The invention discloses a probabilistic load prediction system and method based on a Gaussian process quantile regression model. The method comprises the following steps: inputting a sample data set through an input unit; dividing the sample data set into a training set and a test set, and performing normalization processing on historical power load data and temperature data of the training set; performing correlation factor analysis on the historical power load data and the temperature data of the training set to construct a feature vector; inputting the feature vectors into a Gaussian process quantile regression prediction model; setting a quantile change interval of the model, wherein the model outputs a plurality of conditional quantile load prediction results; inputting the conditional quantile load prediction result into a kernel density estimation function for kernel density estimation to obtain a load probability density function; and calculating a load confidence interval anda load prediction estimation value through a load probability density function. According to the method, the confidence interval and probability density of power load prediction can be effectively constructed, and the accuracy of power load prediction is improved.