A hybrid intelligent soft computing method for ammonia nitrogen prediction in aquaculture

Abstract

Ammonia nitrogen is one of the key parameters in determining the aquaculture water quality condition in pond. The high level of ammonia nitrogen is likely to cause deterioration of water quality and mass death of cultured subjects. Therefore, accurate detection of the cultured water ammonia nitrogen content is crucially important for aquaculture management. While, at present, the accuracy of equipment for measuring the ammonia nitrogen content of aquaculture water in real time cannot meet the requirements for aquaculture. In this paper, the soft computing method is firstly proposed to predict the ammonia nitrogen content in aquaculture water in real time. This method includes empirical mode decomposition (EMD), improved particle swarm optimization (IPSO) and extreme learning machine (ELM). To evaluate the performance of the soft computing techniques, three different statistic indicators were used, including root mean square error (RMSE), the mean absolute error (MAE), and the mean absolute percentage error (MAPE) to compare three artificial soft computing methods. Results showed that the EMD-IPSO-ELM soft computing method showed the best performance among other studied methods in the ammonia nitrogen real time prediction. The EMD-IPSO-ELM model provides moderately and roughly accurately real time prediction value of ammonia nitrogen in aquaculture water.

Keywords

• Ammonia nitrogen prediction
• Extreme learning machine
• Soft computing
• Aquaculture