کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
264130 | 504093 | 2011 | 8 صفحه PDF | دانلود رایگان |
![عکس صفحه اول مقاله: New artificial neural network prediction method for electrical consumption forecasting based on building end-uses New artificial neural network prediction method for electrical consumption forecasting based on building end-uses](/preview/png/264130.png)
Due to the current high energy prices it is essential to find ways to take advantage of new energy resources and enable consumers to better understand their load curve. This understanding will help to improve customer flexibility and their ability to respond to price or other signals from the electricity market. In this scenario, one of the most important steps is to carry out an accurate calculation of the expected consumption curve, i.e. the baseline. Subsequently, with a proper baseline, customers can participate in demand response programs and verify performed actions. This paper presents an artificial neural network (ANN) method for short-term prediction of total power consumption in buildings with several independent processes. This problem has been widely discussed in recent literature but a new point of view is proposed. The method is based on two fundamental features: total consumption forecast based on independent processes of the considered load or end-uses; and an adequate selection of the training data set in order to simplify the ANN architecture. Validation of the method has been performed with the prediction of the whole consumption expressed as 96 active energy quarter-hourly values of the Universitat Politècnica de València, a commercial customer consuming 11,500 kW.
► An artificial neural network (ANN) method is presented for short-term prediction of total power consumption in buildings.
► The method is based on end-uses (EUs) to establish accurate relationships between inputs and each EU.
► Inputs in the ANN architecture have been reduced to physical variables to avoid introducing too much variability.
► A small number of training days, but with consumption close to the day of prediction is chosen to train the ANNs.
Journal: Energy and Buildings - Volume 43, Issue 11, November 2011, Pages 3112–3119