The characteristics and present situation of heating load forecast were summarized, through analyzing and studying the various factors that affect the heating load. The processing approach for influencing factors was proposed by using quantitative of fuzzy data. On this basis, a new type of fuzzy neural network forecasting system was used. In the system was used, the sophisticated BP network method was used as the design core, and the quantitative parameters of fuzzy data for influencing factors were used as input values, and then the forecast network model was gotten. After the model operation parameters were determined, Matlab7.0 was used to simulate and predict. The results show that its accuracy can achieve ideal result under the condition of equal or approximate, and the relative error is under a reasonable range. Moreover, the hybrid algorithm rather than the simple fuzzy algorithm or neural network algorithm has a better prediction precision and stronger generalization ability, which improve the heating quality and save energy.