Please cite this article in press as: K.S. Metallinos, et al., Derivation and evaluation of generic measurement-based dynamic load models, Electr. Power Syst. Res. (2016), http://dx.doi.org/10.1016/j.epsr.2016.06.022 ARTICLE IN PRESS G Model EPSR-4742; No. of Pages 8 Electric Power Systems Research xxx (2016) xxx–xxx Contents lists available at ScienceDirect Electric Power Systems Research j o ur na l ho mepage: www.elsevier.com/locate/epsr Derivation and evaluation of generic measurement-based dynamic load models Konstantinos S. Metallinos a , Theofilos A. Papadopoulos b,∗ , Charalambos A. Charalambous a a Department of Electrical and Computer Engineering, Faculty of Engineering, University of Cyprus, PO Box 20537, 1687 Nicosia, Cyprus b Power Systems Laboratory, Department of Electrical and Computer Engineering, Democritus University of Thrace, 67100 Xanthi, Greece a r t i c l e i n f o Article history: Received 25 January 2016 Received in revised form 10 April 2016 Accepted 9 June 2016 Available online xxx Keywords: Artificial neural networks Generic models Load modelling Measurement-based approach Nonlinear least-square optimization Statistical analysis a b s t r a c t Online recorded responses can be used for aggregate dynamic load modelling, taking advantage of the advent of smart grids and the growing installation of phasor measurement units. Although sev- eral measurement-based dynamic load models have been proposed in the literature, still most network utilities and system operators take advantage of well-known formulations such as the polynomial and the exponential recovery models. However, these types of load models are only valid for a specific range of operating conditions, thus minimizing their applicability and efficiency. This is mainly due to the fact that the model parameter estimation procedure relies on iterative processes. To this extent, the specific scope of this paper is to present a comprehensive identification procedure for evaluating load models under different loading conditions and further to propose two generic modelling approaches that can be used to derive robust load models that are suitable for dynamic simulations over a wide range. Towards achieving the scope of this paper, Monte Carlo simulations are used to train and validate the data of the loading conditions. Finally, several simulations are performed within the DIgSILENT PowerFactory software to assess the accuracy of the proposed models. © 2016 Elsevier B.V. All rights reserved. 1. Introduction Aggregate load models represent the overall coordinated behaviour of individual electric and electronic components, such as motors, lighting, electrical appliances, etc. supplied by a common power system busbar [1]. The impact of the accurate representation of the steady-state and dynamic characteristics of power system loads has been long recognized and investigated both at the trans- mission and distribution network levels, especially considering studies pertaining to voltage and angular stability [2–6]. Neverthe- less, in the last decades the increased penetration of motors and the introduction of new power electronic interfaced loads com- bined with the need for more operational flexibility as well as the application of advanced voltage and frequency control strategies in the distribution network have renewed the scientific interest on load modelling [7]. Although, detailed load models provide very accurate load representation, this method requires signifi- cant computational power and large simulation times for extended ∗ Corresponding author. Tel.: +30 2541079568; fax: +30 2541079568. E-mail address: thpapad@ee.duth.gr (T.A. Papadopoulos). networks. However, detailed information of the real load charac- teristics is rarely available to transmission and distribution utilities. Thus, load busbars are represented by equivalent models repre- senting the aggregation of different individual components, such as static, inductive and capacitive loads, motor driven consumers, etc. [1,6]. In aggregate load modelling there are two main approaches: the component-based and the measurement-based [8]. The former involves the derivation of an aggregate load model based on infor- mation from its constituent parts, including [8]: (a) the load class mix (industrial, agricultural, residential), (b) the composition of each of those classes (heating, cooling, air conditioning, etc.) and (c) characteristics of each load component related to the corre- sponding physical characteristics. The advantages of this approach are that it does not require field-measurements and that it can be easily applied to different bus substations [1]. However, the component-based approach is not considered accurate enough to represent the distinct load characteristics under various system dis- turbances. On the other hand, in the measurement-based approach the model parameters are typically defined a priori and their val- ues are estimated using system identification techniques to fit the input–output data obtained from measurements. Significant http://dx.doi.org/10.1016/j.epsr.2016.06.022 0378-7796/© 2016 Elsevier B.V. All rights reserved.