(246u) Title: An Agent-Oriented Software Framework for Real Time Optimization Solutions Implementation

For the past decades the increasing competition in the global market has lead the chemical process industry to look for the application of tools to optimize processes in order to enhance profitability. Real Time Optimization (RTO) is a Computer Aided Process Engineering (CAPE) technique that continuously reevaluates the process operating conditions to maximize economic benefits under operational constrains. By adjusting selected optimization variables it seeks to improve process performance iteratively using measurement data to drive the operating point toward the actual plant optimum [1].

RTO covers several concerns: real-time process data sampling through interfaces with measurement instruments, Distributed Control Systems and solutions like Plant Information System from OSI Software (www.osisoft.com); gross error detection, data reconciliation and steady-state identification; model parameters estimation and optimization. Once an optimal operational point is calculated for the plant, subjected to operational constraints, its values are sent to the process control to update set-points. Advanced control techniques like model based predictive control (MPC) are responsible for take the plant to the suggested state. Several approaches have been proposed for RTO [1].

Transforming multicomponent CAPE methods like RTO into useful tools requires appropriate software architecture and design [2]. An Agent Oriented Software Architecture (AOA) for the implementation of RTO solutions has been conceived and is presented here.

AOA and Agent Oriented Programming (AOP) are relatively new software paradigms that bring concepts from the Artificial Intelligence field into the domain of distributed systems. According to the AOA approach a software application should be modeled and built as a collection of interacting components called Agents. A definition for the term “software agent” has been addressed in several works [3, 4, 5, 6]. In general, all the definitions converged to the idea that an agent is essentially a special software component situated within and a part of an environment that senses that environment and acts on it over time, has autonomy or behaves like a human agent, working for some clients in pursuit of its own agenda with social or communication abilities.

A variety of application fields are being benefited with the use of multi-agent systems. Examples of industrial domains where this kind of systems have been fruitfully employed include process control [7], system diagnostics [8], manufacturing [9], transportation logistics [10] and network management [11]. The Foundation for Intelligent Physical Agents (FIPA) is an IEEE Computer Society standards organization that promotes agent-based technology and the interoperability of its standards with other technologies. Its specifications contain a collection of standards which are intended to promote the interoperation of heterogeneous agents and the services that they can represent (www.fipa.org).

A software framework following the proposed AOA and using some well known software design patterns and the Object Oriented Programming paradigm (OOP) has been developed also and is presented as well. In the development process, a free an open source implementation of the FIPA standards, named JADE (www.jade.tilab.com), has been used as base platform. Using the agents framework, a prototype was configured to apply RTO to the Williams-Otto study case [12] using a Model Parameter Adaptation approach [13]. For steady-state identification the methodology proposed by Cao and Rhinehart [14] was used. The process model and optimization problem were defined using the modeling language of EMSO [15] and the Ipopt algorithm [16] was used as optimization solver. Another RTO prototype for a PETROBRAS C3 Splitter distillation unit was set as well, also based on an EMSO model [17]. In both cases the framework's worth was proved.

Although proven commercial software products exist for RTO and can be parameterized they usually feature a black box style regarding the way operations are carried out. In addition, the inclusion of new RTO methods is almost always not a feasible option. That motivated the development of the framework as a technically open workbench that can be used for test, research and innovation purposes allowing the implementation of several RTO approaches. It provides a platform for the study of RTO issues, steady state identification methods, gross error detection and data reconciliation techniques. It also creates chances for the interaction with already existing modeling and simulation tools and to test the efficacy of different optimization solvers. Spots for the implementation of different control systems interfaces are also provided. It furnishes a useful environment to study in practice unexplored techniques combination and helps to open new horizons for more sophisticated RTO applications.

(1) B. Chachuat, B. Srinivasan, D. Bonvin, 2009, Adaptation Strategies for Real-Time Optimization, Comput. Chem. Eng., 33, 1557-1567.

(2) B. Braunschweig, R. Gani, 2002, Software Architectures and Tools for Computer Aided Process Engineering, Comput. Aided Chem. Eng., 11.

(3) M.R. Genesereth, S.P. Ketchpel, 1994, Software Agents. Communications of the ACM, 37(7), 48–53.

(4) M.J. Wooldridge, N.R. Jennings, 1995, Intelligent Agents: Theory and Practice. Knowledge Engineering Review, 10(2), 115–152 .

(5) S.J Russell, P. Norvig, 2003, Artificial Intelligence: a Modern Approach, 2nd ed. Prentice Hall.

(6) P.T. Tosic, G.A Agha, 2004, Towards a hierarchical taxonomy of autonomous agents, Proceedings of IEEE International Conference on Systems, Man and Cybernetics, The Hague, Netherlands, 3421-3426.

(7) N. Jennings, 1994, The Archon System and its Applications, Proceedings of the 2nd International Working Conference on Cooperating Knowledge Based Systems (CKBS-94), Dake Centre, University of Keele, UK, 13-29.

(8) M. Albert, T. Laengle, H. Woern, M. Capobianco, A. Brighenti, 2003, Multi-agent Systems for Industrial Diagnostics, Proceedings of 5th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, Washington, DC , 483-488.

(9) H. Parunak, 1987, Manufacturing Experience with the Contract Net. In Huhns, M. (ed.), Distributed Artificial Intelligence, Pitman, London, 285-310.

(10) N. Neagu, K. Dorer, D. Greenwood, M. Calisti, 2006, LS/ATN: Reporting on a Successful Agent-Based Solution for Transport Logistics Optimization. Proceedings of the IEEE Workshop on Distributed Intelligent Systems (WDIS’06), Prague.

(11) D. Greenwood, G. Vitaglione, L. Keller, M. Calisti, 2006, Service Level Agreement Management with Adaptive Coordination. Proceedings of the International Conference on Networking and Services (ICNS’06), Silicon Valley, USA.

(12) T.J. Williams, R.E Otto, 1960, A Generalized Chemical Processing Model for the Investigation of Computer Control, American Institute of Electrical Engineers, Part I: Communication and Electronics, Transactions, Volume 79, Issue 5, 458 – 473.

(13) C. Y., Chen, B. Joseph, 1987, On-line optimization using a two-phase approach: An application study. Industrial and Engineering Chemistry Research, 26, 1924–1930.

(14) S. Cao, R.R. Rhinehart, An Efficient Method for On-line Identification of Steady State. Journal of Process Control 1995, 5, 363.

(15) R.D.P. Soares, A.R. Sechhi, 2003, EMSO: A new Environment for Modelling, Simulation and Optimization, Comput. Aided Chem. Eng., 14, 947-952.

(16) A. Wachter, L. Biegler, 2006, On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming, Mathematical Programming, Volume 106, Issue 1, 25-57.

(17) D.F.M. Muñoz, L.M.P. Garcia, J.E.A. Graciano, C.A.R. Martinez, A.S. Vianna, G.A.C. Le Roux, 2013, Real-Time Optimization of an Industrial-Scale Vapor Recompression Distillation Process. Model Validation and Analysis. Ind. Eng. Chem. Res., 52, 5735-5746.