(288b) Cape-Open Compliant Stochastic Modeling and Reduced-Order Model Computation Capability for Apecs System

Analysis of integrated energy systems for benefit maximization has become essential, and simulation studies constitute an important aspect of the whole process. Aspen Plus simulation models have been extensively used for this purpose.

Aspen Plus simulation models of advanced power systems typically comprise of many separate unit operation blocks that are interconnected by material, heat and work streams. The simulation of such blocks needs to reach the balance of mass and also energy. For advanced power systems, where the given system can be a combination of multiple units, the simulation can be computational intensive. Moreover, incorporation of Fluent computational fluid dynamic (CFD) model for a particular module in the Aspen flowsheet through the APECS (Advanced Process Engineering of Co-Simulations) integrated simulation environment can make the simulations much more computationally intensive than standalone Aspen simulations.

Owing to these computational bottleneckes, these exhaustive models cannot be used for high level decision making. Quite often, an effective approach is to utilize simplified `Reduced Order Models' (ROM) that are computationally efficient. The ROM can operate quickly and allow users to evaluate the effects of different assumptions and options regarding the process performance while retaining much of the rigor of the more detailed ASPEN or APECS simulation models. However, for the success of this approach, these reduced order models must achieve the correct balance of computational efficiency and accuracy. It is important to ensure that these models do not oversimplify the given system, thereby ignoring the salient details of the model. This work proposes to use multiple linear regression to generate the ROM for the energy systems. Multiple linear regression attempts to model the relationship between two or more explanatory variables and a response variable by fitting a linear equation to observed data. Thus, to perform multiple linear regression, one needs to generate a large number of data points. Stochastic simulation of the rigorous model can be used to generate these data points, which are then used for the generation of ROM. Stochastic simulation refers to modeling of the given system by considering certain model parameters to be uncertain. Uncertainty analysis through stochastic simulations is also important in its own sense for advanced energy systems since available performance data are typically scant, accurate predictive models do not exist, and many technical as well as economic parameters are not well established [1]. The idea is to determine the variability in model output with respect to the variation in different model parameter values. The automation of such a stochastic modeling framework along with the generation of ROM will further simplify the analysis. Moreover, creation of a CAPE-OPEN (CO) compliant tool that simplifies the process of setting up a process modeled in the APECS (Advanced Process Engineering Co-Simulator) for stochastic simulation and ROM calculation, will make stochastic simulation and reduced order model calculation more accessible to users. The CO standards provide a set of interfaces which allows seamless integration of Computer Aided Process Engineering (CAPE) modules from various sources (software and equipment vendors, universities, and company generated) into process simulation environments [2]. This enables a process engineer to ?assemble' the necessary computational tools with the minimum effort from a collection of software (in-house, commercial, and/or academic) to achieve a best-in-class solution to various CAPE-related problems. Thus, along with the objective of using muliple linear regression, this work also aims to generate a CAPE-OPEN compliant tool to automate the whole process. The tool is built as a .exe file which can be installed in any computer running Windows OS. This tool automates the process of setting up a stochastic framework to a great extent and guides the user in each step of the process through user-friendly graphical user interface (GUI) windows. The tool has capability for 4 sampling techniues including the efficient Hammersley Sequence Sampling (HSS) and leaped HSS [3]. This work aims to build on this, and automate the process of stochastic simulation based ROM generation.

References:

1. Diwekar U. M., Introduction to applied optimization, Kluwer Academic Publishers, Netherlands, 2003.

2. Koller J. and Tobermann J.C, Global CAPE-OPEN ? D822 Migration Cookbook, October 2002.

3. Kalagnanam, J. R. and Diwekar U. M., An efficient sampling technique for off-line quality control , Technometrics, 39(3), 308-319, 1997.