Monte Carlo and bootstrap approaches for power system adequacy: a study on atmospheric variables correlations

In an electric power system (EPS), the rapid increasingly presence of non-dispatchable renewable energy sources (ND-RES) accompanied by the growing decommissioning of conventional power plants puts its adequacy level at risk. In literature, adequacy analyses are often conducted by simulating the availability profiles of power plants. The limitation of such an approach is that it neglects the stochastic modeling of the impacts of atmospheric variables on energy production and demand. To address this gap, this article aims at the stochastic generation of hourly profiles of atmospheric variables. Following a correlation analysis, using Pearson (r) and Spearman correlation coefficients, a procedure is developed for the stochastic generation of correlated profiles of atmospheric variables. The case studies to which the methodologies are applied are Sicily and Sardinia, two insular market zones of the Italian EPS. It is found that the degree of correlation between atmospheric variables exhibits spatial, temporal variability or both; similarly, for the same atmospheric variable, the probability distribution used for fitting the data varies depending on the time period and location considered. The methodology used to generate synthetic correlated profiles of atmospheric variables, based on Monte Carlo and Bootstrap methods, yielded statistically valid results. Finally, the stochastically generated correlated profiles are used to generate synthetic hourly profiles of (non-dispatchable) electricity production and demand, suitable for adequacy analysis simulations. It is observed that in cases where the correlation between two atmospheric variables is significant (r>0.4), the synthetic production and demand profiles obtained are close to those obtained with historical weather data.