The prediction of the influent load is of great importance for the improvement of the control system to a large wastewater treatment plant. A systematic data analysis method is presented in this paper in order to esti...The prediction of the influent load is of great importance for the improvement of the control system to a large wastewater treatment plant. A systematic data analysis method is presented in this paper in order to estimate and predict the periodicity of the influent flow rate and ammonia （NH3） concentrations： 1） data filtering using wavelet decomposition and reconstruction; 2） typical cycle identification using power spectrum density analysis; 3） fitting and prediction model establishment based on an autoregressive model. To give meaningful information for feedforward control systems, predictions in different time scales are tested to compare the corresponding predicting accuracy. Considering the influence of the rainfalls, a linear fitting model is derived to estimate the relationship between flow rate trend and rain events. Measurements used to support coefficient fitting and model testing are acquired from two municipal wastewater treatment plants in China. The results show that 1） for both of the two plants, the periodicity affects the flow rate and NH3 concentrations in different cycles （especially cycles longer than 1 day）; 2） when the flow rate and NH3 concentrations present an obvious periodicity, the decreasing of prediction accuracy is not distinct with increasing of the prediction time scales; 3） the periodicity influence is larger than rainfalls; 4） the rainfalls will make the periodicity of flow rate less obvious in intensive rainy periods.展开更多
文摘The prediction of the influent load is of great importance for the improvement of the control system to a large wastewater treatment plant. A systematic data analysis method is presented in this paper in order to estimate and predict the periodicity of the influent flow rate and ammonia （NH3） concentrations： 1） data filtering using wavelet decomposition and reconstruction; 2） typical cycle identification using power spectrum density analysis; 3） fitting and prediction model establishment based on an autoregressive model. To give meaningful information for feedforward control systems, predictions in different time scales are tested to compare the corresponding predicting accuracy. Considering the influence of the rainfalls, a linear fitting model is derived to estimate the relationship between flow rate trend and rain events. Measurements used to support coefficient fitting and model testing are acquired from two municipal wastewater treatment plants in China. The results show that 1） for both of the two plants, the periodicity affects the flow rate and NH3 concentrations in different cycles （especially cycles longer than 1 day）; 2） when the flow rate and NH3 concentrations present an obvious periodicity, the decreasing of prediction accuracy is not distinct with increasing of the prediction time scales; 3） the periodicity influence is larger than rainfalls; 4） the rainfalls will make the periodicity of flow rate less obvious in intensive rainy periods.
