A Neuro-Fuzzy Expert System for Flood Forecasting in Real-Time

A Moghaddamnia, ID Cluckie, D Han

Research output: Chapter in Book/Report/Conference proceedingConference Contribution (Conference Proceeding)


A Neuro-Fuzzy expert system has been developed for real-time flood forecasting in the Welland and Glen catchment. An Adaptive Neuro-Fuzzy Inference System (ANFIS) was used for automatically extracting rules to be encapsulated in a real-time expert system. Different prototypes were developed based on the identification of heuristic relationships between forecast points along the river. Each upstream forecast point automatically provides extra knowledge about target downstream forecast points. Importantly, fine-tuning of the membership functions parameters based on the supervised learning method employed in the ANFIS model gave a typically robust performance. The ANFIS structure utilised the past and present knowledge of the upstream forecast points combined with the current downstream state to make water level forecasts (6 hours ahead). During the later stages of development of a prototype expert system, the extracted rules were encapsulated in the COGSYS KBS. COGSYS was chosen for its real-time capability, for its performance and its flexibility. COGSYS KBS is a real-time expert system with facilities designed for real-time reasoning in the industrial context and it also deals with uncertainty aspects. The prototype expert system development process gave promising results and included updating the knowledge base using reliable new knowledge to improve the expert system performance.
Translated title of the contribution"A Neuro-Fuzzy Expert System for Flood Forecasting in Real-Time"
Original languageEnglish
Title of host publication7th International Conference on Hydroinformatics, Nice, France
Number of pages8
Publication statusPublished - Sep 2006

Bibliographical note

Conference Organiser: HIC 2006
Other: In Press


Dive into the research topics of 'A Neuro-Fuzzy Expert System for Flood Forecasting in Real-Time'. Together they form a unique fingerprint.

Cite this