A Neuro-Fuzzy Logic Based Machine Learning Model for Electrical Load Forecasting

One of the critical aspects of meeting the demands of the energy worldwide is the accurate estimate of electrical loads typically connected to inter connected power systems. However, it is a completely non-trivial challenge as the variables which govern the electrical load are extremely fluctuating in nature with high amount of randomness and non-correlation thereby making the forecasting problem difficult in terms of the accuracy. The work presented in this paper is based on the ANFIS regression process. The essence of the algorithm lies in the fact that the proposed approach uses the errors in each iteration as an exogenous input to the system and hence is used to attains faster convergence with lesser error margin for the system. Moreover, the discrete wavelet transform is used as the decimating and smoothening filter. The performance of the proposed system is evaluated in terms of the accuracy of the system and the percentage error. It has been shown that the proposed work clearly outperforms the baseline approach without the DWT filtering thereby attain a higher degree of accuracy in the forecasting approach.