Decision support has two distinct but interrelated aspects: optimisation based on mathematical modelling of physical plants and processes, and the fuzzy information processing by human operators. This research builds on our previous research contributions concerning computation using analogue neural networks and fuzzy sets and systems. A diagrammatic representation of the various components of the project and their interrelationship is presented below.
The solution to this nonlinear optimization problem utilised the Newton-Raphson iterative method. This method has been widely and successfully used in the context of water distribution systems. The state estimates are found by iteratively solving a system of linear equations and adjusting the state vector values.
The novel contribution of this research was the mapping of the state estimation method onto appropriate analog neural network. Several neural networks and the associated optimisation criteria (Least Mean Squares(LMS), Least Absolute Value (LAV) and MinMax (Chebyshev) have been implemented and assessed for their efficiency and robustness. Our solution based on the recurrent neural network [Gabrys, Bargiela, 1995], [Gabrys, Bargiela, 1996], [Cichocki, Bargiela, 1997] has proven itself to offer superior computational efficiency which allows for what-if study of system control scenarios in real time. The network is represented in the diagram below.