Development of a data-driven method for automated water-flow monitoring and short-term assessment in irrigation hydrotechnical systems

Abstract

The object of the study is the process of automated monitoring of water-flow parameters at irrigation hydraulic structures. The problem solved is the transformation of irregular supervisory control and data acquisition (SCADA) records from multichannel Doppler sensors into reliable time series for water accounting and diagnostics, since raw data may contain irregular intervals, duplicates, telemetry gaps, and zero-value artefacts. A data-driven method was tested at the Talas River cross-section near Zhasorken, Republic of Kazakhstan. The method combines four ultrasonic Doppler sensors, SCADA acquisition, VPN transmission, database integration, preprocessing, and Detrended Fluctuation Analysis (DFA). The workflow includes timestamp verification, duplicate handling, gap identification, zero-value analysis, resampling, and interpolation. As a result, 27,296 irregular SCADA records were transformed into 8,424 hourly observations with 20 variables and no missing values. DFA applied to four cleaned water-depth channels produced exponents of α = 1.304, α = 1.306, α = 1.250, and α = 1.245 for depth1-depth4, respectively, with R2 > 0.995 in all log-log approximations. These results indicate a persistent temporal structure, explained by hydraulic inertia, flow coupling, nonstationarity, and possible regulation. The distinctive feature is a complete chain from physical water-flow measurement to analysis-ready diagnostic time series. Thus, the method makes it possible to use imperfect SCADA records for short-term engineering assessment. The results can be used in irrigation systems, concrete-lined river sections, and canal monitoring points with Doppler measurements and SCADA infrastructure for automated water accounting, regime diagnostics, and decision support