A Novel Approach to Design and Develop A Learning Based Framework for Electricity Consumption

Electric energy consumption refers to the amount of power used to meet current demands. Global energy consumption has been rapidly growing as a result of industrialization and increased energy demand. The energy consumption of corporate buildings for a long period of time is kept without being used, similar to the data stored at the electricity supplier for residential buildings. To utilize this data, we developed a learning-based framework for electricity consumption prediction. A dataset consisting of date, time, and energy consumption (MW) as parameters was analyzed. It is separated into two parts: 80 percent of the dataset is used for training and the other 20 percent is the testing dataset. Detailed research utilizing different deep learning-based forecasting models and a methodology employing a Recurrent Neural Network model with LSTM (Long Short Term Memory) network architecture was conducted and presented. Adam optimization is carried out to improve the model's efficiency. The model trained with a dataset is applied to forecast the electricity consumption of an hour/day/month/year/according to the user’s requirement. The output of the process will be the amount of energy that might be consumed in Megawatts and the bill according to the consumption in Rupees is generated. For experimentation purposes, 50000 data points for 500 epochs for training the model and 12874 data points for testing the model was considered and further energy consumption and electricity bill is being predicted. The algorithm was observed to be accurate and efficient. A few more advantages and details regarding implementation are discussed later in this report.