Identification of water polluting enterprises based on neural network analysis

The aim of the work was to solve the problem  of increasing the efficiency of identifying large enterprises that pollute natural water from among many water users of the industrial region. For this purpose, artificial neural networks (INS) were used to detect and evaluate the weights of connections between statistical data, which is important for studying the dynamics of water quality formation in conditions of its spatial and temporal variability. The work was carried out on the example of the available 250 measurements of the  concentration of each  of the  four  priority metals at  the  hydro  chemical gates  of the  Iset  River in  the  area of Yekaterinburg. Neural network analysis made  it possible  to detect  the  interdependencies of individual water quality indicators at neighboring sites. At points  above Yekaterinburg (+5.2 km,  target 1), in the city (target 2) and below it (–4 km,  target 3), in total at 3 × 4 × 250 = 3000 points.  It was found, in particular, that  the effect of the nickel  content  in the water  of the gate 2 on the concentration of other metals of the gate 3, especially zinc, is quite high, so that the correlation coefficient is not lower than 0.6. Such results  made it possible to identify the logistical economic ties of water users and simplify the identification of water pollutants by the “water footprint” left by related enterprises. Thus, it is shown that the INS provides the identification of a man-made decrease in water quality against the background of its natural pollution with the same substances. The reliability of the conclusions is confirmed by the ability to satisfactorily predict the water quality of the section located downstream of the river, according to the data for the section located above, as established in the work by predicting water quality using the INS.

1. Stroganov S.N. Svyaz pitaniya zhitelej i sostava stochnoj vody: Pitanie Moskvy` v 1903-1922 gg. Tr. Soveshchaniya po ochistke stochnyh vod, sostoyashchego pri Kanalizatsionnom otdele M.K.H. // Moskovskoe kommunalnoe hozyajstvo; 1923 g. Vyp. 2.

2. Virender K Sharma, Eric Lichtfouse, Etienne Decroly. COVID-19 epidemiologic surveillance using wastewater. Environmental Chemistry Letters. 2021. V. 19(45). N1. P. 1911-1915.

3. Danilov-Daniljyan V.I. Otsenka sovremennyh podhodov k upravleniyu kachestvom poverhnostnyh vod i ih ohrane / Venitsianov E.V., Adzhienko G.V., Kozlova M.A. // Vestnik RAN. 2019. T. 89. № 12. S. 1248-1259.

4. Gosudarstvennyj doklad «O sostoyanii i ispolzovanii vodnyh resursov Rossijskoj Federatsii v 2020 godu» istina. msu.ru›publications/book/523146645/

5. Rozental O.M., Aleksandrovskaya L.N., Kirillin A.V. Bajesovskij podhod k povysheniyu dostovernosti kontrolya kachestva vod // Analitika i kontrol. 2018. T 22, № 3. S. 334-340.

6. Yakovlev S.V., Voronov Yu.V. Vodootvedenie i ochistka stochnyh vod. M.: ASV, 2002. 704 s.

7. Diffuznoe zagryaznenie vodnyh objektov: problemy i resheniya. Kollektivnaya monografiya pod ruk. V.I. Danilova-Daniljyana. M.: RAN, 2020. 512 s.

8. Martin Anthony, Peter Bartlett. Neural Network Learning: Theoretical Foundations. Cambridge University Press, Cambridge, New York, 2002. 504 p.

9. Borovikov V.P. Nejronnye seti STATISTICA Neural Networks: Metodologiya i tehnologiya sovremennogo analiza dannyh. M.: StatSoft, 2015. 354 s.

10. GOST R ISO 14046-2017 Ekologicheskij menedzhment. Vodnyj sled. Printsipy, trebovaniya i rukovodyashchie ukazaniya.

11. GOST 27384-2002. Voda. Normy pogreshnosti izmerenij pokazatelej sostava i sojstv.