Abstract: To address issues such as water quality management relying on manual experience, insufficient precision in environmental control, and limited intelligence level in existing aquaculture equipment in small-scale aquaculture, an intelligent closed-loop control system for fish farming based on integration of internet of things（IoT）and large language model-assisted decision making has been designed and implemented．System has employed Renesas RA6M5 microcontroller as core, has established a multi-parameter water quality sensing unit to monitor key parameter such as temperature, pH, total dissolved solids（TDS）, and turbidity in real time, and uploaded data to cloud platform via MQTT protocol．By combining an aquaculture knowledge base with a large language model, cloud platform analyzed aquaculture conditions under rule constraints, generated control strategies for temperature regulation, feeding, and water exchange, which were automatically executed by terminal devices, thereby forming a stable closed-loop control process．A 30-day comparative aquaculture experiment has demonstrated that system's stable operation, with water temperature control accuracy reaching ±0.5 °C, and key water quality parameters consistently maintained within appropriate ranges．Statistically significant differences in body length growth were observed, and survival rates exhibited an upward trend．These findings provide an engineering reference for automation and intelligent control of aquaculture environments．