基于SDCQGA优化BP神经网络的岩石可钻性建模

针对智能钻井优化控制过程中岩石可钻性提取存在的建模难、非实时性、精度差等问题,提出基于自适应双链量子遗传算法优化BP神经网络结构的岩石可钻性提取建模方法.依据目标函数在搜索点处的变化率,建立了快速自适应双链量子遗传算法;采用新算法优化BP神经网络结构,以克服BP神经网络受初始权值/阀值影响和泛化能力差的问题.通过对邻近钻井区域的大量测量数据和实验数据的统计分析和预处理,建立岩石可钻性提取模型,有效地解决了复杂地形岩石可钻性提取难的问题.对不同岩性的可钻性参数提取实验结果证明,该建模方法不仅提高了参数提取的精度和模型的泛化能力,而且在相邻实际参数提取时,具有很好的实时性和适应性.

Modeling of rock drillability with BP neural network optimized by SDCQGA

In the control process of intelligent drilling,there are some difficulties in the rock drillability modeling,such as poor real-time,low accuracy of rock drillability extraction,etc.For this reason,a modeling method for the rock drillability extraction is put forward,which is based on the BP neural network optimized by SDCQGA(Self-Adaptive Double Chain Quantum Genetic Algorithm).A fast self-adaptive double chain quantum genetic algorithm is established according to the variation rate of objective function at search point,and then the structure of BP neural network is optimized using this algorithm in order to overcome the shortcomings of easily being influenced by initial weights and poor generalization ability of BP neural network.The model for rock drillability extraction was established according to statistical analysis and preprocessing and analysis shortage of over-fitting,random and of back-propagation neural network which can affect the generalization ability with the subtle changes of the parameters of the network,this paper presents rock drillability extraction modeling methods using an optimization the BP neural network structure which is based on the adaptive double chain quantum genetic algorithm.Finally,the rock drillability extraction model is constructed by using a large number of measurement while drilling data in different drilling areas.The model can be effectively solved difficult extraction of rock drillability in the complex formation.The tests of extraction rock drillability of different lithology prove that this modeling approach not only improves the accuracy of parameter extraction and generalization ability,but also has a good real-time and suitability in the actual rock drillability extraction.