检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:李骏翔 黄亮 吴成铭 吴超 何政达 刘松楠 LI Junxiang;HUANG Liang;WU Chengming;WU Chao;HE Zhengda;LIU Songnan(Zhejiang Post and Telecom Engineering Construction Co.Ltd.;School of Public Affairs,Zhejiang University,Hangzhou 310052,China)
机构地区:[1]浙江省邮电工程建设有限公司 [2]浙江大学公共管理学院,浙江杭州310052
出 处:《软件导刊》2025年第4期101-107,共7页Software Guide
基 金:杭州市重大科技创新项目(2022AIZD0035)。
摘 要:针对电信机房空调能耗高且难以自动调控的问题,深度强化学习算法通过与环境交互的反馈结果优化模型参数,可以在满足机房安全约束的条件下自适应设定空调温度以达到节能的目的。然而,深度强化学习算法需要在生产环境中运行以获得训练样本,在安全性要求以及调控频率的限制下,算法需要较长时间训练才能收敛至最优状态。为此,提出一种基于策略蒸馏的空调节能控制方法,通过策略蒸馏算法将已完成训练的机房模型经验迁移至新机房模型内,使新机房在迭代初期便具备较好的调控效果。实验结果表明,该方法在新部署机房的前30天平均节能率为16.33%,相较随机初始化模型参数方法的平均节能率提高10.8%,且具有更快的收敛速度。In response to the problem of high energy consumption and difficulty in automatic regulation of air conditioning in telecommunica‐tions rooms,deep reinforcement learning algorithms optimize model parameters through feedback results from interaction with the environ‐ment,and can adaptively set the air conditioning temperature to achieve energy-saving goals while meeting the safety constraints of the com‐puter room.However,deep reinforcement learning algorithms need to run in production environments to obtain training samples,and under se‐curity requirements and frequency constraints,algorithm training takes a long time to converge to the optimal state.To this end,a strategy dis‐tillation based energy-saving control method for air conditioning is proposed.Through the strategy distillation algorithm,the experience of the trained data center model is transferred to the new data center model,enabling the new data center to have good control effects in the early stages of iteration.The experimental results show that the average energy-saving rate of this method in the first 30 days of new deployment in the data center is 16.33%,which is 10.8%higher than the average energy-saving rate of the random initialization model parameter method,and has a faster convergence speed.
分 类 号:TP399[自动化与计算机技术—计算机应用技术]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.90