AI-Based Load Forecasting And Resource Optimization For Energy-Efficient Cloud Computing

Abstract

This study presents an AI-based framework for load forecasting and resource optimization to augment energy efficiency for managers of the cloud computing environment. The AI framework uses deep learning techniques Temporal Convolutional Networks (TCNs) and Long Short-Term Memory (LSTM) networks with low prediction errors in user workload patterns. Then, based on the load forecast, we implement a dynamic resource allocation mechanism with reinforcement learning (RL) to reduce energy use (energy consumption) and improve performance. We conduct extensive experiments leveraging benchmark cloud workload datasets to validate the model’s performance. The results indicate a prediction accuracy of 96.3%, which outperformed traditional forecasting models by an average of 8.5%. The results of experimenting with resource optimization to reduce energy consumption (or energy use enhancement) showed a total energy consumption reduction of 15.7% compared to traditional static and heuristic-based methods. Ultimately, the results indicated that the proposed system effectively addressed real-time workload adaptability that optimized resource usage and maintained service quality. The findings suggest that the AI framework can enable the development of intelligent, sustainable, and environmentally conscious cloud computing infrastructures.