M. Jaidane-Saidane

Variability of Predictability of the Daily Peak Load Using Lyapunov Exponent Approach : Case of Tunisian Power System

This study addresses the problem of the expected variability of predictability of the load curves in the Tunisian Company of Electricity and Gas (STEG). For this purpose, the maximum prediction horizon of the daily peak load time series was investigated using the largest Lyapunov exponent estimation. The results may explain the recent observed degradation of forecast accuracy due to the increase in air conditioning load.

The mixture of generalized gaussian model for modeling of the load duration curve: Case of the Tunisian power system

This paper proposes a new estimation technique of the load duration curve (LDC) which is capable to estimate the LDC of any shape and form. In the paper we develop a probabilistic approach, to deal with the heterogeneity and variability of the load distribution based on the estimation of the hourly peak load distribution by mixture of generalized Gaussian distribution (MGG). The model has more flexibility to approximate and to adapt the shape of load data. The update of the model parameters is carried out using an extension of the expectation maximization (EM) algorithm. Numerical results, obtained from the Tunisian power system indicate that MGG technique is accurate, efficient and robust.

The use of Mixture of Generalized Gaussian for trend analysis of the load duration curve: Summer and winter load variability in Tunisia

This paper proposes a probabilistic representation of load duration curve based on the use Mixture of Generalized Gaussian distribution (MGG) which uses eight physicals parameters connected to the load distribution. Based on these MGG parameters the trend evolution of the annual, winter and summer LDC were examined. Load data from the Tunisian Company of Electricity and Gas (STEG) have been used in this paper to investigate historical changes in LDC structure.

Long term multi-scale analysis of the daily peak load based on the Empirical Mode Decomposition

In this paper, an original technique to explore the long term load dynamics using a multi-scale analysis of the daily peak load based on the Empirical Mode Decomposition (EMD) is presented. The signal is decomposed into intrinsic oscillatory components called Intrinsic Mode Functions (IMFs). These modes are derived from the signal itself and not on a specific basis function. In this work, the EMD is used to extract and separate the suitable load component for long term forecast. Physical interpretations and statistical description of the modes are discussed. A comparison is made between load components extracted by the EMD approach and that of a classical multiple linear regression model. The load component predictability was investigated using the mutual information function. Real load data of the Tunisian power systems are used in this study.