Abstract
This study addresses key challenges in intensive aquaculture, such as passive environmental control, high energy consumption, and neglected fish stress, through the development of a multi-objective environmental regulation system for crucian carp utilizing behavioral stress feedback. It combines YOLOv8s-FasterNet for behavior recognition, a specific growth rate model and an energy cost model to form an intelligent decision-making mechanism that maximizes the output–input ratio. In a 25-day experiment, the system showed strong performance. Final body weight and specific growth rate were comparable to the control group. Economically, the system achieved periodic profits that were 8.93, 1.43, and 1.03 times greater than those of traditional threshold control at external temperatures of 2 °C, 8 °C, and 14 °C, respectively, demonstrating significant energy savings. In terms of animal welfare, principal component analysis confirmed significantly lower stress-induced damage in the experimental group, with a comprehensive score (−0.036) closer to the initial healthy group (0.223) versus the control group (−0.348). These results indicate that the system successfully optimized both economic efficiency and fish health, providing a viable solution for intelligent aquaculture management.