One critical national security challenge is maintaining the integrity of national power grids under adverse conditions imposed by natural or human-made causes. Having more renewable power sources on the grid, such as solar and wind power, results in a more uncertain power supply, making disruptions even more difficult to manage. Additionally, a large fraction of distributed generation resources (e.g., rooftop solar panels) are behind meter and are not visible to grid operators.
The ExaSGD project is developing algorithms and techniques to address these new challenges and optimize the grid’s response to many potential disruption events under different weather scenarios. For example, in the Eastern Interconnection grid, there are thousands of contingencies to consider occurring under thousands of possible and impactful weather scenarios. In this case, a high-fidelity security-constrained optimal power flow analysis requires the simultaneous optimization of millions of power grid realizations within a relatively short turnaround time. This analysis will help grid planners and operators assess alternative grid management strategies to best maintain the integrity of the national power grid under extremely complex and uncertain operating conditions but will also require the power of exascale computing to run within the prescribed time.
Energy delivery systems operate by maintaining balance between energy supply and demand. Random equipment failures, attacks via physical or cyber means, severe weather conditions, or other hazards on the grid can create an imbalance between supply and demand. These disruptions can result in drops in voltage and/or frequency, both of which can trip large components and cause cascading failures, leading to a large-scale blackout. To ensure safety and reliability, the grid must operate within narrow voltage and frequency ranges.
Recovering from generation/load imbalance can be achieved by shedding load in the case of generation deficiency (i.e., deliberately cutting off supply to certain consumers and creating a partial blackout) or tripping generation in the case of overgeneration to preserve the functionality of the remainder of the power grid. Emerging technologies make power grids more complex but also more controllable. Examples include advanced cyber-enabled control and sensing, variable generation (e.g., wind and solar), plug-in storage devices (e.g., electric vehicles), and smart meters that can control load at a fine granularity (e.g., throttling home appliances at times of peak demand), to name a few. The emerging technologies provide an opportunity for more efficient grid control strategies. However, to take advantage of that, more elaborate modeling and computational techniques are required. Current computational capability limits grid operators to using simplified models. To compensate for not modeling features such as ramping constraints and stochastic effects, grid operators use heuristic rules, which are based on years of past experiences. However, those experiences might not apply to emerging grid technologies with many renewable resources.
The ExaSGD challenge problem is to optimize the grid’s short-term response (e.g., 30 min per North American Electric Reliability Corporation operating standards) to many potential hazards while satisfying security constraints. The ExaSGD team will use high-fidelity models that capture voltage and reactive power constraints. This replaces more conservative heuristics and allows for higher power flows over existing transmission corridors while simultaneously ensuring that the network is not under the risk of voltage collapse. Next, the team will model stochastic effects (e.g., those caused by weather changes) to minimize the operational risk due to forecast errors. This will allow reserves to be managed more efficiently and improve grid resilience. In particular, it will allow grid operators to balance the likelihood of a disruptive event occurring with the potential impact of the event. Finally, the team will implement a multiperiod analysis to capture ramping constraints correctly (i.e., how fast the new generation can come online). This capability is critical for planning ahead and finding an optimal response based on available forecasts. Each of these layers increases the computational cost combinatorially and very quickly results in computational requirements that are beyond the capabilities of the present power systems analysis tools. The new modeling and analysis capability developed within the ExaSGD project will provide the computational capability needed to meet the needs of grid stakeholders far into the future.