Automated Planning is a significant challenge in Artificial Intelligence (AI), focusing on generating sequences of actions to achieve specific goals. Despite significant advancements, no single planner consistently outperforms all others across different domains. Portfolio-based approaches, which dynamically select and combine multiple planning algorithms, have successfully addressed this limitation by optimizing either computational speed or solution quality. However, these approaches have not yet considered sustainability criteria such as energy efficiency or carbon emissions. Given growing environmental concerns associated with computational resource usage, integrating sustainability into portfolio-based planning is increasingly necessary. This research addresses the critical gap by introducing sustainability metrics explicitly into the portfolio planning selection process. This thesis introduces a sustainability-aware portfolio planning approach that extends the Instance-Based Configured Portfolio (IBaCoP) framework by integrating energy consumption and carbon emissions into both configuration and evaluation. Using the CodeCarbon library, we measure the environmental footprint of planner executions and train predictive models to estimate runtime, energy usage, and emissions. These models guide the selection and scheduling of planners through two newly proposed strategies, Best-N Estimated Energy Allocation (BNEEA) and Best-N Estimated Time and Energy Allocation (BNETA). Both strategies aim to balance computational performance with energy efficiency during portfolio execution. Preliminary experiments on benchmark domains demonstrate that the proposed sustainability-aware portfolios can achieve competitive coverage and runtime performance while reducing estimated energy consumption. Overall, the results demonstrate that sustainability can be effectively integrated into portfolio-based planning, contributing to the development of environmentally responsible AI systems.