CAJ | 학술논문

油井功图是油井工况诊断的重要依据，快速准确地识别油井功图对于提高油田作业效率具有重要意义。传统的人工相面法识别示功图无法实现油井工况的实时在线诊断，而BP神经网络法识别准确率较低，因此提出一种基于概率神经网络的油井功图识别方法。该方法通过提取功图数据的面积特征、特征向量和载荷曲线的傅里叶逼近特征作为油井功图的特征值，PNN网络用特征值作为输入对油井工况进行诊断。实验结果表明与BP网络相比使用PNN网络根据功图提取特征进行油井功图识别时能够达到更高的识别效率。
유정공도시유정공황진단적중요의거，쾌속준학지식별유정공도대우제고유전작업효솔구유중요의의。전통적인공상면법식별시공도무법실현유정공황적실시재선진단，이BP신경망락법식별준학솔교저，인차제출일충기우개솔신경망락적유정공도식별방법。해방법통과제취공도수거적면적특정、특정향량화재하곡선적부리협핍근특정작위유정공도적특정치，PNN망락용특정치작위수입대유정공황진행진단。실험결과표명여BP망락상비사용PNN망락근거공도제취특정진행유정공도식별시능구체도경고적식별효솔。
The pumping diagram is an important basis of pumping condition diagnosis and it is of great significance to identify the pumping diagram rapidly and accurately for improving the production efficiency. Traditional manual identification of pumping diagram can't realize real-time online pumping fault diagnosis and the accuracy of BP network is too low, therefore proposed a method of pumping diagram recognition based on probabilistic neural network (PNN). The method first extracted eigenvalues of area feature eigenvectors of pumping diagram and flourier approximation eigenvalues of load curve as features for pumping diagram identification. PNN used the features as input for pumping condition diagnosis. The experimental result shows that the method of using PNN to identify pumping diagram based on diagram features can achieve better performance than BP network.