Modeling of a full-scale sewage treatment plant to predict the nutrient removal efficiency using a long short-term memory (LSTM) neural network

A long short-term memory (LSTM)-based neural network was proposed and developed to predict the ammonium (NH4-N), total nitrogen (TN), and total phosphorus (TP) removal efficiency of an anaerobic-anoxic-oxic membrane bioreactor (A-A-O MBR) system after data visualization using Python programming. The influent wastewater characteristics, including, total organic contents, NH4-N, TN, TP, chemical oxygen demand, suspended solids, and operating parameters, such as dissolved oxygen, oxidation-reduction potential, and mixed-liquor suspended solids, were considered as inputs, while removal efficiency was taken as an output parameter. First, data analysis and its normalization were conducted to improve the learning speed of the model. Performance criteria of the proposed model were evaluated based on statistical values, including the mean-square error (MSE) and root-mean-square error-observations standard deviation ratio (RSR). Based on the goodness-of-fit values, the proposed LSTM model achieved good performance in the testing dataset; calculation results deviated little, indicated by the MSE values of 0.0047, 0.015, and 0.018 for NH4-N, TN, and TP, and RSR values of 0.0104, 0.129, and 0.141, respectively. The proposed LSTM model predicted the most precise removal efficiency for NH4-N of A-A-O MBR system, while TN and TP predictions were comparatively less accurate, but still acceptable. The proposed LSTM model is promising for predicting the nutrient removal efficiency of the A-A-O MBR system in real-time and can aid in establishing process control strategies. Therefore, the proposed LSTM model is an adequate interpolation tool to predict the nutrient removal efficiency of the A-A-O MBR process in wastewater treatment systems.