The paper studies possibilities to indirectly determine actual values of oil dynamic head in the well and load curves of the bottom-hole pump basing on wattmeter cards (electric power input curves) with application of artificial neural networks.
Load indicator reading systems are commonly used to diagnose the condition of a downhole pump. Such systems consist of two sensor units: a force sensor and a displacement sensor. Currently, motion sensors based on integrated accelerometers are widely used. The use of accelerometers allows to create reliable, inexpensive, simple and convenient displacers. However, such sensors have a significant disadvantage, which is a large error due to the following reasons: a large integration error occurs in double integration; measured acceleration signals have a significant level of noise; currently produced integrated accelerometers have a large intrinsic error. In this paper, various methods are developed and investigated to solve these problems.
