Technical condition of pumping units being reduced during operation as a consequence of normal wear exerts influence on electrical power consumed by the formation pressure maintenance system; it appears in increasing of electrical power losses in the pump.
A method defining optimum structure of the operating pumping units have been developed to reduce electrical power losses caused by the technical condition of pumping units. The efficiency of the developed method to a large extent depends on accuracy of determination of the current technical condition of pumping units.
A diagnostic system based upon the artificial neural network (ANN) and the database obtained from flow sensors of head and temperature at the pump outlet and inlet has been developed for assessing the technical condition of pumping units. For correct operation of the diagnostic system based upon the operation data, neural network training is conducted in order to set the weights between the neurons in such a way, that the INS output signal summary error tends towards zero. Training is executed in Matlab software environment using the Levenberg – Marquardt algorithm. Weights are adjusted depending on the value of the obtained error.
Efficiency factors for two pumps ESP 500-1900 and one pump ESP 180-1900 have been defined using the developed diagnostic system based upon the operational data. The calculation results showed that the estimation error of the efficiency factor obtained using the suggested diagnostic system does not exceed 6%.References
1. Frayshteter V.P., Nissenbaum I.A., Veliev M.K., Improvement of process and
power efficiency of well pad pump stations within oil fields reservoir pressure
maintenance system (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2013,
no. 3, pp. 86-88.
2. Sushkov V.V., Veliev M.K., Procedure for determination of the optimum
structure of operating pump units at well pad pump stations (In Russ.), Neftyanoe
khozyaystvo = Oil Industry, 2013, no. 12, pp. 125-127.
3. Haykin S., Neural networks - A comprehensive foundation, Pearson Education,
Inc, 1999, 832 p.
4. Medvedev V.S., Potemkin V.G., Neyronnye seti. Matlab 6 (Neural networks.
Matlab 6), Moscow: Dialog-MIFI Publ., 2002, 496 p.
