油田产量多变量预测模型的优化

油田开发是一个复杂的多变量非线性动力学系统，为有效地预测油田产量，确保油田生产过程高产稳产，该文提出采用多元线性回归与神经网络相结合的方法对油田产量多变量预测模型进行优化。首先基于回归分析的“后退法”对影响产量的变量进行优选，然后通过神经网络对优选后的变量进行训练得到最终的预测模型，从而实现神经网络与多元线性回归相结合建立多变量预测模型。实际应用结果表明，优化后的模型简洁实用，可以在一定程度上提高模型的预测精度，并减少建模预测所需数据量。

Optimization of Multivariate Model in Oilfield Output Prediction

Oilfield exploitation is a complicated multivariate non-linear dynamic system.To predict oilfield output effectively and insure high yield and steady yield of oilfield,this paper optimizes a multivariate oilfield output prediction model by using multivariate linear regression and artificial neural network(ANN).At first,the prominent factors from many factors which effect variable output by the backpedal method of multivariate linear regression are optimized,and then the ultimate predicting model is built up on the base of training these prominent factors by ANN,and the model is got integrated with multivariate linear regression and ANN.The application results show that the optimized model is more simple and useful,and it can make the prediction precise with less sample data.