Shale gas development has become a global energy hotspot. After large-volume fracturing of shale gas reservoirs, sand production in wellbores has become a frequent phenomenon. To address the sand plugging problem of shale gas wellbores, a wellbore gas-liquid sand-carrying experiment was conducted, and the three-phase flow characteristics of gas, liquid, and sand in horizontal wells were investigated using CFD to analyze the influences of the main factors on the gas-liquid sand-carrying capacity. Within the range of the experimental parameters, the sand-carrying capacity of the wellbore increases with the increase of the gas production rate, and with the increase of the pressure gradient of the wellbore, water-producing gas wells have a stronger sand carrying capacity than non-water-producing gas wells. Based on extensive experimental data, a prediction model for the sand-carrying ratio by gas–liquid flow within wellbores was developed using a backpropagation (BP) neural network algorithm, the model takes gas production rate, liquid production rate, and wellhead pressure as inputs, and outputs the sand-carrying ratio (ε), which characterizes the sand-carrying capacity and risk of sand accumulation in wellbore. Based on the sand-carrying ratio prediction model, a production chart and a production optimization method for sand-control is proposed to support shale gas production.