Location Analytics For Smart Grid Reliability
Image used with permission.
- Jul 25, 2019 6:35 pm GMT
- 560 views
In previous articles I have written about the work of Vivian Sultan in the Energy Informatics field. At the time of those articles, Vivian was a Ph.D. candidate in Information Systems and Technology at Claremont Graduate University (CGU). Recently Vivian obtained her Ph.D. with Location Analytics For Smart Grid Reliability as the topic of her doctoral dissertation. As in her past research activities, Geographic Information System (GIS) technology played a leading role. In particular, Sultan has made great use of the tools within the Esri GIS toolbox.
Unlike her past work where she relied on publicly available data, such as information provide by Los Angeles County and the California Public Utilities Commission (CPUC) as well as other sources, Sultan had the opportunity to utilize information provided by the Electric Power Research Institute (EPRI). The information came from a data mining initiative within an EPRI project. Data sets, provided by an EPRI utility member, included data from Advanced Metering Infrastructure (AMI), Supervisory Control and Data Acquisition (SCADA), Outage Management (OMS), Distribution Management (DMS), Asset Management (EAM), Customer Information (CIS) and other sources. In addition to the EPRI provided data, weather and other data was also obtained from state GIS Clearinghouse and Spatial Data Infrastructure databases as well as from the National Oceanic and Atmospheric Administration (NOAA). The former databases also provided shapefiles for roads and counties.
With this volume of data one of the first activities was a data curation effort. This involved the use of several tools in order to make the data available and usable within Esri’s ArcGIS platform. This involved creating various layers using a common projection and eliminating data that did not pass quality rules. Additionally, various tools were used to conduct analysis of the data to understand its content and usefulness. Figure 1 is an example of a simple analysis which counted the outages by cause. This directed the effort as to the priority of causes to analyze.
Figure 1 Reported power outage events count by cause
Further analysis of the data was done utilizing SPSS. This analysis included descriptive statistics such as percentages and frequencies along with means, standard deviations and min/max values. Correlation analysis was also done to identify areas of interest for further study.
As noted above, some of the data was eliminated as it did not pass quality rules. This is a problem that utilities will need to address as they expand their analytics efforts and one I have seen in my own work with utilities. Examples of data quality issues include blank fields, fields with invalid values, location coordinates that are outside of service territory bounds, location clusters where many assets have the same coordinates (this seems to be especially a problem with AMI meters) and conflicting locations (e.g. a transformer and the pole which it is identified as being on) have widely different coordinates. While some data with quality issues must be rejected, some can be corrected by various techniques that at least make the data usable although still somewhat suspect.
Once the wealth, as opposed to just volume, of data was in usable shape the next step was further detailed analysis. Outages, being a prime indicator of system reliability, were the area of analytical interest in this step. The first analytic was to provide an indication of where outages primarily occurred from a geospatial standpoint. This was accomplished using the ArcGIS Optimized Hot Spot Analysis tool. In layperson’s terms this tool identifies where events, e.g. outages, cluster geographically. Figure 2 is an example of such an analysis.
Figure 2 Optimized Hot Spot Analysis
While this analysis may seem intuitive to the long-time utility employee, “of course that’s where the outages are, that’s all overhead”, it is only a starting point for further analysis. That analysis may start by separating the outage data by overhead areas versus underground areas as the expectation would be that different primary causes would be associated with each. The Hot Spot tool provides the ability to utilize an ‘analysis field’ to further refine the analysis. For her work, Sultan looked at various factors such as temperature, precipitation, equipment age and vegetation management. Figure 3 shows an example of this refinement for outages due to tree related causes.
Figure 3 Tree Related Outages
One more capability of the Hot Spot tool is to indicate trends in what is know as Emerging Hot Spot Analysis. In the case of outages, this would indicate areas where the number of outages is increasing over time. These could then be analyzed against such factors as last tree trimming date or equipment age. In Sultan’s analysis tree related outages emerged as the leading factor in the Emerging Hot Spots followed by weather, equipment failure and system overload.
Having demonstrated that GIS serves as a useful analytics tool to understand factors affecting system reliability, Sultan turned to developing tools that would allow the more casual user do their own analysis and support decision making. For this she turned to Insights for ArcGIS from Esri. Insights provides the ability to create worksheets with multiple pages addressing an analytics/decision making area combining map views with graphic view such as pie and bar charts as well as tables. Figure 4 shows a worksheet developed to address questions related to storm caused outages.
Figure 4 Insights Worksheet for Storm Scenario
This worksheet provides the ability to answer questions such as:
- What type of storm contributes to the greatest number of outages?
- What type of storm contributes to the largest duration of outages?
- Where are the weather related emerging hot spots?
The additional pages of this worksheet address questions related to infrastructure age and precipitation relative to outages. Additional worksheets address questions and decisions relative to outages caused by vegetation management issues and equipment failure.
To assess the viability of her analysis and the decision support tools Sultan conducted a webinar attended by utility personnel where she presented her research and the tool. The response by all attendees was positive with some suggestions for improvements. As a second evaluation, Sultan provided access to the Insights tool to solicit feedback from several participants. Again, the response was positive and supported the conclusion that GIS can be a powerful tool for improving system reliability.