Neural network-based load forecasting and error implication for short-term horizon

Load forecasting is considered vital along with many other important entities required for assessing the reliability of power system. Thus, the primary concern is not to forecast load with a novel model, rather to forecast load with the highest accuracy. Short-term load forecast accuracy is often hindered due to various load impacting factors. Two of the major impacting factors are day-ahead weather forecast and subsequent variation in electricity demand that is independent of weather. To tackle the uncertainty in short-term load forecasting, this paper presents a neural network-based load forecasting technique for short-term horizon based on data corresponding to a U.S. independent system operator. With the real life data, a better understanding of forecasting error is carried out while further identifying the time periods when the load is supposedly to be over- or under-forecast.