Scalable Energy Retrofit Analysis

Realizing retrofits at the scale needed to meet the nation’s energy efficiency goals will depend on a confluence of thousands of decisions, including those by public agencies in planning policy and incentives, utilities in executing energy efficiency programs, industries in providing energy services and technologies, financiers in providing capital to the market, and building owners in deciding to implement energy efficiency measures. Argonne National Laboratory is developing transformative analytic methods that capture energy performance of individual commercial buildings, scale to communities, and correctly evaluate retrofit options in real contexts. The decision tools under development span and integrate building energy performance calculations, uncertainty analysis, and financial risk assessment.

Energy Savings: Argonne, with Georgia Institute of Technology, has developed a building energy analysis tool, EECalc, that provides a scalable, transparent, and affordable platform for benchmarking building energy performance; evaluating energy savings potential of energy efficiency technologies; and analyzing national, regional, and market impacts of energy efficiency initiatives, strategies, codes, and standards.

The tool’s energy calculations are grounded in ISO 13790 – a product of the EU’s Energy Performance in Buildings Directive involving the European Committee for Standardization and the International Organization for Standardization. These calculations estimate thermal energy demands for heating and cooling using a monthly, quasi-steady-state method. Thermal energy demand accounts for heat losses by transmission and ventilation, heat gains from solar and internal sources, and the effect of thermal inertia driven by building mass. Given the thermal energy demand, energy consumption is calculated by end-use: heating, cooling, ventilation, lighting, plug loads, pumps, and domestic hot water systems.

Comparison of EE-Calc energy use predictions with
EnergyPlus simulation results

The energy calculations provide a reasonable level of resolution for evaluating retrofits at the building level and were chosen specifically because they are computationally efficient; use only observable building data with relatively low effort to compile: and require minimal modeling expertise and judgment from the user. These features expand the possibilities for wide-spread, large-scale use in retrofit and uncertainty analysis.

Uncertainty: An often cited barrier to accelerating adoption of building energy efficiency measures is the perceived risk of underperformance. Provided only with deterministic estimates, decision makers lack confidence that predicted energy savings will occur. Robust, credible methods for predicting probabilistic energy savings will allow decision makers to manage risk instead of opting to do nothing. To that end, we apply Bayesian inference as a new approach for calibrating and factoring uncertainty into building energy models.

Bayesian calibration is an alternative to traditional, expert-intensive approaches that require “tweaking” of energy model input parameters to match measured data. Instead, Bayesian statistical methods derive the most likely distributions (posterior distributions) of input parameters from their pre-defined probability density functions (i.e., prior distributions) to explain the measured energy use data. The resulting posterior distributions are then used in Monte Carlo samplings of the energy model to yield probabilistic outcomes of predicted energy savings.

Stochastic methods can be very computationally intensive. An advantage of Bayesian calibration is that lower resolution energy models (e.g., EECalc) can be used without compromising the degree-of-confidence in the outcomes. This advantage has been illustrated in comparison studies.

Bayesian calibration reduces uncertainty in building energy use predictions

Probabilistic energy savings results using EECalc with Bayesian calibration provide insight on under-performance risks

The EECalc energy retrofit analysis tool is available for trial and use. Contact Ralph Muehleisen.

More

September 2012