All posts by Bin Qiu

Improve power utility operational efficiency using smart sensor data and Amazon QuickSight

Post Syndicated from Bin Qiu original https://aws.amazon.com/blogs/big-data/improve-power-utility-operational-efficiency-using-smart-sensor-data-and-amazon-quicksight/

This blog post is co-written with Steve Alexander at PG&E.

In today’s rapidly changing energy landscape, power disturbances cause businesses millions of dollars due to service interruptions and power quality issues. Large utility territories make it difficult to detect and locate faults when power outages occur, leading to longer restoration times, recurring outages, and unhappy customers. Although it’s complex and expensive to modernize distribution networks, many utilities choose to use their capital through the application of smart sensor technologies. These smart sensors are installed in selected locations on distribution networks to monitor various disturbances, such as momentary and permanent outages, line disturbances, voltage sags and surges. The sensors provide analysts with fault waveforms and alerts in addition to graphical representation of regular loads. Different communication infrastructure types such as mesh network and cellular can be used to send load information on a pre-defined schedule or event data in real time to the backend servers residing in the utility UDN (Utility Data Network).

In this series of posts, we walk you through how we use Amazon QuickSight, a serverless, fully managed, business intelligence (BI) service that enables data-driven decision making at scale. QuickSight meets varying analytics needs with modern interactive dashboards, paginated reports, natural language queries, ML-insights, and embedded analytics, from one unified service.

In this first post of the series, we show you how data collected from smart sensors is used for building automated dashboards using QuickSight to help distribution network engineers manage, maintain and troubleshoot smart sensors and perform advanced analytics to support business decision making.

Current challenges in power utility operations

To have a comprehensive monitoring coverage of the distribution networks, utilities normally deploy hundreds, if not thousands, of smart sensors. Similar to any other equipment or device, smart sensors could encounter different issues, such as having defective parts, wearing out over time, becoming obsolete due to technological advances, or suffering loss of communication due to power outages or low cellular signal coverage. Managing such a large number of devices can be challenging.

Furthermore, based on the use case, utilities normally apply sensor technologies from different vendors. Solutions from different vendors can vary, such as data protocols, formats, native connectors, and communication media, which further increases the complexity of managing these smart sensors.

To effectively solve smart sensor management issues and improve operational efficiency, distribution engineers need a BI application that is simple to use and has a powerful data processing and analytics engine. QuickSight provides an ideal solution to meet these business needs.

Solution overview

The following highly simplified architectural diagram illustrates the smart sensor data collection and processing. Smart sensors send data via cellular communication based on a predefined schedule or triggered by real-time events. Data collection and processing are handled by a third-party smart sensor manufacturer application residing in Amazon Virtual Private Cloud (Amazon VPC) private subnets behind a Network Load Balancer. Amazon Kinesis Data Streams interacts with the third-party application through a native connection and conducts necessary data transformation in real time, and Amazon Kinesis Data Firehose stores the data in Amazon Simple Storage Service (Amazon S3) buckets. The AWS Glue Data Catalog contains the table definitions for the smart sensor data sources stored in the S3 buckets. Amazon Athena runs queries using a variety of SQL statements on data stored in Amazon S3, and QuickSight is used for business intelligence and data visualization.

After the smart sensor’s data is collected and stored in Amazon S3 and is accessible via Athena, we can focus on building the following QuickSight dashboards for distribution network engineers:

  • Sensor status dashboard – Analyze and monitor the status of smart sensors
  • Distribution network events dashboard – Analyze the operational information of the distribution networks

Prerequisites

This solution requires an active AWS account with the permission to create and modify AWS Identity and Access Management (IAM) roles along with the following services enabled:

  • Athena
  • AWS Glue
  • Kinesis Data Firehose
  • Kinesis Data Streams
  • Network Load Balancer
  • QuickSight
  • Amazon S3
  • Amazon VPC

Additionally, data collection and data processing are functional blocks of the third-party smart sensor manufacturer application. The smart sensor application solution must be already deployed in the same AWS account and Region that you will use for the dashboards.

This solution uses QuickSight SPICE (Super-fast, Parallel, In-memory Calculation Engine) storage to improve dashboard performance.

Sensor status dashboard

When hundreds or thousands of line sensors are installed, it’s critical for distribution engineers to understand the status of all smart sensors on a regular basis and fix issues to ensure smart sensors provide real-time information for operator decision-making. Assuming a utility has 5,000 smart sensors installed, even if only 1% of the sensors have communication issues (a realistic scenario based on utility experience), distribution engineers need to check and troubleshoot 50 sensors per day on average. The smart sensor communication losses could be caused by low cellular signal strength, low power supply, or planned or unplanned outages. If it takes 10 minutes to analyze one sensor, it will cause the engineering team around 500 minutes per day just to analyze the questionable smart sensors.

Rather than checking smart sensor information from different applications or systems to find answers, a sensor status dashboard solves this problem by aggregating status statistics across all sensors by different attributes, including sensor location, communication status, and distributions in different regions, substations, and circuits.

In the following sensor status dashboard, a hypothetical utility has 102 smart sensors (each location needs three sensors for phases A, B, and C) deployed in five substations and six circuits. During normal operations, smart sensor reports load data every 5–15 minutes, and the event data (different fault events) could come at any time depending on the circuit situation.

Multiple panes are designed to help distribution engineers answer critical questions on smart sensors and facilitate troubleshooting in case communication issues happen to smart sensors:

  • Summary – The top summary pane provides a quick glance of the smart sensor statistics, such as number of substations, circuits, smart sensors with good communications, or smart sensors that have communication issues.
  • Smart Sensor Status By Location – This pane shows the geographical distributions of all the smart sensors. Different colors are used to demonstrate smart sensor operational status. In this case, four of the sensors have communication issues, which are shown in red on the map. The operator can identify the questionable sensors, zoom in, and determine the actual location of these sensors. When operators pick up the questionable smart sensors, the geo-map can auto focus on these smart sensors as well.
  • Sensor Status By Substation and Circuit – This pane gives operators a glance of smart sensors by substation and circuit, such as number of healthy smart sensors and number of sensors with communication issues.
  • Unhealthy Sensor Details – This pane provides information about questionable smart sensor data.
  • Cellular Communication Signal Strength Distribution – Smart sensors transmit data to the cloud using cellular communication. If the signal strength is lower than -100 dBm to -109 dBm (considered poor signal of 1 to 2 bars), the signal might be too weak for the sensor to transmit data. Distribution lines provide power to the smart sensors. If the line current is lower than 5-10 Amps, the sensor may not have enough power to transmit data as well. Therefore, cellular communication strength and circuit loads provide critical information for operators to narrow down the potential root causes of the smart sensor communication loss issues. The Cellular Communication Signal Strength Distribution pane provides this information. Red dots represent smart sensors with either very low signal strength or very low circuit load, orange dots show moderate signal strength and circuit load, and green dots are the sensors with strong signal strength as well as large circuit load.
  • Smart Sensor Health Status Trend – Although real-time information is important to understand the smart sensors’ status live, it’s critical to learn the health trend of smart sensors as well. The Smart Sensor Health Status Trend pane provides a pattern showing whether the overall operations of the smart sensor are better or worse by week or day. Operators can choose the time range, substation, or circuit to learn more granular information.
  • Sensor Distribution by Substation and Sensor Distribution by Circuit – These panes help the operator learn the smart sensor deployment distribution information.
  • Smart Sensor List – This sensor detail pane provides comprehensive information of the smart sensors in a tabular view in case the operator wants to search or sort sensors by detail information.

With aggregated smart sensor data (geo location, cellular signal strength, distributed circuit power flow), operators can quickly identify problematic sensors and narrow down the possible root causes. This approach can save a significant amount of time performing sensor maintenance and troubleshooting—up to 90% or more.

In future posts in this series, we’ll show you how to use the paginated reports function to generate daily reports to improve the operational efficiency even more. The communication pane also shows the smart sensor distribution using a bar chart, and provides insights of smart sensor deployment information based on region, division, substation, and circuit.

Distribution network events dashboard

Smart sensors measure and provide the operational information of the distribution networks. This information is critical for operators to understand the circuit running status and the distribution of different events, such as permanent outages, momentary outages, line disturbance, or voltage sags and swells. QuickSight helps operators quickly configure different views, insights, and calculations on smart sensor information.

When an operator specifies a time range, QuickSight is able to provide smart sensor statistics on various metrics, such as the following:

  • Total number of events compared to a previous time frame
  • Distribution of events across selected regions, substations, or circuits
  • Distribution of events by region, substation, or circuit
  • Distribution of events by event type such as permanent or momentary faults

This information can help operators determine the areas or fault types of interest and study more detailed information. It can also help operators identify the substations or circuits with the most events and take proactive actions to fix any existing or hidden issues. The trend information can also be used to validate the equipment repair or circuit enhancement works.

Conclusion

Many utilities today are experiencing increased integration of distributed energy resources (DERs), such as solar photovoltaic, and power electronics loads such as variable speed drive and electric vehicle battery chargers. However, the existing grid wasn’t originally designed to coordinate these DERs, which can cause hidden issues on the existing networks. A large number of smart sensors are widely used to monitor the distribution networks to improve grid resiliency and stability.

In this post, we showed how QuickSight can help power utility distribution network engineers or operators to visualize smart sensor status in real time and troubleshoot smart sensor issues. We discussed out-of-the-box QuickSight features such as its rich suite of visualizations, analytical functions and calculations, in-memory data engine, and scalability, which will greatly reduce the time, cost, and effort of managing large number of smart sensors and fixing any problems early.

Smart sensors are the eyes and ears of utility distribution networks. With QuickSight BI functions, operators can quickly and easily create circuit event dashboards; search, sort, filter, and analyze different mission-critical events; and help engineers take early action when certain abnormalities occur on the distribution networks.

In the following posts in this series, we’ll show you how to use QuickSight to generate daily paginated reports and use advanced features such as natural language processing to conduct advanced search and analytics functions.


About the Authors

Bin Qiu is a Global Partner Solutions Architect focusing on ER&I at AWS. He has more than 20 years’ experience in the energy and power industries, designing, leading, and building different smart grid projects, such as distributed energy resources, microgrid, AI/ML implementation for resource optimization, IoT smart sensor application for equipment predictive maintenance, EV car and grid integration, and more. Bin is passionate about helping utilities achieve digital and sustainability transformations.

Steve Alexander is a Senior Manager, IT Products at PG&E. He leads product teams building wildfire prevention and risk mitigation data products. Recent work has been focused on integrating data from various sources including weather, asset data, sensors, and dynamic protective devices to improve situational awareness and decision-making. Steve has over 20 years of experience with data systems and cutting-edge IT research and development, and is passionate about applying creative thinking in technical domains.

Karthik Tharmarajan is a Senior Specialist Solutions Architect for Amazon QuickSight. Karthik has over 15 years of experience implementing enterprise business intelligence (BI) solutions and specializes in integration of BI solutions with business applications and enabling data-driven decisions.

Ranjan Banerji is a Principal Partner Solutions Architect at AWS focused on the power and utilities vertical. Ranjan has been at AWS for 5 years, first on the department of defense (DoD) team helping the branches of the DoD migrate and/or build new systems on AWS ensuring security and compliance requirements and now supporting the power and utilities team. Ranjan’s expertise ranges from server less architecture to security and compliance for regulated industries. Ranjan has over 25 years of experience building and designing systems for the DoD, federal agencies, energy, and financial industry.