リカレントニューラルネットワークによるビル空調システムの消費電力推定

―予測に向けた特徴ベクトルの選定―

抄録

This article proposes an estimation method of power consumption for building air conditioning systems with multiple indoor units per outdoor unit. In the proposed method, a recurrent neural network is used to estimate the power consumption from current states of the building. First, the power consumption while the number of indoor peoples is increased is estimated from the people number and both the outside temperature and the room temperature to investigate the influence of these data on the power consumption. The evaluation result showed that estimation accuracy of the power consumption can be improved by considering the number of indoor peoples compared with the case of only the outside temperature and the room temperature are used. Next, by replacing the indoor temperature with the preset temperature, the algorithm is modified to be suitable for predicting the power consumption. It has been also found that the estimation accuracy is kept even if the room temperature is replaced with the preset temperature, and this means the prediction can be feasible.