The invention discloses a multivariable secondary arc characteristic prediction method for single-phase reclosing, and the method comprises the steps: determining input and output sample sets; solving a support vector machine linear regression function through employing a structural risk minimization principle; building a nonlinear regression support vector machine regression prediction model; determining a learning parameter of the nonlinear regression support vector machine regression prediction model and a fitness function for evaluating support vector machine regression, and employing a quick non-dominated ordering genetic algorithm to optimize the learning parameter; and predicting the output of a UHV power transmission system with hybrid reactive compensation through employing the support vector machine regression prediction model with the optimized parameter. The beneficial effects of the invention are that the relative error between a prediction value obtained by a regression support vector machine with the optimized parameters and a result obtained through simulation above is within 2%, thereby indicating the effectiveness of the method.