propose a voltage control strategy comprising a mixed-integer nonlinear programming (MINLP) model considering predictive variables. We introduced a strict time constraint, limited the solution range via scene filtering, and converted discrete variable to continuous ones based on Gaussian penalty function (GPF), which was further solved using the firework algorithm (FWA) to obtain quick and accurate optimization results. Compared to the traditional voltage control strategy, the proposed strategy can increase voltage control range and accelerate convergence speed, and its effectiveness was verified using an actual circuit in the Anhui province.