Date post: | 21-Jan-2018 |
Category: |
Engineering |
Upload: | daniel-coakley |
View: | 161 times |
Download: | 3 times |
Operational Energy Management of the Built Environment Barriers and issues with the creation of calibrated models Presenter: Daniel Coakley
Wednesday, 18 November 2015
Introduction • Introduction
– Building Energy Simulation (BES) Models – Issues with BES Models – Model Calibration
• Barriers to model calibration (and how they can, and are being overcome)
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Building Energy Simulation (BES)
3
Design
• Design assessment and performance evaluation;
• Compliance modelling (e.g. LEED, Part-L Building Regs etc.);
• Building design and system optimisation, Renewables integration etc.
Operation
• Commissioning and retro-commissioning of building systems and controls;
• Fault detection and diagnosis (FDD);
• Simulation-assisted building control.
Retrofit
• Technical and economical evaluation of Energy Conservation Measures (ECM’s);
• Measurement and Verification (M&V) of retrofit savings (IPMVP).
Issues with BES Models Main Issues • Performance Gap: Discrepancy
between design performance and actual performance;
• Uncertainty Quantification: Need to clearly state the prediction confidence, and provide models which are fit for purpose.
Model Calibration Building Model Calibration is the process of improving
the accuracy of simulation models to reflect the as-built status and actual operating conditions
Calibration is widely acknowledged as a complex task (Over-specified and Under-determined)
Most current approaches require significant expert knowledge and numerous manual interventions using different tools/methods to achieve high levels of accuracy
CALIBRATION ISSUES
Barriers to BES Calibration
Standards Expense Complexity Inputs Controls
Faults Uncertainty Identification Integration Automation
Standards • No standard approach for model development and calibration;
• Insufficient standards for calibration criteria;
• Current guidelines only specify nominal acceptable error ranges at whole-building level, but do not account for input uncertainty, meter accuracy, or sub-model discrepancies;
• Concept of a ‘calibrated model’ varies significantly;
• Need for transparent evidence-based model development guidelines;
• Need for accountability within guidelines for model uncertainty and inaccuracies, and inclusion as part of model outputs;
• Need for standard terminologies around model calibration, and clearer communication to building stakeholders;
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Expense (Time + Costs) • Time and resources needed to obtain the required hourly sub-
metered data, system information, etc. can be prohibitive;
• Real system data is often unavailable or difficult to access without extensive effort on the part of the modeler / consultant;
• Lack of data retention and interchange across building life-cycle;
• Cost of metering is continuously decreasing, as is cost of data storage;
• Clients need to be made aware of the value and importance of their building data and system information, as their requirements will drive changes in how information is handled, tagged and retained;
• Better integration between tools, building communication platforms etc. will all help reduce calibration time and expense;
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Model Complexity • Calibration problem is over-specified and under-determined;
• May be thousands of model inputs but relatively few measurable outputs with which to assess model accuracy;
• Increased risk of error in configuring the model, and lack of transparency and ability to validate inputs appropriately;
• Solutions which help simplify the model can reduce noise from unimportant variables, and focus attention on important or influential parameters (e.g. grey-box modelling, free-form profiles);
• Model simplification can reduce time and expense in both model development and calibration;
• Need to maintain model accuracy & outputs which are fit for purpose;
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Input Specification • Lack of availability of high-quality input data for detailed models;
• Inputs often do not take in to account real building operation or system profiles, which can differ significantly from design information (due to changes in operation over time);
• Leads to lack of confidence in model results (GIGO);
• Use of real building inputs, in place of modelled or assumed profiles, where appropriate, can help reduce model development time, reduce noise, and improve model accuracy;
• Availability of stock information from real buildings (i.e. building input library, BIM manufacturer databases etc.) could help improve accuracy of model inputs, and hence confidence over model outputs;
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Controls • Difference between methods of specifying controls within the
simulation environment vs. real building control logic;
• Simulation programs generally geared towards zone-level design load calculations, while control logic focuses on component level operation (e.g. valve and damper positions);
• Ideal scenario offers seamless transition from the design environment to controller (i.e. design -> simulation -> controller);
• Possibility of integration with third-party software, or vendor control modules offers opportunity to exploit strengths of combined tools.
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Faults • Faults in building systems, sensors or components can cause
misleading discrepancies between measured and simulated data;
• Faults may be present in physical components or control algorithms;
• Leads to misdirection during model calibration, and can actually exacerbate model discrepancies if user attempts to mask problem;
• Data gathering phase should incorporate a detailed system audit to identify and rectify (where possible) any pre-existing faults;
• Confidence over model results – do not automatically assume the simulation outputs are incorrect and need re-adjustment – question sources of discrepancies;
• Methods for simulating faults or system degradation over time.
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Prediction Uncertainty • Very few calibration studies account for model
uncertainty (various sources);
• Uncertainty is unavoidable in the field of system modelling, particularly in complex systems which are subject to external scenario influences (e.g. occupant behaviour, weather);
• Need to account for model uncertainty, and provide an expression of the degree of uncertainty associated with model predictions;
• Use of prediction ranges as opposed to traditional expression of explicit numerical values for energy consumption.
Specification
Modelling
Numerical
Scenario
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Identification • Due to model complexity and possibility of inherent faults, it is often
difficult to accurately diagnose the underlying cause of discrepancies;
• Lack of appropriate analysis tools for the purpose of calibration;
• Adjustments are often based on ad-hoc manual interventions based on user judgement, rather than scientific reasoning;
• Need to reduce complexity of calibration problem where possible by splitting problem into tiers for example, thus reducing noise;
• Provide suitable statistical and graphical calibration tools within the simulation environment to improve calibration workflow;
• Use techniques such as sensitivity and uncertainty analysis to provide more structured calibration framework.
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Integration • Issues in integrating data across the BLC (Design-Simulation-Control)
• Design-Simulation: Translation from BIM to BES models is a complex task in itself, with results varying between users and translation tools for identical geometries;
• Industry standard file formats (e.g. IFC, gbXML) are improving but do not provide consistent results, particularly for complex geometries.
• Improvement of overall integration across the BLC is needed to improve data integrity and accuracy of simulation inputs;
• Standard tools and workflows required for file translation from BIM to BEM to ensure consistent results;
• Focus on full BLC integration – from design to simulation to control environment, and back to design. (Close the loop)
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Automation • Complexity of the calibration task, requiring inputs from various
sources and stakeholders, makes the task time-consuming and cost-prohibitive in many cases;
• Lack of a calibration workflow to help guide and automate steps where possible.
• Need for an end-to-end workflow using best available tools and methods appropriate to the calibration problem;
• Use mathematical and statistical optimisation tools, where appropriate, to find parameter solutions, and guide calibration workflow.
• Balance between mathematical methods and user intervention.
Standards Expense Complexity Inputs Controls Faults Uncertainty Identification Integration Automation
Barriers to BES Calibration
Standards Expense Complexity Inputs Controls
Faults Uncertainty Identification Integration Automation