Study on Boiler Combustion Optimization Based on Sparse Least Squares Support Vector Machine

Under the background of seeking high efficiency and low nitrogen oxides (NOx) emissions for the boiler of power plants, this paper used least squares support vector machine (LSSVM) to model the boiler efficiency and NOx emissions of a power plant according to the experimental data acquired from a combustion adjustment test. A pruning algorithm based on active learning was applied to the combustion model built earlier to obtain a sparse LSSVM model. Compared to Suykens standard pruning algorithm for LSSVM, AL-LSSVM (active learning LSSVM) can significantly reduce the complexity of combustion models without degrading much, which provides an effective method for incremental or adaptive learning of combustion models.