Introduction
The Department of Energy (DOE) is at the forefront of integrating advanced AI and machine learning techniques to enhance operational efficiency, safety, and decision-making across its various sectors. According to Deloitte, “50% of oil and gas companies plan to increase investments in analytics, AI/machine learning (ML), automation, IoT, and cloud.” The following use cases highlight the DOE’s innovative applications of natural language processing (NLP), data analytics, and machine learning (ML) tools, demonstrating their significant contributions to information retrieval, environmental data analysis, and system resilience. These projects showcase the DOE’s commitment to leveraging cutting-edge technology to address complex challenges and improve outcomes in energy management, environmental protection, and IT system support.
Use Cases
-
1. Applications of Natural Language Processing and Similarity Measures for Similarity Ranking
This project by the Department of Energy’s EHSS involves developing applications of natural language processing (NLP) and similarity measures for advanced information retrieval and dataset searching. These applications have been used to search various datasets, including SQL databases, CSV files, and reports, and to estimate similarities between records within or across datasets. The similarity search has been successfully applied to the DOE COVID-19 Hotline questions and answers database, DOE annual site environmental reports, and other DOE data systems. The tool, as of October 2021, runs locally on a project basis or as a desktop application. Initial efforts to develop a web-based application were not completed due to a lack of user need and resources.
-
2. Data Analytics and Machine Learning (DAMaL) Tools for Analysis of Environment, Safety and Health (ES&H) data: Similarity Based Information Retrieval
The Data Analytics and Machine Learning (DAMaL) tools project by the Department of Energy’s EHSS focuses on similarity-based information retrieval using natural language processing (NLP) and cosine similarity. These tools leverage AI to enhance the efficiency of finding important records in DOE’s environment, safety, and health (ES&H) datasets, such as occurrence reporting, fire protection, lessons learned, and accident and injury reporting systems. The tool supports unrestricted text queries and offers NLP options like stemming or lemmatization. It aims to improve decision-making in job planning, hazard identification, and insights from operating experience and lessons learned. As of October 2021, the tool is developed and deployed on the DAMaL tools website, with ongoing efforts to maintain, document, improve, and expand data sources.
-
3. Data Analytics and Machine Learning (DAMaL) Tools to enhance the analysis of Environment, Safety and Health (ES&H) data: Classification, Robotic Process Automation and Data Visualization
The EHSS Data Analytics Machine Learning (DAMaL) tools include classification, robotic process automation, and data visualization capabilities. These tools use natural language processing (NLP) and classification algorithms like random forests to automate record classification, visualize trends, and highlight important and risky areas. By leveraging AI, the tool analyzes text from DOE’s environment, safety, and health (ES&H) datasets, such as occurrence reporting, fire protection, and accident reporting systems. It identifies key topics for analysts to explore potential safety issues in DOE operations. Deployed on the DAMaL tools website since October 2021, the tool is expected to be maintained, documented, improved, and expanded with more data sources.
-
4. Data Analytics and Machine Learning (DAMaL) Tools to enhance the analysis of Environment, Safety and Health (ES&H) data: Unsupervised Machine Learning Text Clustering
The EHSS Data Analytics Machine Learning (DAMaL) tools include an unsupervised machine learning clustering tool that uses natural language processing (NLP) and clustering algorithms like k-means, DBSCAN, and dimensionality reduction techniques. This tool leverages AI to analyze text from DOE’s environment, safety, and health (ES&H) datasets, such as occurrence reporting, fire protection, and accident reporting systems. It identifies recurring and important topics for analysts to explore potential safety issues in DOE operations. Partially deployed on the DAMaL tools website as of October 2021, the tool’s development is nearly complete, with a use case in Fire Protection Trending and Analysis under review. Ongoing efforts include maintenance, documentation, improvement, and expansion of data sources.
-
5. Groundwater Modeling
Groundwater modeling involves estimating various parameters to understand and predict groundwater behavior. This process is crucial for managing water resources, assessing environmental impacts, and planning sustainable water use. By accurately estimating parameters, groundwater models can provide valuable insights into the availability and quality of groundwater, helping to inform decision-making and policy development.
-
6. Soil Moisture Modeling
Soil moisture modeling uses multisource machine learning techniques to predict soil moisture levels within a lysimeter embedded in a disposal cell. This approach combines data from various sources to create accurate models of soil moisture dynamics. Understanding soil moisture is essential for managing agricultural practices, assessing environmental conditions, and planning land use. The use of advanced machine learning methods enhances the precision and reliability of soil moisture predictions, supporting better resource management and environmental protection.
-
7. AI-Based Chat Bot
The OCIO EITS Service Desk is exploring the use of AI chatbots to interact with end-users. The goal is to develop a single bot architecture that is highly tuned to IT system languages, ensuring it can handle enterprise-specific terms effectively. The primary benefit is to make knowledge more accessible to end-users in an easily consumable format. Additionally, the chatbot would connect to IT Service Management (ITSM) workflows to automate basic functions like account requests, permission grants, or creating MS Teams sites. The technology also needs to provide significant feedback to the EITS Service Desk on unanswered questions, dropped queries, ineffective responses, and incorrect answers, helping to improve the chatbot’s performance and user satisfaction.
-
8. Adaptive Cyber-Physical Resilience for Building Control Systems
This project uses deep learning models to predict the operation of building energy systems, detect and diagnose health states or cyber attack presence, and optimize the building energy system’s response. The goal is to ensure resilient operation and sustained energy efficiency. By leveraging advanced AI techniques, the project aims to enhance the reliability and security of building control systems, improving their ability to respond to various challenges and maintain optimal performance. This approach supports the development of smarter, more resilient, and energy-efficient buildings.