The traditional way of doing control in industries considers the so-called hierarchical control structure. At the top of this structure, en economic scheduler and planner decides what, when and how much the plant has to produce, taking into account information from the market and from the plant. The outputs of this layer are production goals, prices, economic cost functions and constraints that are sent to a Real Time Optimizer (RTO). The RTO is a model-based system, operated in closed-loop, whose task is to implement the economic decision in real time. So, it provides the setpoints to the control level. It employs a stationary model of the plant and for this reason its sampling time is usually larger than the settling time of the plant. The setpoints calculated by the RTO are sent to the advanced control level, where an advanced control strategy calculates the optimal control action to be sent to the plant, in order to regulate the plant as close as possible to the setpoint, taking into account a dynamic model of the plant and constraints and ensuring stability. These control actions are sent to the low-level controllers, which then regulate the plant. This structure is shown in Figure 1.
Figure 1: hierarchical control structure.
The choice of advanced control strategy is usually MPC, since it considers an optimal criterion and constraints in the formulation of the control problem. The problem of the hierarchical control structure is that, usually, the economic setpoints calculated by RTO may be inconsistent or unreachable with respect to the dynamic layer, and this happens mainly because of the discrepancies between the stationary model of the RTO and the dynamic model used by the MPC, as well as because of the different time scale of these two layers. Hence, the optimal setpoint calculated by the RTO may not coincide with the dynamic steady state of the system. A way to avoid this problem is the so-called two-layer structure: an upper optimization level is added in between of RTO and MPC. This level, referred as the steady state target optimizer (SSTO), calculates the steady state to which the system has to be stabilized, solving a linear or quadratic programming and taking into account information from the RTO. The plant model used in this intermediate level is the same as the MPC one, thus reducing inconsistencies (see Figure 1). However the use of an SSTO does not solve the problem of bad economic performance and, moreover, adds an extra optimization to the control structure.