The project involves the development of a system that can assist operators of large-scale distribution systems in taking control decisions based on qualitative and quantitative evaluation of the system state. State estimation involves the cross-referencing of system measurements by relating actual meter readings and estimated nodal consumptions to the values calculated from the mathematical model of the system. An efficient, numerically stable state estimation procedure for solving these network equations has been developed using a least absolute values regression method.

With reference to water distribution, the consideration of a greater range of nonlinear hydraulic network elements provides the operator with a more realistic view of the network. However the inclusion of these elements causes points of discontinuity and general `unsmoothness' within the nonlinear network equations. The Hermite approximation is shown to be the most suitable smoothing approximation for the hydraulic relationships resulting in an increase in the convergence rate and robustness of the state estimation algorithm.

The ever-increasing size of distribution systems has led to slower convergence of the state estimation algorithm. In order to counteract this effect, parallel state estimation algorithms have been developed and implemented on a network of transputers. The algorithms divide the distribution systems into smaller, more manageable subsystems to be solved on individual processors. The resulting substate estimates are then coordinated to determine the state estimate of the entire system.

Meter readings from the distribution systems invariably include a degree of inaccuracy. This inaccuracy and the inexact approximation of nodal consumptions lead to discrepancies within the state estimate. So the parallel state estimation algorithm has been further developed to show the effect of these discrepancies in the form of state confidence limits.

The implementation of this parallel probabilistic state estimator on a network of transputers has led to the development of a parallel metalanguage which offers a generic tool for programming transputer platforms. The metalanguage enables a logical specification of parallel tasks and intertask communication, making the complexities inherent to a specific physical processor connectivity transparent to the user. This metalanguage complies with the bulk-synchronous parallel processing models, and is seen as a prototype for implementations on other distributed hardware architectures.