Generic dynamic load models using artificial neural networks

The increased availability of measurements in modern power systems, due to the advent of smart grids and the installation of phasor measurement units, has favored the development of dynamic load models using online recorded responses. However, load model parameters are significantly affected by loading conditions and change considerably due to the time-varying and weather-dependent composition of load. Therefore, load model parameters obtained from in-situ measurements are valid only for a narrow range of operating conditions. Scope of this paper is to propose a systematic identification procedure to develop generic dynamic load models, valid for a wide range of discrete operating conditions. For this purpose, two different generic modeling approaches are considered. The first approach is based on statistical analysis, while the second employs Artificial Neural Networks (ANNs). Several simulation scenarios are performed using the NEPLAN software to investigate the accuracy of the derived models over a wide range of different loading conditions, while their generalization capabilities are evaluated using the cross-validation technique.