Authors : Mohammad Reza Ghodsi, Alireza Tavakoli, Amin Samanfar
DOI : 10.5281/ZENODO.11068777
Volume : 12
Issue : 3
Year : 2023
Page No : 71-84
In recent years, with the penetration of renewable energy sources in power systems, electrical power is produced by non-synchronous generators, which use electronic power devices in their structure and do not have inertia like conventional synchronous machines. As a result, the grid frequency fluctuations cannot be effectively damped and therefore, the stability of the power system is seriously threatened. One of the effective ways to overcome this challenge can be the use of virtual inertia, which is usually done by the virtual synchronous generator controller. However, the use of virtual synchronous generator controller leads to unacceptability of transient active power distribution, active power fluctuations and inverter output power fluctuations in case of disturbances. In this thesis, in order to overcome the problem of virtual synchronous generator, the use of a virtual synchronous generator controller based on deep neural network is presented, which increases the inertia and improves the dynamic stability of the microgrid. The proof of neural network stability and its convergence is based on the Lyapunov stability technique. Also, in an island microgrid with renewable resources, load change, wind power fluctuations, disruption in solar radiation power, and on the other hand, the low inertia of inverters, have a negative effect on the performance of the secondary frequency control loop. To deal with this challenge, the virtual inertial controller by Mu-Synthesis method can compensate for the lack of inertia in AC island microgrid clusters. The virtual inertial controller method is used as a supplement to prevent power imbalance in a microgrid, in the presence of traditional load frequency control, and by using a gain factor resistant to structural uncertainties, it can achieve robust performance. and increase the amount of inertia and increase the damping of the microgrid. By modeling and designing a virtual inertial controller based on a deep neural network, it helps to dampen the fluctuations of the inverter output power and distribute the active power properly and restrain transient state fluctuations in microgrids, and also by designing a virtual inertial controller using the Mu-Synthesis method, in line with the regulation The controller coefficients are optimized and robust, and the stability and performance of this method has been proved based on the structured singular value. The simulation of different scenarios in this treatise has been done in MATLAB/SIMULINK software environment. Also, some scenarios have been practically simulated in the laboratory environment and finally confirm the accuracy of the research results.