What is EnergyPlus doing that DOE-2 isn't that requires so much additional computation time? Even when they are both performing hourly simulations, EnergyPlus takes significantly longer.
EnergyPlus is a tightly coupled simulation engine that uses iteration between zones, air systems, water systems, and on-site power generation. EnergyPlus is performing an iterative solution between the simulation domains in order to provide more accurate results. In EnergyPlus, the predicted load in the space is not passed to the air system, but is instead an input to a predictor/corrector algorithm that allows the zone to float for cases where the system does not exactly meet the space load. This is useful for simulating passive technologies such as radiant systems, buildings with demand limiting controls, undersized systems, and systems that properly modulate the flow on and off. The heat balance in the space is then used to compute the correct space temperature, which in turn, informs occupant comfort and surface temperatures. This coupled approach to simulation also shows up in central plant simulation, where the instantaneous fluid temperatures are calculated and then affect the air system by limiting capacity, affecting zone conditions. The iterative approach in EnergyPlus is a major contribution to longer runtimes, however coupling these simulation domains together is key in providing engineers with a tool that better simulates a wide variety of best-practice and cutting-edge technologies.
It would be great to give a description of how DOE-2 works and explain the differences.
This post is actually in answer to the comment asking for a description of how DOE-2 works.
DOE-2 actually consists of two independent programs - doebdl, the input processor, and doesim, the simulation engine.
doebdl reads a user input file, which could be as short as a dozen lines or as long as 10,000 lines, and creates from that two binary building description files containing all the information needed by doesim. doebdl takes virtually no time at all to run, but minimizes the need for doesim to digest the input information.
doesim originally consisted of four modules - LOADS, SYSTEM, PLANT, ECONOMICS - that are run sequentially, although I'm told that in DOE-2.2 SYSTEM and PLANT have been combined. LOADS simulates the heat gains and losses into each space for an entire run-period (year) at a user-specified reference temperature. The simulation is then repeated in SYSTEM, which derives the true temperature in each space, taking into account the heat input or extraction provided by the HVAC system, as well as the effect of outside ventilation air. If there is no HVAC, SYSTEM solves for the space floating temperature. If there is HVAC, SYSTEM simulates the actions of the HVAC and returns the heat input or extraction (up to the system capacity, natch), energy consumption, and final space temperature. For buildings with a central plant, the simulation is repeated once more with PLANT (DOE-2.1E only), which simulates the actions of the big stuff to meet the SYSTEM demand for heating/cooling by fluid type. ECONOMICS allow users to input utility rates and charges,and derive energy costs.
In comparison to EnergyPlus, DOE-2 is more loosely coupled. However, that doesn't mean there's no feedback between the three modules. For example, when SYSTEM solves for the zone temperature, the space loads are recalculated using the zone Weighting Factor, and the conductive loads adjusted by the Zone Conductance. As implied earlier, the effect of undersized HVAC on zone conditions is taken into account. However, up through DOE-2.1E there was no feedback from PLANT to SYSTEM, which was solved in DOE-2.2 by merging the two modules.
DOE-2 was designed in the late 1970's, when computers were much less powerful (I remember a 5-zone building taking 40 minutes on an IBM 370 that would now take 5 seconds on a PC), so computational efficiency, i.e., speed, was imperative. This applied not only to the overall solution technique, but throughout the program design. For example, (1) I/O is handled largely in binary, with the weather data read in two-week chunks, and output reports converted to text once at the end, (2) shading uses a very fast algorithm and done only for one day of each month. I was asked to do some timing comparisons in a 2013 study and found DOE-2.2 to be from 22 to 81 times faster than EnergyPlus (see here, p. 98 (120/152 absolute)).
Of course, there's always the danger of losing accuracy ...
Thanks for the descriptive answer. I appreciate the note about DOE-2 originally taking 40 minutes to run when it was introduced in the 1970s. At that time, I wonder how many people in the industry were concerned with this runtime not being worth the improved accuracy that came along with it. Imagine how the industry would be affected today if, back then, people turned away from DOE2, and it was never adopted in favor of other quick methods in use at the time.
Industry didn't touch DOE-2 until it was ported to the PC in the mid-80's. However, there was general consensus that dynamic hourly simulations were more accurate and insightful than other methods, such as degree-day or bin data calculations, so industry by and large accepted their use for setting building energy standards, etc. The situation now, which you seem to imply by extension, is not exactly the same, since there are now multiple simulation engines arouhd, so the choice boils down to what you need to do and how much effort is required in so doing ?
Other factors not mentioned include EnergyPlus' use of sub-hourly time steps (DOE-2 is hourly), its separate treatment of radiative and convective heat transfer (DOE-2 combines these), and its modeling of HVAC at the component level (DOE-2 models HVAC at the system level). Essentially, DOE-2 combines and compacts phenomena and effects that EnergyPlus handles explicitly. This results in a loss of flexibility and perhaps a loss of accuracy in some situations (although the combined approximations themselves are chosen because they are valid in most cases) but does deliver computational efficiency.
Comments