Indicator diagram recognition method based on regularized attention convolutional neural network

Abstract

The invention relates to an indicator diagram identification method based on a regularized attention convolutional neural network, which comprises the following steps of: 1, establishing a data preprocessing module, and carrying out dimension and grey-scale map processing on a working condition sample set of an oil pumping unit; 2, establishing a regularized attention convolution module, and reinforcing, inhibiting and inactivating the autonomously learned convolution features; 3, embedding the regularized attention convolution module into the convolutional neural network to form a regularizedattention convolutional neural network; 4, establishing an indicator diagram recognition module, and inputting the gray level image of the indicator diagram into the regularized attention convolutional neural network for recognition; 5, establishing an attention loss function, and training a regularized attention convolutional neural network model; 6, inputting the oil pumping unit working condition data collected in real time into the indicator diagram recognition model, and repeating the steps 2-4; 7, taking the indicator diagram identification method based on the RA-CNN as a core, and constructing an intelligent diagnosis system of the working condition of the oil pumping unit. The identification precision of the indicator diagram can be effectively improved.

Description

[0001] 1. Technical Field: [0002] The invention relates to an intelligent diagnosis system for pumping unit working conditions, and specifically relates to an indicator diagram recognition method based on a regularized attention convolution neural network. [0003] 2. Background technology: [0004] In the oil production process of the oil field, the indicator diagram is a closed continuous curve composed of the load and displacement of the reciprocating movement of the donkey head of the pumping unit. It contains rich device status information, which can reflect the gas, oil, water and sand in real time. The influence of factors in wells such as, wax, etc. on the working condition of the pumping unit, so the analysis and identification of the indicator diagram is an important means of diagnosis of the working condition of the pumping unit. Due to the large number of pumping units and the frequent collection of displacement and load data, manual analysis methods are difficult to mo...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide the dynamometer recognition method based on regularized attention convolutional neural network

Images