Designing large-scale balanced-complexity models for online use

Model-based online applications such as soft-sensing, fault detection or model predictive control require representative online models. Basing models on physics has the advantage of naturally describing nonlinear processes and potentially describing a wide range of operating conditions. Implementing adaptivity is essential for online use to avoid model performance degradation over time and to compensate for model imperfection. Requirements for identifiability and observability, numerical robustness and computational speed place an upper limit on model complexity. These considerations motivate the design of balanced-complexity physical models with adaptivity for online use. Techniques used in the design of balanced-complexity models are given with examples from offshore oil and gas production. Despite potential benefits, the effort required to implement balanced-complexity models, particularly at large scales, may deter their use. This paper presents a Modelica-based approach to reduce implementation effort by interfacing exported Modelica models with application code by means of a generic interface. The suggested approach is demonstrated by parameter estimation for a subsea well-manifold-pipeline system.