Abstract: To solve the problem of low efficiency in industrial computing first network construction caused by the contradiction between the limited resources of computing first accessing servers and the instability of the system structure of computing power resources,a new quantum genetic algorithm was proposed.Based on the multi-objective optimization model,the deployment of computing first access servers was analyzed.Subsequently,the ε-constraint method was used to transform the problem into a computing first access server deployment problem based on the constraint method,and it was theoretically proved that the two problems were equivalent to some extent.When designing the algorithm,quantum gene loci were used to replace traditional gene loci to construct quantum chromosomes.Afterwards,the entire search space was divided based on the constraints of the problem,and the population was evenly distributed through the size relationship between the probability amplitudes of quantum gene loci in the quantum chromosome.Based on the optimal quantum chromosome,the remaining quantum population was iteratively updated based on individual equilibrium and through quantum rotation gates.For each selected quantum chromosome in the constraint space,a random quantum non gate was used to complete the mutation operation.The experiment verified the effectiveness and convergence of the algorithm from four aspects:the total error rate of computing resources,the load balancing deviation rate,the algorithm convergence rate and the optimal solution deviation rate.

Key words: industrial computing first network, computing first access server, ε-constraint method, quantum genetic algorithm

摘要： 为解决由算力接入服务器的资源有限性与算力资源的系统结构的不稳定性之间的矛盾所产生的工业算力网络建设效率低下问题,提出一种新型量子遗传算法。首先,基于多目标优化模型分析了算力接入服务器部署问题。随后,利用ε约束法将该问题转化为基于ε约束法的算力接入服务器部署问题,并从理论上证明了两个问题在一定程度上是等价的。算法设计时,采用量子基因位代替传统基因位构建量子染色体。之后,基于问题的约束条件划分整个搜索空间,并通过量子染色体中的量子基因位复数概率幅之间的大小关系完成种群均匀分布。接着,以最优量子染色体为基准,剩余量子种群基于个体均衡度并通过量子旋转门完成种群迭代更新。最后,对约束条件空间中每个被选中的量子染色体利用随机量子非门完成变异操作。实验从算力资源总量误差率、负载均衡偏差率、算法收敛率以及最优解偏差率等4个方面验证了该算法有效性和收敛性。

关键词: 工业算力网络, 算力接入服务器, ε约束法, 量子遗传算法