ENERGY PREDICTION IN BUILDINGS WITH DEEP LEARNING METHODS

Abstract

Energy consumption is rising because of the change in life style of people. Managing energy of buildings and utilisation of renewable energy resources are very much important. Detecting the energy demand from the historical data set helps to manage the energy consumption pattern of a building. Hybridisation of two methods DWT and LSTM, transformer model and 1D convolutional neural network models are used here. Comparison of these four models is conducted in which the obtained results show that transformer model is efficient. DWT is used to get important characteristics from the non-stationary time series data given in the analysis. A long short term memory method (LSTM) used for the management of energy demand in buildings. Transformer model and 1D Convolutional network are also used for prediction of energy. This method enables monitoring and controlling of energy demand pattern of a building. Thus, this forecasting methods represents to predict building energy consumption.