Abstract
Introduction: Houston's wastewater infrastructure is vast and complex, serving over 2.3 million residents through 38 wastewater treatment plants, approximately 380 lift stations, 5,800 miles of gravity sewers, and 134,000 manholes. In addition to the operational challenges inherent in maintaining such a large system, Houston must also comply with strict regulatory requirements under a federal consent decree, which mandates specific actions to prevent sewer overflows and improve aged infrastructure for system reliability. While addressing the compliance of consent decree, the city is also preparing for the period beyond the consent decree, it is adopting a forward-thinking approach to infrastructure management that emphasizes operational efficiency, regulatory compliance, and resilience. To address these demands, Houston is applying advanced technologies such as data integration, real-time monitoring, and predictive analytics through the Advanced Infrastructure Analytics Platform (AIAP), a 'Platform of Platforms' built on AWS Data Lake. AIAP integrates multiple operational and planning systems to enhance data-driven decision-making and optimize the management of the city's wastewater infrastructure. This paper explores how Generative AI, with a specific focus on Retrieval-Augmented Generation using AWS Bedrock and Amazon Q, is transforming Houston's wastewater management, providing critical insights into operational performance, predictive maintenance, and long-term infrastructure planning. LS plays a vital role in areas where gravity alone cannot transfer wastewater effectively. LS pumps act as the heart of the system, maintaining flow and preventing backups. Any disruption can ripple through the system, impacting public health and safety. By applying predictive failure methods, we can better protect these essential components and the communities they serve. Objective: The primary goal of this study is to: - Determine the status of LS operations at Risk using SCADA data - Develop pump downtime risk profiles for each LS to serve as a baseline - Track and evaluate the risk profiles against the baselines - Suggest a predictive maintenance plan This study presents a data-driven method using SCADA pump status data to monitor and evaluate the performance and risk of LS. By detecting the early signs of potential failures, it enables proactive maintenance and necessary actions. The approach provides a structure risk analysis, helping to prioritize maintenance for inefficient LS, effectively focused on the most critical issues to prevent high-risk incidents. Data preparation/methodology: We conducted EDA on historical pump performance data and developed a risk analysis approach that categorizes lift station conditions into three levels (Normal, Medium, and High Risk) based on SCADA data, considering the number of pumps in each station and their downtime, with traffic light colors representing each level. This method pinpoints high-risk stations, forecasts future risks based on current conditions, and culminates in a risk matrix to evaluate the station's condition based on pump functionality. To streamline the approach, a modern data architecture was implemented, with AWS data lake features being used to collect, store, and process large datasets. Scalability and reliability in analytical workflows were ensured by this setup, which efficiently managed extensive data volumes. Results/Key Findings: By analyzing SCADA data on pump operational hours, downtime frequency and the total number of operating pumps at each LS, the system identifies how many pumps are functioning well and how many are not. To assess LS risk based on pump performance, a new method is established, using traffic light indicators to classify LS conditions into three categories: normal operation (Green), medium risk (Yellow), and high risk (Red). Historic data revealed that stations transitioning from Green to Yellow often progress to Red without prompt intervention, highlighting the potential for early risk forecasting. Our method was validated using Sanitary Sewer Overflow (SSO) data, which commonly occurred near lift stations with higher pumps downtime risks. Discussion/Implications: The findings of this study contribute significantly to the field of wastewater infrastructure management by offering a proactive, data-driven approach to LS health assessment. The predictive risk model facilitates timely interventions, reducing the likelihood of pump failures and system overflows. This approach also enables better resource allocation by prioritizing high-risk LS for further assessment or renewal. Conclusion: In conclusion, this study introduced a novel approach to assess the health of LS and predict early warning signs of risk using SCADA pump status data, enabling intervention before the stations reach the danger zone. To achieve this goal, EDA techniques were applied alongside AWS data lake capabilities. The study enables proactive maintenance strategies by identifying underperforming LS, while systematically maintaining and upgrading assets to promote long-term sustainability.
This paper was presented at the WEF/WEAT Collection Systems and Stormwater Conference, July 15-18, 2025.
Author(s)Maymandi, Nahal, Rabbi, Fazle, Pradhan, Pratistha, Islam, Jinia
Author(s)N. Maymandi1, F. Rabbi2, P. Pradhan3, J. Islam2
Author affiliation(s)IMS Engineers, 1City of Houston, 2PNA Technical Services, 3City of Houston, 2 ,
SourceProceedings of the Water Environment Federation
Document typeConference Paper
Print publication date Jul 2025
DOI10.2175/193864718825159863
Volume / Issue
Content sourceCollection Systems and Stormwater Conference
Copyright2025
Word count17