Minhas, Daud Mustafa; Khalid, Raja Rehan; Frey, Georg: Short Term Load Forecasting Using Hybrid Adaptive Fuzzy Neural System: The Performance Evaluation. 2017 IEEE PES PowerAfrica Conference (PES Africa 2017), PP 468 - 473, Accra, Ghana, Jun. 2017.


In this paper, an evaluation theory of hybrid model for short-term electricity load forecasting is presented using simple soft-technique of predicting data. A model that integrates fuzzy system with neural network database is demonstrated and eventually compared with a traditional statistical method of linear regression. Power load forecasting errors especially for weekends, which is much higher than that of weekdays, is reduced using the probabilistic and stochastic natured Hybrid Adaptive Fuzzy Neural System (HAFNS) method. Neural network database uses temperature and power loads as predictors to train the data sets and then use fuzzy system to develop membership functions forecasting future power load demands for subsequent hours. HAFNS model is made using recent power load and temperature figures of 2015. The training and testing set of HAFNS is composed of  yearly data, which may be decomposed on monthly, daily and hourly basis for comparison. The simulation results of the forecasted data including error distribution graphs are demonstrated.


Smart grid, Load forecasting, Fuzzy logic, Neural network