Machine Learning based Load Prediction

Load forecasting is very important tool for energy suppliers and also for other participants in electric energy generation, transmission and distribution systems. It plays an important role in the power system planning and operation. Load forecasting has great impact on power system applications such as energy purchasing, energy generation and infrastructure development. Many mathematical methods are used for load forecasting. The load forecasting reduces capital investment on the equipments to be installed. In this project, the focus is on Short-Term Load Forecasting (STLF) which is an hourly load forecasting on next day. In our proposed project STLF will be carried out using one of the conventional methods, Multiple Linear Regression (MLR) and modern method, Artificial Neural Network (ANN). The ANN and MLR here use the data such as past load and weather information like humidity and temperatures. Once the Artificial Neural Network is trained for the past set of data, it can give prediction of future load. Finally, there is a comparison drawn between MLR and ANN on the basis of their Mean Average Percentage Errors (MAPE).