基于强化学习的煤层气井螺杆泵排采参数智能决策

为了实现煤层气井螺杆泵排采参数的连续决策和连续控制,使煤层气井长期高效稳产,以煤层气井螺杆泵生产周期内最大累积产气量为优化目标,提出了一种具有动作自寻优能力的螺杆泵排采强化模型的框架和Q学习及Sarsa、Sarsa(lambda)算法。研究通过与环境的交互式学习,对动态环境进行灵活奖惩,实现智能体在复杂环境下智能决策和参数优化,可有效获取煤层气螺杆泵排采最优协调控制,从而解决传统方法不能根据环境变化迅速做出调整而降低排采效果的问题。实验分析表明,给定煤层气井产气量的变化曲线,以螺杆泵的频率为单一控制变量,应用Q学习方法能有效得到螺杆泵排采变频控制的最优策略,具有一定的应用潜力。     

Simulation of liquid meniscus formation in the ionic liquid electrospray process

Though tremendous efforts have been made to investigate electrospray, some aspects, such as the evolution of the menisci on the micropores of porous emitter tips and the transient response of the meniscus during the polarity alternation, need to be further understood. This paper presents a computation fluid dynamics (CFD) model to describe the meniscus formation in the ionic liquid electrospray process. The CFD model, based on the Taylor–Melcher leaky dielectric fluid theory and the volume of fluid (VOF) method, is validated by experiments. The evolution of the meniscus on the basis of a micropore is presented using two typical ionic liquids, EMI-BF4 and EMI-Im. The influences of the pore size, flow rate and applied voltage on the formation of the meniscus have been studied. Results show that a larger pore is more likely to start emission, and the time consumed for liquid meniscus formation decreases with increasing applied voltage and flow rate. Further, it is found that alternation of polarity does not destroy the structure of the meniscus but retards the formation process, and a faster polarity alteration leads to a shorter delay in meniscus formation time.