期刊
CLEAN, EFFICIENT AND AFFORDABLE ENERGY FOR A SUSTAINABLE FUTURE
卷 75, 期 -, 页码 3014-3026出版社
ELSEVIER SCIENCE BV
DOI: 10.1016/j.egypro.2015.07.616
关键词
short-term load forecasting; time series; ARMA; transfer function; neural network
Short term load forecasting, ranging from a few hours ahead to a few weeks ahead has great importance in the operations and planning of the electric power system. As forecast accuracy increases, the overall system can be operated closer to its optimal point, directly affecting its profitability and stability. In this paper, measured hourly weather variables (temperature, specific humidity, Global Horizontal Irradiation and wind speed) were used for modelling and forecasting the electricity load for the city of Abu Dhabi, UAE. A Transfer Function (TF) model was developed and its average accuracy measured using 30 one-week forecasts generated every day over a period of one month. The accuracy of the TF method was compared to an Autoregressive Integrated Moving Average (ARIMA) model and to an Artificial Neural Network (ANN) model based on the same exogenous variables. When perfect knowledge of the exogenous variables over the forecasting horizon was assumed, the TF model had better accuracy for one-and two-day forecasts, while the ANN was more accurate for one-week ahead forecasts. With a more realistic scenario, where the exogenous variables are not known over the forecasting horizon and have to be forecasted before being used in the load forecast, the TF model had better accuracy than the ANN approach for all three tested forecasting horizons. Average accuracy of the preferred Transfer Function method is better than 1.5% for 24-hour horizon, better than 2.5% for 48-hour horizon and better than 4% for 168-hour horizon. With the added uncertainty of forecasted weather drivers, the accuracy of the proposed method degrades only slightly, while the neural network approach degrades rapidly and becomes unusable beyond a two-day horizon. (C) 2015 The Authors. Published by Elsevier Ltd.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据