{"title":"Deep learning framework for short term power load forecasting, a case study of individual household energy customer","authors":"Khursheed Aurangzeb, Musaed A. Alhussein","doi":"10.1109/AECT47998.2020.9194153","DOIUrl":null,"url":null,"abstract":"Due to the seamless benefits of the integration of Distributed Energy Resources (DERs) for the residential customers, the forecasting of the short term power load of individual household energy customer is becoming an essential task for the future operation and planning of the smart grids. Recently, different studies concluded that due to lack of fast connectivity and awareness, the energy customer were not able to exploit the benefits of the DERs to the full extent. Nevertheless, with the rapid advancement in connectivity, data analytics, internet of things, artificial intelligence and machine/deep learning, the prospective benefits of the DERs can fully be explored. But both the short term power load of the individual energy customer and the power generated through DERs is dependent on the weather conditions and seasonality. In this paper, our focus is on forecasting the short term power load of the end energy customer using a deep learning framework. The proposed deep learning framework is based on a pyramid architecture of convolutional neural network. We developed and trained/evaluated the model for forecasting the short term power load of the individual household customer based on a large database of energy data from Australia. Our analysis indicates that forecasting the individual household power load is highly unpredictable. More than 57% of the customers (40 out 0f 69) have more than twenty outliers in the daily energy consumptions (which means highly unpredictable power load). The results show that our pyramid-CNN based deep learning approach is successful in predicting the individual household power consumption.","PeriodicalId":331415,"journal":{"name":"2019 International Conference on Advances in the Emerging Computing Technologies (AECT)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"8","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2019 International Conference on Advances in the Emerging Computing Technologies (AECT)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/AECT47998.2020.9194153","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 8
Abstract
Due to the seamless benefits of the integration of Distributed Energy Resources (DERs) for the residential customers, the forecasting of the short term power load of individual household energy customer is becoming an essential task for the future operation and planning of the smart grids. Recently, different studies concluded that due to lack of fast connectivity and awareness, the energy customer were not able to exploit the benefits of the DERs to the full extent. Nevertheless, with the rapid advancement in connectivity, data analytics, internet of things, artificial intelligence and machine/deep learning, the prospective benefits of the DERs can fully be explored. But both the short term power load of the individual energy customer and the power generated through DERs is dependent on the weather conditions and seasonality. In this paper, our focus is on forecasting the short term power load of the end energy customer using a deep learning framework. The proposed deep learning framework is based on a pyramid architecture of convolutional neural network. We developed and trained/evaluated the model for forecasting the short term power load of the individual household customer based on a large database of energy data from Australia. Our analysis indicates that forecasting the individual household power load is highly unpredictable. More than 57% of the customers (40 out 0f 69) have more than twenty outliers in the daily energy consumptions (which means highly unpredictable power load). The results show that our pyramid-CNN based deep learning approach is successful in predicting the individual household power consumption.