Lossless Compression of Synchrophasor Measurement Unit Archives
Grid Protection Alliance
Data from synchrophasor measurement units, or PMUs, promise to increase wide-area situational awareness of grid conditions like no previous technology. However, as PMUs become more widely deployed, and as their reporting rate grows from 30 measurements per second to higher levels, the amount of data that must be archived and retrieved for later analysis and reporting requirements will grow to cumbersome levels. The purpose of this work was to develop a compression technique for PMU data that preserves the data’s meaningful signal content. The technique needed to be tailored to the unique characteristics of PMU data, and characterize the bits of PMU data in terms of signal content and possible noise. It also needed to extract meaning from PMU data even when metadata identifying the nature of individual data streams are lacking. We observed that the temporal and spatial coherencies associated with power system data are similar to the continuities that exist in images, characteristics used by image compression techniques. Similar to PNG image compression, the compression algorithm we created estimates the next value of a measurement based on its previous value, and the change seen in a neighboring measurement. The algorithm stores the difference between the expected value and the actual value. If these differences tend to be small, they will lie within a narrow range, and the resulting stream will therefore be more compressible. An analysis of the bits in each PMU value revealed some interesting characteristics related to randomness and noise. The least-significant bits of the data in our test streams passed NIST randomness tests, suggesting that they are not part of the meaningful signal content. Our compression approach filters these bits. Using our approach, compression ratios of up to 20 to 1 were seen with actual PMU data.