基于反距离加权法的DTW-RNN神经网络吸水剖面预测模型

Water Absorption Profile Prediction Model with DTW-RNN Neural Network Based on Inverse Distance Weighting Method

基于同位素吸水剖面法和分层注水量计算法在应用中存在这样或那样的不足,创新了一种油水井多层合注合采液量劈分技术方法。该方法是在实测吸水剖面资料的基础上,选取油层有效厚度、有效渗透率、措施系数、原油粘度等共8个因素作为建立预测模型的主要影响因素,以注入井的注水数据为输入值,通过反距离加权法以及时间规整算法对输入值进行距离计算,得到相关性较大的输入,最终通过循环神经网络进行预测建模。应用结果表明,该模型得到的各小层累积吸水量比传统KH计算值更接近实测吸剖结果,说明其具有一定的有效性。

Based on some deficiencies in the application of the isotope water absorption profile method and stratified water injection calculation method,a new method of splitting the amount of multi-layer combined injection and production of oil and water wells was invented The method is established on the measured data of injection profile to select the reservoir effective thickness,effective permeability,stimulation coefficient,crude oil of crude oil,a total of eight factors as the main influence factors of the prediction model;the data of water injection in an injection well are used as input values,and time to distance calculation of the input values is conducted to obtain the larger correlated inputs through the inverse distance weighting meth-od and the dynamic time warping algorithm(DTW);and finally,the prediction model is built by the recurrent neural network.The field results show that the cumulative water absorption of each layer obtained by the model is closer to the measured results than those calculated by the conventional KH,which indicates that the model has certain effectiveness.