Use of Neural Networks for Simulation of Pump

2017 2nd International Conference on Electrical and Electronics: Techniques and Applications (EETA 2017) ISBN: 978-1-60595-416-5

Use of Neural Networks for Simulation of Pump Equipment Operation for a Feed Preparation Complex of an Oil-and-gas Production Enterprise Iakov S. KOROVIN Southern Federal University, Scientific Research Institute of Multiprocessor Computer Systems, (SRI MCS SFEDU), Taganrog, Russia Keywords: Electic submersible pumps, Modelling, Artificial neural networks, Intelligent simulation, Hybrid approach.

Abstract. Creation of a system of computer-aided analysis of pump equipment functioning is an actual problem of design and development of nodes and units for increasing of estimation reliability of products operation, for real-time diagnostics of operation, and for support of real-time decisions for elimination of faults. Owing to wide approximation abilities of a neural network computational architecture it is possible to create multi-parametric simulation mathematical models of complex technical systems intended for identification of operation parameters and criteria of operability. Applying this approximation it is possible to develop an optimization strategy of pump equipment control, upgrade of diagnostic and emergency protection systems. Introduction Nowadays the situation in Russian oil industry is severe difficult. Oil companies have faced two basic challenges: • the fall of world crude oil prices at least in two times; • sanctions against Russian key oil sector companies, initiated by United States, United Kingdom of GB and NI and the European Union countries, that lead to restrictions of applying foreign equipment and technologies. Such circumstances have left no choice except to develop domestic technologies and further to apply them in oil production process. One of the basic problems to be solved, taking into consideration the era of cheap crude oil prices is the decrease of oil production primary cost. Solution of this problem will directly cause the raise of oil extraction profit and will stabilize the situation is industry in whole. One of the most obvious ways to reduce the cost oil extraction process –is to reduce the number of failures of the major and the most expensive and popular equipment. Further on we consider a novel approach of modeling of the key well equipment functioning. Electric Submersible Pumps–the Key Oil Well Equipment Our choice was made to the side of electric submersible oil well pumps (ESP). The reasons to choose a submersible oil pump as an object for research activities and further on as a controlled objects were: • Electric submersible pumps are widely implemented in the oil extraction process; • The price of this type of equipment is comparatively high; • The failure of such equipment causes great economic wastes because of: • The necessity to install and adjust new pump equipment (especially in cases when the pumps fall into the well column); • The volume of unextracted hydrocarbons (the amount of extracted oil is strictly planned and firmly controlled) • There are significant amounts of retrospective data, depicting the modes of ESP functioning, stored in the corporative databases (this moment is vital cause we are applying data mining methods) • The level of ESP automation is rather high in comparison with the other types of the equipment. The values of some 300 parameters are stored in databases and can be handled as 101

in online, so as in offline mode, section headings are in boldface capital and lowercase letters. Second level headings are typed as part of the succeeding paragraph (like the subsection heading of this paragraph). Application of Artificial Neural Networks The methodology of development of neural network models of pump equipment operation processes, created by our scientific team, is based on analysis of available data of telemetry observation. We take in to account all relevant information for the whole period of electrical submersible pumps functioning. The input parameters, whose variation influences pump equipment operation, and whose values are measured during operation, are data from existing automatic system. The output parameters, which characterize operability of a unit, are both estimation of trouble-free operation, and the values set by the user or by the top levels of a hierarchical automatic control system. Direct measuring of these values is rather difficult or is not possible, and further they are used as control parameters in a safety system. Owing to the neural network mathematical model which simulates functional dependence of the operability criteria from the input parameters in the operating range of their variation, we can increase quality of experimental data analysis, and can timely detect defects and problems of operation. One of the main requirements to the created multi-parameter “portraits” of pump units operation is high precision of the model results for decreasing of allowances for operability parameters detection during diagnostics and control of technical condition [1][2][3][4][5]. Intelligent Simulation of Oil Equipment The One of the ways of simulation of centrifugal pumps operation is use of an analog method (particularly electrohydraulic), as a base for integration of scientific knowledge from various areas of science for creation of models of objects and processes of the surrounding world. Application of the fundamental circuit theory is a well-known mechanism. Taking into account use of analogy, we can say that this theory claims to be a generalized theory for simulation of subsystems of various nature (electrical; a mechanical subsystem of progressive motion; a mechanical subsystem of rotary motion; hydraulic (pneumatic) and thermal) [6]. To optimize mathematical simulation of pump equipment with invariable design parameters, in this paper we suggest to consider the following types of pumps: an idealized centrifugal pump (ICP), a theoretical centrifugal pump (TCP), and a real centrifugal pump (RCP) [6]. The ICP is a singlestage and single-entry pump with the infinite number of infinitely thin vanes for pumping an ideal fluid with no power loss. The theoretical centrifugal pump is similar to idealized centrifugal pump, but it contains a wheel with the finite number of vanes of a certain thickness, and it has no volume, hydraulic and mechanical loss. The RCP is a real analog of the TCP with power loss, which pumps a heterogeneous fluid. We have developed the theory of simulation of an idealized hydraulic machine with the help of the method of electrohydraulic analog and the principal notions of the unified circuit theory. To this purpose we have used a modified pressure balance equation for the idealized centrifugal pump with the set-up geometriсs at the constant frequency of the wheel rotation (n = const). , (1) where , are the current values of thrust and fluid delivery rate at the idealized centrifugal = (XX) is the thrust of the idealized centrifugal pump in the mode pump output, respectively, of no-load operation (closed output gate valve), which is similar to the electro-moving force in the direct current electric circuit H =



D

D

= const,

(2) 102

Rt is the internal hydraulic resistance of the ICP, which is a constant value and does not depend on the pump mode, and which can be obtained as follows: R =



= const.

(3)

Here D2, D1 are the external and internal diameters, respectively; b2, b1 are the output and input width of the vane; β2V, β1V are the output and input vane angles of the operating wheel of the idealized centrifugal pump; ρ, g are the density of the pump fluid and the acceleration of gravity. The equivalent circuit diagram corresponds to the modified equation (1), where Rphs is the hydraulic resistance of the pressure piping of the hydraulic system =pg / . (4) Use of the unified circuit theory for description of centrifugal pump units opens new aspects of their simulation and allows us to set new electrohydraulic analogs, which exist between electric machines and centrifugal pumps. Such analogs are based on similarity of design of these machines, which have a fixed (stator) and rotating (rotor) parts. Both types of machines have the input and the output of the energy carrier, and energy increment at the output is provided owing to mechanical rotation energy supplied through the pump shaft. Besides, both of the machines can change the direction of the energy carrier, i.e. can operate in the generator (pump) mode or the engine (turbine) mode. Using analogy of intervals operating modes of a centrifugal pump, we suggest to use the theoretic coefficients of thrust γH∞, the pump delivery rate γQ∞, the power γN∞ and the resistance γR∞ of the ideal centrifugal pump, reduced (normalized) in the range [0,1]

;

;



;



(5)

1.

An alternative approach to solution of the problem of real equipment analysis by means of neural network methods is the one, concerning development of a working research model of the BCS complex [7] for collection of log data in conditions, which provide precise accounting of the working fluid components. This entails considerable organizational complications and requires application of expensive measuring equipment which cannot be used in industry. So, owing to the developed neural network mathematical model which reproduces functional dependence of the operability criteria from the input parameters in the operating range of their variation, it is possible to increase quality of experimental data analysis, and to detect defects and problems of operation in time [8]. Conclusion Creation of a system of computer-aided analysis of pump equipment functioning is an actual problem of design and development of nodes and units for increasing of estimation reliability of products operation, for real-time diagnostics of operation, and for support of design decisions for elimination of faults. Owing to wide approximation abilities of a neural network computational architecture it is possible to create multi-parametric simulation mathematical models of complex technical systems intended for identification of operation parameters and criteria of operability, and on their base it is possible to develop an optimization strategy of pump equipment control, upgrade of diagnostic and emergency protection systems. Acknowledgement This paper is published due the financial support of the Russian Science Foundation (RSF) via the grant No 15-19- 00196.

103

References [1] Iakov S. Korovin, Igor A. Kalyaev, Modern Decision Support Systems in Oil Industry: Types, Approaches and Applications, International Conference on Test, Measurement and Computational Methods (TMCM) 2015, Chiang Mai, Thailand; Advances in Computer Science Research. ISBN 978-94-6252-132-2. ISSN 2352-538x, pp. 141-144. [2] Iakov S. Korovin, The Importance of New Approaches Development and their Implementation in the Oil and Gas Industry in Russian Federation―the Current Situation Analysis, International Conference on Advances in Energy, Environment and Chemical Engineering (AEECE) 2015, Changsha, Peoples R China; Advances in Engineering Research. ISBN: 978-94-6252-109-4. ISSN: 2352-5401. Atlantis Press, Volume 23, 2015, p. 94-97. [3] Iakov S. Korovin, Anatoly I. Kalyaev, Maksim V. Khisamutdinov, Data Mining Methods Application to the Problem of Handling Corporative Dataset on Heavy Oil Production, International Conference on Intelligent Control and Computer Application (ICCA), Zhengzhou, PEOPLES R China: Jan 16-17, 2016 Proceedings of the 2016 international conference on intelligent control and computer application, ACSR-Advances in Computer Science Research Volume 30, pp. 387-389. [4] I.S. Korovin, M.V. Khisamutdinov, G. Schaefer, A. Kalyaev, Real-time diagnostics of oil production equipment using data mining, 2016 5th International Conference on Informatics, Electronics and Vision (ICIEV), 2016, pp. 1196 – 1172, DOI: 10.1109/ICIEV.2016. 7760184. [5] I.S. Korovin, M.V. Khisamutdinov, G. Schaefer, A. Kalyaev, Application of hybrid data mining methods to increase profitability of heavy oil production, 2016 5th International Conference on Informatics, Electronics and Vision (ICIEV), 2016, pp. 1149-1152, DOI: 10.1109/ICIEV.2016.7760179. [6] V.S. Kostyshin. Simulation of operation modes of centrifugal pumps on the base of the electrohydraulic analog. – Ivano-Frankovsk, 2004. – pp. 42. [7] R.A. Badamshin, A.R. Taneyev, K.F. Tagirova. A refined mathematical model for real-time control of oil production technological process // Mechatronics, automation, control. [8] Booster pump stations [On-line resource]. Information on: http://petrolibrary.ru/ustanovkapredvaritelnogo-sbrosa-vodyi-upsv-dozhimnaya-nasosnaya-stancziya-dns.html.

104