Abstract: It is proved that the even order Gaussian wavelet function satisfies the condition of translation-invariant kernel function of support vector machine(SVM).The Gaussian wavelet SVM is constructed with wavelet kernel function,and cloud theory-based genetic algorithm(CGA) is used to optimize the parameters of SVM and its kernel function.The experiments are conducted using the proposed method,the conventional Gaussian kernel SVM and Morlet wavelet kernel SVM respectively. It is shown that the proposed method is efficient and superior by nonlinear function approximation and short-term load forecasting of power system,and has some practical value.