To solve the voltage control problem caused by the high penetration of photovoltaics (PVs) in distribution networks, a day-ahead two-stage stochastic reactive power scheduling optimization method is proposed. This uses the reactive power regulation capability of flexible loads (FLs) and considers PVs and load uncertainties. In the first stage, the discrete decision variables reflecting operation of capacitor banks and scheduling of FL reactive power are optimized based on the predicted PV generation and load power. In the second stage, the continuous decision variables reflecting PV reactive output and FL scheduled reactive power are optimized considering the uncertainties from PVs and loads. Two objective functions are set in the proposed model. They minimize the voltage deviation and the amount of FL reactive power scheduling simultaneously. The ε-constraint method is adopted to tackle the multi-objective optimization model, and the entropy weight TOPSIS method is used to select the optimal compromise solution. Case studies of the modified IEEE 33-bus system show that the proposed method can effectively alleviate the voltage over-limit problem caused by PV integration.