Energy saving strategy of cloud data computing based on convolutional neural network and policy gradient algorithm.

Area of Science:

• Computer Science
• Artificial Intelligence
• Energy Management

Background:

• Cloud Data Computing (CDC) requires efficient energy management for user data centers.
• Real-time monitoring of Information Technology (IT) equipment energy consumption is crucial.
• Existing methods struggle with the variable energy consumption, load, and power demands of CDC.

Purpose of the Study:

• To develop intelligent energy-saving strategies for safe, visualized energy management in CDC.
• To design an energy-saving model using Convolutional Neural Network (CNN).
• To create a task scheduling model for CDC addressing uncertainty and volatility using Policy Gradient (PG) algorithms.

Main Methods:

• Utilized Convolutional Neural Network (CNN) for an intelligent energy-saving model.
• Implemented Policy Gradient (PG) algorithms for CDC task scheduling.
• Analyzed neural network model performance on energy consumption and load optimization.
• Simulated CDC task scheduling using PG algorithms to assess demand.

Main Results:

• CNN-based energy-saving model outperformed Elman and ecoCloud algorithms in energy consumption.
• CNN reduced virtual machine migrations by 9.30% compared to the Elman algorithm.
• Deep Deterministic Policy Gradient (DDPG) algorithm showed superior performance in task scheduling with an average response time of 141.
• Deep Q Network (DQN) algorithm performed poorly in task scheduling.

Conclusions:

• Deep Reinforcement Learning (DRL) and neural networks effectively reduce CDC energy consumption.
• These methods significantly improve the completion time of CDC tasks.
• The study provides a valuable research reference for CDC resource scheduling and energy optimization.