时序化生产预警有效影响因子的获取方法研究

Research on Acquisition Method of Effective Impact Factors in Production Early Warning by Time Series

在生产预警有效影响因子的筛选过程中,为了达到降低维度,增强影响因子集的有效性,从而提高生产异常预警准确率的目的,选取和分析所有原始项目,应用模糊综合评价法量化模糊限制语,采用TRIMMEAN内均法排除极端评估值。运用特征选择技术发现敏感特征因子,借鉴混合智能方法定义影响因子集的逻辑表达结构,基于粒度分析处理时序化数据,同时利用激剧判定函数摒弃无效元素完成对数据的降维以及筛选,得到高精细化的有效影响因子集。以此达到辅助深度挖掘数据内部潜在规律,解决信息杂乱等现象,运用于生产异常分析,提高预警准确率的目的。最后针对大庆油田某采油厂生产历史数据,完成时序化生产预警有效影响因子的获取。

In the screening process of effective impact factors for early warning,in order to reduce dimension,enhance the effectiveness of the impact factor set and improve accuracy in early warning of abnormal production,all original items are selected and analyzed,and fuzzy constraints is quantified based on fuzzy comprehensive evaluation method,using TRIMMEAN to eliminate extreme values. Then,sensitive feature factors are determined by using feature selection techniques. At the same time,the logical expression structure of the influence factor set is defined via the hybrid intelligent method,and the time sequence data is manipulated based on granularity analysis. Next,it finishes dimensionality reduction and selection of data through the dramatic function to achieve the effective impact factors of high precision. To reach the purpose that excavates potential lawin the data deeply,and solves the phenomenon of information clutter,using the method in the production of abnormal analysis to improve accuracy in early warning. Finally,in combination with the history data of an oil production plant in Daqing Oilfield,the effective impact factors acquisition of the production early warning by time series is achieved.