Urban Energy Systems - Google Summer of Code 2021 Projects
The UES Lab at Empa is currently supporting four projects for Google Summer of Code 2021:
Project 1: Automated Segmentation of Google Street Images for Estimating the Window to Wall Ratio of Building Façade.
Project 2: Interactively mapping the energy needs of rural households in developing countries.
Project 3: Geovisualization and model preparation of 3D open data for urban-scale energy simulation.
Project 4: Graphical User interface for Combined Energy Simulation and Retrofit Tool CESAR-P.
More details on each of the projects is given below. Please follow the instructions on the Google Summer of Code website if you are interested in applying to any of the projects.
Project 1: Automated Segmentation of Google Street Images for Estimating the Window to Wall Ratio of Building Façade
Mentor: Fazel Khayatian
The Urban Energy Systems Lab (UESL) of the Swiss Federal Laboratories for Materials Science and Technology (Empa) resorts to the in-house open-source tool CESAR-P for estimating current and future energy demand profiles of buildings. The tool uses building footprint and height to set up a geometric representation of the buildings, which is necessary for energy demand estimations. Given that building windows play an important role in the accuracy of energy demand estimations, CESAR-P uses rough estimations of the Window to Wall ratio to calculate the energy demand. However, imprecise estimates of the glazed area of building façades propagates into inaccurate predictions of the energy demand. There is a potential to improve estimations of buildings' window area by segmenting Google Street Images, and extracting a more accurate percentage of glazed area for each building façade.
The goal of this project is to build on existing façade parsing and image segmentation literature, and create a pipeline of modules for estimating the window to wall ratio of building façades. Using segmented and labeled images of building facades from existing datasets, a CNN (most likely a ResNet) is trained, and then validated on Google Street Images. Most likely, there is a need for further training (and fine-tuning) the model with additional (manually) labelled Street Images, as the angle, perspective, and obstructions in the training data might not fully represent the variety in Google Street Images.
The project will complement the codes developed by Lucerne University of Applied Sciences and Arts during Open Data Hackdays 2020, sponsored by the UESL of Empa. The outputs of the project will be used as inputs for the open-source CESAR-P tool to improve the accuracy of building demand estimations.
Project 2: Interactively mapping the energy needs of rural households in developing countries
Mentor: Cristina Dominguez
The Urban Energy Systems Laboratory (UESL) at Empa and the Chair of Building Physics (CBP) at ETH Zurich are developing a geospatial data-driven model based on open-source survey, GIS, and satellite data to estimate the electricity needs of rural households in developing countries to improve the planning and design of electrification projects and energy access strategies. The results are validated using real electricity consumption data from mini-grids and field studies in the regions of Sub-Saharan Africa, South Asia, and Latin America. In addition, the estimated electricity needs are forecasted based on different scenarios and projected at a village level in selected countries from the studied regions.
The aim of this project is to build an interactive tool with two purposes:
- To develop a front-end user interface to visualize in a geospatial form the simulated results by the model for different regions in the world. As the model is data-driven, large sets of georeferenced time-series data are required to be structured in order to query the data based on the user’s inputs on the interface. For this, an entity-relationship model has to be developed.
- As the simulated results by the model are difficult to validate for every rural settlement in the world due to the lack of data, a crowdsourcing calibration method has to be developed. For this, key parameters that are required for calibrating these results for a specific geographic location will be asked to users that voluntarily wish to participate in the data collection process. Users will select their geographic location within a fixed radius, then, the entered parameters will be stored in a GIS based Postgre-SQL database for that specific location, a factor will be calculated and then the results simulated by the model will be corrected by this factor and the new results will be geospatially visualized.
Project 3: Geovisualization and model preparation of 3D open data for urban-scale energy simulation
Mentor: James Allan
The Urban Energy Systems Laboratory (UESL) at Empa have developed software to simulate energy demand profiles of buildings at an urban scale. This includes predicting the demands for heating, cooling and electricity. This software is used to evaluate energy efficiency strategies, such as retrofitting, and the design of energy supply systems, such as decentralized energy hubs and district heating networks. One of the main inputs for these simulations is 3D geometries that have recently been released openly across Switzerland.
We would like to build a front-end interface and visualization tool that allows the user to visualize open datasets that contain 3D geometries for simulations in commonly used formats used by EnergyPlus. The user will be explore the datasets using a 3D interface. For visualization we propose either a Python-based or a Node.js-based web-client that makes use of the Cesium 3D geospatial platform. The CityGML building data will be converted and stored in a postGIS database.
Project 4: Graphical User interface for Combined Energy Simulation and Retrofit Tool CESAR-P
Mentor: Léonie Fierz
The Urban Energy Systems Lab (UESL) at Swiss federal laboratories for materials science and technology (EMPA) has developed an open source tool, CESAR-P, to simulate the energy demand of a district with a bottom up per-building approach. The tool prepares input data for the EnergyPlus simulation engine and post-processes those results. It further supports analyzing different retrofit strategies. As an input the footprint of the buildings, year of construction, building type (e.g. single family home, office, school) as well as weather data has to be provided. Results include annual or hourly heating, hot water, cooling and electricity demand results per building. There are many in-detail options the user can specify if needed, such as the constructional properties or operational profiles.
The goal of the project is designing and implementing a graphical user interface for CESAR-P to support the workflow for a simulation run with CESAR-P including option configuration and results visualization. A smart input data cleanup, such as checking and simplifying the building footprints, as well as a plausibility check for the result are further ideas to facilitate the process.