Application of dynamic liquid level prediction model based on improved SVR in sucker rod pump oil wells

This study built a dynamic liquid level prediction model based on improved SVR (Support Victor Regression) in sucker rod pump wells. This modeling method adopted sliding window to limit the number of samples and applied genetic algorithm to realize automatic optimization of C ande which are parameters of SVR. Through the simulation experiment, we verified the effectiveness of the modeling method and improved the precision of the model. After a period of actual operating in some oilfield, good results were obtained and the precision could perfectly meet the requirement of oil production.