Application of Gaussian processes for black-box modelling of biosystems

Area of Science:

• * Nonlinear dynamic system identification
• * Machine learning applications in environmental engineering

Background:

• * Traditional models for nonlinear dynamic systems have limitations.
• * Gaussian process models present a new, promising alternative.
• * Key features include confidence measures and parameter efficiency.

Purpose of the Study:

• * To present a framework for dynamic system identification using Gaussian processes.
• * To illustrate the framework with a simulated bioreactor example.
• * To apply the framework to real-world environmental case studies.

Main Methods:

• * Application of Gaussian process modeling for system identification.
• * Simulation of a bioreactor system.
• * Case studies on wastewater nitrification and Venetian Lagoon biomass growth.

Main Results:

• * Successful identification of dynamic systems using Gaussian processes.
• * Demonstrated utility in complex environmental modeling scenarios.
• * Highlighted the importance of model validation using prediction variance.

Conclusions:

• * Gaussian process models are effective for nonlinear dynamic system identification.
• * The proposed framework is applicable to environmental engineering problems.
• * Prediction variance is a valuable tool for Gaussian process model validation.