Production scheduling in dynamic real-time optimization with closed-loop prediction

Highlights

• Novel framework for integrated scheduling and dynamic RTO (DRTO) that accounts for transition dynamics.

• Utilizes predicted closed-loop response of plant under action of constrained MPC.

• Determines optimal economic production sequencing and set-point trajectories.

• Communication of optimal decisions via set-point trajectories permits standard MPC formulation to be retained.

Abstract

Process plants are operating in an increasingly dynamic environment, fueled largely by globalization and deregulation of energy markets, resulting in fluctuating market conditions and large variations in electricity prices. Such conditions pose challenges for traditional hierarchical plant decision-making systems, leading to efforts toward integration across the decision-making layers. This paper proposes a formulation for integration of production scheduling decisions within a dynamic real-time optimization (DRTO) framework. The DRTO formulation utilizes a closed-loop prediction of the plant response under the action of constrained model predictive control (MPC). The integrated scheduling and DRTO system communicates decisions to the underlying MPC system through time-varying set-point trajectories, thereby permitting the standard MPC implementation to be retained. The efficacy of the prosed system is illustrated through application to both single-input single-output (SISO) and multi-input multi-output (MIMO) case studies.

Introduction

Decision making in large chemical and petrochemical plants typically follows a hierarchical structure, with basic PI/PID controllers at the lowest level, an MPC system that provides set-points to the lower-level controllers, a real-time optimization (RTO) layer that provides economics-based set-points to the MPC system and possibly also inputs directly to the plant, and a scheduling and planning layer that provides production targets [12], [31].

Increased globalization and deregulation of energy markets have created a dynamic plant operating environment, requiring frequent transitions in order to meet varying product demands and respond to fluctuating energy prices. Backx et al. [2] recognized the impact on process plants of an increasingly transient and competitive marketplace, and advocated a strategy of “intentionally dynamic operation” for enhanced profitably within this setting. Traditional plant decision making systems are more suited to steady-state production of commodity products, rather than the current highly dynamic plant environment. Integration across the various levels of decision making has been recognized as a key challenge in enterprise-wide optimization [22]. Such considerations have led to efforts toward integration of the decision-making layers, or elements thereof, including economic MPC [1], [16], dynamic RTO (DRTO) [25], [27], [42], and integration of scheduling and control [5].

This paper extends a recently proposed DRTO scheme based on closed-loop prediction of the plant response under constrained MPC [25], [28] to include production scheduling decisions. The production scheduling includes the sequencing of products at different specification levels, with the overall objective of maximizing the economics of production over a prediction horizon, where the economics of the transition dynamics are accounted for. The optimal decisions are communicated to the MPC system through time-varying set-point trajectories, thereby allowing the standard regulatory MPC formulation to be retained.

The remainder of this article is organized as follows. Section 2 presents a brief review of literature on DRTO and different approaches for combining scheduling and control. Section 3 presents the detailed formulation of the integrated DRTO and scheduling framework implemented in this work, and the solution approach followed. Section 4 demonstrates the performance of the method through application to SISO and MIMO case studies. Section 5 concludes the article with a summary of findings and identifies future research directions.