Short-term building load forecasting based on similar day selection and LSTM network

In order to improve the accuracy and efficiency of traditional short-term building load forecasting, this paper proposes a new load forecasting method that is based on similar day selection and LSTM network model. The algorithm is divided into two parts. The first part is to establish similar day sets and complete pattern matching. The purpose is to select and optimize the input variables of the prediction model for improving the prediction efficiency. For a start, the similarity index ( the human comfort index) of the historical data and the forecast day is calculated, and then K-Means clustering algorithm based on particle swarm optimization is made to establish similar day sets. The next, the forecast days are taken to perform the similar day matching according to the Euclidean distance method. The second part is to establish the LSTM prediction model, train the variables that are filtrated in the first part and complete the prediction. Taking the actual data of a hotel building in Shanghai as an example, the results prove that the method is more accurate and effective than the traditional method.