An Association-Rule Method for Short-Term Electricity Demand Forecasting and Consumption Pattern Recognition

Electric demand forecast is an important tool for the effective operation of electrical power systems. The lack of accuracy in the estimation of the electric demand may cause an excessive electric supply to the power system, resulting into an excessive amount of electricity purchase. Consequently, disturbances in the power lines appear which can damage energy generation equipment. On the other hand, an underestimation of the electricity demand may cause undesired power outages. Electricity demand is influenced by human behavior on a day-to-day basis. Modeling such behavior can enhance the accuracy of the electricity demand forecast. Towards that end, in this paper we investigate a modeling approach based on association rules (association rules are useful to describe a model in terms of cause and effect). The proposed approach aims at predicting electricity demand for two hours ahead in 8 periods of 15 minutes. The data set is from a representative zone of Mexico, which is 15-minute load demand measures. To evaluate the proposed method, we compare its performance with a popularly used method (ARIMA). Experiments show that the methodology presented in this work does not outperforms the popularly used method, but helps to locate the most frequent patterns of electricity consumption.