Self-learning adaptive-network-based fuzzy logic power system stabilizer

Power System Stabilizer Design Using an Online Adaptive Neurofuzzy Controller With Adaptive Input Link Weights

“…However, when considering the use of an FLC working as an adaptive fuzzy PSS, investigators have usually focused on the parameter optimization of the MFs and the rule base [8]- [11], and little attention has been given to the SFs of the controller. By modifying the SFs related to a certain variable, the corresponding working range will enlarge or reduce, producing a change in the sensitivity of the controller for the input variables or in the gain for the output variable.…”

“…4, and consists of two main components: an adaptive neuroidentifier to track the dynamic characteristics of the plant, and an ANFPSS to provide the appropriate control actions for damping the electromechanical oscillations of the system. Some configurations found in the literature are ANFPSSrecursive least squares identifier (RLSI) [9]- [11], ANFPSSadaptive neuro-fuzzo identifier (ANFI) [8], and ANNPSS-ANNI [5]. An ANNI is used in this paper because a neuralnetwork-based plant model with online adaptation has the capability to cope with plant complexity, uncertainty, nonlinearity, and variations with time (either expected or due to failure) [16].…”

“…A common practice in the design of this type of PSS to obtain an optimum performance is to adjust the input membership functions (IMFs) and consequent parameters (CPs) [8]- [11]. However, SFs also play a very important role in the successful design of the controller [12].…”

Comparative Performance of Neuro-Fuzzy PSS Architectures with Adaptive Input Link Weights and Nonlinear Functions

“…Besides, it has been proven that by using a series-parallel model a feed-forward network of a single layer with a finite number of nodes can implement any continuous nonlinear function (Nguyen and Widrow, 1990). For the proposed configurations, as the ANNI simply functions as a black box, there is no need to use a fuzzy system as in Hariri and Malik, 1996.…”

“…A fuzzy logic controller (FLC) based on an adaptive network architecture provides a suitable medium to optimize its parameters by applying a gradient descent training algorithm (Jang et al, 1997). Therefore, it has been used to develop PSSs with enhanced performance and increased robustness (Hariri and Malik, 1996;You et al, 2003;Barton, 2004). The common practice in the designs has been the tuning of input membership functions (IMFs) and consequent parameters (CPs) of the rules.…”