Reliability assessment of renewable energy integrated power systems with an extendable Latin hypercube importance sampling method

The increasing integration of renewable energy sources has brought about a great computational burden for the traditional methods that assess the power system reliability. To reduce the computational costs, an extendable Latin hypercube importance sampling (ELHIS) method that combines importance sampling (IS) and Latin hypercube sampling (LHS) is proposed in this paper. First, the challenge of combining IS and LHS is analysed, and a customized sampling process of LHS is designed accordingly. Then, in the IS part of ELHIS, the cross entropy theory and Gaussian mixture model are adopted in the stage that constructs the quasi-optimal probability distribution of random variables. In addition, the samples in this stage are also utilized in the estimation of reliability indices to reduce the waste of computational efforts. In the LHS part, an extendable LHS approach is used to make ELHIS adaptive and flexible in determining the sample size to reach the required accuracy. Finally, numerical tests are performed on the modified IEEE-RTS 79 test system, where the real historical data from three wind farms and a photovoltaic station in Northwest China are employed. The results show that ELHIS is obviously faster than recent IS methods when used to assess the power system reliability.