To solve the problem of consensus control of industrial cyber-physical systems (ICPS) composed of multiple controlled objects under denial of service (DoS) attack and uncertain external disturbance, optimal H∞ consensus controller is designed by combining multi-agent with ICPS and employing state observer and zero-sum game methods. Firstly, for the ICPS composed of multiple controlled objects, a fixed topology multi-agent ICPS model is constructed based on the information interaction characteristics of agents. Secondly, considering the effects of DoS attacks causing communication channel blockages when acting on sensor channels and actuator channels, the state observer is used to reconstruct the blocked state information, and the multi-agent ICPS model under DoS attack based on the state observer is constructed. Then, the interference suppression level is introduced, and the controller and the disturber are considered as participants in a zero-sum game. Based on the optimization objective of interference suppression level and the condition of system H∞ performance, an optimal design framework based on zero-sum game is constructed. Finally, by imposing constraints on disturbance suppression levels and solving linear matrix inequalities that satisfy H∞ consensus, an optimal consensus controller is designed. Using the control problem of four uninterruptible power supplies as a simulation object, Matlab simulation results demonstrate that compared to design methods without state reconstruction and parameter constraints, the proposed approach achieves shorter consensus time for multi-agent ICPS under dual-channel DoS attacks and exhibits good relative stability.