混沌粒子群优化的RBF神经网络在热油管道仿真中的应用

Hot Oil Pipeline Simulation Based on Chaotic Particle Swarm Optimized RBF Neural Network

针对加热输油管道仿真中管道内油流温度和压力变化,本文基于神经网络算法,利用管道数据采集与监控(SCADA)系统获取的500组历史运行数据,建立了管道沿线温度和压力的预测模型.提出的预测模型采用混沌粒子群改进的RBF(CPSO-RBF)神经网络算法.对RBF神经网络的参数(中心和宽度)、连接权重进行优化,通过与其他方法对比可知提出的CPSO-RBF预测模型具有精度高、收敛快等特点.在日照-仪征热油管道实际运行方案中验证了提出的CPSO-RBF预测模型的可行性.

Regarding the variation of temperature and pressure in hot oil pipeline simulation, a forecast model of temperature and pressure along the pipeline based on neural network is established using 500 sets of historical operation data from supervisory control and data acquisition （SCADA） system. The fore- cast model is proposed based on chaotic particle swarm optimized radial basis function （CPSO-RBF） neu- ral network algorithm in this paper, through which parameters （the center and the width） and connection weight are optimized. Compared with other algorithms, CPSO-RBF based forecast model shows higher precision and faster convergence speed. Its application in Rizhao-Yizheng hot oil pipeline verifies the fea- sibility of CPSO-RBF based forecast model.