基于BP神经网络算法的电力系统惯量评估方法

Inertia Evaluation Method of Power System Based on BP Neural Network Algorithm

中文关键词:惯量评估  BP神经网络  频率偏差  频率变化率

keywords:inertia assessment  BP neural network  frequency deviation  rate of change of frequency

中文摘要:针对当前电力系统惯量评估存在精度不足、计算效率较低的问题，提出一种基于反向传播（back propagation，BP）神经网络算法的电力系统惯量评估方法。首先，通过设置不同新能源出力比例、不同惯性时间常数的多种场景开展仿真，获取频率相关运行指标，并据此计算对应场景下的惯性时间常数；其次，构建以频率运行指标为输入、惯性时间常数为输出的BP神经网络模型，利用仿真数据训练模型，并建立两者之间的非线性映射关系；最后，通过训练完成的模型，根据实际或仿真场景中的频率运行指标，快速估算出电力系统惯量。算例实验结果表明：所提方法能够准确地计算出电力系统惯性时间常数，提升对新型电力系统动态特性的评估能力。

Abstract:To address the low accuracy and poor computational efficiency of existing power system inertia evaluation methods，this paper proposes an inertia evaluation method based on BP neural network. First，simulations are carried out under various scenarios with different renewable energy output ratios and inertia time constants. The frequency-related operating indicators are collected，and the corresponding inertia time constants are calculated. Second，a BP neural network model is established，which takes frequency indicators as inputs and inertia time constant as output. The model is trained with simulation data to build the nonlinear mapping relationship between inputs and outputs. Finally，the trained model can rapidly estimate power system inertia according to frequency indicators in actual or simulated operating scenarios. Case study results verify that the proposed method can accurately calculate the inertia time constant and improve the evaluation capability for dynamic characteristics of new power systems.