Modeling of Overcurrent and Short Circuit Characteristics Curves in Digital Protection by ANFIS Method

Abstract

The application of renewable energy can produce electricity such as the use of solar cells that are applied at home. This system has the potential to experience disturbances such as overcurrent and short circuit, therefore a protection system is needed. To be able to secure the system quickly, accurately, and according to standards, digital protection is needed. Digital protection requires a standard characteristic curve modeling with the Adaptive Neuro-Fuzzy Inference System (ANFIS). In this paper, we will discuss the modeling of characteristic curves using the ANFIS method, the curves to be modeled are the inverse and definite curves. To produce suitable curve modeling, training data are carried out with 8 different types of Membership Function (MF) inputs. The results of the training that have been carried out the MF Pimf input type has the lowest error, 0.9168%, and the Trimf type has the highest error, 92.3214%. Then do training data, testing data, and checking data for Time Multiplier Setting (TMS) 1; 0.8; 0.6; 0.4; 0.2; and 0.1 to get the value of the FIS modeling. To see the suitability of the reference data and the results of the Fuzzy Inference System (FIS), an average percentage error calculation is calculated. The results of the average percentage of error values for all TMS values from training, testing, and checking are 0.4829%.