A Confirmation Technique for Thermal Performance Simulation Models

摘要

This paper deals with the problem of empirical validation ofrnthermal performance computer programs. It begins with arnbrief review of a number of techniques which have been usedrnas a measure of the goodness-of-fit between measured andrnpredicted data in a variety of empirical validation exercises.rnSeveral inadequacies inherent in existing techniques arernidentified as,rna) no attempt is made to take into account thernseverity of the validation test.rnB) none give a single measure of the success (orrnotherwise) of the test.rnC) isolation of sources of error are difficult.rnD) tests cannot be used easily for internalrnvalidation and/or algoriffim "tuning".rnAn argument is presented that an objective technique forrnestablishing the accuracy of simulation predictions and whichrnaddresses these inadequacies is required. To satisfy thisrnrequirement a confirmation factor Cs based on an inequalityrncoefficient is defined. Following from Cs it is shown that arndegree of confirmation D may be evaluated.rnMultiple input variables may be taken into account byrncalculating Cs and D from the estimates of the PrincipalrnComponents derived from independent linear fúnctions ofrnthe original variables.rnThe confirmation technique is illustrated with an examplerncomparing the predictions of the computer programrnTEMPAL with internal temperatures derived fromrnmonitoring a house. Possible sources of program error arerninvestigated and the improvements made with a newrnprogram EnCom2 are shown.