<p>Resource allocation plays a crucial role in the reliable operation of large-scale multi-agent systems, particularly when the operating environments and the optimization goals are time-varying. This paper investigates a distributed real-time resource allocation problem over a directed communication network. The global objective is to minimize the sum of time-varying local costs subject to a time-varying resource allocation constraint. The time-varying nature of both the local cost functions and the resource allocation constraint is modeled by an exosystem. To address the combined difficulties of time-varying optimization and directed information flow, a distributed algorithm is designed for each agent that only utilizes the information of its own cost function and the information obtained through a network represented by a strongly connected and weight-balanced digraph. The distributed algorithm contains a distributed estimator that estimates global information involving all agents. It is proven that the proposed method achieves exponential convergence of all agents’ decisions toward the time-varying optimal solution with any pre-specified level of accuracy. Simulation studies involving distributed energy resources in a virtual power plant further demonstrate the practical effectiveness of the algorithm.</p>