Stochastic voltage estimation for islanded DC grids

The rapid growth and adoption of renewable energy sources have increased the appearance of DC-microgrids. However, the state of islanded DC-microgrids is highly sensitive to the variations of generated and demanded powers. In consequence, DC-microgrids require to be monitored in real-time to guarantee a stable state during their operation. This paper proposes and explores three different linear stochastic approaches to estimate the nodal voltages solely from measurements of the generated and demanded powers. In this study, two of the proposed approaches are data-driven, one based on a multivariate Gaussian distribution and the other on a linear multilayer neural network. Meanwhile, the third proposed approach is based on the Kalman filter. The performance of the proposed approaches is validated over four test systems using MatLab software. The experimental results demonstrated that the Kalman filter approach performs better when compared to the other proposed methods according to the root mean squared error and the correlation coefficient values.

Automated summary

This paper proposes and validates three different approaches for estimating nodal voltages in DC-microgrids. The experimental results indicate that the Kalman filter approach outperforms the other methods in terms of performance.