Use of Machine Learning in Processes Optimization for Drinking Water Treatment Plant Butoniga (Istria, Croatia)

Abstract

Drinking water treatment plant Butoniga is located in central Istria (Croatia) downstream of the Butoniga reservoir. The Butoniga reservoir is an artificial lake created in 1987 with two main objectives: 1) protection from adverse water impacts, and 2) drinking water supply. The operation of the drinking water treatment plant is mainly related to the tourist season, and the 5.000.000 m3 of produced and distributed drinking water annually, 3.000.000 m3 is produced and distributed during June 15 to September 15, when the raw water quality in the Butoniga reservoir is the worst. Regarding this, main problem with Butoniga reservoir and thus drinking water treatment plant appear in summer months when water temperature is the most critical parameter and raw water for the treatment process must be captured from the lowest layer of the reservoir which have increased concentrations of Mn, Fe, NH4 and lower pH values and thus influence on the treatment processes. To deal with this problem, model predicting Mn, seven days in advance is build using machine learning approaches. Build model have high accuracy compared to the measured data, with a good prediction of the peak values. As such, obtained model can help in optimization of the treatment processes which are depending on the quality of raw water, and overall, in sustainability and management of the drinking water treatment plant Butoniga.