Comparison of different artificial intelligence-based models for prediction of dissolved oxygen rates at fish growing pools- Case Study: Ahvaz city

High energy consumption and worker costs caused to fish growers don’t have highly motivated to buy aerators. Dissolved Oxygen (DO) is one of the importance parameters which can be applied for time of starting and finishing activity of aerators however its sensor is expensive and should be estimated by other parameters. The aim of this research was to create a suitable model to estimate DO rate at growing pools. Input parameters were pH and temperatures of water, moisture and temperature of air and wind speed. Along growing period total of parameters were measured and with three models were estimated. The first one was Artificial Neural Network (ANN). The results showed that the maximum R2 was achieved by logsig-purelin transfer function with 17 neuron at hidden layer with 0.70. Second one was an aggregated ANN-AG model and its result showed that the maximum of R2 was 0.41 and finally, Third one was Adaptive Nero Fuzzy Inference System (ANFIS) model and the results showed that at condition of 3 Membership function per layer with gaussian type and constant output could estimate DO with R=0.87 accuracy. Therefore ANN was the best method for evaluation of DO at growing pools for the region under study.