Forecasting Gas Usage for Big Buildings using Generalized Additive Models and Deep Learning
Permanent Link
http://hdl.handle.net/11603/11040Metadata
Show full item recordDate
2018Type of Work
8 pagesText
conference papers and proceedings preprints
Rights
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please contact the author.Subjects
gas forecastingdeep learning
long short term memory
generalized additive model
forecasting
Time series behavior of gas consumption is highly irregular, non-stationary, and volatile due to its dependency on the weather, users’ habits and lifestyle. This complicates the modeling and forecasting of gas consumption with most of the existing time series modeling techniques, specifically when missing values and outliers are present. To demonstrate and overcome these problems, we investigate two approaches to model the gas consumption, namely Generalized Additive Models (GAM) and Long Short-Term Memory (LSTM). We perform our evaluations on two building datasets from two different conti-nents. We present each selected feature’s influence, the tuning parameters, and the characteristics of the gas consumption on their forecasting abilities. We compare the performances of GAM and LSTM with other state-of-the-art forecasting approaches. We show that LSTM outperforms GAM and other existing approaches, however, GAM provides better interpretable results for building management systems (BMS).
Abstract
Time series behavior of gas consumption is highly irregular, non-stationary, and volatile due to its dependency on the weather, users’ habits and lifestyle. This complicates the modeling and forecasting of gas consumption with most of the existing time series modeling techniques, specifically when missing values and outliers are present. To demonstrate and overcome these problems, we investigate two approaches to model the gas consumption, namely Generalized Additive Models (GAM) and Long Short-Term Memory (LSTM). We perform our evaluations on two building datasets from two different conti-nents. We present each selected feature’s influence, the tuning parameters, and the characteristics of the gas consumption on their forecasting abilities. We compare the performances of GAM and LSTM with other state-of-the-art forecasting approaches. We show that LSTM outperforms GAM and other existing approaches, however, GAM provides better interpretable results for building management systems (BMS).
Related items
Showing items related by title, author, creator and subject.
-
An Analysis of NCEP Tropical Cyclone Vitals and Potential Effects on Forecasting Models
Trahan, Sam; Sparling, Lynn (American Meteorological Society, 2012-06-15)This study analyzes the Tropical Cyclone Vitals Database (TCVitals), which contains cyclone location, intensity, and structure information, generated in real time by forecasters. These data are used to initialize cyclones ... -
Geomagnetic secular variation forecast using the NASA GEMS ensemble Kalman fllter: A candidate model for IGRF 2020
Tangborn, Andrew; Kuang, Weijia; Sabaka, Terence J.; Yi, Ce (Springer Nature, 2020-07-21)We have produced a 5 year mean secular variation (SV) of the geomagnetic field for the period 2020-2025. We use the NASA Geomagnetic Ensemble Modeling System (GEMS), which consists of the NASA Goddard geodynamo model and ... -
International propagation of shocks: a dynamic factor model using survey forecasts
Lahiri, Kajal; Zhao, Yongchen (Towson University. Department of Economics, 2018-09)This paper studies the pathways for the propagation of shocks across G7 and major Asia-Pacific countries using multi-horizon forecasts of real GDP growth from 1995 to 2017. We show that if the forecasts are efficient in ...