PERCEPTRON CLASSIFIER OF THERMAL COMFORT

Abstract

Algorithmic control of automated systems by human thermal comfort for the establishment of optimal thermal comfort with the use of classification and computational capabilities of the simplest single layer neural network – perceptron. Classification  of  the  hyperspace  of  the  states  of  the  thermal  comfort  system  by  an artificial neural network, mathematical analysis of the classification equations of the hyperplanes, formed by the internal periphery perceptron neurons, optimization of the thermal comfort state by determining of the coordinates of the projection of the thermal comfort vector-image on the hyperplane. Findings. Development of methods and models of neural networks for forming of the algorithm of control  of  human  thermal  comfort  system.  The  use  of  classification  and  computational  properties  of  the perceptron as an instrument of geometric interpretation of the transition from the real to the desired state of thermal comfort s in the multidimensional hyperplane space has been studied. Such an approach made it possible to abandon of the formulation of the analytical optimization function.
Originality. The classification properties of the artificial neural network have been developed and extended to the class of systems of thermal comfort. For the first time, the method of Kačmaža was used to further formalize the algorithm of environmental parameters change. Practical value. Energy efficiency of complex multifactorial dynamic technical systems is ensured not  only  by  advanced  modern  equipment,  but  by  rational  management  models  also.  Neural  network technologies  allow  to  use  the  classification  and computational  capabilities  of  artificial  neural  networks optimally to form commands for the devices of thermal comfort systems managing.