改进交替方向乘子法求解冷水机组负荷分配群智能优化问题

Improved alternating direction method of multipliers for solving optimal chiller loading problem in swarm intelligent control system

群智能控制系统中的多台冷水机组负荷优化分配问题是一个多块优化问题,传统分布式方法难以获得其收敛解.文中将交替方向乘子法(ADMM)引入冷水机组负荷分配群智能优化问题中,并通过一种有效的高斯罚函数(GPF)更新策略改进了交替方向乘子法收敛特性.同时,建立了一种基于ADMM–GPF–GBS双层分布式计算框架的冷水机组负荷优化分配模型,该模型仅利用相邻节点间的局部信息传递,即可求解得出最优运行策略.最后,通过两个典型算例对比分析了所提优化方法的有效性,并在实际硬件系统中进一步对该算法进行应用与验证.结果表明,所提算法适用于群智能控制系统下的多台冷水机组系统,且具有比传统分布式算法更好的寻优能力和收敛性,可以取得显著的节能效果.

As a multi-block optimization problem,the optimal chiller loading(OCL)in the swarm intelligent control system is difficult to obtain a convergent solution by using the conventional distributed algorithm.To solve this problem,an improved alternating direction multiplier method(ADMM)was used in this article.In the method,the convergence characteristic was improved by an effective Gaussian penalty function(GPF)update strategy.An optimal chiller loading model based on ADMM–GPF–GBS double-layer distributed computing framework was established,and the global optimal solution can be obtained by parallel computing only by using the information transmission between adjacent nodes.The effectiveness of the proposed optimization method was compared and analyzed by two numerical examples,and the algorithm was further applied and verified in the actual hardware system.The results indicate that ADMM–GPF–GBS is suitable for the central air conditioning multi-chiller systems under the swarm intelligent control framework and has excellent optimization ability with good convergence.The energy-saving effect is remarkable.