Using Machine Learning to Optimize Building Energy Consumption Prediction and Generate Energy-saving Strategies

With the increase of global energy consumption, optimizing building energy consumption has become a key step to achieve sustainable development. In this study, the long-term and short-term memory network (LSTM) model is used to predict building energy consumption and formulate energy-saving strategies. In order to reveal the influencing factors and study their correlation with energy consumption, this paper determines the key influencing variables based on the building energy consumption data set. In this paper, data preprocessing, including normalization and sequence processing, is carried out to ensure the quality of input data. Then the topological structure of the model is designed, including input layer, LSTM layer, full connection layer and output layer, and the model is trained and optimized. The experimental results show that the LSTM model has good performance in building energy consumption prediction. Compared with the traditional gradient propulsion machine (GBM) and support vector machine (SVM), the LSTM model has higher prediction accuracy and stability. When evaluating the energy-saving strategy, according to the suggestion of LSTM model, this paper implemented energy-saving measures for 30 buildings, with the highest energy-saving rate of 60% and the lowest energy-saving rate of 30.5%. The experimental data also support the advantages of LSTM model, which can improve the accuracy and stability of prediction and help to verify the energy-saving strategy formulated in this study.