Pattern-based Short-Term Load Forecasting using Optimized ANFIS with Cuckoo Search Algorithm

Abstract

Accurate short-term load forecasting (STLF) depends on proper data selection and model development. The research addresses the problem of data selection based on the energy consumption pattern using correlation analysis and hypothesis test. It also employed the use of Cuckoo Search Optimization algorithm (CSO) to improve Adaptive Network-based Fuzzy Inference System (ANFIS) by replacing the Gradient Descent (GD) algorithm in the backward pass of the classical ANFIS model. The aim is to improve the forecasting error and enhance the forecasting time. Based on the conducted experiment CSO parameters for optimal performance of ANFIS were determined and utilized. Based on the results obtained it is observed that CSO-ANFIS with proposed data selection produced low Root Means Square Error (RMSE) and Mean Absolute Percentage Error (MAPE).