Data Mining Application in Assessment of Weather-Based Influent Scenarios for a WWTP: Getting the Most Out of Plant Historical Data

Abstract

Since the introduction of environmental legislations and directives, the impact of combined sewer overflows (CSO) on receiving water bodies has become a priority concern in water and wastewater treatment industry. Time-consuming and expensive local sampling and monitoring campaigns are usually carried out to estimate the characteristic flow and pollutant concentrations of CSO water. This study focuses on estimating the frequency and duration of wet-weather events and their impacts on influent flow and wastewater characteristics of the largest Italian wastewater treatment plant (WWTP) located in Castiglione Torinese. Eight years (viz. 2009–2016) of historical data in addition to arithmetic mean daily precipitation rates (PI) of the plant catchment area are elaborated. Relationships between PI and volumetric influent flow rate (Qin), chemical oxygen demand (COD), ammonium (N-NH4), and total suspended solids (TSS) are investigated. A time series data mining (TSDM) method is implemented with MATLAB computing package for segmentation of time series by use of a sliding window algorithm (SWA) to partition the available records associated with wet and dry weather events. According to the TSDM results, a case-specific wet-weather definition is proposed for the Castiglione Torinese WWTP. Two significant weather-based influent scenarios are assessed by kernel density estimation. The results confirm that the method suggested within this study based on plant routinely collected data can be used for planning the emergency response and long-term preparedness for extreme climate conditions in a WWTP. Implementing the obtained results in dynamic process simulation models can improve the plant operational efficiency in managing the fluctuating loads.