A Comparative Analysis of Various Stochastic approaches for Short Term Load Forecasting

One of the biggest challenge that distribution system operators are facing today is unpredictable load demand. Various research methodologies including deterministic as well as stochastic techniques are used to determine the load. Although deterministic methods are effective but in certain new cases they fail to predict load accurately. This paper aims at determining load using various stochastic approaches for Short Term Load Forecasting (STLF). A historical data of load for 5 years is taken using various dependent parameters. This data is trained using various classifiers techniques. These classifier techniques include K Nearest Neighbours (KNN), Naive Bayes, Decision tree, Support Vector machines(SVM) and Ensembles. The best classifier technique is fed to various regression learners and the most effective regression learner is determined. The algorithms are implemented in PYTHON and MATLAB Platforms. The best method for both classifier and Regression is obtained as Ensemble with 95.1% accuracy and root mean square error of 0.02.