关键词: 过热器温差, 偏互信息(PMI), 支持向量回归(SVR), 数据驱动, 建模

Abstract: In order to reduce the temperature deviation between different superheater areas in large coal-fired boiler, and to improve the operating safety and stability, a data-driven modeling method PMI-SVR is proposed to describe the superheater temperature deviation. The main factors that affect the superheater temperature deviation are chosen from many on field operation data based on the partial mutual information (PMI) criteria. Then the data driven temperature deviation is modeled using support vector regression (SVR) algorithm. The influences of the algorithm parameters of PMI-SVR are discussed in detail to obtain the optimal model. Simulation results on a 350 MW unit show that the feature selection method based on PMI can effectively obtain the main factors that affect the temperature deviation of the superheater. Based on these variables, the data driven model drawn from the data has high accuracy, and it will be useful for further temperature deviation control.

Key words: Superheater temperature deviation, partial mutual information (PMI), support vector regression (SVR), Data driven, modeling