Improved Utility Asset Management and Maintenance using AWS IoT and GenAI Technologies

Average worldwide household electricity use is expected to rise about 75% between 2021 and 2050 (ExxonMobil Report, 2024) . Electric Vehicles (EV) adoption is expected to drive 38% of the domestic electricity demand increase by 2035 (Ross Pomeroy – RealClear Science). In addition, Distributed Resources (DER) deployments, such as solar PhotoVoltaic (PV) systems, will increase infrastructure complexity for utilities. All of these factors could put a major strain on the utility electric grid.

Utilities are beginning to use smart sensor-based Internet of Things (IoT) technologies to monitor utility assets, such as electrical transformers. These sensors can also detect issues with power quality, and underlying transmission and distribution lines. To develop a sustainable and scalable IoT solution for utilities, it is critical to collect, manage, and process large volumes of data in a timely and secure manner. This data can then be analyzed to deliver meaningful insights using artificial intelligence (AI) and machine learning (ML) technologies, for instance generative AI (GenAI). This blog describes how to collect and analyze utility data with AWS services, such as AWS IoT Core, Amazon Kinesis Data Streaming, Amazon TimeSeries, and Amazon DynamoDB. We also use transformer monitoring as an example to illustrate an end-to-end data flow.

Current challenges in monitoring a transformer

Transformers play a vital role in residential power distribution by efficiently stepping down high voltage levels to safer and usable levels. They enable reliable and safe electricity supply to our homes, promoting energy efficiency and reducing power loss during transmission. Distribution transformers are designed and rated to perform at specific load and temperature ranges. When the internal operating temperature exceeds the specified ranges for extended periods of time, these transformers can be damaged and disrupt the electrical supply grid. This can also cause increased maintenance cost and customer frustration. Even worse, it could cause fires and endanger the surroundings.

The number of transformers scale with the size of the utility company and its service population. Major utilities can operate hundreds of thousands of transformers. To cover their service area, the transformers are distributed throughout their geographic regions. Maintaining and replacing transformers represents a major part of the utility’s operating budget and capital investment. It’s crucial to monitor the distribution transformers’ operating conditions, such as internal temperature and load. If an issue is detected, the solution must generate alarms in a timely manner.

However, monitoring a large number of distribution transformers is a complex task. AWS offers services to meet your business requirements. For small to medium-sized transformers with a limited number of measurement points, AWS IoT Core is a good option. For large and complex transformers, you can use AWS IoT SiteWise and AWS IoT TwinMaker to model and monitor the digital asset. Additionally, you can apply Machine Learning (ML) to analyze the data and detect potential behavioral issues for effective predictive maintenance.

Solution overview

The following diagram illustrates the proposed architecture for transformer temperature monitoring and analysis. It includes: data sensing and collection, transmission, data processing, storage, analysis, AI/ML, and data presentation.

Data sensing and collection: There are different transformers that have specific purpose, size, and capacities. These transformers require different sensors to measure data parameters, such as transformer temperature, ambient temperature, vibration, and load. These sensors must have a good balance between measurement precision, data collection cost, and battery life when applicable.

Sensor communication: Depending on the transformer, sensors can be installed in the substation, utility poles, and remote locations. It is important for transformer sensors to support diverse communication networks (multi-channel), including LoRaWAN, 4G/5G cellular, or even satellite communication. Communication can be facilitated by AWS services, such as AWS IoT Core for LoRaWAN and AWS IoT Core for Amazon Sidewalk.

Sensor data transmission: AWS IoT Core is a managed cloud service that allows users to use message queueing telemetry transport (MQTT) to securely connect, manage, and interact with transformer sensors. The AWS IoT Rules Engine processes incoming messages and can support connected devices to seamlessly interact with AWS services. It’s recommended to store raw data for auditing and subsequent analysis purposes. To achieve this, you can use Amazon Data Firehose to capture and load streaming data into an Amazon Simple Storage Service (Amazon S3) bucket.

Sensor data processing: When data arrives in AWS IoT Core, an AWS Lambda function preprocesses the message in near-real-time. This preprocess removes unwanted data, converts sensor readings to usable measurements, and formats the raw sensor data into a standard message. This standardized message is then sent to Amazon Kinesis Data Stream for further downstream processing through AWS Serverless services. This flow follows the AWS best practice outlined in the event- driven architecture model.

The following items provide examples of message processing:

• Near-real-time alerts: These alerts indicate that the transformer may be overheated or under certain abnormal conditions. Lambda identifies issues and generate alerts if the readings are outside a specific threshold. This notification is sent to Amazon Simple Notification Service (Amazon SNS). The Amazon SNS service issues email, or SMS messages to notify operators/engineers for human intervention. Based on the IEEE guidance model, the Lambda function compares the near real-time temperature measurements with the calculated values that are based on the transformer model, load, and ambient temperature. An alert is created when the transformer’s temperature is outside the expected parameters.
• Time series transformer sensor data storage: This data is processed by Lambda functions and stored into Amazon Timestream. Amazon Timestream is a purpose-built, managed time series database service that makes it easy to store and analyze billions of events per day. It’s designed specifically to solve time series use cases and has over 250 built-in functions using standard SQL queries, which eases the pain of writing, debugging, and maintaining thousands of lines of code.

User interaction through GenAI: GenAI through Amazon Bedrock can detect behavioral deviations in equipment and predict potential failures. GenAI can also generate multiple detailed reports, including identifying regions with a higher risk of fire or power outages. These predictions allow engineers and technicians to rapidly access technical information about transformers, and receive best practices for repair and maintenance. With these advanced analytics capabilities, the system can proactively address issues before they lead to service disruptions.

Dashboards and reports: AWS provides different services for you to view transformer time series or event data and data with a certain time period, such as overall trend and percentage of overheat. These services include Amazon Managed Grafana, Amazon Q in QuickSight, and Amazon Q. Amazon Managed Grafana is a fully managed service based on open-source Grafana that makes it easy for users to visualize and analyze operational data at scale. Amazon QuickSight is a business intelligence (BI) solution and Amazon Q provides new generative BI capabilities through executive summaries, natural language data exploration, and data storytelling.

Predictive maintenance: Capturing equipment failures as they happen is crucial. However, taking proactive measures to predict failures before they manifest is even more important. Proactive maintenance helps to minimize unplanned downtime and reduce maintenance costs. Amazon SageMaker helps to empower businesses to leverage ML and predictive analytics to monitor equipment health and detect anomalies. You can develop custom models or utilize existing ones from the AWS Marketplace to identify anomalies and promptly issue alerts.

Other services: The story does not end here, when an overheating transformer is identified, a work order can be created and issued to the SAP application. The repair/replacement ticket can then be created and tracked, and generative AI can create detailed steps to troubleshoot and complete the repair.

Conclusion

The growing demand for electricity and the increasing complexity of the power grid present significant challenges for utilities. However, AWS IoT and analytics services offer a comprehensive solution to address these challenges. By leveraging smart sensors, diverse communication networks, secure data pipelines, time series databases, and advanced analytics capabilities, utilities can effectively monitor asset health, predict potential failures, and take proactive measures to maintain grid reliability.

The architecture outlined in this blog demonstrates how utilities can collect, process, and analyze transformer data in near real-time, enabling them to rapidly identify issues, generate alerts, and inform maintenance decisions. The integration of generative AI further enhances the system’s capabilities, allowing for the generation of detailed reports, technical insights, and predictive maintenance recommendations. The same architecture can be used in for other industries that need to manage and monitor a complex and diverse network of assets.

As the electric grid evolves to accommodate growing electricity demand and distributed energy resources, including the growth of renewable energy sources like wind and solar, this AWS-powered solution can help utilities and stay ahead of the curve, optimizing asset management, improving operational efficiency, and ensuring a sustainable and reliable power supply for their customers. By embracing the power of IoT and AI/ML, utilities can transform their operations and better serve their communities in the years to come.