输电线路导线覆冰AMPSO-BP神经网络预测模型

输电线路覆冰严重危害电网安全运行,因此,有必要开展线路覆冰预测研究。随着人工智能技术的不断发展,其在电网覆冰监测中的优势逐渐凸显。现有的基于覆冰增长物理模型和统计回归模型覆冰预测方法,一定程度上实现了通过微气象等因素预测覆冰增长的效果,但大都针对短期覆冰周期,对数据采集频率有很高的要求,实际工程中实现较为困难。因此文章统计分析了重庆市送变电公司2015—2019年线路观冰数据,得到了西南地区高湿环境下输电线路覆冰特性及规律,并依据覆冰增长物理过程选取了工程可测量气象参数作为覆冰影响因素,提出了一种基于自适应变异粒子群算法(adaptive mutation particle swarm optimization algorithm,AMPSO)优化BP神经网络的人工智能覆冰厚度预测模型,优化了BP神经网络的权值阈值选取,优化后的模型在预测精度上要强于单一BP神经网络与已有研究中提出的小波神经网络,具有良好的工程适用性。

Prediction of Transmission Line Icing Thickness Applying AMPSO-BP Neural Network Model

Icing of transmission line seriously threatens the safe operation of power system. Therefore, it is necessary to investigate icing prediction of transmission lines. With the development of its performance, artificial intelligence technology gradually shows advantages in power grid icing monitoring. To a certain extent, the existing statistical regression model of ice thickness prediction for transmission line can partly predict the ice accretion growth. However, these traditional models are only suitable for short icing periods and difficult to realize in actual engineering because of their requirement of high data acquisition frequency. This research collected the transmission line's ice observation data from 2015 to 2019 got by Chongqing Transmission and Transformation Engineering Co, Ltd.. By analyzing the data, the characteristics and rules of transmission line icing under high humidity environment in southwestern China are obtained. Then, according to the ice growth's physical process on transmission line, the research selects measurable parameters in practical work as the impact factor of ice accretion growth. On this basis, an artificial intelligence ice-thickness prediction model based on adaptive mutation particle swarm optimization (AMPSO) is proposed. According to the training results, the AMPSO-BP neural network is more accurate and reliable on ice thickness prediction, compared to the traditional BP neural network.