Adaptive neural fuzzy inference system for feeding decision-making of grass carp (Ctenopharyngodon idellus) in outdoor intensive culturing ponds

Feed is the most main expenditure in outdoor intensive culture system. The improvement of feeding efficiency has great significance for increasing production and reducing costs. Presently, with the development of precision agriculture, automatic adjustments of the feeding amount according to the demands of the fish has become a developing trend. The objective of this paper was to develop an automatic feeding decision-making system based on the water quality parameters to solve the problem of inefficiency in artificial feeding control. In this study, we proposed an effective control method using adaptive neural fuzzy inference system (ANFIS) to achieve this purpose. First, two input variables (dissolved oxygen saturation [DO]; temperature [T]) and one output variable (feeding percent [FP]) were selected and defined. Second, the model of the linguistic variables and the optimal fuzzy rule base were obtained by the training and learning by utilization of hybrid learning methods. Finally, an ANFIS controller for on-demand feeding was developed and the performance was compared with Fuzzy logic control (FLC) and artificial control (AC) by the Nash-Sutcliffe efficiency coefficient (NS), the root mean squared error (RMSE), and fish growth parameters. The results indicated that the NS and RMSE of the ANFIS model were 0.8539 and 0.0541, respectively, and were better for forecasting feeding decisions compared with the FLC and AC methods. Compared with the AC, there was no significant differences in promoting fish growth (P > 0.05), whereas the feed conversion rate (FCR) was reduced by14.35%. In addition, the mean of ammonia nitrogen concentration decreased by 22.59%, and the mean of turbidity increased 5.5 cm to 28.9 cm, reducing eutrophication and pollution of water in pond. Therefore, applying those approaches based on ANFIS control to the feeding decision system in outdoor intensive culturing is flexible and effective, and has potential for the design of fine feeding equipment and to guide this practice for other species.