To address the complex interdependencies among demand forecasting errors, uncertainties in user response delivery, and asymmetric deviation settlement in multimarket electricity trading, this paper proposes a forecast-driven, risk-aware optimization framework for electricity procurement and sales. A CNN–LSTM model is employed to extract local load features and temporal dependencies, and produces demand forecasts for the following 24 hours. By integrating both price-based and incentive-based demand response, the framework characterizes user electricity consumption adjustments and the uncertainty in actual response delivery. The study formulates a two-stage distributionally robust optimization model to coordinate electricity procurement, retail pricing, flexibility resource allocation, and real-time corrective actions, with Conditional Value-at-Risk (CVaR) used to quantify tail risks. Based on hourly load data, the study conducts forecast performance comparisons and operational scenario analyses. The results demonstrate that the integrated model follows the observed fluctuations in electricity demand; furthermore, price-based responses and energy storage scheduling can reshape the temporal distribution of electricity procurement and consumption—though peak shaving and price reductions do not necessarily lead to a decrease in trading deviations.