Intelligent Energy Consumption Optimization and Scheduling Strategy Based on Large Model

Abstract

At a time when energy resources are becoming increasingly scarce, achieving intelligent energy consumption optimization and efficient scheduling has become a key path to alleviating energy pressure. This paper focuses on the research of intelligent energy consumption optimization and scheduling strategies based on large models and proposes an adaptive dynamic programming collaborative reinforcement learning algorithm (ADP-CRL). After an in-depth analysis of the current energy dilemma and the limitations of existing research, the unique design idea of the ADP-CRL algorithm that combines the global planning ability of adaptive dynamic programming with the dynamic environment adaptation characteristics of reinforcement learning is explained in detail. In order to verify the performance of the algorithm, a simulation environment containing a variety of energy-consuming devices (such as servers, air conditioners, etc.) and diversified task loads (computation-intensive, data transmission, etc.) was constructed and compared with traditional greedy algorithms and rule-based scheduling algorithms. Experimental data show that under the same task load conditions, the system energy consumption of the ADP-CRL algorithm is reduced by 25.3% compared with the traditional algorithm, the average task completion time is shortened by 18.7%, and the resource (CPU, memory, etc.) utilization rate is increased by 22.1%. This fully demonstrates that the ADP-CRL algorithm has significant advantages in intelligent energy consumption optimization and scheduling and provides a practical new solution for improving energy utilization efficiency, which is expected to play an essential role in actual energy management scenarios.