#### Improved PMU Design for Off-Nominal Conditions in Power Distribution Systems

Students: Salish Maharjan and Jorge E. Martinez | Advisor: Jimmy C.-H. Peng | Collaborator: James L. Kirtley Jr., Massachusetts Institute of Technology | Project Duration: 2013 -2015

The notion of a smart grid is established based on the innovation of a sensor technology known as Phasor Measurement Units (PMU). PMU usually implements Enhanced Discrete Fourier Transformation (EDFT) to compensate leakage and signal discontinuity errors at off-nominal frequencies. However, EDFT results in a high total vector error (TVE) at large off-nominal frequency and worsens with harmonics in power signal. Although sample value adjustment (SVA) technique addressed such demerits, it generates an asymmetrical TVE profile. Additionally, the frequency estimator was not taken into account in the previous publication on SVA.

Figure 1. Overview of the interpolation technique used by ISVA to convert the input signal of unknown frequency to an output signal of the reference frequency.

Figure 2. Performance comparison between the proposed ISVA against its predecessor SVA and the commonly used EDFT. TVE refers to the total vector error that is used to quantify the estimation accuracy.

Figure 3. TVEs of the prototyped PMU operating at various grid frequency.

Figure. Prototyped PMU design with ISVA installed for phasor computation.

This research developed an improved sample value adjustment (ISVA) as a pre-DFT technique to further enhance the phasor estimation (Figure 1). The improvement was achieved by reducing the interpolation error while transforming the input signal to the nominal frequency. Better interpolation has been achieved by updating the displacement factor and the reference sample index for each samples of the input signal. As a result, the ISVA generates a symmetrical TVE profile and performed better than EDFT, SVA and Taylor weighted least square (TWLS) method at steady-state conditions (Figure 2). Subsequently, the ISVA is integrated with a recursive least square frequency estimator, which enabled it to track dynamic signals as well.

The proposed algorithm has been validated in MATLAB simulation and in a prototyped PMU. Static and dynamic test signals are based on IEEE C37.118.1a-2014 standard (Figure 3). The simulation results demonstrate the proposed ISVA method complies with the P-class PMU under both static and dynamic conditions. Subsequently, a prototype PMU has been fabricated and tested. During the hardware experimentations, phasors were computed in real-time at every sampling instance. Their corresponding TVEs were found to be higher than those in the simulated model due to the limitation of the computing power, and bandwidth of data communications.

This work has sparked a number of commercialization possibilities, especially in the context of developing the next generation smart meters for advanced metering infrastructure (AMI). Further hardware enhancements using advanced processors with floating point capability for the computation and Ethernet-based communication for a higher data bandwidth are part of the ongoing work.