Modeling and Estimating of Load Demand of Electricity Generated from Hydroelectric Power Plants in Turkey using Machine Learning Methods

Abstract

In this study, the electricity load demand, between 2012 and 2021, has been estimated using the load demand of the electricity generated from hydroelectric power plants in Turkey between 1970 and 2011. Among machine learning algorithms, Multilayer Perceptron, Locally Weighted Learning, Additive Regression, M5Rules and ZeroR classifiers are used to estimate the electricity load demand. Among them, M5Rules and Multilayer Perceptron classifiers are observed to have better performance than the others. ZeroR classifier is a kind of majority classifier used to compare the performances of other classifiers. Locally Weighted Learning and Additive Regression classifiers are Meta classifiers. In the training period conducted by Locally Weighted Learning and Additive Regression classifiers, when Multilayer Perceptron and M5Rules classifiers are chosen respectively, it is possible to obtain models with the highest performance. As a result of the experiments performed using M5Rules and Multilayer Perceptron classifiers, correlation coefficient values of 0.948 and 0.9933 are obtained respectively. And, Mean Absolute Error and Root Mean Squared Error value of Multilayer Perceptron classifier are closer to zero than that of M5Rules classifier. Therefore, it can be said the model performed by Multilayer Perceptron classifier has the best performance compared to the models of other classifiers.

Keywords

• electricity load forecasting
• machine learning
• multilayer perceptron
• rule based learning
• time series prediction