Hot Corrosion Monitoring in Boilers Using a Nonlinear Estimator and Electrochemical Noise-Based Corrosion Sensors

Abstract

A kinetic model is developed to calculate the thickness of the corrosion scale formed due to hot corrosion in the waterwall section of a boiler in coal-fired power plants. The proposed model is validated by using the data collected from an operating industrial boiler and using a novel electrochemical noisebased corrosion sensor. As corrosion dynamics are significantly slow compared to the dynamics of temperature and gaseous species concentration, a multiscale model is developed where datadriven models are developed for algebraic states that affect the corrosion rate. The novel corrosion sensors can be costly and may be infeasible to place at all locations, especially due to the high temperature and harsh operating conditions in the waterwall section. This paper investigates opportunities for corrosion estimation by placing sensors for the algebraic state variables, the measurement of which is comparatively more mature and considerably cheaper than that of the corrosion sensors. To this end, a nonlinear estimator is investigated. Unlike many systems that are represented by differential-only or differential algebraic equations, in this particular system, differential equations do not affect the algebraic states but the algebraic states affect the differential states. For nonlinear state estimation of this specific class of differential algebraic equation system, a modified unscented Kalman filter algorithm is developed. The performance of the estimation algorithm is evaluated with different configurations of the sensor network by simulating the boiler with various operating conditions, as expected during load-following operation.