The fast development of cloud computing has increased global energy consumption and carbon emissions, which has led to a pressing need of smart and sustainable to the environment workload management techniques. Despite the recent studies on energy-conscious and reinforcement learning-driven task scheduling, the current strategies usually do not have the ability to predict carbon-intensity, integrate renewable energy, optimize workflow, and provide real-time flexibility of cross-cloud and edge computing. To counter these drawbacks, this paper gives the Intelligent Green-Cloud Scheduling Framework (IGCS), which is a multi-objective, deep reinforcement learning-based system to reduce carbon emissions at the cost of the highly-reliable performance and services. The suggested architecture combines the carbon-intensity prediction, renewable energy forecasting, thermal-aware resource consolidation, and dynamic scaling to be able to place workloads proactively and dynamically and context-sensitive. Multi-objective DRA scheduler assesses the information of system state, in terms of estimated carbon values, workload parameters, thermal fluctuations and renewed conditions, to optimize the assignment of tasks in heterogenous cloud areas. Decision logic Workflow-aware is used to ensure that the scheduling of dependent tasks is efficient and the overhead of cross-region communication is reduced. The adaptive control of migration and resource in response to varying workloads and carbon conditions is made available through real-time monitoring. Empirical investigations with actual-world carbon measurements, renewable-energy measurements, and Google cluster workloads prove that IGCS can result in large carbon emission and energy savings in contrast to baseline RL, heuristic and metaheuristic scheduling techniques. The outcomes also indicate better workflow stability, less thermal load, greater efficiency of the resource use, and better SLA adherence. In general, the Intelligent Green-Cloud Scheduling Framework creates a scalable and practical channel towards sustainable cloud computing through the integration of predictive modeling, deep reinforcement learning and carbon-sensitive resource management.