Modeling method for wind-light-fire coupling system frequency response model based on GAN and GRU neural networks

Abstract

The invention aims to provide a modeling method of a system frequency response model based on a generative adversarial network (GAN) and a gate cycle unit (GRU) neural network, aiming at complex dynamic characteristics such as time varying, nonlinearity, uncertainty, intermittency and the like of a wind-light-fire coupling system, and discloses a modeling method of a system frequency response model based on the GAN and the GRU neural network. The generative adversarial network is utilized to solve the problem that training samples are deficient in the building process of the system frequency response model based on data driving. In addition, the problem that an existing modeling method cannot accurately describe nonlinearity, uncertainty and the like of system frequency response is solved by utilizing a gate circulation unit neural network.

Description

technical field [0001] The invention relates to the field of modeling of a frequency response (system frequency response, SFR) model of a power system, in particular to a modeling method for a frequency response model of a wind-solar-fire coupling system with less sample data. Background technique [0002] The grid-connected operation of large-scale wind turbines and photovoltaic power plants brings challenges to the active power and frequency regulation of the power system. The random fluctuation of the output of wind turbines and photovoltaic power plants increases the risk of large fluctuations in the steady-state frequency of the system. And the access of a high proportion of power electronic equipment leads to the reduction of system inertia, which increases the possibility of frequency instability when the system is disturbed by severe active power imbalance. In order to ensure the safe and stable operation of the power grid with a high proportion of wind power and ph...

AI Technical Summary

Problems solved by technology

Using generative confrontation network to solve the problem of lack of training samples in the process of building a data-driven system frequency response model

In addition, using the gated recurrent unit neural network solves the problem that the existing modeling methods cannot accurately describe the nonlinearity and uncertainty of the system frequency response.

Images