As global competition intensifies, manufacturers are increasingly challenged to adapt quickly to shifting customer demands and shorter product life cycles. In light of rapid technological development, the incorporation of intelligent forecasting into production systems has become a strategic necessity. In today’s dynamic manufacturing environment, efficiency and sustainability represent core factors in modern industrial strategies. In this context, this paper proposes a novel approach for a Reconfigurable Cyber-Physical Production Systems (R-CPPS) based on accurate demand forecasting. The proposed approach aims to simultaneously optimize critical sustainability performance metrics, including labor cost efficiency, energy efficiency, and average operator workload. Specifically, the proposed approach consists of two main steps. First, a hybrid Long Short Term Memory (LSTM) and Support Vector Regression (SVR) model is employed to forecast product demand. Then, using real-world data from an electronic product manufacturer, various demand scenarios are generated. The scenarios assess scheduling and reconfiguration strategies to optimize the three sustainability dimensions under variable demand across four product families. Assuming possible reconfiguration at the end of each day, the results demonstrated significant performance gains of the R-CPPS when applying a strategy over a 14-day forecasting window. This approach effectively enhances the system’s adaptability to demand variability while maintaining a balance across economic, environmental, and social sustainability objectives.