基于改进型BP神经网络的油井产量预测研究

Research on Single Well Production Prediction Based on Improved BP Neural Networks

为了保证油田生产持续稳定地发展,针对油田单井产量提出了基于改进型BP神经网络的预测模型。对传统的BP神经网络的结构和训练算法进行了研究,发现它存在易于陷入局部极小,收敛速度慢等问题。提出了使用LM算法的改进型BP神经网络。最后给出了基于改进型BP神经网络的单井产量预测模型仿真实验。结果证明该算法的实用性和可行性,在油井产量预测方面有一定的实用价值。

In order to ensure oil field production for sustainable development, aim at oil field production, the predictive model based on improved BP neural networks is put forward. The structure of traditional BP neural networks and its training algorithm are studied, some weak points of it are found, such as easy to fall into local minimum value and slow convergence. The BP neural networks improved by L-M algorithm is put forward. At the end, simulation experiment of the predictive model based on improved BP neural networks is illustrated. The result of which proves the practicability and the feasibility of this algorithm and high use value in the oil well production prediction.