Short-term heat load forecasting in district heating systems using artificial neural networks

Abstract

With the advent of sustainable energy systems based on renewable energy sources (RES) and the development of a new generation of district heating systems (4GDH), it has become imperative for cogeneration and RES plant operators, as well as district heating (DH) operators, to apply new tools that lead to improvements in production planning, energy efficiency, and at the same time, reduce costs of heat generation. In recent years, machine learning (ML) methods used for the estimation and forecasting of energy demand have drawn considerable attention due to their advantage over linear and nonlinear programming models. In this context, the paper presents an artificial neural network (ANN) approach for the prediction of short-term heat load in a district heating system. The ANN model is trained with past heat load data, weather data and social behavior components. The predictive performance of the neural network model is measured by the mean absolute percentage error (MAPE) and the root mean square error (RMSE).