Research on Building Energy Consumption Prediction based on BP Neural Network

Abstract

The construction department is an important energy consumption department in the contemporary society. Building energy conservation is an important means to promote the energy transformation and realize the strategic goal of “carbon peak and carbon neutrality”. Accurate building energy consumption prediction is the support technology to build building energy management system to achieve building energy saving. The building energy consumption is not only affected by the building physical structure, the actual function and use mode of the building also determine the actual energy consumption of the building, so the building energy consumption has significant volatility and uncertainty.Based on this paper, we designed a BP neural network for predicting building energy consumption with an mse of 0.0043 on the experimental dataset, and subsequently changed the loss function to a mish function based on the neural network, reducing the mae value to 1.03e-05. It shows that the prediction results of the energy consumption prediction model of BP neural network are slightly biased from the original data, and it performs well in the prediction performance to ensure the accuracy of the prediction.