Original scientific paper

https://doi.org/10.17559/TV-20200721133924

Complex Hydrological System Inflow Prediction using Artificial Neural Network

Petar Matić* orcid.org/0000-0002-1799-5257 ; University of Split, Faculty of Maritime Studies, R. Boskovica 37, 21000 Split, Croatia
Ozren Bego ; University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture (FESB), R. Boskovica 32, 21 000 Split, Croatia
Matko Maleš orcid.org/0000-0002-9834-3962 ; University of Split, Faculty of Maritime Studies, R. Boskovica 37, 21000 Split, Croatia

Abstract

Artificial neural networks have been successfully used to model and predict water flows for a few decades. Different network types have proven to work better in different cases and additional tools and algorithms have been implemented to improve those neural models. However, some problems still occur in certain cases. This paper deals with the limitation of complex hydrological system inflow prediction using artificial neural network and inflow time series. This limitation is called the prediction lag and it disables the model from giving accurate predictions. To eliminate the prediction lag and to extend prediction horizon an alternative input variable named forecasted precipitation frequency is proposed in addition to antecedent inflow time-series. Simulation results prove the efficiency of the proposed solution that enables time series neural network model for 7th-day inflow prediction. This represents important information in operational planning of the hydrological system, used for short-term optimization of the system, e.g. optimization of the hydroelectric power plant operation.

Keywords

artificial neural network; complex hydrological system; forecasted precipitation frequency; inflow prediction; prediction lag

Hrčak ID:

269496

URI

https://hrcak.srce.hr/269496

Publication date:

15.2.2022.

Visits: 1.364 *