A Directed Weather Data Filter

It is important to reduce the computational burden of energy-analysis computer programs (such as those that compute expected heating and cooling loads and the energy required to meet these loads) since the input may include, for example, hourly measurements of several weather variables over periods of time of up to ten years. It is possible to generate a small set of weather data to represent a larger, more comprehensive set by using computer programs called weather data filters. However, existing filters that are generally able to preserve statistical properties of the original data offer no a priori method for controlling or even estimating the errors they introduce in the output of the energy analysis programs. This problem exists because these filters deal only with weather data and do not utilize any information regarding the building to be analyzed, its occupancy schedule, its mechanical systems, or the energy analysis function itself. This paper describes a weather filter that preserves the “energy content” of the data; i.e., given a particular building design and usage schedule, this filter generates a small set of weather data which, when properly weighted, produces the same total heating and cooling loads as the original data. Because the error caused by using the filtered weather data falls well within acceptable bounds even when changes are made in the building design, the filter described allows the designer to test a large number of design or retrofit alternatives for energy efficiency at very low computational cost.