Heterogeneous networks are prominent and promising architectures for extended usage and real time communication processes. The placement of small cell Base stations the densely populated to increase the services offered to multiple users simultaneously and hence constitute heterogeneous networks. Energy requirements computation of HetNets is challenged by continuous demand, leading to be deployment of substantial energy harvesting mechanisms. Energy utilization and resource allocation determine the performance of heterogeneous networks, apart from user scheduling and implementing hybrid energy sources. This research article contemplates a model of users scheduling and a load balancing resource allocation scheme, in turn improving the energy efficiency of overall network. The properties of wireless channel communications and environmental conditions are un-predicted and stochastic in nature, reinforcement learning technique is implemented to update the optimal performance in every iteration. Identified problem in our proposal is to acknowledge the continuous valued actions and States of a hybrid model, which is approached by a user-demand deterministic algorithm. The users are derived based on Gaussian distribution, and defining their demands in a continuous state as actions. A gradient ascent methodology is implied to derive every attribute with respect to stochastic actions identified from continuous States. Function approximation analyses the performance of load distribution and resource allocation according to the demands of users. The proposed model is evaluated with simulations to acknowledge the performance of HetNets in terms of improvised energy utilization, optimized users scheduling and conservation of energy through deploying green energy.

异构网络是用于扩展使用和实时通信过程的突出且有前途的架构。部署密集的小小区基站以增加同时提供给多个用户的服务，从而构成异构网络。HetNets 的能源需求计算受到持续需求的挑战，导致部署大量的能量收集机制。除了用户调度和实施混合能源之外，能源利用和资源分配决定了异构网络的性能。本研究文章设想了一个用户调度模型和一个负载均衡资源分配方案，进而提高整个网络的能源效率。无线信道通信的特性和环境条件本质上是不可预测的和随机的，实施强化学习技术以在每次迭代中更新最佳性能。我们的提案中确定的问题是确认混合模型的连续值动作和状态，这是由用户需求确定性算法处理的。用户是基于高斯分布推导出来的，并将他们在连续状态下的需求定义为动作。隐含梯度上升方法来推导与从连续状态识别的随机动作相关的每个属性。函数逼近根据用户的需求分析负载分配和资源分配的性能。