In this paper, an indirect dual ascent method with an exponential convergence rate is proposed for a general resource allocation problem with convex objectives and weighted constraints. By introducing the indirect dual variables, the dual dynamics can be executed in a decentralized manner by all nodes over the network. In contrast to the conventional methods, consensus on all the dual variables is not required. This further leads to the fast convergence, reduced communication burden and better privacy preserving. Moreover, the exponential convergence rate of the proposed algorithm is established through the Lyapunov method and the singular perturbation theory. Application of the dynamic power dispatch problem in smart grid verifies the effectiveness and performance of the proposed algorithm.

针对具有凸目标和加权约束的一般资源分配问题，提出了一种具有指数收敛速率的间接对偶上升方法。通过引入间接对偶变量，可以由网络上的所有节点以分散方式执行对偶动力学。与常规方法相比，不需要对所有双重变量达成共识。这进一步导致了快速收敛，减轻了通信负担并更好地保护了隐私。此外，通过Lyapunov方法和奇异摄动理论建立了该算法的指数收敛速度。动态功率分配问题在智能电网中的应用验证了该算法的有效性和性能。