A multivariate analysis of the daily water demand of Skiathos Island, Greece, implementing the Artificial Neuro-Fuzzy Inference System (ANFIS)
Nikolaos Mellios, Dimitris Kofinas, Elpiniki Papageorgiou, Chrysi Laspidou
Thursday 2 july 2015
9:45 - 10:00h
at North America (level 0)
Themes: (T) Special session, (ST) FP7 ICT and water
Parallel session: 10I. Special session: FP7 ICT and water
Considering the worldwide extensive demand in optimizing the water distribution networks, in terms of leakage detection and pressure management, as well as the need to reduce urban water consumption, a lot of effort is made in the past decade in order to define accurate, long term and micro time scale water demand forecasting methods. Linear regression models, such as ARIMA, and Artificial Neural Networks have been used, as well as different hybrid approaches. In this paper, a multivariate analysis of daily water demand of Skiathos Island, Greece and an investigation on the benefits of the Artificial Neuro-Fuzzy Inference System (ANFIS) forecasting method are presented. Skiathos is a touristic island at the Aegean Sea with typical Mediterranean climate and seasonally intense population fluctuations due to the touristic activity. These parameters form a highly periodic water demand, with the summer demand surpassing by far –almost six times- the winter demand. The applied method considers an investigation among weather, touristic activity and hydraulic variables. Firstly, the seasonality and the trend are extracted from the variables’ values. The formed residual data of a two year time series are used for training and a single year time series is used for validating. The seasonality and trend are added to the forecasted residuals in order to produce the final forecast. The method benefits among others, in performing adequately when the water demand includes non-linear parts and also provides a Fuzzy Rule Base, giving the researcher a handful tool on interpreting the physical aspect of the interrelationships.
