Seawater Ammonia Nitrogen Concentration Modelling via RS-SCN

Aiming at the complex nonlinearity of ammonia nitrogen concentration detection in the process of marine aquaculture, a soft sensing modeling strategy called RS-SCN combining rough set theory and stochastic configuration networks is proposed. This method utilizes the attribute reduction advantage of rough set in the processing of sample data to analyze the main parameters affecting ammonia nitrogen concentration in seawater. By using the advantage of SCN model in dealing with non-linear problems, the more representative variables and sample data set are selected, and the corresponding soft sensing model is established. At last, the proposed model is applied to the marine aquaculture process for soft measurement of ammonia nitrogen concentration. Experiments with comparisons on the prediction effect of RS-SCN, SCN and BP models are carried out. Results show that the proposed model has better prediction accuracy and generalization performance than other methods in marine aquaculture process.