An optimization method for environmental parameters in factory farming of juvenile spotted knifejaw based on deep learning and dual-objective optimization
摘要
The growth of cultured fish is influenced by various environmental factors. Understanding the growth performance of fish under different environmental factors can help find a balance between energy consumption and productivity. However, there is limited research on optimizing energy consumption and fish growth under different environmental factors. To address this issue, we focus on the specific growth rate, feed conversion rate, and power consumption of juvenile spotted knifejaw under different temperature and flow velocity conditions. Power consumption across flow velocities was modelled using a linear function to identify the optimal flow rate. Furthermore, a Transformer algorithm is employed to predict power consumption at different temperatures, which is subsequently integrated with the non-dominated sorting genetic algorithm II (NSGA-II) to derive a Pareto front. The ideal point method is then applied to this front to determine the optimal temperature parameter. Experimental results identify the optimal growth temperature and flow velocity for juvenile spotted knifejaw as 26.3 ~ 26.4 °C and 0.5 BL/s, respectively. The proposed EP-Transformer model demonstrates strong predictive performance, with a root mean square error (RMSE) of 0.108 kWh and a coefficient of determination (R2) of 0.978 when predicting the power consumption required to maintain different water temperatures for 1 day. The Transform-NSGA-II model achieves bi-objective optimization, maximizing growth rate and minimizing power consumption costs for juvenile spotted knifejaw. The optimal temperature is selected from the Pareto front by the ideal point method, which provides a practical solution for the efficient rearing of juvenile spotted knifejaw.