Residential Electricity Demand Forecasting using Data Mining

摘要

In this paper, the proposed system is designed which predicts the electricity demand. Data mining techniques are used such as data cleaning, data smoothing to get the data required for prediction.The Artificial Neural Network (ANN) plays a great role in forecasting the electricity consumption. The existing methodology used to find the electricity consumption and demand prediction for household used ANN, data mining and data preprocessing. The context features like weather, temperature, humidity and public holiday are used as input for the prediction system. Along with context features seasonwise electricity consumption forecasting to achieve improved accuracy is done using proposed system which is based on Support Vector Regression (SVR) and Linear Regression (LR). LR and SVR gives better accuracy than the existing system. LR produces the MAPE value of 0.59% and SVR produces MAPE value of 0.11%. The RMSE (Root Mean Squared Error) performance metrics is used to evaluate the system performance. The RMSE value for LR is 0.73 and for SVR it is 0.34.