Time-series analysis approach to the characteristics and correlations of wastewater variables measured in paper industry

Abstract

Advanced wastewater treatment technologies and protocols are required to minimize the overall water footprint and environmental cost of paper industries. Time-series analysis provides powerful methods to analyze industrial wastewater data with the aim of understanding and predicting the behavior of relevant wastewater variables such as chemical and biochemical oxygen demand. In this study, we introduce a variety of novel computational methods for complex systems and demonstrate their applicability to data obtained from a wastewater treatment plant, including the preprocessing of the raw data, time-dependent characteristics of individual measured parameters, as well as the mutual correlations between multiple parameters. These methods include the Potts model for preprocessing, empirical mode decomposition for periodic component extraction, detrended fluctuation analysis for noise characterization of the data and finally time-lagged windowed cross-correlation, which is utilized to obtain time-dependent couplings between the measured parameters. The results provide valuable insights into the underlying processes in the wastewater treatment plant. The Potts model effectively processes noisy data, from which periodic variations can be successfully removed with empirical mode decomposition, while preserving the relevant characteristics. Further, detrended fluctuation analysis shows prospects for indicating periods of abnormal behavior in the data. Finally, the correlation analysis reveals the characteristic time delays between influent and effluent chemical oxygen demand, giving an average delay of one day, which has implications in plant control. In conclusion, our methods show promising prospects for further applicability in wastewater analysis.