Multi-scale building energy modeling
Abstract:
Multi-scale building energy modeling (MBEM) is a powerful computational tool to simulate performance of buildings across scales: from individual buildings to district of buildings to entire building stock in a city or region. MBEM can provide insights into prioritization of strategies to reduce building energy use, decarbonize buildings, and improve their climate resilience. This talk will introduce the MBEM research at LBNL including data, tools, model calibration methods, and use cases.
Bio:
Dr. Tianzhen Hong is a Senior Scientist and Deputy Head of the Building Technologies Department of Lawrence Berkeley National Laboratory. He leads the Urban Systems Group. He is an IBPSA Fellow, ASHRAE Fellow, and 2021 Highly Cited Researcher. He published more than 160 journal articles on various topics of buildings, energy, resilience, occupant behavior, machine learning, modeling and simulation. He received B.Eng. and Ph.D. in Building Science, and B.Sc. in Applied Mathematics from Tsinghua University, China.
Summary:
We want our buildings to provide comfort and protection from extreme weather conditions at a low energy cost
Challenges
Global warming
Urban activities generate heat
Changing energy demand
Solar panels
Electric vehicle charging
Energy-intensive devices
Power grid is changing (e.g. duck curve: power demand peaks sharply in evening)
US buildings compute 39% of primary energy (28% transport, 33% industry), similar for carbon emissions
Varies by city (San Diego is transport-heavy, Chicago and New York are building-heavy)
New building designs are net zero energy (hundreds are being built today)
Can we deploy these at scale?
Need an Integrated Design approach:
HVAC, lighting, daylighting, architecture
Costs may be paid back via energy efficiency
Occupant behavior is key (Are they closing the windows? What temperature is the thermostat set to?)
Need to choose energy efficient appliances
Public policy needs to set appropriate incentives to encourage adoption of energy efficient technologies
Need building simulations to capture every point in the lifecycle of a building
Program
Design
Build
Operate
EnergyPlus: Department of Energy flagship building energy simulation tool
21 years old
> 1 million lines of C++ code
Physics-based: takes into account
Building structure
Occupant behavior
Ventillation
Model inter-building effects in the urban environment
Building <-> Building: radiant heat exchange
Atmosphere -> Building: induce requirements for heating / cooling
Building -> Atmosphere: release of heat / cold to environment
Analysis:
Small office buildings emit 3.7 times heat emission relative to their direct use
Large ones are 1.7: much more efficient
Suite of modeling tools from DOE based on EnergyPlus
Can model scales from whole countries, cities, buildings to individual rooms/floor within buildings
Require additional data: weather, building design, occupant behavior, etc.
City-scale modeling
Requite city data
Some cities provide databases on their building stock
Land use, building permits, electrical permits, boiler permits
Need to simplify and standardize building info the make it usable within EnergyPlus
2D, 3D box, rough building layout, detailed building CAD design
Have collected building records for 6 major US cities
CityBES: web application that enables energy analysis for whole cities, using EnergyPlus
Use-cases
Retrofit analysis for 940 office/retail buildings in San Francisco
LED lighting, plug load controllers, occupancy sensors, daylighting sensors, temperature setpoint, add economizer
Buildings can achieve 30-40% reduction in energy use
Centralizing heating/cooling for all buildings in a campus
Leverages more efficient large facilities and uncorrelated demand among different buildings
Modeling heat resilience of King’s neighborhood in Fresno
Active air conditioning: temperature can still reach 30°C
Power outage: 40-45°C
Passive heat management measures: Roof insulations, cool roof coating, window shades, etc.
US National building stock analysis: 80% reduction in building CO2 emissions by 2050 via a suite of retrofits and new building improvements
Sensitivity to occupant behavior
Sensitive for smaller buildings where occupants have choices
Less so for large buildings with central HVAC control or non-openable windows
Can track occupants to improve model accuracy
E.g. via Nest sensors, wifi signals, PurpleAir sensors
Accuracy of EnergyPlus is constrained by lack of data about building structure and occupant behavior
Currently working on explicitly documenting and modeling all the major error factors
International efforts to pool modeling resources and software and observational data
Technical/physical: climate, building envelope, building equipment
Human: Operation, maintenance, occupant behavior
Wide range of energy data to calibrate models:
Surveys
Distribution of building energy use
Aggregated energy use for many buildings over time
Annual/monthly energy use for individual buildings
Smart meter data (15 mins whole building or individual users) for individual buildings
Calibrating models at all spatial/temporal granularities
Looking for error patterns (depend on season, building type, occupant type)
Comprehensive modeling of energy in society
Exascale computing problem: major city is 10E6 buildings, 10E6-10E7 people, 10E6 vehicles, 10E6-10E8 sensors and devices
DOE has created a comprehensive simulation of city transport, buildings, etc.
IM3: national models
Use-cases for machine learning
Infer occupant behavior
Detection of faults in sensors and building equipment
Prediction of
Heating/cooling loads,
Energy production from solar panels
Infer building structure from street, satellite photos
Use-cases for better weather forecasts
Predict solar power generation in next few hours
Buildings are connected to grid, so need to use power when power is cheapest
EnergyPlus can be used for
Long term: analyzing design of buildings
Short term: real-time to plan the operation of a building dynamically
Additional tools based on Modelica that provide more accuracy