首页|期刊导航|山东电力技术|类噪声扰动下基于物理信息神经网络的电网惯量在线评估

类噪声扰动下基于物理信息神经网络的电网惯量在线评估OA

Online Evaluation of Grid Inertia Based on Physics-informed Neural Networks Under Noise-like Disturbance

中文摘要英文摘要

针对高比例新能源电力系统中电网惯量在线评估问题,提出一种基于物理信息神经网络(physics-informed neural networks,PINN)的电网惯量和阻尼系数在线评估方法.通过将物理方程约束与数据驱动相结合,利用电力系统中持续存在的类噪声扰动数据进行惯量评估.以时间序列和功率扰动为输入,频率偏差为输出构建PINN模型,神经网络嵌入转子运动方程残差约束,建立包含数据拟合、物理正则化及初始条件的复合损失函数.通过物理方程微分项的自动微分计算,强制网络输出满足系统动态规律,实现了惯量和阻尼系数的同步辨识.仿真结果验证了该方法的高精度和快速性,测试数据平均误差1.84%,最大误差4.846%,频率预测曲线拟合度达95.6%.本文方法能够有效避免对特定扰动和精确模型的依赖,显著降低数据依赖性.将物理方程约束融入神经网络训练,有效避免过拟合,增强模型泛化能力与物理可信度.对实时跟踪新能源出力波动下的电网等效惯量、促进高比例新能源消纳具有重要研究意义.

To address the problem of online evaluation of grid inertia in power systems with high-proportion renewable energy,this paper proposes an online evaluation method for grid inertia and damping coefficients based on physics-informed neural networks(PINN).By combining physical equation constraints with data-driven approaches,the inertia evaluation is carried out using persistent noise-like disturbance data in the power system.A PINN model is constructed with time series and power disturbances as inputs and frequency deviations as outputs.The neural network embeds residual constraints of the rotor motion equation to establish a composite loss function including data fitting,physical regularization,and initial conditions.Through the automatic differentiation calculation of the differential terms of physical equations,the network output is forced to satisfy the dynamic laws of the system,realizing the synchronous identification of inertia and damping coefficients.Simulation results verify the high accuracy and rapidity of the method,with an average error of 1.84%for test data,a maximum error of 4.846%,and a frequency prediction curve fitting degree of 95.6%.The method can effectively avoid dependence on specific disturbances and precise models,significantly reducing data dependency.Integrating physical equation constraints into neural network training effectively avoids overfitting and enhances the model′s generalization ability and physical credibility.It is of great significance for real-time tracking of the equivalent inertia of the power grid under fluctuating renewable energy output and promoting the consumption of high-proportion renewable energy.

李雅然;张宁宇;邹小明;乔勇;陈中

江苏省电力试验研究院有限公司,江苏 南京 211103||国网江苏省电力有限公司电力科学研究院,江苏 南京 211103江苏省电力试验研究院有限公司,江苏 南京 211103||国网江苏省电力有限公司电力科学研究院,江苏 南京 211103江苏省电力试验研究院有限公司,江苏 南京 211103||国网江苏省电力有限公司电力科学研究院,江苏 南京 211103东南大学电气工程学院,江苏 南京 210096东南大学电气工程学院,江苏 南京 210096

信息技术与安全科学

新型电力系统惯量评估物理信息神经网络类噪声扰动在线评估频率动态响应

new power systemsinertia evaluationphysics-informed neural networksnoise-like disturbancesonline assessmentsfrequency dynamic response

《山东电力技术》 2026 (8)

84-99,16

江苏省电力公司科技项目(SGJSDKSYNYJS2400586). Science and Technology Project of Jiangsu Electric Power Company(SGJSDKSYNYJS2400586).

10.20097/j.cnki.issn1007-9904.250404

评论