Кey Words: Neural classifiers; state of river waters; Struma river; presence of inadmissible pollution.
Abstract. The aim of the present work is to study the possibilities for developing a system for monitoring the state of the water on the Struma River and the presence of unacceptable pollution. The development of the algorithms of the monitoring system is in accordance with the normative documents of the Republic of Bulgaria and is built on the basis of monitoring measurements for the quality of the river waters from six monitoring points for the period 2010-2019. The following variables characterizing water quality were analyzed:
ammonium nitrogen (NH4-N), nitrate nitrogen (NO3-N), nitrite nitrogen (NO2-N), pH, biological oxygen demand (BOD), electrical conductivity (EC), total nitrogen (TN), total organic carbon (TOC), total phosphorus (TP), orthophosphate phosphorus (PO4-P), dissolved oxygen (DO), temperature (T) and chemical oxygen demand (COD).
The section of the river Struma from before Blagoevgrad to the border (bridge for the village of Topolnitsa) was chosen for the research. The available measurement data from six monitoring points in the considered section of the Struma River are as synchronized as possible.
Correlation analyzes were performed for the possibility of combining the available measurements from the six monitoring points and for the presence of a correlation dependence between the measured variables characterizing the water quality in the Struma River. The presence of correlations between the variables directs the use of the principal components method as a mathematical apparatus for building the monitoring system. Procedures have been developed for classifying the status of river waters depending on current measurements / analyzes. Expressions for the main components for eleven of the measured variables characterizing water quality are derived.
Research has been carried out to develop a procedure for classifying these measurements into the terms “excellent”, “good” and “moderate”.
Possibilities for classifications with neural networks have been studied. Measuring some of these variables is a lengthy procedure and can hardly be used for operational actions. For this reason, the possibilities for developing a procedure for monitoring and classifying the status of river waters on the basis of variables, which in principle can be measured automatically (pH, EC, DO and T), have been explored. They also apply the principal components method and classification with neural networks in the same classification categories.
The developed approaches for monitoring the quality of the waters of the Struma River can be applied for each measurement in the six monitored monitoring points.
