Development of energy saving technique for setback time using artificial neural network

This study is desired to form an adaptive artificial neural network (ANN)-based model for optimal setback period application, thus predicting the suitable time for setback temperature deployment. A TRNSYS-MATLAB co-simulation platform is formulated to simulate a given building envelop with the application of desired control. Here, a single building model was simulated under five different climatic conditions of Taxila (Pakistan), Mogadishu (Somalia), Darwin (Australia), Phoenix (USA) and Havana (Cuba). Five influencing parameters were selected for the study, and using analysis of variance test, it was found that inside air condition is the most influencing parameter for the setback time period. Initially, one model was used for all five climates. Parametrical tuning optimisation was used for making ANN model optimal for all five climatic regions. This made five optimal ANN models separately for all five weathers. Three peak summer months were selected for all five climatic zones in which first two months were used for ANN training and last one month was used for ANN learning. ANN models prediction accuracy was computed by determination coefficient (R-2) and by coefficient of variation for root mean square error (CV for RMSE) . R-2 of ANN control algorithm was as follow: 0.9754 for Taxila, 0.9721 for Mogadishu, 0.8934 for Darwin, 0.9609 for Phoenix and 0.9978 for Havana, while CV for RMSE for Taxila, Mogadishu, Darwin, Phoenix and Havana was 9.88%, 8.83%, 19.49%, 11.34% and 3.98%, respectively. The ANN control algorithm appeared to be energy efficient, saving 12.5%, 2.4%, 8.4%, 2.0% and 18.2% (in kW h) for Taxila, Mogadishu, Darwin, Phoenix and Havana region, respectively, as compared to basic control algorithm.

Automated summary

This study establishes an adaptive artificial neural network model for different climate conditions, optimizing the model through parameter tuning to achieve prediction and energy-saving control in various climates.