Abstract
Prediction of water flow rate in a photovoltaic water pumping system (PVWPS) is of high importance for investors who wish to achieve an efficient management of water demand in remote and desert areas. In this paper, different prediction methods based on Artificial Neural Networks (ANNs) have been investigated and compared. Data used to predict and estimate the hourly water flow rate have been acquired from an experimental PVWPS installed at Madinah site (Saudi Arabia). Results show that developed models can predict accurately the hourly flow rate based on measured hourly air temperature and solar irradiation, as input parameters. They can be used first to control the PVWPS by making a comparison between measured and predicted hourly flow rate, second to investigate the economic feasibility of the system to supply water in desert areas or isolated sites that have no access to an electric grid depending on water demand and finally fault detection based on the unexpectedly changing of delivered water amount. Operators can benefit from the proposed models. In fact, once their own PVWPS model is designed, they can predict its flow rate given the weather forecasts for the following day.