Embodiments herein relate to improving a stochastic forecast for uncertain power generations and demands to quantify an effect on an electrical
power grid. To improve the stochastic forecast, a methodincludes fitting marginal distributions to data of the uncertain power generation and demand by power generation and demand nodes of the electrical
power grid. The power generation and demand nodes provide corresponding uncertain power generation and demand based on a
renewable energy source. The method also includes determining a correlation structure between the power generation and demand nodes by transforming the data from marginal distributions to a second distribution and by fitting a multivariate
time series on transformed data. The method also includes simulating multivariate stochastic forecast with an improved correlation structure.