基于改进差分进化算法的神经网络管道泄漏检测

Pipeline Leak Detection Based on Improved Differential Evolution Algorithm and Fuzzy Neural Network

提出了一种基于神经网络与差分进化算法的天然气泄漏预测方法,该方法采用RBF神经网络作为泄漏预测模型,引入改进的差分算法对网络的初始连接权值进行优化。为了在全局搜索和局部搜索之间取得最佳平衡,提出了一种自适应变异因子和交叉概率的改进算法,并将其应用于泄漏预测神经网络模型优化。将所提出的方法与原始算法的前向网络预测方法进行了比较。结果表明:所提出的方法收敛速度快,所得模型的预测误差小、准确率较高、迭代次数少、泛化能力强,对天然气的泄漏预测有很好的参考作用。

A natural gas leakage prediction method based on neural network and differential evolution algo- rithm was proposed, which uses RBF neural network as leakage prediction model and employs the improved differential algorithm to optimize initial connection weights of the network ; meanwhile, in order to get the opti- mum balance between global search and local search, an improved algorithm based on adaptive mutation factor and crossover probability was presented to optimize leakage prediction neural network model. Having it compared with original algorithm＇s forward network prediction method shows that the proposed optimization algorithm has fast convergence speed, low prediction error, less iteration and strong generalization capability. It provides good reference for the prediction of natural gas leakage prediction.