Research On Oilfield Intelligent Fault Diagnosis System Based On Optimized RBF Neural Network

With the construction and development of digital oil fields,the indicator diagram data has entered the era of big data.The indicator diagram diagnostic method using neural network as the core algorithm has become a research hotspot and development trend in the field of pumping unit fault diagnosis.RBF neural network is one of the most widely used neural networks,but the limitations of its kernel function parameters have a great impact on the performance of the algorithm.Starting from optimizing the kernel function parameters of RBF neural network,this paper establishes PSO-RBF neural network and improves the performance of RBF neural network.A new oilfield intelligent fault diagnosis system is developed using PSO-RBF neural network as the core algorithm for fault diagnosis of pumping system.Firstly,based on the principle of matrix graying,this paper designs and writes a set of grayscale program of indicator diagram,and performs grayscale processing on various types of indicator diagrams,and extracts the characteristics of gray matrix of indicator diagram.A indicator diagram gray matrix characteristic information database is established as a source of fault diagnosis sample data.Aiming at the limitations of RBF neural network in kernel function parameter selection,an optimization strategy for its kernel function parameters is proposed.After comparing the optimization methods,the K-Means algorithm is selected to optimize the center value of the RBF network kernel function,and the improved The PSO algorithm replaces the gradient descent method,optimizes the weight and width of the RBF network kernel function,and constructs the POS-RBF neural network.The data in the indicator diagram feature information database and the Winequality data in the UCI database is used to compare the performance of the improved algorithm.The comparison results show that the optimized PSO-RBF neural network is superior to the traditional RBF neural network in training error and fitting error.Finally,based on the established indicator diagram information database,the traditional indicator diagram diagnosis method and PSO-RBF neural network fusion algorithm are proposed,and a new oilfield intelligent fault diagnosis system is developed to realize the indicator diagram.Accurate diagnosis of the type of fault.