It is extremely difficult to realize real-time active-reactive optimal power flow (RT-AR-OPF) in distribution networks (DNs) with wind stations (WSs) due to the conflict between the fast changes in wind power and the slow response from the optimization computation. To address this problem, a new lookup-table-based RT-AR-OPF framework is developed in this paper. According to the forecasted wind power for a prediction horizon, scenarios are generated based on its stochastic distribution. The corresponding mixed-integer nonlinear programming (MINLP) problems are solved online which simultaneously optimize the active and reactive power dispatch of WSs, active-reactive reverse power flow, and discrete slack bus voltage, resulting in a lookup table. Based on the actual wind power available in a sampling time, one of the solutions will be selected and realized to the DN. A new reconciliation algorithm is proposed to ensure both the feasibility and optimality of the realized operation strategy. The applicability of the proposed framework is shown using a medium-voltage DN. (C) 2018 Elsevier Ltd. All rights reserved.