Performance prediction model for desalination plants using modified grey wolf optimizer based artificial neural network approach

Abstract

Desalination represents an effective method for alleviating water scarcity, applying algorithmic techniques to predict the performance of reverse osmosis (RO) desalination plants, Modified Grey Wolf Optimizer (MGWO) based Artificial Neural Networks (ANN) can predict the performance of membrane distillation (MD) equipment. Four experimental inputs are selected: feed salt concentration(35–140 g/h), feed flow rate(400–600 L/h), evaporator inlet temperature (60–80 ℃), and condenser inlet temperature (20–30 ℃). The permeate flux (L/h m2) is selected as the experimental output. Ten prediction models were proposed and compared with the existing models (ANN, WOA-ANN, and GWO-ANN). The results showed that the MGWO-ANN model-5 showed the best regression results: R2 = 99.3 %, mean square error (MSE)= 0.004. This model outperformed the existing ANN model (R2 =98.8 %, MSE=0.060), WOA-ANN model (R2 =99.1 %, MSE=0.005) and GWO-ANN model (R2 =98.9 %, MSE=0.007). Model-5 has a single hidden layer (H=1), 13 hidden nodes (n = 13), 10 search agents (SA=10), and 75 %− 20 %− 05 % dataset division. Its residual error is within acceptable limits (spanning −0.1 to 0.2). Optimizing the number of hidden layer nodes (n) and the number of search agents (SA) can improve the training efficiency and prediction accuracy of the model, and the MGWO-ANN model is capable of more accurately predicting the performance of desalination plants.