The invention discloses a heat supply load prediction method and system based on Box_cox transformation and UFCNN, and the method comprises the steps: determining a heat supply load prediction region,collecting the temperature and climate data and heat supply data in a prediction region, and constructing a sample database; carrying out the Box-Cox transformation of the input and output data of each sample in the sample database, and dividing the data after transformation into a training dataset and a verification dataset; constructing a UFCNN, inputting the data in the training dataset into the network for multiple training, obtaining a trained network, and carrying out the verification and testing of the learning effect for the verification dataset through the network; inputting to-be-predicted data into the UFCNN, and calculating a heat supply load prediction result with the same dimension as the Box-Cox transformation; carrying out the inverted Box-Cox transformation of the heat supply load prediction result with the same dimension as the Box-Cox transformation, and obtaining a final heat supply load prediction value. The method can achieve the correction of the skewed distribution of heat supply load data, enables the heat supply load data to accord with the normal distribution, and improves the data prediction accuracy.