Prediction in amounts of photovoltaic generation and building energy consumption with recurrent neural network

Abstract

Currently, the machine learning model method for energy prediction of ultrahigh resolution, which is the prediction of consumption of building energy in domestic and overseas, is difficult to approach. At the same time, prediction performance decreases due to the increase of randomness, and the cyclic regression network using weather data. The photovoltaic power generation prediction model utilized also meets the domestic and international social backgrounds where the use of renewable energy is increasing. This study suggests a paradigm of a three-dimensional energy management system that manages the energy consumption of the building itself by producing the energy itself in a local energy management method that manages the consumption of the building.