دانلود رایگان مقاله انگلیسی مدلسازی و کنترل ریزشبکه مبتنی بر اینورتر - الزویر 2018

Abstract:

Assuming the most common control structure for zero and primary control of inverter-based microgrids, i.e. three cascades with the highest one being droop control, the potential benefit of optimizing the control parameters is investigated. A detailed nonlinear plant model is derived that compactly describes the dynamics in local dq-coordinates. Then, the design of the decentralized, cascaded controllers is converted into the problem of designing one centralized static controller with structural restrictions. To tune the controller parameters, a direct method for pole-assignment is used. The simulations show that the oscillations in the transient response can be reduced greatly by choosing appropriate control parameters, while the speed of the system is restricted due to the low-pass filtering of the power for primary control.

9. CONCLUSION AND OUTLOOK

It has been shown that the oscillations in a microgrid can be reduced considerably just by tuning the controller parameters. The overall system speed is restricted by the low-pass filter parameters ωc, i.e. by the higher oscillations that were not modeled in this work. Still, it might be possible to enhance the system behavior even further by generating more degrees of freedom. To investigate this possibility, the commonly used controller structure with three cascaded controllers should be reconsidered. Instead, the application of general multiple-input multipleoutput controllers, one for zero-level control and another for primary control, should be investigated. Especially breaking the structure of the zero-level controller seems promising. First of all, there is no need for two cascaded PI controllers. Therefore, the dynamical order of the controller could be reduced. At the same time, the degrees of freedom could easily be increased from the five control parameter available with the common control structure KpC, KiC, KpV, KiV, FpV to twelve degrees of freedom of a controller matrix with six inputs and two outputs plus the degrees of freedom due to the dynamics of the controller. Since the zero-level controller uses unfiltered measurements, it is much faster than primary control. Therefore, one might also be able do dampen the faster oscillating modes, which should reduce the content of harmonics in the network and might allow to raise the bandwidths ωc of the low-pass filters.

Also, the optimization of the control parameters had predefined eigenvalue locations as objective, which often are not crucial. Additional degrees of freedom from point of view of the optimization can be generated, when an area is specified in which the eigenvalues have to lie, instead of precise locations, e.g. Konigorski (1987a).

We designed the controllers simultaneously. This has the advantage that each controller is designed considering the other controllers. And yet, this is not necessarily the best approach, since the optimization problem that must be solved for eigenvalue assignment is not convex. With the size of the network, this drawback becomes more severe. Considering this, a better approach might be to design zero and primary controllers independently based on matched models. Besides this time-scale decomposition, other decomposition techniques like the -decomposition and the overlapping decomposition, e.g. Siljak (1991), should be considered for larger networks.