Forecasting short-term grid electricity load in homes that were connected to the smart grid by a novel method: recurrent linear regression

The short and long-term forecasting of grid electrical energy of a province or region is important for the management of the conventional electricity transmission and distribution network. Nowadays, the amount of electrical energy that each residential building has taken from the grid has gained importance within the scope of smart grids. Residential buildings that take electricity from the smart grid can generate electricity with alternative energy sources such as solar energy. Considering this situation, it has been tested in this project with linear regression and artificial neural networks, which are the classical methods for estimating the electrical energy of such a residential building, but, the desired success rate was not achieved. For this reason, a new method is needed to estimate the amount of electricity of a residential building that can generate electricity with solar energy. For this reason, in this study, linear regression method, which produces successful results in many regression problems, has been developed and proposed a new method called recurrent linear regression which is able to model dynamic systems and increase the estimation success in estimating the amount of electricity drawn from the grid at any time and taking historical data into consideration.In order to test and validate the proposed approach, the Sundance dataset, which is shared in the U Mass Trace Repository according to Smart Project was used. To confirm the success of the proposed method, the linear regression and extreme learning machines methods were used for each different data set. Obtained results, which were taken from 59 different residential buildings, showed that lower rooth mean square error (RMSE) values were achieved by recurrent regression compared to linear regression and extreme learning method. The reason for this success is the ability to model the system more successfully thanks to the dynamic modeling capabilities of recurrent methods in time-sequential data sets and signals.