Research on Parameter Prediction of Heating System in a Middle School Based on Optimized Sample Features

In order to optimize the parameters of the secondary network process of the heating system, a data collection system for heating based on the Internet of Things was built in a middle school in Tianjin. After the system was built, heating season data from 2020, 2021, and 2022 was collected. After analysis, it was found that there were significant differences between the daily flow, supply and return water temperatures of the secondary network of the heating system and the curve obtained by the formula. Afterwards, the water temperature and daily flow of the three heating seasons were optimized in MATLAB by combining outdoor meteorological data. Multiple days and multiple meteorological factors were used as sample features, and a random forest was used to build a prediction model. The importance of sample features was highlighted using visualization tools, and they were screened. Then, according to the CS optimizttion, a BP neural network was established to predict the supply and return water temperatures and daily flow using the screened sample features. The results showed that in the random forest prediction model, the correlation coefficients (R2) of supply and return water temperatures and daily flow were improved from 0.46, 0.4, and 0.55 to 0.70, 0.69, and 0.64, respectively. In the BP neural network prediction model, the determination coefficients (R) of supply and return water temperatures and daily flow were 0.76, 0.83, and 0.91, respectively, indicating that this sample feature optimizttion method can effectively improve the prediction accuracy.