HYBRID LEVENBERG-MARQUARDT AND CUCKOO SEARCH ALGORITHM FOR SHORT-TERM LOAD FORECASTING IN ELECTRICAL POWER SYSTEMS

Abstract

Short Term Load Forecasting (STLF) of electricity supply is important for optimal planning and operation of an electric power system. However, finding an appropriate STLF model for a specific electricity network has become imperative as in-appropriate model yields sub-optimal solution and may cause a huge power loss to electrical power system. This study carry out STLF of electricity supply on Nigerian electric power system using Lavenberg-Marquardt (LM) based Back Propagation (BP) algorithm trained with Cuckoo Search (CS) algorithm. An Artificial Neural Network (ANN) forecasting model that can forecast electricity supply for one day ahead was designed using daily historical hourly load data from Ayede transmission sub-station feeder Ibadan. The initial weight parameter of the designed forecasting model was optimized with CS algorithm. Then, LM based BP algorithm was used to train the forecasting model and simulation was done in MATLAB R2021a. The objective function is to improve the prediction accuracy and minimize the forecasting model error value. The results showed that, the application of CS with LM based BP algorithm for STLF of electricity of Ayede transmission sub-station feeder improved the forecasting accuracy of the feeder, reduced the error value to minimum and provide a reliable tool to optimize the operation of power system. Hence, this study provides a viable solution to the load growth or expansion problem of electric power system