Intellectual Merit: Since the inception of feedback control, there has always been a strong, but vague, relationship between control system performance (in a regulatory sense) and the time-averaged profit of the plant (which is a function of nominal operating conditions). The transformative concept underlying Profit Control is the simple notion of quantifying this relationship, which will enable control system design while simultaneously maximizing plant profit. Such an approach represents a paradigm shift in controller synthesis, in that the generated controllers will be tailored not only to the dynamics of the plant but also to the specific economic situation.
The project objectives are twofold. First, the PI will extend the theoretical foundation of the scheme to include the following disturbance classes: non-stationary (or steady-state) disturbances, model uncertainty and peak-to-peak measures. The result will be a unified approach that simultaneously addresses the new disturbance classes as well as the original stochastic (or H2) disturbances. The entire scheme will be guided by a measure of economic performance common to the Real-Time Optimization (RTO) literature: the notion of minimizing the backed-off operating point from that suggested by the RTO.
The second objective is to apply Profit Control to systems from a variety of disciplines. These include; the multi-product, multi-echelon supply chain problem (management science), a fluidized catalytic cracking unit (chemical processes), and the power system of a hybrid electric vehicle (electrical systems). These application studies will illustrate the broad and applicability of this approach.
Broader Impact: The broader aspects of the project are threefold; multidisciplinary interaction, student outreach and software dissemination. Due to the broad spectrum of topics covered, achievement of the primary project objectives will require multidisciplinary input. As such, collaborations with individuals from outside the field have been scheduled. The PI is planning to recruit both graduate and undergraduate students of all backgrounds, genders and ethnicities. Toward broad dissemination of the technology, a software tool aimed at those versed in control theory, but non-experts in linear matrix inequalities and branch-and-bound calculations, will be developed and freely distributed via the internet.
Intellectual Merit: The operation of a manufacturing plant is always restricted in some way; equipment capabilities, safety concerns, and environmental regulations are just a few examples. In most cases, the condition of greatest profit is associated with operation at one or more of the plant limitations. Unfortunately, operation at such a point is precluded by the influence of uncertainty – the unpredictable aspects of plant operation will inevitably result in a violation. Control systems are critical to mitigating the impact of uncertainty. For example, the controller of a heating and air conditioning system will respond to the uncertainty of outside temperatures and mitigate their impact on inside temperatures. In more complicated systems, the impact of uncertainty is a multidimensional issue. That is, there are many controllers available and each will result in unique uncertainty mitigation characteristics. In this NSF-supported research, Professor Chmielewski and his team at the Illinois Institute of Technology have developed a new computational technique that turns this relation around. Rather than a trial-and-error approach - where one selects a controller and then tests to see if resulting mitigation properties are sufficient - they can specify mitigation characteristics up front, and the new algorithm will generate a controller to achieve the desired mitigation properties, if one exists. Professor Chmielewski reports that this finding has paved the way for a profit based approach to control system design. In essence, the scheme identifies a controller to minimize the impact of uncertainty in directions most influential to profit. This new paradigm seems to be a game changing concept, in that the Edisonian approach (tune the controller, check the profit and then repeat) can be replaced by a single step procedure. The main development of the project is a reformulation of the approach, which reduces computational effort (by more than 98%) and greatly expands the class systems one may address. The main discovery of the project is that combining the approach with a similar method achieves massive reductions in on-line computational effort (more than 99%). The impact of this discovery is that profit based controllers can be applied to more complicated and challenging systems. An additional discovery is an extension of the method so that it can be used within smart grid applications – ranging from building HVAC systems with thermal energy storage to electric power networks with renewable sources and massive energy storage facilities. Broader Impacts: This NSF-funded research was also integrated with numerous educational activates targeting K-12 students, engineering students and engineering professionals. The project resulted in over 800 pre-college students being exposed to science and engineering. These day-long events (4 in total) were organized by Professor Chmielewski in collaboration with the Chicago Local Section of the American Institute of Chemical Engineers (AIChE) and the IIT Student Chapter of the AIChE. Training and mentoring of engineering students was also a primary goal of the project. The project supported 1 engineering graduate student directly and 3 other graduate students indirectly. The project also involved 3 undergraduate engineering students in the research activities. Due in part to this experience, one of these students is currently pursuing her PhD in chemical engineering and has a career objective of teaching at the college level. The project also promoted the engagement of engineering professions. To this end a number of tutorial sessions were organized at professional society conferences. Another effort, focused on multidisciplinary interaction, was the organization of an AIChE Workshop on Smart Grid for the Chemical Process Industry. This workshop featured 15+ presenters from government, academia and industry and explored the opportunities and challenges associated with smart grid participation from the chemical process industry.