Abstract: According to the characteristics of architecture heat supply and the demands for energy-saving control, load forecasting based on the maximum entropy method is proposed. By using the least square fitting and the pretreatment approach, the historical random load series collected from the heat supply station are separated into the certain part and the random part. Then the load series is dealt with by the auto-correlation method and the maximum entropy method respectively. Comparing the results of these two methods shows that the load forecasting based on the maximum entropy theory can meet the demands for heating energy-saving control in terms of the forecasting accuracy, auto-adaptive and real-time ability better.

Key words: maximum entropy, random series, load forecasting, heat metering