District heating systems load forecasting: a deep neural networks model based on similar day approach

Abstract

ABSTRACT Accurate heat load forecasting is an important issue to ensure the reliable and efficient operation of a district heating system. In this paper, a hybrid model that combines similar day (SD) selection and Deep Neural Networks (DNNs) to construct SD-DNNs model for short-term load forecasting (STLF) is presented. A new Euclidean Norm (EN) weighted by eXtreme Gradient Boosting (XGBoost) is used to evaluate the similarity between the forecasting day and historical days. In this EN, the outdoor temperature, wind force and day-ahead load are simultaneously considered. And eight features are chosen as inputs of the DNNs to predict the heat load. The Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and Mean Percentage Error (MPE) are used as accuracy evaluation indexes. The experimental results demonstrate that the SD-DNNs model can accurately forecast the heat load.