Summary

This paper provides a review of the literature on AI/ML applications in the power grid, a review of the factors affecting data quality, an evaluation of how the limitations of current instrumentation impact data quality, and a novel approach to efficiently obtain high quality, synchronous waveform or Continuous Point On Wave (CPOW) data from the distribution grid at scale.

Abstract

It is well understood that the quality of data used in any Artificial Intelligence (AI) or Machine Learning (ML) application is critical to the accuracy and effectiveness of the results. This is true for the data that is used to train the AI/ML model as well as for the data that is fed to the model once it is deployed (serving data). Once a model has been deployed, serving data is often used to refine and improve the model continuously. Since the results generated from deployed AI/ML models may be used in planning and operational decision making, getting the data right (both in training and in deployment) is essential for making the best planning and operational decisions. This is particularly true when considering the use of AI/ML applications in the distribution grid that are necessary to “keep the lights on”.

While the use of big data analytics and AI/ML for the grid has been a topic of research since at least 2015, the primary challenge preventing wide scale adoption of AI/ML methods in planning and operations has been consistent access to high-quality data across the distribution grid. The current instrumentation available to distribution grid operators, e.g., Distribution-level Phasor Measurement Units (D-PMUs, a.k.a. micro-PMUs or µPMUs), Harmonic Phasor Measurement Units (H-PMUs), Waveform Measurement Units (WMUs), and Advanced Metering Infrastructure (AMI) are not able to provide consistent, high-quality data required for training and deployment of AI/ML methods at scale in planning and operations of the distribution grid. A primary challenge in making use of this instrumentation is the lack of a secure and reliable communications infrastructure with sufficient bandwidth to carry the serving data from the instrumentation to the centralized platforms that execute the deployed AI/ML models.

This paper provides a review of the literature on AI/ML applications in the power grid, a review of the factors affecting data quality, an evaluation of how the limitations of current instrumentation impact data quality, and a novel approach to efficiently obtain high quality, synchronous waveform or Continuous Point On Wave (CPOW) data from the distribution grid at scale.

Contact Us to receive white paper

---