Bayesian Inference Using Expected A Posterior Estimation for Prediction of Thermal Index and Power Consumption of Air Conditioner Based on Mean-field Theory

Based on statistical mechanics of information due to the mean-field theory (MFT), we predict environmental variables and electric power consumed by air conditioner in small-scale systems by making use of Bayesian inference using the expected a posterior (EAP) estimation. Here, we clarify that the present method succeeds in providing an optimal thermal index due to the temperature-humidity index (THI) at each sampling point and power used due to air conditioner, if we set parameters under optimal conditions and that this fact is confirmed by exact proof via equality. Also, we find that the provided THI is further improved, if we introduce correlations of environmental quantities between sampling points into the conventional model.