The use of radial basis function networks (RBFN) to predict critical water parameters in desalination plants

The quality of seawater is very important in the design of the pre-treatment unit of any desalination plant. It is worth noting that the operational parameters (salinity and TDS) depend on the seasons (dry or hurricane) such that they are not constants along all year. Recently, new studies indicate that the RBFN is a good tool for prediction of parameters such that the aim of this paper is to apply a RBFN approach as predictive model focus in the context of the Seawater in the Caribbean. The model is obtained taking into account the behaviour of salinity and total dissolved solids content according to the season (dry or hurricane season). The methodology is based on redistributions of centres to locations where input training data possess significant effects can lead to more efficient RBFN. The proposed method herein is based on clustering of input space vectors and computing weights of Euclidian distances and histogram equalization within each cluster will determine the centre and width of each receptive field. The results of this paper indicate that the parameters can change up to 30% along year and the accurate of the prediction obtained can be around 96.7%.