(295f) Fundamentals of Wave Propagation and Development of a Control Strategy for Pump Control in Hydraulic Fracturing

A stimulation job entails high pressure pumping of fluid and proppant into a well-bore to induce hydraulic fractures in the subsurface formation, typically along a horizontal well. There may be 10-25 reciprocating pumps in simultaneous operation. The pressure pulses from the pumps induce wave-guided acoustic modes in the pipes that travel at the wave speed along the pipe. When these bounce off a reflecting surface (such as a valve or a bend in the pipe) they generate standing waves that may combine constructively or destructively. When the piping system comprises elbows, tees, or diameter changes, pressure pulsations can lead to piping vibrations.

As a reduction in the peak-to-peak pressure pulsations will lead to a concomitant reduction in vibration, an active control strategy is presented that reduces the pulsations from the reciprocating pumps by operating in a regime where the stationary waves combine destructively at a majority of points in the piping system.

A 10-pump Flowmaster® CFD simulation model was developed and calibrated using experimental data obtained from a physical 2-pump system. This model was incorporated in an optimization framework designed to minimize the peak-to-peak (P2P) pressure differential observable at multiple locations with active pump speed and phase control. The procedure commences from a given starting condition, the prevailing state of the pump phases and speeds. This solution is then perturbed within specified limits over a number of consecutive steps and the variations are recorded. The process serves to qualify the lower and upper bounds, and the distribution properties, of the P2P merit function in the Sampling Mode, and then aims to detect good configurations yielding low P2P values in the Control Mode. When a solution within the designated tolerance of the lower bound estimate is found, the incumbent design configuration (pump phases and speeds is retained as the active solution). The process reverts to the Sampling Mode if the conditions change appreciably. Results for implementation of this control method will be presented. A solution close to the estimated lower bound is established, that is considerably far from the upper bound. This demonstrates a robust and active control strategy, one that can manage any number of pumps and indeed, is applicable for active real-time control in the field, with its added complexity.