Abstract: In order to guide the stable and safe operation of power systems,a short-term load probabilistic forecasting approach based on conditional generative adversarial network （CGAN） curve generation is proposed.Firstly,an adaptive integrated forecasting model of key values of daily load based on bi-directional long short-term memory neural network is constructed by taking features of date,temperature and historical load as input.Secondly,the maximal information coefficient method is used to weight the load features,and the similar curve data set is constructed based on weighted K-nearest neighbor algorithm and weighted resampling.Then,taking the load key values and the similar curve data set as the conditions and training set,respectively,a load curve generation model based on CGAN is constructed.The numerical deviation and curve form deviation are proposed to correct the loss function.Finally,considering the uncertainty of the model and noise,the mapping from noise to the probability distribution of the model output is constructed,and the short-term load probabilistic forecasting is carried out.Taking the load data of a power grid in a certain region in East China as a case,it is verified that the proposed method has higher forecasting accuracy than the traditional methods.