In this paper artificial neural networks (ANN) are applied to modeling of electronic circuits. ANNs are used for application of the black-box modeling concept in the time domain. Modeling process is described, so the topology of the ANN, the testing signal used for excitation, together with the complexity of ANN are considered. The procedure is first exemplified in modeling of resistive circuits. MOS transistor, as a four-terminal device, is modeled. Then nonlinear negative resistive characteristic is modeled in order to be used as a piece-wise linear resistor in Chua's circuit. Examples of modeling nonlinear dynamic circuits are given encompassing a variety of modeling problems. A nonlinear circuit containing quartz oscillator is considered for modeling. Verification of the concept is performed by verifying the ability of the model to generalize i.e. to create acceptable responses to excitations not used during training. Implementation of these models within a behavioral simulator is exemplified. Every model is implemented in realistic surrounding in order to show its interaction, and of course, its usage and purpose.

The tremendous rise of electrical energy demand worldwide has led to many problems related to efficient use of electrical energy, consequently posing difficult challenges to electricity consumers of all levels—from households to large companies’ facilities. Most of these challenges could be overcome by the accurate prediction of electricity demand. Additionally, balance responsibility includes the penalty-based financial mechanism causing extra expense for badly estimated consumption, above the allowed imbalance limits. In this paper, a method for electricity consumption prediction based on artificial neural networks is proposed. The electricity consumption dataset is obtained from a cold storage facility, which generates data in hourly intervals. The data obtained are measured for a period of over 2 years and then separated to four seasons, so different models are developed for each season. Five different network structures (ordinary RNN, LSTM, GRU, bidirectional LSTM, bidirectional GRU) for five different values of horizon, i.e., input data (one day, two days, four days, one week, two weeks) are examined. Performance indices, such as mean absolute percentage error (MAPE), root mean square error (RMSE), mean absolute error (MAE) and mean square error (MSE), are used in order to obtain qualitative and quantitative comparisons among the obtained models. The results show that the modifications of recurrent neural networks perform much better than ordinary recurrent neural networks. GRU and LSTMB structures with horizons of 168h and 336h are found to have the best performances.

Electronic devices are complex circuits, consisting of analog, switching, and digital subsystems that require direct current (DC) for polarization. Since they are connected to the mains delivering alternating current (AC), however, AC-to-DC converters are to be introduced between the mains and the electronics to be fed. A converter is an electric circuit containing several subsystems, the most important being the switch-mode power supply, drawing power from the mains in pulses hence it is highly nonlinear. That happens, in reduced amplitude, even when the electronics to be fed is switched off. The process of AC-to-DC conversion is not restricted to feeding electronic equipment only. It is more and more frequently encountered in modern smart-grid facilities giving rise to the importance of the studies referred hereafter. The converter can be studied (theoretically or by measurements) as two-port network with reactive and nonlinear port-impedances. Characterization is performed after determining the port electrical quantities which are voltages and currents. Based on these data power and power quality parameters - power factor and total harmonic distortion- may be extracted. When nonlinear loads are present, one should introduce new ways of thinking into the considerations due to the existence of harmonics and related power components. In that way the power factor can be generalized to total or true power factor where the apparent power, involved in its calculations, includes all harmonic components. After introducing a wide range of definitions used in contemporary literature, here we describe our measurement set-up both as hardware and a software solution. The results reported unequivocally confirm the importance of the subject of characterization of small nonlinear loads to the grid having in mind their number which is rising without saturation seen in the near and even far future.

ICT and energy are two economic domains that became among the most influential to the growth of modern society. These, in the same time, due to exploitation of natural resources and producing unwanted effects to the environment, represent a kind of menace to the eco system and the human future. Implementation of measures to mitigate these unwanted effects established a new paradigm of production and distribution of electrical energy named smart grid. It relies on many novelties that improve the production, distribution and consumption of electricity among which one of the most important is the ICT. Among the ICT concepts implemented in modern smart grid one recognizes the artificial intelligence and, specifically the artificial neural network. Here, after reviewing the subject and setting the case, we are reporting some of our newest results aiming at broadening the set of tools being offered by ICT to the smart grid. We will describe our result in prediction of electricity demand and characterization of new threats to the security of the ICT that may use the grid as a carrier of the attack. We will use artificial neural networks (ANNs) as a tool in both subjects.

Two solutions to the polynomial filter’s transfer function synthesis problem are considered for comparison in the frequency and time domain: the broad class of filters with a critical monotonic amplitude characteristic (CMAC) in the passband and filters which use Chebyshev (C) polynomials. To complete the synthesis procedure for linear phase applications, group delay correctors are considered, for which a convenient approximation procedure is proposed here. Comparisons of the original functions and the corrected ones are performed in the frequency and time domain. It is shown that when CMAC and C are compared as such, the latter is by no means preferable from the selectivity point of view, while the opposite stands when the comparison is based on passband amplitude distortions. When phase-corrected filtering functions are compared, based on circuit complexity and time domain performance, the CMAC are shown to be preferable.