AbstractAt present, oil companies are committed to applying the theory and means of mathematics or data science to the research of oilfield data rules. However, for some old oil wells, aging equipment, complex environment and backward management, cause the authenticity and accuracy of the data collected by the equipment cannot be determined. According to the actual engineering demand of the old wells, this paper proposes a method based on principal component analysis, cluster analysis and regression analysis to mine and analyze the data of polished rod load of old oil wells, so as to judge the working conditions of the oil wells. Combined with the application of this study in several operation areas of some oilfields, the findings of this study can help for better understanding of the working condition information hidden in "big data" of oilfield. Meanwhile, the PCA method can reduce the complexity of the original data, the regression equation can calculate the size of the polished rod load more accurately, and the prediction model can effectively judge the working conditions of the old oil wells on site.