Office Building Energy Consumption Prediction Based on LSTM and SVR

According to the actual operating energy consumption of office buildings, this paper proposes an energy consumption prediction method suitable for office building operations in order to efficiently save energy. To achieve more reasonable and accurate predicting results, this method separates prediction into short-term forecast and long-term prediction. Through analysis and selection of energy consumption predictors, we establish an LSTM neural network model for short-term energy consumption prediction and a SVR algorithm based on support vector machines for longterm energy consumption predictions. The prediction results support decision making in building energy-saving operation.