关键词: 软测量, 支持向量机, 遗传算法, 氨氮浓度

Abstract: As ammonia nitrogen concentration measurement in the intensive aquaculture process is difficult online, large investment and low precision, a soft sensor modeling method based on GA and SVR is proposed. Firstly, according to the analysis of aquiculture water quality influence factors, water temperature, dissolved oxygen, pH and conductivity are collected and used as auxiliary variables. Then, GA is used to optimize penalty parameter C and kernel function parameter g in SVR. Finally, the SVR model is used to predict the ammonia nitrogen concentration in water during aquaculture process. The predicted results were compared with BP, RBF neural network and SVR model based on grid search. The experimental results show that the soft-sensing method based on GA-SVR has better prediction accuracy, it can also provide effective operation guidance for the control and optimization of the intensive aquaculture process.

Key words: Soft sensor, support vector regression, genetic algorithm, ammonia nitrogen concentration