Prediction, comparison and analysis of comfort level and energy consumption of a passively remodeled dwelling based on BP neural network computation

Abstract

Listen

Translate article

Many ordinary dwellings are not well built in the vast rural areas of China, with poor comfort level and high energy consumption. The China Resources Charity Foundation funded the renovation of more than 200 farmhouses. Compared with the BP neural network model established with the measured data, the BP neural network model established with the simulated data can more quickly estimate the energy saved by the renovation measures of rural dwellings and the improved indoor environment, which is very necessary for the government and enterprises to make decisions to help. In this paper, one of the renovated houses (No. 285) was selected for tracking measurement before and after renovation to obtain real indoor environment and energy consumption data. Simulation software (DesignBuilder) was used to simulate the other 36 renovated houses in order to train the BP neural network model. The energy consumption of No. 285 was calculated by using the trained BP neural model, and the calculated results were compared and analyzed with the simulated data and measured data. Results show that the BP neural network model after training is better than the simple energy consumption simulation software in energy consumption prediction, but there are still errors from the actual measurement values. Regardless of software predictions or actual measurements, the results show that the renovation measures have indeed exerted a very positive impact. As to the overall trend, the software predictions are the same as of the actual measurements, and the variance of day and night temperature variance and the annual extreme values are greatly reduced.