Machine Learning Based Short-Term Load Forecasting for Smart Meter Energy Consumption Data in London Households

The eolved deployment of smart meters has enabledan extensive authority and monitoring on both electricity operating companies and customers. Based on smart meters, it is possible now to establish viable load prediction techniques. These techniques are playing a critical role in modern smart grids by establishing a mechanism of advanced knowledge of energy demand. This information is beneficial for both electricity operating companies and end consumers. The electricity providers can take early decisions to manage an economical and reliable supply of electricity. It can also helps the customers to effectively plan their energy use and thereby reduce their overall consumption and bills. This paper employs machine learning models for building relationships between historical energy consumption readings to build and test a short-term load forecasting. A publicaly available dataset, about London city households consumption information, is used to examine the performance of suggested method. The datasets are organized and preprocessed with the help of Google Colaboratory. The used machine learning models are k-nearest neighbor (KNN) and support vector machine (SVM). The performances of these models are quantified by using metrics such as percentage root mean square error (PRMSE) and mean absolute percent error (MAPE). The lowest MAPE value of 4.13% and PRMSE value of 1.08% are secured.