A novel GA-based neural modelling platform for nonlinear dynamic systems
| Authors: | Li Kang, Queen's University Belfast, United Kingdom
Peng Jian-Xun, Queen's University Belfast, United Kingdom |
|---|
| Topic: | 3.2 Cognition and Control ( AI, Fuzzy, Neuro, Evolut.Comp.) |
|---|
| Session: | Intelligent Modelling and Identification I |
|---|
| Keywords: | Genetic algorithms, neural networks, optimisation, neural modelling, nonlinear system, software development, power generation |
|---|
Abstract
A novel genetic algorithm based neural modelling platform is developed for nonlinear dynamic systems. In this platform, the selection of neural inputs, optimisation of the neural network structure and identification of the optimal connection weights are formulated as an integrated optimisation problem. Both one-step-ahead and long-term prediction performances of the neural model are incorporated into the performance index. Genetic algorithms are then used for this mixed integer nonlinear optimisation problem. The platform provides user-friendly environment that gives users maximal flexibility in defining the modelling project and detailed visualisations of the optimisation process. This platform is then applied to the NOx emission modelling and prediction of a coal-fired power generation plant to confirm its effectiveness.
