Predicting Monthly Electricity Demand Using Soft-Computing Technique

Electricity is an essential commodity for all. The generation, transmission and distribution (GTD) of electrical energy needs much planning since every megawatt transmitted for distribution depends on consumers demand. An accurate estimation (prediction) of future demand helps prevent a power shortage and load scheduling (“Dumsor”). This study proposed a soft-computing technique based on multi-layer perceptron (MLP), support vector machine (SVM) and decision tree (DT) algorithms to predict a 30-day head electrical energy demand. Using three years of real-world historical weather and electrical energy demand data from Bono region of Ghana, we experiment with the proposed model. The obtained accuracy of 80.57% for DT, 95% for MLP and 67.2 for SVR and RMSE values of 0.064221, 0.021184 and 0.100776 for DT, MLP and SVR respectively revealed the efficiency of the proposed model in predicting future electrical load.