基于分布式神经网络的同步合闸时间预测方法研究

Synchronous Closing Time Precalculation Based on Distributed Neural Network

阐述同步开关技术在抑制投入电容器组时产生的过电压和过电流的有效性,深入分析了同步控制过程及其控制算法。鉴于精确预测合闸时间在同步控制中的重要作用,根据不同环境温度下,断路器合闸时间与控制电压、环境湿度等因素之间的复杂关系以及断路器的初始动作电压特点,提出了分布式径向基神经网络的合闸时间预测模型。该模型分2层结构,第1层根据不同的环境温度将输入变量映射到分布式径向基神经网络的某个网络模型,第2层运用映射出的网络子模型预测断路器合闸时间。实验测试结果表明,该方法预测速度快,精度高,完全能够满足同步控制对合闸时间预测精度的要求。

Synchronously controlled switching to suppress transit overvoltage and overcurrent resulting from when the circuit breakers on medium voltage systems are closed is described. This paper deeply analyses the new controlled switching algorithm to minimize the total delay time and the importance of precisely precalculating closing time. Distributed radial basis function neural network is proposed. The model is divided into two layers. The first layer maps the input vectors to their relational neural network. The second step precalculates the closing time. Experimental results show the precalculated closing time can completely satisfied appliance with high precision.