Frequency Risk Assessment for Power Grid with High-penetration Renewable Integration Based on Fast Kernel Learning

Abstract

In the aftermath of power disturbance in power grid with high-penetration renewable integration， the frequency response trajectory is more likely to present large frequency deviation and high rate of change， which leads to a significant effect on frequency protection action. In order to accurately predict the frequency response characteristics of the system under the strong random operation mode， this paper uses the fast kernel learning algorithm based on the pseudo-inverse of the kernel matrix and integrates the key feature quantities obtained from the common mode frequency analysis to propose a frequency risk assessment method for power grid with high-penetration renewable integration. This method uses a set of mutually independent sample data to directly construct the regular term function through the pseudo-inverse operation of the kernel matrix， avoiding the convergence problems caused by the iterative solution of general machine learning algorithms without reducing the generalization ability of the learning results. The analysis of an example in the IEEE 39-node system verifies the effectiveness of the proposed method.

Keywords

• frequency characteristics
• kernel learning
• kernel matrix
• pseudo-inverse
• risk evaluation