The transition to net-zero greenhouse gas (GHG) emissions presents a major challenge for the manufacturing sector, which relies heavily on electricity. Grid electricity has a time-dependent carbon intensity that fluctuates with the energy sources used for its production. While on-site renewables offer a greener alternative, their availability is limited and variable. To achieve net-zero, dynamically adjusting electricity consumption to align with fluctuating grid carbon intensity and renewable availability is crucial. Carbon-aware scheduling offers a promising solution by optimizing electricity use to minimize GHG emissions, considering real-time grid carbon intensity and renewable availability. In this project, we will develop a robust framework for carbon-aware scheduling in real-world manufacturing environments. First, we will create a standardized instance benchmark including energy-related data. Then, we will design advanced optimization techniques to address the complexities of flexible carbon-aware job-shop scheduling, incorporating time-dependent job and setup power requirements. To manage daily variability in renewable availability and grid carbon intensity, we will develop tuning-free, self-adaptive algorithms using deep reinforcement learning, enabling rapid deployment across diverse energy profiles. Finally, a closed-loop rescheduling framework will ensure robustness by dynamically adapting schedules to evolving energy forecasts throughout the planning horizon.