资源约束下绿色算力系统风险响应模型  

A Risk Response Model of Green Computing Power System under Resource Constraints

作　　者：余通 黎新 山成英 汪金祥 杨冰[1] 刘思扬 YU Tong;LI Xin;SHAN Chengying;WANG Jinxiang;YANG Bing;LIU Siyang(Qinghai Branch of the People's Bank of China,Qinghai Xining 810000,China;Electric Power Research Institute of Guangxi Power Grid Co.,Ltd.,Guangxi Nanning 530000,China;Haidong Branch of the People's Bank of China,Qinghai Haidong 810001,China)

机构地区：[1]中国人民银行青海省分行,青海西宁810000 [2]广西电网有限责任公司电力科学研究院,广西南宁530000 [3]中国人民银行海东市分行,青海海东810001

出　　处：《广西电力》2024年第3期9-15,共7页Guangxi Electric Power

基　　金：国家自然科学基金项目(6217111);广西电网有限责任公司科技项目(047000KK52210036)。

摘　　要：绿色算力系统各节点间相互依赖,从底层架构到上层应用,节点关联程度高,呈现出智能联动、一体化等特征,体系内节点风险具有向关联节点传播和扩散的特点,对整个算力系统体系稳健运营带来威胁,并且对风险点的溯源抑制和风险链路的分析阻隔需要花费较高技术和资源成本。因此,提出资源约束下经济最优的绿色算力系统风险区域响应模型。首先,设计混合风险节点功率弹性压缩和过负载节点退出机制的响应策略,缓解高负载系统节点在局部区域内的聚集度。其次,在资源约束视角下,构建计及经济性最优和安全性最高的绿色算力系统区域风险评价指标,据此设计绿色算力系统风险区域应急策略,并基于群体自适应全局优化思想利用差分进化算法对响应模型进行求解,实现最优经济成本下的系统节点区域隐患抑制。最后,进行仿真实证分析,验证了所提方法的有效性。The nodes of the green computing power system are interdependent from the underlying architecture to the upperlayer application,the nodes are highly correlated,which shows the characteristics of intelligent linkage and integration.The node risk in the system has the characteristics of propagation and diffusion to the associated nodes,which poses a threat to the stable operation of the entire computing power system.The traceability and suppression of risk points and the analysis and isolation of risk links require high technical and resource costs.Therefore,an economically optimal risk regional response model for green computing power system under resource constraints is proposed.Firstly,the response strategy of power elastic compression of mixed risk nodes and the exit mechanism of overload nodes is designed to alleviate the aggregation degree of nodes in the local area of high-load system.Secondly,from the perspective of resource constraints,the regional risk evaluation index of the green computing power system with the best economy and the highest security is constructed,and the regional emergency response strategy of the green computing power system is designed,and the response model is solved by differential evolution algorithm based on the idea of group adaptive global optimization,so as to realize the suppression of hidden dangers in the system node region under the optimal economic cost.Finally,the simulation empirical analysis is carried out to verify the effectiveness of the proposed method.

关 键 词：绿色算力 资源约束 系统风险 经济最优 应急策略 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]