Goal-Oriented Intelligence in Optimization of Distributed Parameter Systems

Abstract

Models of complex systems can be differentiated by their ability to reproduce or generate system behavior, by their prediction power, by their robustness, or, conversely, by their sensitivity to inputs and parameters; by their uncertainty (if captured); and by their intelligence. Even the term "prediction" is not unique. First, a first-principle (physically based) distributed parameter model could be an excellent predictor if (a) it captures the main system behavior, and (b) its parameters and inputs are known accurately; otherwise, it would fail, possibly drastically. Second, predictive power depends on the data, on the goal, and on the time scale. For example, scheduling of pumping and injection in an oil field for maximum profit over the next 5 years; or pumping from a contaminated aquifer in order to maintain certain (low) concentration at a compliance point for the next 20 years, vs. prediction of plume migration in groundwater towards a nearby river, over time: in each case, the model has a slightly different expected function, as well as different intelligence type. The paper reviews the recent developments in subsurface fluid flow management such as optimization of oil production and groundwater remediation (both sharing similar practices, though for different purposes) as a continuous struggle to increase intelligence by (a) adapting new tools such as artificial intelligence and dynamic stochastic control; (b) attempting to integrate these tools; and (c) reducing uncertainty. Although the systems discussed seem specific to the (mathematical) geosciences (specifically to oil reservoirs and contaminated aquifers), and although these systems are very different from man-made machines, similar rigid structure and reliance on differential-integral calculus, as well as the serial processing, knowledge evolution, and uncertainty propagation from one discipline to the next exist in most science and engineering fields, and so does the need for a paradigm shift.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 2004
Accession Number
ADA516072

Entities

People

  • Shlomo Orr

Tags

Communities of Interest

  • Energy and Power Technologies
  • Human Systems

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Bayesian Networks
  • Computational Fluid Dynamics
  • Computational Science
  • Computer Programming
  • Computers
  • Cost Reductions
  • Differential Equations
  • Dynamic Programming
  • Fluid Dynamics
  • Fluid Flow
  • Geophysics
  • Information Processing
  • Kalman Filters
  • Mathematical Filters
  • Operations Research
  • Optimization

Readers

  • Artificial Intelligence
  • Calculus or Mathematical Analysis
  • Computational Fluid Dynamics (CFD)

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - DoD AI Strategy
  • AI & ML - Machine Learning Algorithms