Abstract
Sand production constrains artificial-lift reliability, shutdown management, and energy efficiency. This study develops an integrated experimental framework linking particle settling, bulk sand transport, shutdown-related plugging, and electrical demand using 12 anonymized Kumkol wells. The database contains 4380 daily records, 720 laboratory tests, 84 shutdown/restart events, 3600 energy points, 4800 high-frequency restart samples, and 132 maintenance events. In 480 settling column tests, Stokes yielded R2 = 0.937 and MAPE = 24.8%; one-parameter calibration improved R2 to 0.981 and MAPE to 11.3%, with leave-one-well-out MAPE of 11.4%. In 240 flow loop tests, the full-data non-unstable threshold was 1.052 m/s; nested held-out well accuracy was 87.9%, and the strict stable criterion yielded 83.8%. Shutdown duration was the dominant field predictor: plug odds rose 3.87-fold per 10 h, and nested threshold validation yielded 78.6% accuracy. For ESPs, the ratio-based SEC minimum occurred near 0.773 m/s. A denominator-free active power model controlling for pressure rise, VFD frequency, sand concentration, and well effects achieved R2 = 0.851 and retained a positive velocity coefficient in all leave-one-well-out fits. The results define a locally calibrated energy–transport operating window; laboratory transport thresholds are not claimed as direct field setpoints without hydrodynamic scaling and on-well verification. The flow loop threshold block comprises only four independent wells; accordingly, its held-out results are treated as small-cluster evidence, and the numerical velocities remain laboratory reference constraints rather than direct field settings.