As a building energy modeller, it can be difficult to know what combination of building upgrades to recommend for a particular building. Building performance optimization is a new method that helps modellers select a package of upgrades from a plethora of options. Optimization has become more widely available in recent years as powerful cloud computing resources which allow thousands of simulations to be run over short periods of time have come online.
Why is optimization needed? Let me explain.
A typical goal of energy modelling is to inform design decision-making by quantifying the benefits of building upgrades. Examples of these upgrades are increased insulation thickness, better windows, air source heat pumps, condensing boilers, solar panels, or low flow water fixtures. These upgrades may be applied to new or existing buildings. A typical approach is to run energy models on each potential upgrade to calculate the potential energy savings. The annual energy cost savings can be balanced against the cost of implementing the upgrade by calculating a net present value or payback period. The client will typically use this information to select a package of upgrades based on their energy performance and financial goals.
This system can work reasonably well, but there is a problem with selecting several upgrades from a list of upgrades that have been modelled individually. The problem is that this process ignores interactions among the upgrades. These interactions could mean that the predicted energy and/or cost savings are much higher or lower than the actual energy and/or cost savings. Take for instance upgrades to the wall insulation and HVAC efficiency in a poorly insulated and inefficiently heated building. Each of these upgrades will show high energy savings on their own, but the energy savings of the upgrades modelled together will be less than the sum of the energy savings of the upgrades modelled separately. Wall insulation reduces the heating load, and the improvements in HVAC efficiency mean that the smaller heating load is met more efficiently. These interactions make it hard to predict the potential energy savings in a package of multiple upgrades, which makes it hard to decide which package of upgrades would offer the best performance on cost and energy. This is where optimization comes in.
Optimization automates the process of testing hundreds or thousands of different combinations of upgrades. It is more consistent than the typical modeller-directed, educated-guessing process, because it uses sophisticated search algorithms that have been tested by researchers to search through thousands of combinations for the optimal packages of upgrades. A commonly used algorithm in building optimization is the Non-domination Sorting Genetic Algorithm Two (NSGA-II or NSGA2), which is used to do multi-objective optimization. NSGA-II can search for any two objectives simultaneously (like energy savings and cost savings). That means that optimization tools can consistently find the lowest cost upgrade package for a given energy or carbon saving goal.
Most of the optimization tools in the past have been created and used by researchers and have been relatively inaccessible to energy modellers until recently. Right now, I know of only three tools that are accessible for use: NREL’s BEopt, DesignBuilder Optimization, and NREL’s OpenStudio-analysis-spreadsheet. Optimization tools like these are the only way for a modeller to do optimization. The time it takes to write an optimization tool is prohibitive. In the future, building energy modellers will demand optimization capabilities from their modelling tools and optimization will be much more common. If it can be modelled, it can — and should — be optimized. More optimized buildings means more cost-effective and more efficient buildings. Building performance optimization is better for clients and for the environment.