IIIT Hyderabad Publications |
|||||||||
|
Development of a Surrogate Model that estimates Building Energy Consumption by extracting Top Characteristic Feature VectorsAuthor: Sai Abhilash Reddy Sangireddy Date: 2020-05-27 Report no: IIIT/TH/2020/61 Advisor:Vishal Garg AbstractIn early stage of any building design, the design team has to probe several building models to analyze the building energy consumption. This requires simulation of numerous different models using building energy modelling software. The number of combinations increase exponentially with increasing parameters. This requires a lot of computation to simulate energy consumption for all the input combinations. In a scenario considering five building variables, with ten design choices for each variable, the design team will be required to simulate a hundred thousand different models using building energy management software. This study presents a machine learning based solution to estimate energy consumption of a building in the early phase. The existing research in this domain doesn’t focus on determining the best sample to train the regression models. This study uses a clustering based sampling method to identify the training data. Using clustering, top characteristic feature vectors are derived. The annual energy consumption is simulated for the extracted samples using EnergyPlus. A relationship is then established between the input samples and their corresponding energy consumption through various regression models. These regression models are in turn used to estimate the energy consumption for the large input data sets. This method of surrogate modeling saves a lot of computation by reducing the number of computations by a 100-fold. This method is tested for data from Jaipur and Hyderabad cities of India. Approximately hundred thousand simulations are performed for each location using cloud based parallel computation. By simulating approximately one percent of the input combinations, annual energy consumption for the large set of combinations are estimated using SVR and k-means clustering for Jaipur with accuracy greater than 93% for 99.8% of the input combinations. When the same model is trained for Hyderabad, it produced accuracy greater than 93% for 98% of the input combinations. Using a python based REST framework, an open source REST based web API service - FASTSIM API is developed which can be integrated into any existing software using HTTP requests. The data that we generated for Hyderabad and Jaipur cities is also made publicly available for the research community. Full thesis: pdf Centre for IT in Building Science |
||||||||
Copyright © 2009 - IIIT Hyderabad. All Rights Reserved. |