基于自适应神经网络的电压快速估计

Voltage Speediness Estimation Based on Adaptive Neural Network

随着电网规模扩大、复杂度加深,对在线潮流计算确定节点电压提出严峻挑战。同时由于传统潮流计算依赖于模型建立的精度,其负荷模型建立存在较大难度,因此通过电压对无功的响应数据来快速精确估计电压发展趋势具有重要意义。本文提出一种基于自适应神经网络的节点电压快速估计方法,以系统负荷水平、无功激励为输入,节点电压为输出,离线训练神经网络,并在训练中加入遗传算法以提高网络收敛性和参数寻优能力,最终得到用于在线估计节点电压的隐性模型。通过IEEE 24节点系统标准算例验证表明,该方法具有较强的电压拟合能力和外推能力,计算结果相较于传统神经网络更加精确,不易出现过拟合。

With the expansion of grid scale and complexity, it is difficult to determine the node voltage by online power flow calculation. Also, traditional power flow calculation relies on the precision of model, however, there are still many problems in the establishment of load models, so it is very important to estimate the trend of voltage development quickly and accurately by the voltage response data. This paper presents a fast estimation for voltage based on adaptive neural network, inputs are the system load level and reactive power, the output is the node voltage, neural network is trained with them. Genetic algorithm is added to the training to improve network convergence and parameter searching ability. Finally, a hidden model for online estimation of voltage is obtained. The results of standard IEEE 24-bus system justified that the proposed methed has better ability of voltage fitting and extrapolation than traditional neural network.