jEPlus is work-in-progress, and is likely to remain so in the near future. It was developed to meet (our own) research requirements; therefore it will evolve with its users knowledge and skill. It is a highly specialized tool to fill in a gap in the existing modeling field. Where the next gap is depends on many things, among which the trend of building simulation research is most certainly a crucial factor.
In a typical design project, many options need to be evaluated. These design options can be the thickness of an insulation layer, the size of a window, the construction of the walls, the selection of HVAC system and its operation strategy, or the building form. Of course you can define all design options in jEPlus, using the “Parameter Tree”. A design solution is subsequently a combination of the alternative values of the parameters. If a building design contains a dozen parameters, the number of design solutions can easily add up to millions. In these cases, optimisation techniques have to be used to identify a “good” (NOT “the best”) design solution.
jEPlus' data structure makes it highly suitable for optimisation purpose. When you create the jEPlus project, you are really defining and discretising the search space of your optimisation problem. In our recent paper, we demonstrated an Evolutionary Algorithm (EA) based optimisation procedure integrated with jEPlus can consistently find near optimum solutions in a 115K+ solution space within 200 simulations.
An EA package will be incorporated in the next major release of jEPlus.
Life is full of uncertainties. Why should a building model be an exception? Well, it is not! Most of the modelers just try to ignore uncertainties, because they cannot do anything about it. Monte Carlo simulation has been used in many engineering fields; but it is rarely seen in building simulation.
You can already perform Monte Carlo simulation with the current version of jEPlus, by assigning a sampling profile (probabilistic distribution function and sample size) to each parameter, then running a random sample of the project. After simulation jobs have finished, further manual post-processing is necessary to get the results you want.
There are some significant limitations in the current process, with the lack of Latin Hyper-cube Sampling method, and jEPlus' memory limit being the major issues. The next release will address these issues, and implement suitable post-processing, so that you can get probabilistic distribution of the output variables straight after the simulations.
The implementation of some sensitivity analysis methods, such as Morris Method and regression analysis, are currently being considered for the next release.
We have always assumed jEPlus' users to be 'proper' researchers, who would not find the stock charts that simulation software produces sufficient. They would always want to use 'proper' tools to visualize data. As a result, not much effort has been put into data visualization. However, a quick way to inspect results will be of interest to even advanced users, especially there is not yet a tool for visualizing parametric result without substantial programming/scripting. Since we are doing this in our research anyway, we will make new tools available in due course.
We expected to release jEPlus v2.0 in 2013. However it did not happen as focus was shifted to online simulation and optimisation algorithms. V2.0 is now planned for 2014. It should incorporate key components for Monte Carlo simulation, optimisation, and uncertainty/sensitivity analysis. We are also working on more post-processing flexibilities and results visualization, although whether these will be integrated into jEPlus' GUI, or as a separate utility, is yet to be decided. Beyond that where jEPlus should go, we don't know. We will rely on building simulation researchers to guide its future.
Why not leave your thoughts using the discussion board below? If you want a function urgently, we may be able to do something before the major release.