Over the years, olefins plants have evolved into highly integrated, highly flexible processing systems that can profitably adjust to ever changing landscape of raw material availability and market demand for high purity olefins products. Advanced process control technologies such as model predictive control (MPC) are commonplace in olefins plants and have greatly improved the consistency of product quality and constraint protection, and have typically resulted in quick payback on investment. Another advanced technology, that of on-line optimization promises even greater benefits. Application of traditional on-line optimization technology however remains a challenging task primarily because of long process dead times, mixed process dynamics and frequent disturbances. The steady state models used in traditional optimization are inadequate representation of the on-line behavior of most commercial olefins plants. A unique approach to on-line optimization of dynamic processes is presented in this paper. In this approach plant steady state is not a requirement for on-line optimization. This approach fully utilizes the MPC dynamic models for optimization. Application of this on-line optimization technology has resulted in continuous savings for some very dynamic olefins plants. This paper briefly describes this dynamic optimization technology and summarizes experiences from several completed and in-progress olefins plant optimization projects.