基于柔性神经树的智能变电站过程层网络流量预测

Process level network traffic prediction for smart substations based on flexible neural tree

从智能变电站过程层网络流量监测系统的需求出发,提出了一种基于柔性神经树模型的智能变电站过程层网络流量预测解决方案.柔性神经树模型利用遗传规划开发与优化,其参数由粒子群优化算法做优化.首先初始化随机的柔性神经树结构和相应的参数,不断地学习优化柔性神经树结构,完成树结构的优化之后,再对参数进行优化.最终通过反复循环获得满足要求的柔性神经树.实验结果表明,柔性神经树模型能有效地预测过程层网络流量,再现真实的流量测量的统计特性;其预测精度较高,误差集中在0附近.预测结果可以有效地监控网络流量状况,保障电网安全运行.

From the demand of process layer network traffic monitoring system for smart substation,this paper proposes a process level network traffic prediction solution based on flexible neural tree model for smart substation.Flexible neural tree model is developed and optimized by using genetic programming;and its parameters are optimized by particle swarm optimization.Firstly,the random flexible neural tree structure and its corresponding parameters are initialized;and then the flexible neural tree structure is optimized;continually and its parameter afterwards.Finally,after repetitive cycle we can get the flexible neural tree which meets the requirements.The experimental results show that the proposed method can effectively predict the process level network traffic and reproduce the statistical properties of the real traffic measurement.It has high prediction accuracy whose deviation is nearly zero.The prediction results effectively monitor network traffic,so as to ensure the safe operation of power grids.