The report describes the activities carried out in the first year of “iTesla Power System Tools” open source project, which followed the iTesla EU FP7 project. The iTesla project has developed a methodology and a software platform for the evaluation of the dynamic security of the electrical system for in-line applications, and up to a few hours ahead of the real time, taking into account the uncertainties in the forecasting errors related to the power absorption of loads and power injections from non dispatchable renewable energy plants.

The first part of the report recalls the methodology of validation for the in-line platform modules, with particular attention to the MCLA (Monte Carlo Like Approach) module, underlying its potentialities and limitations. In order to cope with its limitations, an accurate statistical analysis of the stochastic variables is performed: in particular, an algorithm identifying the multimodality of a variable is proposed. Considering that the MCLA procedure, which exploits the Nataf transformation, provides better results in case of unimodal variables, different techniques for clustering (hierarchical clusterization, clusterization using variable names) are compared to combine multimodal variables to get unimodal variables. The analysis carried out on the French grid data allows to tune the parameters of these clusterization techniques and to identify few tens of pairs of multimodal variables -over several thousands of variables- which can be combined to get “smoother” unimodal variables which can be better treated by the MCLA workflow.

The second part of the report is dedicated to describe the improvements to the performances of the uncertainty model generated by the MCLA module and its thorough verification on the historical data referring to a real world grid. In particular, the upgrades suggested in the present report intend to identify a good tradeoff between the model accuracy and the sharpness of its uncertainty clouds, in order to improve the generalization capability of the model, i.e. its capability to predict the injection uncertainties in future hours starting from the knowledge of historical data.

In particular, a separate sampling of unimodal and multimodal variables is proposed, so that each pair of forecast and snapshot related to a multimodal injection is modeled using a two-dimension Gaussian mixture model, while the snapshots for the unimodal injections are sampled exploiting the Gaussian conditional sampling formulation of the conventional MCLA module. Moreover, a simplification in the dependence structure of the uncertainty model may improve the generalization capability of the module. The verification of the uncertainty model highlights the benefits brought by the proposed upgrades to the performances (accuracy and sharpness) of the MCLA module.