Big Data Techniques to Address ADMS Data Quality Issues

August 17, 2018

The IEEE Big Data and Analytics committee held a panel session on data quality last week.

Terry Nielsen describes how new techniques can be used in ADMS deployments.

I attended the IEEE General Meeting last week and participated in the super session panel on data quality sponsored by IEEE Big Data and Analytics subcommittee. The panel sessions raised awareness of issues related to data quality in the industry today. My presentation was about how data quality issues are being addressed in the repository that is part of the ARPA-E GRID DATA project.

It is obvious to me that the use of big data techniques in the power industry is already here, areas with multiple successful deployments include the use of AMI data and Phasor Measurement Unit (PMU) data. In fact, two of the panelist talked about these specific applications. The predominate application of AMI data analytics has been to look deep into large amounts of AMI data and finding new insights such as identifying theft. On the PMU side, the industry has been deploying many situational awareness and disturbance detection applications using PMU data, especially at the transmission level. However, there are many more power systems problems that could potentially be addressed using data analytics and there is some new industry research that is investigating these challenges.

One challenge that has not yet been addressed in the industry is how to use data analytics to support operational applications such as those found in an Advanced Distribution Management System (ADMS). One promising approach was presented by Yingchen Zhang who is working at the National Renewable Energy Laboratory (NREL) on a project that is part of the DOE ENERGISE (Enabling Extreme Real-time Grid Integration of Solar Energy) program. Part of the NREL project is to develop a distribution state estimation solution that uses both time series measurement data from PMUs combined with a physical network model to produce better state estimation results than either a traditional pure model-based state estimation approach or a data analytics approach alone would produce.

I have spent many years, working on projects where cleaning up physical network models for power system applications in EMS, OMS and DMS became the most critical roadblock to success. The approach being pursued by NREL when deployed in commercial ADMS solutions, appears to me that it could significantly reduce the needs for the physical network model to be extremely accurate and could significantly reduce the model clean-up effort in an ADMS deployment.

The industry is desperately in need of ways to reduce the cost of addressing data quality issues in ADMS deployments and this could be just one of many that are applied to the problem.

Terry Nielsen, GridBright EVP of Utility Solutions