Machine Learning-Based Assessment of Hydropower Integration and Regional Net Exchange in East Kazakhstan
Year: 2026
Type: conference-paper
Abstract: Hydropower is a key renewable electricity source in hydro-rich regional power systems, where it contributes to local demand coverage, operational flexibility, and regional exchange conditions. This paper presents a data-driven clustering-based assessment of hydropower integration and regional net-exchange regimes in the East Kazakhstan regional power system. The study uses operational dispatch measurements for 2024, including regional demand, load, hydropower generation from major hydropower plants, thermal generation, and operational saldo indicators. The original 10-minute measurements were quality-controlled and aggregated to hourly resolution, resulting in 8784 hourly observations. To obtain an interpretable regime classification, K-means clustering was focused on the hydropower share in the hydro-thermal generation mix, while demand coverage and net-exchange intensity were used for post-clustering interpretation. Three generation-mix regimes were identified: a hydro-dominated high-coverage regime, a low-hydro thermal-supported regime, and a mixed hydro-thermal regime. The K-means clustering configuration achieved a silhouette score of 0.6160, a Calinski-Harabasz index of 30520.1188, and a Davies-Bouldin index of 0.4845. A Gaussian Mixture Model baseline produced strong agreement with the K-means assignments, with an Adjusted Rand Index of 0.9167 and a Normalized Mutual Information score of 0.9057. The results indicate a stable and interpretable generation-mix regime structure. The findings suggest that hydropower is one of the key operational factors associated with demand coverage and net-exchange intensity in East Kazakhstan. However, the results should be interpreted as operational dispatch-level profiles rather than deterministic physical or commercial exchange states.
Institution: Astana IT University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203634544