An Investigation Into the Level of Fidelity Required to Model Heat Exchangers of a Heat Recovery Steam Generator

Process modelling has traditionally been used in the design and development of Combined Cycle Gas Turbine power stations. In recent years it has been applied across the complete process lifecycle, from initial concept testing, through design, development, and operation. In the operation of these plants the process model may be used for offline or online applications. Offline refers to the model being solved in isolation from the plant. Such studies typically look at optimising the operation of the plant, or investigating the effect of plant modifications. Online operation means the process model is run in parallel with the actual plant, receiving measured data from the plant information system. In this role the model can be used to monitor plant performance, allowing degradation of equipment to be tracked, and maintenance scheduled accordingly. In addition, model estimations can be used to reconcile sensor data, improving knowledge about the plant and aiding troubleshooting. Combined Cycle Gas Turbine (CCGT) power stations consist of a gas turbine and steam cycle, linked by a heat recovery steam generator (HRSG). The HRSG extracts heat from the gas turbines exhaust gases by boiling water for the steam cycle. This facilitated by a series of heat exchangers within the HRSG. Given the central role of heat exchangers within CCGT power stations, the performance and accuracy of heat exchanger models in the process model is of highest importance. A review of the current literature revealed two competing methods for modelling the variation in the overall heat transfer coefficient of a heat exchanger. The first utilises semi-empirical correlations developed specifically for a given heat transfer problem. The alternative method estimates the variation in the overall heat transfer coefficient with fluid conditions, by applying weights to a nominal value, usually chosen to be the design value for overall heat transfer coefficient. The weights are determined using the ratio of design condition fluid properties to those at the fluid conditions of interest, raised to some constant power. This constant is chosen based on the particular heat transfer problem. A final possibility is to assume the value of the overall heat transfer coefficient does not vary significantly from design conditions and maintain this value constant. These three methods for estimating the overall heat transfer coefficient represent three levels of modelling fidelity which can be used in heat exchanger models. To determine the level of fidelity required to for modelling HRSG operation, each method was implemented into a heat exchanger model designated the Correlation, Weighted, and Basic heat exchangers. These were then incorporated into a model of the HRSG in the Otahuhu B CCGT power station located in Auckland, New Zealand. The model was then used to simulate the operation of the HRSG at four different load points, spanning the operating range of the plant. Results from the model were then compared to measured plant data. Overall the Weighted heat exchanger model was found to be the most accurate in following the variation in heat exchanger behaviour in the actual plant, as the load was varied. The Correlation model accurately followed trends in the measure plant data, however, the results were often offset from the data due to an inaccurate estimate of the initial value for overall heat transfer coefficient. The Basic heat exchanger model was the least accurate, and was unable to follow trends in the measure plant data as the load varied. In addition the Correlation model, in general, took over ten times longer to solve than either of the other models