{"title":"基于外部关注LSTM和质量驱动损失函数的风电区间预测","authors":"Hao Quan, Wei Zhang, Tao Zhou","doi":"10.1109/ICCSI55536.2022.9970599","DOIUrl":null,"url":null,"abstract":"The uncertainties of the wind power forecasting should be qualified effectively, a higher quality prediction interval (PI) is able to provide more valuable forecasting information. In this paper, a new model based on external attention long short-term memory network (EA-LSTM) is proposed, and a simplified quality-driven loss function is used to train the models. A new evaluation index is designed to obtain the optimal parameters of the loss function. Two 15 minutes time resolution datasets are used and two benchmark models are compared in case studies. The results demonstrate that the PI coverage probabilities (PICPs) generated by the proposed model can always reach the PI nominal coverages (PINCs), and the high quality PIs can be obtained. The EA-LSTM is also proved to be effective.","PeriodicalId":421514,"journal":{"name":"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Wind Power Interval Prediction Based on External Attention LSTM and Quality-Driven Loss Function\",\"authors\":\"Hao Quan, Wei Zhang, Tao Zhou\",\"doi\":\"10.1109/ICCSI55536.2022.9970599\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The uncertainties of the wind power forecasting should be qualified effectively, a higher quality prediction interval (PI) is able to provide more valuable forecasting information. In this paper, a new model based on external attention long short-term memory network (EA-LSTM) is proposed, and a simplified quality-driven loss function is used to train the models. A new evaluation index is designed to obtain the optimal parameters of the loss function. Two 15 minutes time resolution datasets are used and two benchmark models are compared in case studies. The results demonstrate that the PI coverage probabilities (PICPs) generated by the proposed model can always reach the PI nominal coverages (PINCs), and the high quality PIs can be obtained. The EA-LSTM is also proved to be effective.\",\"PeriodicalId\":421514,\"journal\":{\"name\":\"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)\",\"volume\":\"22 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-11-18\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICCSI55536.2022.9970599\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCSI55536.2022.9970599","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Wind Power Interval Prediction Based on External Attention LSTM and Quality-Driven Loss Function
The uncertainties of the wind power forecasting should be qualified effectively, a higher quality prediction interval (PI) is able to provide more valuable forecasting information. In this paper, a new model based on external attention long short-term memory network (EA-LSTM) is proposed, and a simplified quality-driven loss function is used to train the models. A new evaluation index is designed to obtain the optimal parameters of the loss function. Two 15 minutes time resolution datasets are used and two benchmark models are compared in case studies. The results demonstrate that the PI coverage probabilities (PICPs) generated by the proposed model can always reach the PI nominal coverages (PINCs), and the high quality PIs can be obtained. The EA-LSTM is also proved to be effective.