This paper presents recent results in applying a multi-agent system combined with extremal optimization to deal with integrated scheduling for "steelmaking, continuous cast and hot strip mill", a multistage, hybrid and real-world production system in steel industry. The main goal of this research aims to minimize the energy losses through the optimization, cooperation, coordination and communications between the multiple knowledge-based agents. The problem formulation, system architecture and optimization algorithms are described. The simulation results with production size data show the applicability of the proposed solutions.