Hybrid nonlinear homotopy BP neural network and genetic algorithm for oil-pumping control system

In this paper, hybrid nonlinear homotopy BP neural network and genetic is applied to intermittent oil-pumping control to solve under loading and empty pumping problems, which are associated with pumping unit of oil wells and cause waste of energy and inefficient usage of equipment. Furthermore, a quick nonlinear homotopy is proposed to greatly improve the convergence speed of conventional BP and overcome its drawback of getting stuck at local minima. And also, an improved GA with enforced mutation and crossover and mutation adaptive parameters is developed to prevent premature and realize global-optimization effectively. The fundamental idea is to identify the pumping model through nonlinear homotopy BP neural network with a nonlinear normalization method, and optimize the downtime through GA. The proposed algorithm is validated with computer simulations to provide guidance for actual oil-pumping application.