Water quantity prediction based on particle swarm optimization and evolutionary algorithm using recurrent neural networks

Stormwater pollution is one of most important issues that the District of Columbia faces. Urban stormwater pollution can be a large contributor to the water quality problems of many receiving waters, as runoff transports a wide spectrum of pollutants to local receiving waters and their cumulative magnitude is large. Therefore, evaluations of stormwater runoff quantity are necessary to enhance the performance of an assessment operation and develop better water resources management and plan. However, some computational intelligence methods that have most successful applications on time series prediction have not yet been investigated on water quantity prediction. Only a limited number of neural networks models were applied to the water quantity monitoring. Therefore, we proposed an Elman style based recurrent neural network on the water quantity prediction. A hybrid learning algorithm incorporating particle swarm optimization and evolutional algorithm was presented, which takes the complementary advantages of the two global optimization algorithms. The neural networks model was trained by particle swarm optimization and evolutional algorithm to forecast the stormwater runoff discharge. The USGS real-time water data at Four Mile Run station at Alexandria, VA were used as time series input. The excellent experimental results demonstrated that the proposed method provides a suitable prediction tool for the stormwater runoff monitoring.