Multi-Agent Stochastic Simulation of Occupants’ Behaviours

Slides:



Advertisements
Similar presentations
Precision and thermal comfort
Advertisements

Institute of Energy & Sustainable Development De Montfort University
SAVE-ODYSSEE MONITORING TOOLS FOR ENERGY EFFICIENCY IN EUROPE Energy efficiency index ODEX B Lapillonne,, K Pollier, Enerdata D Bosseboeuf, ADEME Septembre.
University of Liège Faculty of Applied Sciences Thermodynamics Laboratory Workshop “Commissioning and Auditing of Buildings and HVAC Systems” Use of a.
'A policy framework for the governance of integration across the energy system' Professor Brian Collins Professor of Engineering Policy, UCL.
Introduction Air conditioning Adaptive comfort Adaptive design
Daylighting Prediction Tool Online Presented By: Eng.Reham Mostafa Mohamed Mohie El-Din By: Christoph Reinhart Web Address:
SINTEF Energy Research 1 Remodece meeting January 2007 Nicolai Feilberg.
Do Try This at Home!. Reducing energy consumption at home has many benefits.
1 S HORT METHODOLOGIES FOR IN - SITU ASSESSMENT OF THE INTRINSIC THERMAL PERFORMANCE OF THE BUILDING ENVELOPE Rémi BOUCHIE, CSTB Pierre BOISSON, Simon.
So it all boils down to ‘$ and Sense’ No matter what your motivation energy conservation can be argued form both the dollar and sense position.
Sustainable Homes and Communities Program. Saving Energy and Water at Home Presentation Outline Why is it important to save energy? Where is energy used.
College of Management & Economics, Tianjin University Projections of energy services demand for residential buildings: Insights from a bottom-up methodology.
Domestic Energy Demand: Projections to 2030 Diana Dixon.
Continuous Probability Distributions Continuous random variable –Values from interval of numbers –Absence of gaps Continuous probability distribution –Distribution.
THERMAL INERTIA FOR SMALL SCALE RESIDENTIAL BUILDING STIJN VERBEKE UNIVERSITY OF ANTWERP UNIVERSITY COLLEGE BAUSIM 2010 CONFERENCE.
Nina H. Fefferman, Ph.D. Rutgers Univ. Balancing Workforce Productivity Against Disease Risks for Environmental and Infectious.
Predator/Prey Simulation for Investigating Emergent Behavior Jay Shaffstall.
ELECTRICITY KNOWLEDGE AND USAGE SURVEY BY: NINA NEL; KATHERINE VAN WYNGAANRDEN, MIA ROBINSON, SABRINA SHAW.
Monitoring Results – Ethelred Estate Chris Dunham, Carbon Descent 13 th July 2010.
4. INTERIOR LAYOUT  Good interior layout will facilitate many of the passive strategies recommended in this toolkit, in particular thermal mass, lighting.
Stats Term Test 4 Solutions. c) d) An alternative solution is to use the probability mass function and.
Energy Mythbusters!. For most schools hot water is provided by your boiler – which is in turn controlled by your TREND Building Management System It’s.
© 2007, Itron Inc. Statistically Adjusted End-Use Model Overview & Thoughts about Incorporating DSM into a Forecast May 4, 2009 Frank A. Monforte, Ph.D.
The Importance of Energy Efficiency for Public Power
Skills Training in India: Market or Privilege?
Generation of Domestic Electricity Load Profiles
Building Environmental Systems
Discrete-time Markov chain (DTMC) State space distribution
Adnan K. Chhatriwalla, MD Saint-Luke’s Mid America Heart Institute
Math 4030 – 4a More Discrete Distributions
Designing JITAI for Skylar
More kinds of FSM ENGR 110 #
WP2 INERTIA Distributed Multi-Agent Based Framework
GENESYS Redevelopment Strawman Proposal
Comparison of THREE ELECTRICAL SPACE HEATING SYSTEMS IN LOW ENERGY BUILDINGS FOR SMART LOAD MANAGEMENT V. Lemort, S. Gendebien, F. Ransy and E. Georges.
Modelling the impact of integrated EV charging and domestic heating strategies on future energy demands Nick Kelly, Jon Hand, Aizaz Samuel ESRU, University.
WP2 INERTIA Distributed Multi-Agent Based Framework
Economic Operation of Power Systems
V5 Stochastic Processes
International Conference on Sequence Analysis and Related Methods
RealValue H2020 Overview RealValue project funded under H2020 LCE 8 – 2014: Local / small-scale storage Commenced 1st June 2015 (duration 36 months) EU.
The future of cooling – where it is needed, how it is used
The UK’s changing cultural influence through media and food .
Nick Kelly, Jon Hand, Aizaz Samuel
CPSC 531: System Modeling and Simulation
Environmental house project
Experiment Basics: Designs
System Performance: Queuing
Techniques for Data Analysis Event Study
Considering impacts of PEVs in planning optimal hybrid systems
Math CC7/8 – Be Prepared On Desk: Pencil Calculator Math Journal
Shifting the timing of energy demand: A stochastic modelling approach
Unlocking Demand Contribution to Distribution Network Management
R. W. Eberth Sanderling Research, Inc. 01 May 2007
The impact of occupants’ behaviour on urban energy demand
Eoghan McKenna & Murray Thomson CREST, Loughborough University
Responsive Architecture
WESTERN REGIONAL WORKSHOP
Neural Networks ICS 273A UC Irvine Instructor: Max Welling
Arslan Ahmad Bashir Student No
Prof. Fionn Stevenson Sheffield School of Architecture
Development of a household occupancy state simulation model for multi-energy load profile generation Residential buildings account for around 30% of the.
Occupancy data analytics and prediction: A case study
Evolutionary Ensembles with Negative Correlation Learning
Discrete-time markov chain (continuation)
Frédéric Haldi, Darren Robinson, Claus Pröglhöf, Ardeshir Mahdavi
CS723 - Probability and Stochastic Processes
Amir Marcovitz, Yaakov Levy  Biophysical Journal 
Understanding Experimental Methods
Presentation transcript:

Multi-Agent Stochastic Simulation of Occupants’ Behaviours Darren Robinson Sheffield School of Architecture

Stochastic simulation Peoples’ decisions depend on both deterministic and random responses to stimuli: they are stochastic in nature. The same occupant may respond differently, on different occasions, even in response to identical stimuli. We may also encounter considerable differences in response between individuals to identical stimuli. This randomness can have significant implications for comfort and for buildings’ energy and other resource demands. 29/05/2018 © The University of Sheffield

Methods Three modelling tools: Bernoulli process Discrete time random process: Markov chain Continuous time random process: Survival analysis Applying: Forward selection (Cluster analysis) k-fold cross validation

Presence Short-term Presence profile: Pij(t) Current models have duration of a single presence as 4hours (am or pm)….often less than this. Red = simulated, blue=measured. Page, Robinson, Morel and Scartezzini, Energy & Buildings 40(2), 2008 (5th most cited paper: 2008-13)

Activities Aggregate activity model: Pj(t) & Dj(t) [UK] Simulated and observed plot for one model only as an example for the aggregated population. WP means washing appliance.

Is culture important? American TUS! Measurements in France Simulated and observed plot for one model only as an example for the aggregated population. WP means washing appliance. Measurements in France

Model robustness: model applied to Germany, France, & Spain

Appliance Activity-dependent appliance modelling: D(t) & Pij(t...t+D) | P(t)=1 Duration and power transition given that appliance is on

Windows Window openings: Pij(occ), Dj | P(t)=1 Haldi and Robinson, Building and Environment : 44(12), 2009 Best Paper Prize: 2009

Windows: diversity beware! Conventional behaviour Actions increase with qin and qout. Predicitve thermal behaviours Similar, but decreased actions for high qout to avoid overheating. Non-thermal behaviours almost independent of thermal stimuli.

Blinds Blind position: Pij(t)… Haldi and Robinson, JBPS : 3(2), 2010 Best Paper Prize: 2010-2011

Lights Lights (Lightswitch 2002): Pj(t)… Hunt Pigg On arrival switch-on probabilities Measured switch-off probabilities as a function of absence duration Pigg Reinhart Within-day switch-on probabilities

No-MASS framework Synthetic population generator Appliance allocation / use Large small Activities (homes) Short absences (workplaces) Long absenses Location Metabolic gains Heating use (machine learning) Hot water use Use of shading Use of window Use of lights Adaptive comfort Social Interactions BDI rules Extension to DSM (and LVN) Chapman and Robinson, JBPS (under review), 2017

Example results: two collocated office occupants [distributions] The left graph presents stochacity due to variations in models, for a house and office in two locations (Nottingham, Geneva) On the right we have variations due to archetypical window models. Monthly heating energy demands: office ♯window openings for different agents

Example results: interacting collocated occupants Managing negotiations Weighted Voting System Window opening Agent a Likes window open Agent b Opens window less Action with most votes takes place AGENT B CURTAILS WINDOW OPENING DURATION WHEN MOBILE. The archetypical behaviours create conflict between agents that need to be managed. We employ a weighted voting conflict system. Each agent is asigned a power, power wieghted votes are counted and the action with most votes wins the action Here agent a prefers the window open, even with the heating on. Agent b prefers the window closed most of the year. Different weights give different results, however the voting system appears to do a good job at managing the conflict with results generally a compromise between the two. a b a>b a<b

Example results: DSM Sancho-Tomás, Chapman and Robinson, Proc. Building Simulation 2017

In conclusion… Good progress has been made in modelling: Synthetic population generation and attribution Presence chains and activities Behavioural actions (aggregated): envelope + personal characteristics Appliance ownership and use (homes) Social interactions We still have lots to do: long-term absences Use of electrical appliances in workplaces Completion and validation of DSM framework Rigorous empirical basis to negotiated behaviour modelling Population diversity (rigorously) Adaptive comfort and overheating Ensemble validation But this stuff is fun!