超深井泵深影响因素的神精网络分析模型

A neural network model for influencing factors of well depth in ultra-deep well

邢庆河 张士诚 赵晓

XING Qinghe ZHANG Shicheng ZHAO Xiao

中国石油大学石油工程教育部重点实验室，北京　102249 中石化研究院海外研究中心，北京　100083

Key Laboratory for Petroleum Engineering of the Ministry of Education, China University of Petroleum, Beijing 102249, China Sinopec Exploration and Production Research Institute Overseas Research Center, Beijing 100083, China

以塔河油田奥陶系碳酸盐岩储层为代表的西部油藏具有超深、高温、高压、缝洞发育等复杂的地质特征，伴随地质储量的不断扩大和动用，深井、超深井数量正迅速增长，因地层能量递减快、补充能量困难导致泵挂深度不断增大的问题日益突出。应用神?网络理论建立了有杆泵泵深影响因素分析模型，可得到不同影响因素的权重排序，从而确定出关键因素。塔河油田的应用实例表明,该模型所得到的预测结果具有较高精度，可有针对性地指导举升工艺设计，最大限度地提高下泵深度，以保证超深低能油井的正常生产。

Carbonate rock reservoirs in the Ordovician of Tahe oilfield representing western reservoirs are characterized by ultra-deep, high temperature, high pressure and fracture-cavity system. The number of deep or ultra-deep wells is increasing obviously with the deep exploitation of geological reserves. Formation energy decreased dramatically but hard to compensate, which led to the issue of continuously deepening setting depth of pump. This paper presents the analytical model of sucker rod pump depth based on the neural network, which can obtain the weight sequence of the influencing factors and determine the key element. Application in Tahe oilfield indicates that the prediction results have a high accuracy and the model can guide the design of artificial lift technology, deepen the pump depth utmost to guarantee the normal production of the ultra-deep well with low energy.