Support vector machines (SVM) based short term electricity load-price forecasting

This paper presents a support vector machine based combined load-price short term forecasting algorithm. The algorithm is implemented as a classifier and predictor for both load and price values. The implicit relationship between price and load is modeled employing time series. A pre-classification technique is applied to reject the unwanted data before starting the process of the data using the proposed model. In the implemented model, support vector machine plays the role of a classifier and then acts as a forecasting model. Principle component analysis (PCA) and K nearest neighbor (Knn) points techniques are applied to reduce the number of entered data entry to the model. The model has been trained, tested and validated using data from, Pennsylvania-New Jersey-Maryland. The results obtained are presented and discussed.