The concentration of ammonia nitrogen is one of the key indexes in the process of marine aquaculture. Excessive levels of ammonia nitrogen in the water produce strong neurotoxins, leading to large-scale death of aquatic organisms. Therefore, it is very important to monitor the concentration of ammonia nitrogen in water in real time and accurately. Due to many factors affecting seawater quality, and the complex factors often affect each other, there is no instrument to realize the real-time detection of seawater ammonia nitrogen concentration at present. Firstly, the current research status of ammonia nitrogen monitoring in water of aquaculture was reviewed. Then, the formation and nitrification process of ammonia nitrogen in marine aquaculture water was analyzed，and the parameters (temperature, conductivity, pH value and dissolved oxygen concentration) related to ammonia nitrogen concentration were selected as auxiliary variables. A soft measurement model of ammonia nitrogen concentration was established by using a stochastic configuration networks with high convergence speed and strong generalization ability. In order to verify the effectiveness, the proposed method was compared with other neural network modeling methods by using the measured data of the turbot intensive marine aquaculture system independently established by the laboratory. The results showed that the proposed method had higher generalization ability, higher prediction accuracy and faster running speed. Finally, the aquaculture water quality monitoring system was developed, and this method was embedded in the upper computer WinCC software to realize online monitoring of ammonia nitrogen concentration.