Data mining of WWTP data

Data mining of WWTP data to optimze plant operation

On wastewater treatment plants (WWTP), much effort and money is spent to measure operating parameters, particularly online instrumentation for process monitoring and control is extensively used. The amount of data gathered is steadily growing due to the increasing complexity of the plants and due to new developments in sensor and information technology. Often, several hundred signals are measured in parallel, each signal having its own characteristics. In order to cope with the massive data, modern plants use process information systems which record the data and ensure a homogenous dataset.

Because exploring and analyzing the massive and dense data sets is difficult, operators nowadays still rely on basic statistical methods to investigate the gathered measuring data and therefore do not fully exploit the information contained in the data. Hidden information, which cannot be discovered by classic approaches and human eyes only, is lost. Consequently, optimal plant control, together with successful water pollution control, relies on decisions that are based on the operator’s long-term experience and intuition.

Advancements in the field of information technology allow the development of novel data mining methods and algorithms specialized on the analysis of massive data sets and are capable of the extraction of implicit, previously unknown but potentially useful patterns.

Within the scope of a PhD project, new data mining algorithms shall be developed and existing methods shall be adapted in order to assist WWTP plant operators with information extracted from the large data set available. The systematic data exploitation will have positive impact on the environment and on treatment costs as well as on process understanding.

The methods developed in this project are implemented as ready to use software in order that they can be used by WWTP plant operators on-site during their day-to-day task.

Focus of research

• Methods to generate new, previously unknown signals (software sensors) and methods to check existing probes using operating data

• Methods to model the hydraulic and biological processes in activated sludge tanks using information extracted from data collected in simple and inexpensive experiments or from operating data

• Development of innovative visualization concepts to illustrate the current state of a WWTP and its comparison with past states