Accurate forecasting plays a crucial role in strategic planning across diverse domains, enabling informed decision-making and proactive adaptation to uncertainty. While short-term forecasts can be relatively reliable, their accuracy declines over longer periods as unpredictability increase. To address this, probabilistic and stochastic approaches provide a more refined way of capturing uncertainty, offering data-driven insights that help in making more adaptive and informed choices. Integrating these advanced forecasting methods into decision-making processes enhances flexibility, ensuring that strategies remain adaptive in the face of evolving conditions. In this study, we propose a comprehensive framework integrating probabilistic predictions with stochastic modeling to enhance long-term energy system planning. The proposed framework leverages Bayesian Neural Networks (BNNs) to forecast energy demand while quantifying uncertainty. The uncertainty bounds are then used to generate data-driven scenarios, incorporated into a multi-period stochastic programming based on mixed-integer linear programming (MILP), allowing for the exploration of adaptive investment and planning strategies within the Texas energy market. To evaluate the value of incorporating uncertainty, the performance of the proposed stochastic framework is benchmarked against a traditional deterministic model. The results demonstrate that incorporating uncertainty not only leads to more flexible planning outcomes but also provides a more realistic foundation for long-term decision-making in complex energy systems.