Load -Price Forecasting Model Employing Machine Learning Techniques

Short term load forecasting is always an important study from operational and planning point of view. But short term price forecasting is a new topic. In this study, with the implementation of machine learning techniques, a new algorithm is proposed to predict both load and price values. A machine learning techniques such as Principle Component Analysis, and K nearest neighbor points, are applied as preprocessing techniques to assist the Support Vector Machine. The proposed model is implemented to build a closed loop dynamic model to forecast both 24 hours a head short term load and price values, which is a common study in markets. Because of leak of data, data proposed in this study are load, price time series only which are available in markets sites. Load and price volatilities time series are also needed to be obtained. The model has been trained and validated using data from, Australia electricity market. Moreover, the model has been tested using data from North England