This post is co-written with Imran Rafique at Deloitte Consulting and Charlie Zha at PG&E.
Amazon Quick Sight is a serverless, fully managed business intelligence (BI) service that enables data-driven decision-making at scale. Quick Sight meets diverse analytic needs with modern interactive dashboards. In Part 1 and Part 2 of this series, we showed how Quick Sight can improve power utility operational efficiency and how to create, schedule, and share highly formatted multi-page reports based on different dashboard requirements. In this post, we focus on implementing generative artificial intelligence (AI) in Quick Sight and how generative AI can help operators quickly analyze and identify circuit faults to improve power utility operational efficiency.
The introduction of generative AI has significantly impacted the global landscape, demonstrating an impressive ability to analyze data and produce innovative outcomes. Generative AI technology advances resourcefulness, introduces novel approaches to problem-solving, and enhances productivity. Goldman Sachs has estimated that generative AI could contribute to a 7% increase in global GDP, amounting to nearly $7 trillion, and potentially boost productivity growth by 1.5 percentage points over a decade. Additionally, generative AI enhances the capabilities of existing AI systems and technologies, driving transformation across a wide range of industries and enabling businesses to achieve new heights in productivity.
In April 2024, AWS announced the general availability of the generative BI capabilities of Amazon Q in Quick Sight. These capabilities combine Amazon Bedrock large language models (LLMs) with existing Amazon Q in Quick Sight features, such as natural language understanding and visual insights. The new functionalities create new, user-friendly interactive experiences with reliable analytics delivered through the Quick Sight analytics engine. The enhancement enables Quick Sight to generate visualizations through natural language prompts, allowing business analysts to swiftly produce maps, charts, and dashboards from conversational queries.
The introduction of generative BI capabilities within Quick Sight empowers business analysts to run routine tasks through natural language, including:
- Rapidly generate dashboards using enhanced visual authoring capabilities
- Adjust and format visualizations through simple natural language commands
- Effortlessly calculate metrics with natural language, eliminating the need for knowledge of functions in Quick Sight
- Present intelligent insights and share findings with the data storytelling feature
- Incorporate security, governance, and responsible AI with generative BI automation controls
Current challenges in power utility operations
Power systems have three primary components: generation, transmission, and distribution. Generation produces electricity from coal, natural gas, nuclear energy, and renewable sources like hydroelectric, solar, and wind power. The transmission system transports electricity from power plants to distribution substations, and the distribution system delivers electricity to end-users. As advanced technologies elevate power demand, maintaining uninterrupted power supply becomes crucial, though complete continuity in power transmission is challenging.
Within distribution networks, outages stem from various faults. Momentary faults, often due to temporary short circuits, and permanent faults, resulting in prolonged power interruptions, are common, along with line disturbances—abnormalities or disruptions within distribution lines. Prompt detection and accurate diagnosis of these faults can improve energy supply reliability, stability, and quality. For instance, analyzing patterns of momentary faults can identify potential issues like vegetation encroachment, where tree branches close to power lines may cause brief outages that, under strong winds, could escalate into permanent faults, leading to extended service disruptions.
Furthermore, business users frequently extract data and insights from outage analysis dashboards to prepare presentations for stakeholders to take predictive maintenance or repair tasks. This process typically involves copying and pasting outage visuals into various documents for external sharing beyond their BI system. This practice poses challenges, because governance protocols aren’t upheld outside the BI system, leading to rapidly outdated insights.
Solution overview
Amazon Q in Quick Sight greatly helps engineers and operators address these challenges. In addition to the existing dashboards built by developers, at any time, operators can use natural language queries to pinpoint the issues in certain circuit segments for certain faults within certain time frames. Operators can also combine load information in that circuit and study the potential cohesion between load pattern and fault pattern. Operators just need to focus on the business logic and can solve the problems in real time, without the need to learn how to build up dashboards or request the development team to build new dashboards. For critical insights and findings, the executive summary function can help operators generate reports and share with other teams or leadership to quickly identify and solve power system issues.
Amazon Q brings three experiences to Quick Sight customers:
- AI-answered questions of data inform business decisions for business users when and where they come up
- AI-assisted data storytelling enables business users to discover and share findings from their data to drive team decisions
- AI-accelerated dashboard authoring helps business analysts quickly build dashboards and reports
In this post, we mainly focus on how utility operators (non-technical users) can take advantage of these new capabilities to quickly get their business insights, drive business outcome, and create compelling data stories using generative BI capabilities within Quick Sight using Amazon Q.
Executive summary capability
Executive summaries allow users to generate a high-level overview of the key insights from a dashboard without having to analyze individual visualizations. As shown in the following video, utility operators can now generate executive summaries of their dashboards highlighting key operational patterns. This is helpful for concisely sharing findings with executives and stakeholders.
Storytelling capability
The new storytelling capability in Quick Sight helps users put together their findings and share insights through compelling narratives. Stories combine data-driven insights, real-world expertise, and AI—all framed in engaging design. Persuasive data stories can help teams reach conclusions and drive business decisions faster.
Stories in Quick Sight generate cohesive, powerful, and insightful narratives by analyzing data with only a few words. Users can enter the story description in natural language and choose the visuals they want the narrative to use, and Quick Sight will generate a story quickly. In addition, users can control AI verbosity, customize narrative text, and apply stunning visual themes to bring content to life. Furthermore, users also have complete flexibility to create stories from scratch or modify stories with point-and-click options, enabling them to quickly and effortlessly share and update data any time.
In the following video, we use power utility distribution line event data captured by line sensors to show the Quick Sight storytelling capability. Based on the provided natural language prompt, Quick Sight analyzes event data, identifies the event spikes across the distribution network, and weaves a compelling story about circuit events, supported by visuals and their high-level summary. We run event data to understand the root causes and mitigate further issues. In the real world, distribution line events can be combined with the loads on these lines to identify abnormalities for predictive maintenance use cases.
Natural language query capability
Traditionally, BI engineers or analysts create dashboards based on business needs. However, business users might have further questions that the dashboard creator might not have predicted. For instance, business users reviewing a dashboard that outlines substation outage events may want to know the outage distribution per circuit in the substation, or compare the outage event trend to last week, last month, or a specific time frame. By using the natural language processing (NLP) capabilities of Amazon Q in Quick Sight, business users don’t need to learn how to create a dashboard or write complex queries. Instead, users type a question using natural language, such as “Show event distribution per circuit in substation ABC” or “Show event distribution in substation ABC compared to the same time last year.”
Amazon Q in Quick Sight will provide multi-visual answers with narrative insight summaries that explain the answer context. Additionally, Amazon Q in Quick Sight supports vague questions and provides users with “did you mean” alternatives that enable iterative factfinding. Users can get started with AI generated questions that show what can be asked based on the content in the dataset. End-users are enabled to quickly get insights out of data that facilitates data-driven decision-making on the fly.
Conclusion
Integrating advanced renewable technologies like microgrids, Distributed Energy Resources (DERs), and EV charging infrastructure introduces increased complexity to the power grid, heightening its susceptibility to disruptions. The growing frequency and severity of natural disasters—such as wildfires, exacerbated by climate change—further contribute to power grid vulnerabilities, leading to more frequent service disruptions. Promptly interpreting power outage data and disturbances and identifying key insights and actionable recommendations is crucial for engineers tasked with diagnosing grid issues and enhancing grid resilience.
This post highlighted how Quick Sight uses generative AI technologies to assist network engineers and operators. Quick Sight empowers professionals to derive meaningful insights through NLP and its integrated generative AI capabilities by utilizing compelling narratives, generating executive summaries, and offering predictive maintenance advice.
If you’re interested in further exploring Quick Sight and its generative AI functions, check out the Quick Sight Community. Explore the community, ask questions, learn, and share knowledge with peers, including access to additional AWS resources.
About the Authors
Karthik Tharmarajan is a Senior Analytics Specialist Solutions Architect with extensive experience in enterprise data analytics and business intelligence (BI). He specializes in seamlessly integrating BI solutions with business applications, enabling data-driven decision-making. Karthik’s deep understanding of data analytics and BI empowers organizations to leverage their data assets effectively, driving strategic insights and informed decision-making processes.
This is a companion discussion topic for the original entry at https://aws.amazon.com/blogs/business-intelligence/improve-power-utility-operational-efficiency-using-smart-sensor-data-and-amazon-quicksight-part-3-generative-ai-assisted-data-storytelling-and-executive-summary/