The types of buildings which need the heating supply are complex in universities

and there is no effective method for forecasting the heating load in a short term. All these result in a waste of energy. In this paper

the BP neural network model was established in a 2-10-1 structure based on historical data of 2015-2016 heating load in a university which is located in cold regions. The main influential factors were the highest and lowest temperature. The results showed that the training

verifying and testing accuracy of BP neural network were respectively 0.048

0.054 and 0.096

and R was 0.975 5. The method could be used for forecasting short-term load in college. At the same time it could also provide a scientific mean to solve the problem of the imbalance between energy supply and demand.