An Efficient Approach to Power System Uncertainty Analysis With High-Dimensional Dependencies

Abstract

The integration of high penetration of renewable energy brings greater uncertainties for the operation of future power systems due to its intermittency and lack of predictability. The uncertainties brought by wide scale renewables might have dependencies with each other because their outputs are mainly influenced by weather. However, an analysis of such uncertainties with complex dependencies faces the "curse of dimensionality". This challenges the power system uncertainty analysis in probabilistic forecasting, power system operation optimization, and power system planning. This paper proposes an efficient approach that is able to handle high-dimensional dependencies. The approach uses the high-dimensional Copula theory and discrete convolution method to conduct a high-dimensional dependent discrete convolution (DDC) calculation. A recursive algorithm is proposed to decompose the computation of DDC into multiple convolutions between each pair of stochastic variables so that the "curse of dimensionality" is solved. The computational complexity of the proposed method is linear with respect to the number of dimensions and guarantees computational efficiency. Finally, illustrative examples of power system reserve requirement evaluation and wind power capacity credit assessment analysis are used to verify the effectiveness and superiority of the proposed approach.