基于遗传算法优化的神经网络在长输管道泄漏检测系统中的应用

针对由于输油管道泄漏在生产生活中造成的诸多损失与危害,以吉林油田的一段长输管道为研究对象,利用小波包变换提取管道泄漏压力信号的特征向量,将得到的特征向量作为神经网络的输入,根据输出对管道的运行状态进行识别,再应用负压波定位法对泄露点进行定位。在此基础上,提出了一种基于遗传算法优化的RBF神经网络输油管道检测的方法。该方法把遗传算法应用于神经网络的参数确定中。实验结果表明,优化的RBF神经网络模型的误差为1%左右,提高了泄漏检测的精度与效率。

Research on the System of Pipeline Leak Location Based on RBF Network

For many losses and harm caused in production from the pipeline leak.In this paper, taking a long-distance pipeline of Jilin Oilfield as the departure point.Firstly,extract pressure pipeline leakage eigenvectors with wavelet packet as the neural network input,using output to recognize pipeline work station.Then point the leak orientating use the neg-ative pressure wave theory.On this basis, a new method of pipeline detecting based on optimized RBF neural networks using genetic algorithm was improved.Genetic algorithm was applied to optimize position of data centers.The result shows that the error rate of optimized RBF neural network model is less about 1%, so the accuracy and efficiency of leak detection had been improved.