基于平行系统的游梁式抽油机采油系统建模与控制方法研究

In worldwide oilfield production, the beam pumping unit plays a very important role. It is formed by two main parts: ground and underground, so it is a complicated industrial system. Improving its continuous production capacities, decreasing its production costs and using advanced automatic control methods and technologies are development tendency of the future oilfield production. However, there are many problems at present, such as: difficult to build the whole system model, easily affected by non-project factors and difficult to find better control strategies, and so on. To solve these, our research uses the conception of the parallel system to build system framework, and uses the hybrid modeling method combined with mechanism model and data model to solve problems that many down-hole production parameters in mechanism model are difficult to be set exactly and many model parameters under different reservoir conditions are different, and so on; for the influences of the non-project factors, their function modeling, behavior description modes and computational experiments are studied; the control methods based on reliability and stability analysis are studied to structure effective feedback control strategies, and the real-time evaluation methods of the system are studied also. According to our research, it has important theoretical significance to improve the modeling and control theories for complicated industrial system and enrich the control methods of it; it also has important practical application value for development of control method and technology of completed and integrated automation system in oil production.

游梁式抽油机采油系统在国内外油田生产中占有重要的作用，由地上和地下两部分组成，是一类较为复杂的工业系统。提高系统的连续生产能力，降低生产成本和能耗，采用先进的自动控制方法和技术是未来生产方式的发展趋势。针对目前生产中存在的难以建立系统的整体模型、易受到非工程因素的影响以及较难建立有效的控制策略等问题。本课题采用平行系统的概念搭建系统框架，采用机理模型和数据模型相结合的混合建模方法，解决传统的机理建模中所存在的井下生产参数不易准确设定和不同油藏条件下模型参数存在差异等问题；对系统中非工程因素的影响，研究其功能模型建立、行为描述和计算实验方法等；研究基于可靠性和稳定性分析的系统控制方法，构建有效的反馈控制策略，并对系统进行实时评价。通过本课题的成功研究，将对完善复杂工业系统建模与控制理论，丰富其控制方法有重要的理论意义，对石油生产综合自动化系统控制方法和技术的发展具有重要的实际应用价值。

游梁式抽油机采油系统是油田的主要生产方式，由于井下生产环境的动态性和不确定性，现有的机理建模方法存在较大误差，且难以进行整体建模。本项目在平行系统理论架构下，采用计算实验、平行执行等相关方法解决游梁式抽油机采油系统建模与控制的若干问题。具体工作如下：针对难以掌握电动机驱动力矩与负载变化之间准确关系的问题，提出一种基于串行结构的混合模型；研究负载动态变化下电动机负载扭矩的在线建模新方法，提出基于OS-ELM的在线混合模型；针对井下工况对系统生产效率的影响，提出一种基于IFOA-KELM-MEA的井下工况的短期预测模型；提出一种基于8方向链码和OS-ELM的游梁式抽油机采油系统井下工况的新的预测模型；针对油井油液含水率参数，提出一种基于自动谱聚类与多极端学习机的油井油液含水率软测量模型；提出了基于IBH-LSSVM的时间序列预测模型，并将其用于抽油井动液面的短期预测；针对抽油井的间歇生产方式，提出一种基于平行系统的间歇时间预测模型，采用计算实验、平行执行的反馈控制等方法解决相关问题。进一步，本项目还完成了以下研究工作：提出了一种基于无监督学习的自动聚类方法，将其应用于抽油井井下故障诊断中；为了进一步提高示功图特征提取的准确度，提出了一种基于16方向链码和K均值聚类的方法；提出一种基于权重优化的D-S证据融合方法，将其用于井下故障诊断中。本项目的研究工作，既解决了游梁式抽油机采油系统建模与控制的一些特性问题，又对相关领域中的一些共性问题提出了解决方法，具有重要的理论意义和应用价值。