Hybridisation of artificial neural network with particle swarm optimisation for water level prediction

Abstract

Accurate water level (WL) prediction is essential for the efficient management of various water resource projects. The creation of a reliable model for WL forecasting is still a difficult task in water resource management. This study applies an artificial neural network (ANN) integrated with the particle swarm optimisation algorithm (PSO-ANN) for simulating monthly WL of the Tigris River in Alkut City, Iraq. Data pre-treatment methods are utilised for improving raw data quality and detect the optimal predictors. Monthly WL and climatic variables from 2011 to 2020, were used to construct and validate the proposed technique. The results showed that singular spectrum analysis (SSA) is a high-performance technique for denoising time series. The PSO-ANN model produces good results coefficient of determination (R2) of 0.85.