An Intelligent Dual Optimization Approach for Improved Load Following of Supercritical Power Unit Based on Condensate Throttling

Abstract

In modern real-time unit load scheduling, it is unavoidable for large-capacity supercritical (SC) units to participate in peak-load regulation with automatic generation control (AGC), which raises the requirements for load control of a large SC power unit. With only the traditional coordinated control system, it is easy to fail in meeting the requirements of rapid load-changing and cause large fluctuations for main steam pressure and temperature. Condensate throttling technique, as an emerging new technology for rapid adjustment of unit load, has attracted much attention in recent years. This paper develops a neural network based unit load prediction model which takes condensate throttling into account for a 600 MW SC power unit. An intelligent dual optimization approach is then developed, which uses the load prediction model twice, first to optimize the deaerator valve opening during the initial load-changing stage for fast load following, and then to optimize the turbine valve opening during the later condensate flow recovery phase to keep the load-following accuracy. Simulation tests show that the proposed approach can greatly improve the unit load dynamic response in speed and control accuracy, thus effectively improve the unit load adaptability to AGC.