Iberdrola reduces incidents at power distribution facilities using AWS IoT and Edge services

Iberdrola is a global leader in electricity generation, distribution, and commercialization, pioneering wind power and renewable energies. i-DE is Iberdrola’s entity responsible for the electric distribution grid in Spain and Portugal, accountable for millions of operational assets with a very heterogenous nature, which makes it challenging to have a live and consolidated view of each asset health and its correlation to the overall distribution performance. In this post we discuss SmartPoint, an assets management platform that the i-DE innovation team has developed to monitor the health of components such as line supports/poles, transformation centers, substations, lines themselves, and field workers.

i-DE business challenges

As owners of Iberia’s power distribution grid, i-DE has faced diverse challenges that have been addressed by incorporating technological solutions to better manage different assets types, such as the following:

Substations are key components of the electric power system that transform voltage and make sure of reliable power distribution. Given that the assets in the facilities are exposed to the open environment, video analysis is required to detect physical threats and anomalies. Achieving line of sight to the assets with fixed cameras requires a vast amount of supporting infrastructure and leaves blind spots.

A transformer station is a facility that steps down high-voltage electricity from transmission lines to lower voltages suitable for distribution. Data telemetry from legacy sensors is the key requirement for this type of installation to monitor a wide range of environmental variables: smoke, intrusion, temperature, flooding, etc.

Transmission and distribution lines are the critical infrastructure for delivering electricity. They require the incorporation of video analysis to detect fires, and use satellite images to detect potential risks derived from vegetation.

Line supports and utility poles are the physical structures that hold and support the electrical lines. For these assets, modern Internet-of-Things (IoT) sensors are incorporated to measure different parameters, such as slope changes derived from landslides. Additionally, thermal IoT cameras are deployed to detect fires as soon as they appear.

Field workers are responsible for maintaining and repairing the electrical grid infrastructure. With the aim of securing fieldwork, SmartPoint enables the usage of sensors for fall detection, geofencing, position control in substations, smart keys, etc.

In summary, i-DE needs a platform to centrally collect data from such a broad diversity of assets, detect anomalies, report incidents, and remediate by collaborating with on-site personnel.

Functional requirements

i-DE was looking for a solution that can connect to both legacy and modern IoT sensors, as well as derive insights from video to detect and act upon anomalies. Through a web app, Iberdrola’s operators need to receive alerts about the events that take place in the substations. In turn, the platform should help them correlate and cross-analyze the data from the systems that are related to the event, and so an incident is reported for on-site operators to solve the problem.

Image 1. SmartPoint alerts dashboard

Therefore, the platform should display data in two different ways: first in the form of an alert of events that occur in real-time, but also in charts that should be refreshable on-demand to reflect the current assets state. In addition, there are KPIs on the performance of the substation that, in order to be analyzed and visualized, entail the correlation between multiple elements of the installation.

Image 2. Visual discrepancy detection

Image 3. Inclinometer chart

To achieve this, the Iberdrola and AWS Solutions architects defined the following design principles to rule the solution:

• Agnostic and standardized device connectivity: every new use case must be able to plug into SmartPoint seamlessly. For this, SmartPoint needs to support MQTT and other connectors for industry protocols that Iberdrola can list to manufacturers at acquisition time and/or share with equipment development teams.
• Purpose-built data repositories: SmartPoint needs to connect with the substation’s control and telemetry systems using different protocols, and then route the data and events to the most appropriate data repositories in the cloud, which can be tackled for specific data-driven use cases. There should be databases for IoT use cases that SmartPoint can query in real-time, and also a central repository of data to query with business intelligence and data science purposes without impacting transactional databases. This decoupling would make the platform support analytics use cases economically, while performing in real-time by using production data bases only when necessary.
• Scalability: using only serverless services in the AWS Cloud would reduce time-to-market, lower the operational load, and lower the total cost of the platform. Serverless is a good strategy to scale cost-effectively, because it allows for the elimination of infrastructure management complexity, while guaranteeing quality of service in critical moments.
• Intuitiveness: this allows the UX/UI team to work independently from the backend and data teams, following best practices in design and usability so the platform can be exploited at its maximum potential by Iberdrola’s operators. SmartPoint’s frontend hosting should be fully decoupled from the backend infrastructure, enabling the front developers to deploy changes autonomously to test usability and iterate faster toward a more intuitive experience.
• Extensibility and future proofness: the platform should allow for easy integration and enable new workflows such as digital twins, virtual and mixed reality, generative AI, automated building information modelling (BIM), quantum computing, and other capabilities.

Solution overview

With these design principles defined, Iberdrola created SmartPoint, which can connect to a diversity of field equipment and sensors using the AWS IoT stack. Then, it can apply streaming analytics to act based on the ingested data in real-time and persist it on purpose-built data repositories for further analysis. This enables business intelligence and data science capabilities. SmartPoint’s data pipeline uses AWS serverless services to guarantee scalability and performance of the system while devices continue to be added. Due to the nature and number of installations, edge computing capability is used to capture the unit data from the sensors and make on-site decisions when necessary. These edge gateways need connectivity with SmartPoint for both northbound telemetry and southbound management of business rules, updates, etc.

Image 4. SmartPoint’s architecture diagram

The platform architecture was designed to use the same connectivity and data processing pipeline for all types of assets:

1. Subscription on MQTT enabled sensors to publish messages to the AWS IoT Core broker.

2. Sync logic coded in a scheduled and scalable AWS Lambda reads from legacy sensors that don’t support MQTT and inject the readings on the AWS IoT Core to have a consolidated view on all of the devices.

3. Machine learning (ML) video inference running on autonomous robots was integrated by a startup partner, Star Robotics. This enables i-DE to detect anomalies, publish events to MQTT broker, and upload objects (images and video) to Amazon Simple Storage Service (S3), an AWS object storage service with industry-leading scalability, data availability, security, and performance. AWS IoT Greengrass, an open-source edge runtime and cloud service for building, deploying, and managing device software, is used with over-the-air deployment (OTA) for the video inference and the autonomous navigation runtime processes in the robot. See the next section for more details.

4. Rules route topic messages to purpose-built data persistence services.

5. Raw events are stored in an S3 bucket for historical storage, analytics, and ML usage.

6. AWS Glue performs extract, transform, and load (ETL) jobs over the raw events and makes them suitable to be queried by Amazon Athena in a serverless fashion, be queried directly on Amazon S3 using SQL ad hoc, and feed Amazon QuickSight BI dashboards. This stack enables the correlation of multiple systems and establishes benchmarks for the overall performance of the distribution facilities, supporting business decision making. Moreover, this is done without impacting transactional databases that are dedicated to real-time production use cases.

7. Amazon IoT Events ingests data from multiple sources to detect the state of processes and devices. It is used to trigger alarms for failures or changes in operations, visualize performance and quality of operations, and recognize more complex patterns by correlating longer time-windows.

8. The frontend is built with Angular, deployed to an S3 bucket, and served through an Amazon CloudFront content delivery network.

9. The backend is defined as a serverless application with persistence in Amazon DynamoDB, a purpose-built key-value pair database, so that events are available for the user application in real-time. And there is persistence in Amazon Timestream, a managed time series database that enables SQL to query the sensor data with low latency and using rolling time windows. APIs to interact with the backend are hosted in a serverless fashion by Amazon API Gateway and secure authentication is enabled by Amazon Cognito.

Video inference at the edge

Star Robotics is a Spanish startup offering modular technologies in mechanics, electronics, autonomous navigation, and artificial intelligence (AI) to configure tailor-made solutions for different use cases and industries. SmartPoint uses a two-stage detection solution for intelligent surveillance and inspection tasks enabled by Star Robotics. The initial inference stage happens at the edge using Nvidia Edge tools and AWS IoT Greengrass with OTA of models trained in the cloud using Amazon SageMaker, as well as the AWS services umbrella to build, train, and deploy ML models for use cases with fully managed infrastructure, tools, and workflow.

This approach allows for immediate inference on captured data from the 360 degree camera module mounted in the robots, making sure of prompt and localized decision-making.

Image 5. 360 degree view through robot

Once an anomaly is detected, alerts are generated and sent northbound through AWS IoT Core, so that IoT Rules process the messages, storing them in DynamoDB and Timestream. Simultaneously, when an event occurs, the pan-tilt-zoom camera (PTZ, which can pan horizontally, tilt vertically, and zoom), aims at the target and captures several images, uploading them to an S3 media bucket. The second inference stage in the cloud is triggered and results are sent to the application backend for user review.

Image 6. People detection with robot

By implementing this dual-tiered inference pipeline, the platform makes sure of both the agility of edge computing and the depth of cloud-based analysis. Additionally, this approach allows for incremental model training, incorporating user feedback and false positive (FP) feedback from previous alerts to refine the system’s accuracy over time.

Results and benefit

SmartPoint deployment in production enabled a 40% reduction in reported incidents at distribution facilities, with the most significant impact on i-DE business operations in the following categories:

• Reduced to 50% of the number of trips to facilities, whenever an incident occurs.
• 30% lower unauthorized access to risky areas by using geofencing and location tracking solutions.
• Close to zero unauthorized access to facilities with the implementation of smart keys.
• 50% reduction of thermal failures in the primary substations due to the early detection of environmental threats with daily inspections using on-site cameras and/or robots.
• Reduced power supply lockdown times in case of emergencies, due to the deployment of fire detectors, fall detectors, and other sensors.

Additional resources

If you want to know more, please explore AWS Solutions for Energy to drive operational efficiency and sustainability. Also discover AWS IoT and Edge Computing to use real-time data and predictive analytics. Contact our AWS Energy Experts to take the first step toward a smarter, more efficient energy future.