A novel prediction method for down-hole working conditions of the beam pumping unit based on 8-directions chain codes and online sequential extreme learning machine

In the oilfield operation, the beam pumping unit is a very important artificial lift method. As the down-hole parts work at hundred and thousand meters underground, they are hard to be found immediately when failures come out. If we can predict down-hole working conditions and master its continuous operation states in time, great improvement of the oil well production will be developed. In this paper, a novel down-hole working conditions prediction method for the beam pumping unit based on the chaos time series prediction is proposed. First, curve contour of the dynamometer card is redrawn by 8-directions chain codes, and then eight feature vectors are extracted to construct eight feature vector time series; then, the online sequential extreme learning machine (OS-ELM) method is used to build the prediction model, which can realize fast updating with dynamic work condition changes; finally, the grey interval relational degree between the predicted feature vectors and feature vectors of each fault type is calculated to determine the predicted fault type. Actual production data of an oil well are used for example verification, and both online diagnosis and offline diagnosis illustrate the effectiveness of the method.