Abstract: In view of the difficulty in modeling the mechanism caused by the complex and strong coupling nonlinearities between the multiple variables in the limiting conditions of leading phase operation, a novel method is proposed in this paper to optimize the leading phase operation limit of generator based on data-driven support vector machine regression (SVR). The limit calculation of generator leading phase operation is converted to the minimization of reactive power subject to the multiple constraints of leading phase. Based on the generator power angle equation, the objective function equation of reactive power is established. In order to formulate the constraint equation model, the nonlinear mapping relationship between constraint variables and independent variables in the objective function is constructed based on SVR data-driven model. The improved second-order oscillation particle swarm optimization algorithm is then applied to solve the optimization model. The case studies show that in addition to its modelling simplicity, the proposed method has exhibited high accuracy and strong adaptability, for the purpose of fast calculation of the generator leading phase limit under the conditions of any given active power output. Therefore it can be used for the online modeling and monitoring of the margin of generator leading phase operation.

Key words: support vector machine, data driven, leading phase limit, optimization, second-order oscillation particle swarm