Building Energy Consumption Forecasting based on K-Shape clustering and CNN-LSTM

Abstract

The management and dispatch of building energy consumption plays a significant role in ensuring energy dispatch, reducing energy consumption, and alleviating the challenges of climate change. Accurate building energy forecasting is the basis for establishing an energy management system. To improve the accuracy of building energy consumption forecasting, this paper proposes ultra-short-term building energy consumption forecasting based on K-Shape clustering and convolution and a long short-term memory network (CNN-LSTM). First, the clustering algorithm is used to cluster the energy consumption of single apartment buildings into multiple apartment building clusters to obtain the energy consumption component features; then a multi-feature time series data set consisting of clustered data of building energy consumption component features and temporal and climatic information is constructed. The data are then fed into a CNN network based on a one-dimensional convolutional layer to extract the nonlinear relationships between different feature variables; the data are then fed into an LSTM network to extract the time-series characteristics of the data in the time dimension. And finally the short-term power consumption forecasting results are output by the fully connected layer. The experimental results show that the proposed method has higher forecasting accuracy compared with other methods.