Abstract
<title>Abstract</title> <p>Optimizing Progressive Cavity Pump (PCP) artificial lift systems is limited by the challenge of observing downhole fluid submergence. Physical downhole sensors for direct measurement are often susceptible to failure in harsh environments while surface flowmeters can be economically and technically infeasible in heavy-oil, multiphase applications. This research introduces the design, implementation and field validation of a sensorless Industrial Internet of Things (IIoT) novel architecture that employs an Extended Kalman Filter (EKF) as a real-time virtual sensor to address this operational challenge. A reduced-order, nonlinear state-space model was derived to correlate surface-measured mechanical torque and electrical parameters with the hydrostatic fluid load. The EKF was integrated with supervisory logic recognizing that purely statistical observers are prone to model mismatches during hydrodynamic transients, multiphase slugging and zero-speed periods. This hybrid architecture incorporates a state correction rate limit, friction torque interlock and anomaly detection to address these periods. This proposed architecture was implemented on a heavy-oil production asset (Well X). The EKF submergence estimates were evaluated against independent acoustic surveys and a Root Mean Square Error (RMSE) of 54.3 m (178.16 ft) was calculated. This reflected an error margin of 3.82% across the 1,420.4 m (4,660 ft) wellbore, establishing a novel benchmark for sensorless estimation accuracy. The ability to continuously estimate fluid levels enables operators to optimize pressure drawdown. This operational optimization unlocks substantial and scalable commercial value. When assessed over a one year operational cycle for a full-scale fleet of 24 wells, this architecture yields a net one year revenue of USD $7,647,675.59 (TT$ 51,851,237.84). An estimated 5% increase in production results in aggregate Year 1 operating revenue of USD $7,726,320.00 (TT$ 52,384,450.00). Additionally, the deterministic interlock effectively mitigates the risk of catastrophic dry-run stator failures, leading to a cost avoidance of USD $3,000,000.00 (TT$ 20,340,000.00) per well. Moreover, transitioning to this self-hosted LoRaWAN architecture delivers a consolidated annual savings on operational expenses amounting to TT$ 669,580.80 (USD $98,758.00). This is achieved by eliminating telemetry subscription fees and changing manual acoustic diagnostics to a semi-annual schedule, ultimately saving 1,920 labor hours each year. While the system successfully achieved its design objectives, it is bounded by specific constraints. The reliance on static parameter approximations may introduce estimation bias during significant reservoir shifts or long-term component wear, necessitating periodic recalibration. Furthermore, the telemetry update rate is restricted to five minutes and live closed loop speed modulation was limited to open loop validation to avoid inducing transient mechanical stress on the active rod string. Despite these challenges, the project successfully integrates nonlinear stochastic estimation with practical industrial automation thereby providing a resilient, lucrative solution for the intelligent control of PCPs.</p>