Non-residential buildings are responsible for more than a third of global energy consumption. Estimating building energy consumption is the first step towards identifying inefficiencies and optimizing energy management policies. This paper presents a study of Deep Learning techniques for time series analysis applied to building energy prediction with real environments. We collected multisource sensor data from an actual office building under normal operating conditions, pre-processed them, and performed a comprehensive evaluation of the accuracy of feed-forward and recurrent neural networks to predict energy consumption. The results show that memory-based architectures (LSTMs) perform better than stateless ones (MLPs) even without data aggregation (CNNs), although the lack of ample usable data in this type of problem avoids making the most of recent techniques such as sequence-to-sequence (Seq2Seq). Graphical

Pre-processing and analysis of sensor data present several challenges due to their increasingly complex structure and lack of consistency. In this paper, we present TSxtend, a software tool that allows non-programmers to transform, clean, and analyze temporal sensor data by defining and executing process workflows in a declarative language. TSxtend integrates several existing techniques for temporal data partitioning, cleaning, and imputation, along with state-of-the-art machine learning algorithms for prediction and tools for experiment definition and tracking. Moreover, the modular architecture of the tool facilitates the incorporation of additional methods. The examples presented in this paper using the ASHRAE Great Energy Predictor dataset show that TSxtend is particularly effective to analyze energy data.