A Prototype for predicting energy consumption in buildings: a case of commercial office buildings
Date
2019
Authors
Wachira, Paul Manasse Macharia
Journal Title
Journal ISSN
Volume Title
Publisher
Strathmore University
Abstract
Energy consumption remains the highest cost areas for businesses together with facilities, people and equipment but unfortunately, it is the only one that is not carefully monitored. For businesses to be able to manage energy consumption they must first be able to predict future energy consumption so as to aid in budgeting and planning for cost reduction strategies. This study proposed an energy consumption prediction prototype to help predict future consumption of energy in commercial office buildings thus aiding proper budgeting and cost reduction. To develop the prediction model the study used the 2012 CBECS (Commercial Buildings Energy Consumption Survey) dataset hosted by the Energy Information Administration (EIA) of the United States of America. After cleaning and reviewing the data set, 26 Features were selected for Features Engineering. Features Engineering enabled the research to choose the best 4 Features which were used for training and testing different Regression Based Machine Learning Algorithms. Using R2 (R Squared), MAE (Mean Absolute Error) and RMSE (Root Mean Square Error) to determine performance, the study selected Gradient Boost Machines as the best algorithm for the prototype with an Accuracy of 97%. Python packages Pandas, NumPy, Matplotlib, Seaborn and Scikit-leam were used indata cleaning, descriptive statistics, features engineering, data visualization and training and testing the machine learning algorithms for the energy consumption prediction model. The prototype was developed using Flask (a Python micro web framework) to enable the Building owners provide the prototype with data via web browser related to the 4 features selected for energy consumption prediction. Usability Test was done with 48.1% of the users strongly agreeing and 44.2% agreeing to use the prototype in future for prediction of electricity consumption in their buildings.
Description
Submitted in partial fulfillment of the requirements for the degree of Master of Science in Information Technology at Strathmore University