Evaluation of mid-term load forecasting case study based on adaptive neuro-fuzzy inference system (ANFIS) and artificial neural networks (ANNs)

Abstract

This paper presents the different techniques for Thailand's medium-term load forecasting. It expected that load forecasting can effectively improve the electrical load efficiency of the Electricity Generating Authority of Thailand (EGAT). In addition, the accuracy of load forecasting is an important part of decision-making for power plant investment and the planning of the power distribution system. In this study, the input data has been trained by several predictive models, which have been artificial neural networks (2,3 and 4 hidden layers) and adaptive neuro-fuzzy inference systems. Learning and prediction depends on three important factors, including Thailand's peak load history (simple moving average of 12 months, 9 months, 6 months and 3 months), month codes and Quarterly Gross Domestic Product (QGDP). The results show that training ANN with two hidden layers produces the best predictive performance. The most accurate load forecast using this method is 1.1527% of MAPE and 14.45 minutes of learning time.