Application of dynamical system theory for short-term water load demand prediction

This paper first presents the concept of embedding a time series. It then presents a method for determining the dimension of the model of the load demand time series based on correlation dimension of observed load data. The procedure to approximate the time series model by tessellation and the application of the natural neighbour interpolation technique to the tessellation to perform water load demand predication is described. The suitability of this approach is illustrated through an application example using actual water consumption figures supplied by the Water Authority of Western Australia