Data driven and fuzzy techniques for wind speed calculation inside urban canyons

Wind speed characteristics in the centre of a city differthan the ones in its suburban. Street canyon effects result in weak airflows in the centre of highly dense citiesbecause of the combination of inertia and gravitationalforces. A huge experimental campaign took place in the centre of Athens in order to create a database of the main parameters that affect wind speed characteristics.Previous studies aimed to calculate wind speed in urban canyons by using deterministic techniques. The agreement between experimental and calculated valued present a low accuracy, for most cases of ambient incidencewind flow, because of the incomplete description of the physical phenomena that determine wind flow insidestreet canyons. This paper present alternative data driven and fuzzy techniques able to estimate the wind speed in urban canyons. Four different relationships were estimated between inertia and gravitational forces and that were the criterion of clustering the input-output data. For each cluster using statistical analysis, the more probable wind speed inside the canyon, was calculated. Three different data driven prediction models were developedin order to calculate wind speed inside street canyons. These were a 3D graphical interpolation method,a decision tree method as well as a linear regression model. Using the results of the graphical interpolation model, a fuzzy estimation model has been developed as well. All methods have been compared against the experimental data. The values for wind speed derived from the fuzzy model followed closely the experimentaldata and predicted with sufficient accuracy the most probable wind speed in urban street canyons.