Midterm Energy Forecasting using Fuzzy Logic: A comparison of confidence interval estimation techniques

The modern methods for power system load and energy prediction are usually based on artificial neural networks and fuzzy logic, which present satisfactory results. However, the estimation of the confidence intervals can not be applied directly, unlike to the classical forecasting methods. The objective of this paper is to present an optimized fuzzy logic method for midterm energy forecasting, which can use different techniques for the estimation of the confidence interval, such as the statistical calculation based on the forecasting method errors of the training set, the re-sampling technique and a novel analytical mathematical method based on the membership functions. Finally, the next annual energy demand of Greek interconnected power system is estimated analytically. Simultaneously, the standard deviations through the aforementioned techniques are calculated and compared. Key-Words: confidence interval, fuzzy logic, re-sampling technique, midterm energy forecasting