Neural Networks Based Decision Support in Presence of Uncertainties

A new approach, based on the examination of patterns of state estimates by a general fuzzy min-max neural network (GFMM) has been proposed and evaluated, and an analogy between the information processing by the GFMM and by human operators has been identified and highlighted in this context.

Abstract

This paper addresses the problem of efficient and effective interpretation of water distribution network state estimates that are typically calculated on the basis of measurements and pseudomeasurements (consumption estimates) that have significant uncertainties associated with them. The task of the system state interpretation is particularly relevant to the diagnosis of leakages and other operational faults occurring in water distribution networks. A new approach, based on the examination of patterns of state estimates by a general fuzzy min-max neural network (GFMM) has been proposed and evaluated. The GFMM classification and clustering has been incorporated into a two-level fault diagnosis system. The proposed diagnostic procedure builds on the concept of confidence limit analysis of state estimates and estimation residuals. An extensive leakage detection and identification study in a small test system for a complete 24-h period of operation has been carried out. An analogy between the information processing by the GFMM and by human operators has been identified and highlighted in this context.