Evaluating the specification requirements for (N)EPSAC-MPC implementation on a Programmable Logic Controller (PLC) J. Camila Espitia Duarte University of Ibague, Colombia Faculty of Engineering - Research Area camila.espitia@gmail.com Andres Hernandez and Robin De Keyser Ghent University, Belgium Department of Electrical Energy, Systems and Automation (EeSA) Andres.Hernandez@ugent.be, robain.dekeyser@ugent.be Abstract—Programmable logic controllers (PLC) are by far the most common hardware used in industry for process automation. As observed in many electronic devices, in the recent years, processing capabilities have increased, thus making possible to implement complex algorithms in such embedded devices. Among the different control methodologies, Model Predictive Control (MPC), stands out due to its ability to deal with constrained control problems. MPC is becoming the standard control strategy in process control as it meets the current industry demands of quality, production, resource optimization, low hardware cost, among others. The aim of this work is to develop a methodology to implement in a straightforward manner advanced control algorithms on a PLC, while providing a guideline on the mini- mum specifications required to properly choose a PLC reference, thus reducing the gap between the theoretical contributions and industrial practice. An experimental validation is achieved by implementing and comparing the performance of a classical PID, a linear (EPSAC) and nonlinear (NEPSAC) approach to constrained MPC on the Festo MPS PA workstation. I. I NTRODUCTION The PID Family controllers (P, PI, PD, PID) until a few years ago were the most control strategy used in the industry, because of its undoubtedly simplicity, low implementation cost and acceptable response. As a result, they have sev- eral advantages over more robust control strategies [11] [6]. However, industry requirements are constantly increasing [18]. Currently, several works have presented the implementation modern strategies and robust control, such as, MPC (Model Predict Control) which exceeds the performance of traditional PID, applied to different types of industrial processes [1] [4] [12] [16] [2] [13]. In order to implement modern control strategies in real form, certain characteristics are required for the hardware of the controller where the implementation is performed. The controller is usually a PLC - controller standard of the industry [19] [3], PLCs manufacturers are introducing devices with better characteristics related to the memory capacity, programming languages and the processing speed instruction [14] [21]. On the other hand, the software specialized in control like MatLab include interesting tools that represent a link to migrate from theoretical developments to real implementations [15] [17] [22]. Work starts of the fact that is possible, with available tools, to implement advanced control strategies on PLCs, but its focuses are the needs of to provide a clear-simple procedure and shown the minimum PLC hardware requirements for it. The detail of the implementation of predictive control strategies: EPSAC and NEPSAC on Siemens PLC 1615-3 is presented. The experimental validation is performed using a pilot FESTO MPS PA with pressure system - SISO linear system, and level system - SISO nonliear system. Matlab was used for data processing and as an interface for programming the PLC, with tool PLC Coder and the traditional PID was used like benchmark to results. II. METHODOLOGY DESCRIPTION The methodology used as an interface for PLC program- ming and experimental validation of their performance under implementation of robust control strategies, it begins with the formulation of the control strategy within a block function in Matlab-Simulink (Stage 1) and the translation of this using appropriate language through PLC Coder tool (Stage 2), the file result is loaded, compiled and downloaded along with other data acquisition subroutines to the PLC (Stage 3) using a IDE (Integrated Development Environments). Finally the scheduled will run a control strategy to maintain a plant process in desired states (Stage 4). The instructions for data acquisition are executed each sampling time, in order to allow the collecting, storing in a global data base. (Stage 5) The data is displayed data into a HMI (Human Machine Interface), from which they are exported as a text file (*.txt) for later analysis (Figure 1). III. ELEMENTS OF METHODOLOGY A. Control Strategy - (N)EPSAC The Extended Prediction Self-Adaptive Control (EPSAC) (Keyser and Cuawenberghe, 1985) belongs to the large family of Model Predictive Control (MPC) algorithms. MPC is a set of control strategies, which most attractive feature is its ability to handle constraints, because MPC has detailed knowledge of the dynamic behavior of the system (including critical units) represented in a mathematical model. As it is stated above MPC seems to be a good option to face the industry