High uncertainties in load and renewable energy forecast errors can lead to violations of electrical system safety limits. It is therefore important to accurately assess and manage forecast uncertainties during the operational planning phase.

The report compares the performance of methods for modeling probability distributions of forecast errors based on their quantiles, to identify the most suitable method for integration into a probabilistic control with safety constraints, which is the focus of the report. This control relies on the redispatching of dispatchable units and, if necessary, the curtailment of renewables to ensure that the probability of violating safety limits in states N and N-1 is below predefined thresholds: probabilistic constraints are expressed in terms of quantiles of stochastic input distributions.

The added value consists in modeling forecast uncertainties while considering correlations and characteristics (e.g., asymmetry) of probability distributions. To this end, an innovative combination of techniques and an efficient problem formulation based on linear programming are adopted.

The implementation of the problem in the GAMS-MATLAB environment is then described: this framework leverages GAMS’s capabilities to easily formulate the optimization problem and interface with efficient solvers for large-scale problems, while retaining the advantages of the MATLAB modeling environment. The exchange of information between the two tools and the problem’s “ready-to-implement” formulation are explained, including the listing of parameters and optimization variables. An appendix provides the code in both tools for implementing the control in GAMS-MATLAB.

The control is applied to an IEEE network. First, the control in GAMS-MATLAB is tested with a specific parameter set: the results obtained (e.g., redispatched power patterns, involved units, etc.) match those obtained with the initial implementation carried out entirely in MATLAB. The GAMS-MATLAB tool will be used for further research on large electrical networks. Finally, the results of a sensitivity analysis performed on the control for various values of some parameters (e.g., probability of violation and correlation between forecast errors) are discussed.