This paper proposes a fuzzy inference method to control the air volume by predicting DO(dissolved oxygen)based on LSTM(long short-term neural network)which solve the problem of unstable DO in the aeration process of wastewater treatment plants using AAO process caused by some unstable factors,such as the delay of DO changes in the aerobic zone of the biological tank,sudden changes of influent water quality and low frequency of chemical method for measuring COD.The method predicts the DO value of the aerobic zone in the biological tank and the fluctua-tion of COD in the influent water by LSTM,then inputs them to the next layer of LSTM network to predict the final DO value.The fuzzy inference system is used to control air volume to achieve stable control of the evaluation index of oxygenation process,DO.The method significantly im-proved the stability of DO in the aerobic zone of the biological tank compared to manual control and PID-based control methods while meeting energy consumption requirements when implemented in a wastewater treatment plant in Wuhan,and proved has a good robustness when COD measurement was not timely after a mutation as the same time.
LSTMFuzzy inferenceSudden changes of CODThe stability of DO
维普
万方数据
