Real-Time Operating Limits
Grid operation challenges
Every major electric power utility manages its grid with more than one type of software package. These software solutions are typically referred to as "dynamic planning" models and "operational" models. Although each type of model provides a different type of functionality, they are also distinct in the rigorousness with which they calculate the grid's dynamic responsiveness, with the dynamic planning software providing the highest degree of sophistication, but at the cost of computational speed. The planning software runs on a PC platform, taking about 10 minutes to solve about 25 seconds of grid response for a WECC-sized system. Due to the complexity of this software, these best-in-class software modules are not able to run in real-time, and are not useable in an operational context.
Grid operators rely on rules that are established using this sophisticated software. Such rules dictate operational limits based on dynamic and voltage stability constraints of the power grid. The onset of problems associated with exceeding these limits are difficult to recognize in real-time, and can lead to unrecoverable situations; therefore, operators must ensure that sufficient margin exists so as not to encroach upon these limits.
Since the operating rule sets are established off-line they must be designed to apply to a broad set of operating conditions, implying that excessive operating margin likely exists at various moments and locations, but one cannot safely verify this nor utilize this spare capacity. The corollary is that during moments of grid distress, there is excess capacity of the grid that could be used, but it cannot be identified in the short period of time necessary to use it.
Pacific Northwest National Laboratory is performing research that is moving these sophisticated dynamic planning models to the operational realm. We are working toward taking these software solutions from 10 minutes to less than a second, much faster than real-time. This is being accomplished by re-writing specific algorithms in this software, enabling the software to operate on a high-performance computing platform. Although Pacific Northwest National Laboratory has access to various computing platforms, we are currently using an Altix Shared Memory HPC platform and software formulated with Fortran 95 with OpenMP Application Program Interface. As research progresses, the methodology and hardware platform for parallelization may be changed.