基于BAS-BP的钻井机械钻速预测模型

Prediction Model of Mechanical ROP during Drilling Based on BAS-BP

针对现有的机械钻速预测理论模型缺少对工程实际数据的应用而难以满足现场需求的问题,建立一种以人工智能算法与神经网络相结合的机械钻速预测新模型。首先,利用小波滤波方法对钻井现场实测数据进行降噪处理,并依据互信息关联分析优选机械钻速预测模型的输入参数,降低模型冗余。其次,利用天牛须搜索(Beetle Antennae Search,BAS)实现BP神经网络初始权值、阈值的优化,以此建立机械钻速预测新模型。最后,将建立的BAS-BP新模型与标准BP、PSO-BP及GA-BP等3种模型进行对比分析。结果表明:BAS-BP机械钻速模型预测值与实际测量值误差最小,同时具有良好的收敛性和搜索能力。

The existing theoretical models for mechanical ROP prediction lack the application of practical engineering data and are difficult to meet the field demand.Therefore,a new prediction model of mechanical ROP based on artificial intelligence algorithm and neural network is established.Firstly,the drilling field measured data is denoised using wavelet filtering method,and the input parameters of the mechanical ROP prediction model are optimized according to the mutual information correlation analysis to reduce the redundancy of the model.Secondly,the initial weight and threshold of BP neural network are optimized by using Beetle Antennae Search(BAS),so as to establish a new model for mechanical ROP prediction.Finally,the new BAS-BP model is compared with standard BP,PSO-BP and GA-BP models.The result shows that the error between the predicted values of mechanical ROP by BAS-BP model and the actual measured values of mechanical ROP is the smallest,and its algorithm has good convergence and search ability.