For parallel state estimation, parallel sparse linear solvers are used solve state estimation problem and tested with real utility data.
For massive contingency analysis, a dynamic load balancing scheme was implemented to achieve optimal speedup, minimizing the impact of different solution time for different contingencies. The contingency execution is based on the availability of cores. Therefore, while the number of cases assigned to each core might be different, the computational time on each core is roughly same. The same scheme can be applied to look-ahead contingency analysis.
For look-ahead contingency analysis, smart sampling algorithm is needed to have a reduced, but representative data set to cover the uncertainty brought by forecast errors and significantly reduce computational time. One year real wind forecast data and actual generation data extracted from state estimation snapshots are used to test the performance of smart sampling algorithm, in conjunction with actual and forecast load data. After the simulation is completed, probabilistic analysis techniques, including extreme value distribution, are used to predict system future conditions to allow users foresee system’s future conditions and take actions ahead of time. The algorithm has been integrated with GE Grid Solutions’ energy management system tool, leveraging GE SMTNET multiple-timepoint study, as a proof-of-concept for a seamless integration
For visualization, advanced visualization techniques were used to represent grid status with different colour schemes for an intuitive view. More supporting features, such as contingency ranking, interactively corrective action comparison are also implemented to further provide effective decision-making support to operators and engineers. Both stand-alone tool and web browser based visualization tool are implemented.