基于代理辅助野草猴群算法的井位优快决策方法

Optimal and fast decision making method for well location based on surrogate-assisted weed monkey algorithm(WMA)

作为油藏开发管理的核心步骤,合理的井位布置对提高油藏开发效率至关重要。基于野草猴群算法,结合径向基函数神经网络模型提出了一种井位优快智能决策方法,充分发挥径向基函数神经网络模型在小样本下的高效学习效率和野草猴群算法兼顾局部和全局的搜索能力,以最大净现值为目标,采用径向基函数神经网络模型作为代理辅助,利用野草猴群算法对合理的井位部署进行搜索。以SPE10油藏模型上布置一注四采5口井和某复杂断块油藏布置八注十六采24口井为实例对该方法进行应用,相较于标准方案,优化后的布井方案分别提高了17.11%和18.01%的经济效益,且与井位优化常用的粒子群算法和差分进化算法相比,基于代理辅助野草猴群算法的井位决策方法能在相同的迭代次数下得到更优的解。研究成果为油藏井位的优快决策提供了一种高效合理的新思路。

As a core step of reservoir development management,rational well deployment is essential to improve efficiency of reservoir development. Based on weed monkey algorithm(WMA)and radial basis function neural network(RBFNN)model,an optimal and fast well location decision making method is proposed. By making full use of high learning efficiency of RBFNN model in small samples and combined local-global search capability of WMA,taking maximum NPV as the objective,RBFNN model is used as surrogate assistant and WMA is used to search rational well deployment. The method is applied by taking 5 wells including 1 injector and 4 producers in SPE10 reservoir model and 24 wells including 8 injectors and 16 producers in a complex fault-block reservoir as examples.Compared with standard scheme,economic benefits of optimized well deployment schemes are increased by 17.11% and 18.01% respectively. And compared with particle swarm optimization algorithm and differential evolution algorithm commonly used in well location optimization,the method based on surrougate-assisted weed monkey algorithm(WMA)can obtain better solution with the same number of iterations. The research achievements provide a high-efficiency,rational and new idea for optimal and fast decision making of reservoir well location.