Predicting Irregular Power Consumption According to the Indoor Environment with Sensor of Home Appliances
宮澤 重明 ,
志田 匠 ,
一色 正男杉村 博
13 , 2018-03-20 , 神奈川工科大学
We propose a method for predicting a future pattern of the power consumption by mechanically classifying multiple patterns of the power consumption based on sensor information collected from home appliances. Conventional methods for predicting the amount of power consumed by general households or similar small areas are effective for a regular change in power consumption. However, it is difficult to predict an irregular change by those methods. In our method, representative patterns are mechanically extracted from the past changes in the power consumption recorded each day, and a future pattern is predicted from those representative patterns. The past representative patterns are extracted by clustering. The future pattern is predicted by creating a model of the correlation between the consumption pattern and indoor conditions using a classifier. The prediction by the classifier requires data on the indoor conditions. Therefore, we constructed a sensor network using networked home appliances. The use of the home appliances enabled the construction of a sensor network that is easier to maintain and lower in cost than conventional sensor networks. The result of an experiment on the maximum absolute error confirmed the effectiveness of the prediction based on the information collected from home appliances.