Realtime Thermal Power Plant Feedwater Chemistry Control Using Artificial Intelligence

Abstract

This paper presents an enhanced, publication-ready version of the original work on AI-driven real?time control of boiler feedwater chemistry. Feedwater chemistry control is critical for ensuring reliable and efficient thermal power plant operation. Traditional practices rely heavily on fixed dosing, delayed manual actions, and limited parameter interactions. These methods often fall short under transient operating conditions. This paper introduces a robust AI framework capable of forecasting short?term chemistry variations and optimizing chemical dosing, resulting in improved stability, reduced chemical consumption, and proactive corrosion prevention.

Introduction

The text discusses the importance of boiler feedwater chemistry control in power plants and proposes an AI-based, real-time predictive control approach to overcome the limitations of conventional practices. Poor chemistry control can cause corrosion, deposits, and silica transport, leading to reliability, safety, and maintenance issues—problems that are amplified in modern plants operating under cycling conditions and variable makeup water quality.

Traditional chemistry control methods rely on periodic sampling, fixed chemical dosing, and operator judgment, resulting in slow responses to transients, limited insight into parameter interactions, and frequent excursions during startups, shutdowns, and load changes. These shortcomings increase the risk of equipment damage and operational instability.

The proposed AI-based system acts as an intelligent decision-support layer integrated with existing control systems. It uses real-time analyzer data and machine learning models to predict short-term changes in key chemistry parameters such as pH, conductivity, dissolved oxygen, and sodium. Supervised learning techniques, including Random Forest and Gradient Boosting, forecast near-term chemistry behavior and guide adaptive dosing decisions.

An AI-assisted dosing strategy replaces fixed chemical feed rates with incremental, predictive adjustments, reducing over- and under-dosing, improving responsiveness during load swings, and minimizing chemistry alarms. Performance evaluations show improved chemistry stability, reduced excursion severity, optimized chemical usage, and enhanced corrosion protection.

Overall, the approach represents a shift from reactive to predictive feedwater chemistry management, offering a more adaptive, reliable, and cost-effective solution for modern power plants, particularly those with frequent operational transients.

Conclusion

This paper presents an advanced AI?enabled strategy for real?time boiler feedwater chemistry control. Through predictive modeling and adaptive dosing, the framework enhances chemical stability, minimizes excursions, and supports proactive corrosion management. Future research will focus on real?plant validation and integrating the system directly into plant DCS for automated operation.

References

[1] Dooley, R.B., Flow?Accelerated Corrosion in Power Plants, PowerPlant Chemistry, 2016. [2] IAPWS, Technical Guidance Documents on Power Cycle Chemistry, 2019. [3] Adam, E.J., et al., Artificial Intelligence for Boiler Performance Optimization, Journal of Cleaner Production, 2023. [4] Lee, K.J., et al., Machine Learning Applications in Boiler Control Systems, Energies, 2024. [5] Applied Thermal Engineering, Feedwater Chemistry Monitoring and Failure Prevention, 2013.

Copyright

Copyright © 2026 Deepak Shankhdhar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.