Reservoir is underground with strong uncertainty, so the optimization of reservoir production is a complicated systematic engineering. It is necessary to form an integrated design process through production dynamic analysis, numerical simulation and historical fitting, production program optimization and other links to ensure the effect of well pattern adjustment, layer subdivision, injection and production optimization, and optimization of measures. Traditional methods mainly rely on the experience of researchers, using reservoir engineering qualitative analysis, reservoir numerical simulation quantitative decision to complete the optimization of the scheme. The research group integrates machine learning, intelligent optimization, reservoir numerical simulation and reservoir engineering, aiming at the key problems of integrity, uncertainty, strong nonlinear and efficient calculation in the design of oil and gas field development schemes. Through theoretical and technological innovation, Effectively improve the efficiency of the whole process of reservoir production "dynamic analysis of reservoir production, automatic history fitting of numerical simulation, optimization of production plan and risk assessment of program implementation", and form a set of new methods of reservoir production optimization of intelligent oil fields under the cloud computing framework