A neural network based several-hour-ahead electric load forecasting using similar days approach

This paper presents a practical method for short-term load forecast problem using artificial neural network (ANN) combined similar days approach. Neural networks applied in traditional prediction methods all use similar days data to learn the trend of similarity. However, learning all similar days data is a complex task, and does not suit the training of neural network. A Euclidean norm with weighted factors is used to evaluate the similarity between the forecast day and searched previous days. According to similar days approach, load curve is forecasted by using information of the days that are similar to weather condition of the forecast day. An accuracy of the proposed method is enhanced by the addition of temperature as a major climate factor, and special attention was paid to model accurately in different seasons, i.e. Summer, Winter, Spring, and Autumn. The one-to-six hour-ahead forecast errors (MAPE) range from 0.98 to 2.43%. Maximum and minimum percentage errors, and MAPE values obtained from the load forecasting results confirm that ANN-based proposed method provides reliable forecasts for several-hour-ahead load forecasting.