This presentation discusses our experiences and techniques for data recording during simulations and HIL experiments at FSU-CAPS that leverage various RTDS components including the GTNET, GTFPGA, and RSCAD runtime environment.
Recording data generated during simulations and HIL experiments is critical for understanding and reporting results. The RTDS runtime environment provides facilities for recording data (e.g., Plot in the RSCAD runtime). However, these built-in facilities are not practical for recording large amounts of data and recording data over long time intervals. One issue is the loss of interaction with the Runtime during captures, which prevents the collection of results through sequences of Runtime inputs as a single, continuous sequence of events. Another issue is the inability to vary the sampling rate for different signals and to vary these over the duration of the capture. Additionally, limitations on the number of points that can be captured may restrict how data are collected. Other difficulties stem from the need to capture relatively infrequent events, such as faults. While means exist for internally triggered plots, this can be impractical for capturing infrequent events in some cases, due to the need to disable Runtime interaction while waiting for the event to trigger the plot capture. Thus, a more flexible means for capturing data with the RTDS has considerable utility.
A complimentary tool for capturing data leveraging the GTNET and GTFPGA cards is presented. This tool provides flexibility to accommodate the needs described above, as well as a way to capture high-resolution data continuously for long periods of time—down to the time step—while not encumbering the RTDS Runtime. Recording a large number of signals at a high sampling rate (e.g. at every time-step) throughout an entire simulation or HIL experiment can help reduce the probability of missing data for important events but also results substantial amounts of data. Fortunately, commodity computer hardware including persistent storage devices are very economical and enable much larger amounts of data to be recorded. However, recording data in real-time comes with inherent timing constraints, which results in design challenges to ensure such timing constraints are met and data is not lost. These challenges are especially prevalent when leveraging commodity hardware and software that is not specifically designed to meet timing constraints. With the help of the Linux kernel and the PREEMPT_RT modifications, these challenges are significantly reduced and have made developing a general-purpose data recording capability practically achievable.
Mark Stanovich ~ Center for Advanced Power Systems, Florida State University