Feed is the most important cost of aquaculture. Precise feeding based on feeding welfare is the focus and difficulty of aquaculture research. The fundamental way to solve the above-mentioned problems is to clarify the characteristic parameters such as feeding demand and feeding rhythm, and present them to decision makers in an intuitive form to guide production. On the basis of systematic literature analysis, the paper deeply analyzes the research progress and development trend of precision feeding technology in aquaculture at home and abroad, and summarizes the specific implementation process and application scenarios of precision feeding research methods based on feeding behavior, non-feeding behavior, self-demand feeding from the perspectives of operating principles, applicable models, and technical characteristics. Based on the scenarios, the key technical points and limitations involved in each method are analyzed in depth. In view of the complex and changeable breeding environment and the uncertainty of fish behavior, realizing accurate feeding of fish still faces a series of challenges. Some problems are still outstanding, such as the difficulty of simultaneous improvement of quality and efficiency, over generalization of feeding decision model, and insufficient acceptance of implementation costs. In the future, from the three aspects of monitoring quality and efficiency, decision-making model, and implementation cost, it is necessary to further improve the sensor detection accuracy and expand functions, to optimize the processing efficiency of algorithm models, and to integrate the advantages of various precision feeding technologies. Systematic research should be carried out based on the logic line of “when to eat (feeding rhythm)-how much to eat (feeding demand characteristics)-how to eat” to achieve efficient and unified system of breeding object-feed-equipment, which can scientifically formulate the feeding mode based on breeding objects, farming models, and farming environment so as to maximize the breeding and environmental benefits, and help the high-quality and healthy development of aquaculture in China.

Feeding systems can dominate the primary efficiency and cost in aquaculture. The high management efficiency of feeding can also alleviate ecological environmental pollution. However, artificial identification has been widely used for fish feeding in the pond so far, leading to the time-consuming and laborious. Empirical feeding management cannot fully meet the large-scale pond aquaculture, in terms of production efficiency, environmental pressure, and aquaculture risk. In this study, a set of precision feeding systems was developed using fish feeding welfare. Four modules included water quality monitoring, decision control, drive execution, and remote monitoring. A systematic investigation was also implemented to clarify the effects of environmental factors on the physiology and energetics of fish. The affecting factors were then determined by the feeding demand of fish. The dissolved oxygen, temperature, and body growth were taken as the input parameters, whereas, the target demand was as the output parameters of the feeding system. The fuzzy PID control was combined with the particle swarm optimization (PSO) to realize the precision feeding operation. A comparative test of pond culture was carried out with the intensive grass carp as the research object, in order to verify the practicability and effectiveness of the system. A comprehensive evaluation was made from the aspects of regulation, growth performance, and economic and environmental benefits. The results showed that a stable and reliable control performance was achieved in the constructed precision feeding system, where the control error was less than 7%. The nash-sutcliffe (NS) index increased to 0.913 in the decision-making performance of the feeding system. whereas, the root mean squared error (RMSE) was reduced by 16.10, compared with the traditional. The feed conversion rate was significantly reduced by 11.73 % (P<0.05). Importantly, the aquaculture income increased by about 14 600 yuan/hm², whereas, the pollution was reduced by 241.40 kg per ton of fish produced. When the feed coefficient was 1.0, per ton of aquatic products brought about 500 kg of organic waste, 26 kg of nitrogen, and 13 kg of phosphorus. When the feed coefficient was 2.5, per ton of aquatic products produced 1 625 kg of organic waste, 117 kg of nitrogen, and 38 kg of phosphorus. By contrast, the grass carp produced per ton was equivalent to about 1 250 kg of organic waste, 86.70 kg of nitrogen, and 29.70 kg of phosphorus under the traditional feeding model. The total production of national grass carp in 2021 was 5.76×10⁶ t. If all grass carp farming was adopted as the precise feeding system, it was equivalent to reducing the pollution emissions of 1.17×10⁶ t organic waste, 9.43×10⁴ t nitrogen, and 2.59×10⁴ t phosphorus. The developed system can be expected to alleviate the environmental and ecological pollution caused by aquaculture. Strong comprehensive application performance can provide the theoretical reference and technical support for the research and development of other aquaculture modes and farmed fish. The feeding system should be further improved from three aspects in the future: monitoring quality and efficiency, decision-making model, and implementation cost. The sensor detection accuracy can be improved to optimize the processing efficiency of models. Systematic research of multiple feeding can be integrated using the logic line of “when to eat (feeding rhythm)-how much to eat (feeding demand characteristics)-how to eat”, in order to achieve an efficient and unified mode of breeding objects, feed, and machine.