The time-domain simulation is fundamental for power system transient stability analysis. Accurate and reliable simulations depend on accurate dynamic component modeling. In practical power systems, dynamic component modeling has long faced the challenges of model determination and model calibration, especially with the rapid development of renewable generation and power electronics. In this paper, based on the general framework of neural ordinary differential equations (ODEs), a neural ODE module with external inputs and a neural differential-algebraic equations (DAEs) module are proposed for power system dynamic component modeling. Autoencoder-based frameworks of the proposed modules are put forward to improve the performance of trained models. The methodology of integrating neural dynamic models trained by the proposed neural modules into transient stability simulations is also demonstrated. With datasets consisting of sampled curves of input variables and output variables, the proposed modules can be used to fulfill the tasks of black-box modeling, physics-data-integrated modeling, parameter inference, etc. Tests are carried out in the IEEE-39 system to prove the validity and potentiality of the proposed modules.
翻译:时间域模拟是动力系统瞬时稳定性分析的基础。 精确和可靠的模拟取决于准确的动态元件模型。 在实用的动力系统中,动态元件模型长期面临模型确定和模型校准的挑战, 特别是可再生能源和电力电子的快速开发。 在本文中, 根据神经普通差分方程(ODEs)的总框架, 为动力系统动态元件建模提议了一个包含外部输入和神经差分数方程(DAEs)模块。 以自动编码为基础的拟议模块框架用于改进经过培训的模型的性能。 也演示了将由拟议神经元模块培训的神经动态模型纳入瞬时稳定性模拟的方法。 由输入变量和输出变量抽样曲线组成的数据集可以用来完成黑箱建模、物理数据集成模型、参数推断等任务。 在 IEEE- 39 系统中进行了测试,以证明拟议模块的有效性和潜力。