It is a general knowledge that the Type-2 fuzzy logic control (T2FLC) provides an extra degree of freedom to model fuzzy uncertainties, unlike the Type-1 counterpart that models crispt uncertainties, this freedom comes at the expense complex mathematical calculations that increase the computation time for the control system. While the T2FLC is able to handle these uncertainties better, it still remains an expert-dependent control strategy that requires a proper understanding of the relationship between the input(s) and output(s) variables for efficient mapping. It becomes important to look into model-free and self-learning control strategies that require less dependence on expert knowledge. Also, critics of the T2FLC for power systems application argue that uncertainties associated with power system dynamics are continuos-definite such that a properly designed T1FLC would perform as efficiently as the T2FLC with less computational time.

Thanks so much for the brilliant explanation Anuoluwapo Aluko . You are very much right on your submission, and I quite agree with you entirely. The type-1 FLC performs better than the type-2 FLC even in terms of rise-time, settling-time, and integral of absolute error. Even the robustness of the conventional PID controller can be enhanced using different methods. As you know that we may need to adjust the operating point (like active and reactive power set-points) of our control at times in order to mitigate the effects of grid fault/disturbance and intermittency (uncertainties). Hence, are there any tested model-free control frameworks one can apply in such a power system application.

It is important to point out that the concept of model-free controllers for power system applications does not exclude the knowledge of the system for which the control is being designed, rather, it focuses on the adaptive learning of the controller as operating points change. To design model-free controllers, adaptive and reinforcement learning methods come to mind.

To justify Omar H. Abdalla

Looking at the problem from system architecture side of solution and predictive learning process solution, I suggest that you introduce a Bayesian based fuzzy logic approach to get several prior probabilities and possibilities of the power system controllers and build upon a predictive Bayesian learner . The Bayesian based uncertainties functions use for fuzzy rules building can easily handle the uncertainties problem computation and process time. You can test this solution as you want with future controllers.