:::

臺灣能源期刊論文全文

臺灣能源期刊第5卷第4期內容

出刊日期:December, 2018

題目
動態迴歸模型於短期區域電力負載預測之研究
Title
A Study of Short-term Regional Power Load Forecasting By Dynamic Regression Models
作者
李宜馨、陳彥銘、李明峯 
Authors
Yi-Hsin Li, Yen-Ming Chen, Ming-Feng Lee
摘要
本研究分析臺灣本島負載與溫度變化的相關性,並將臺灣本島分成北部、中部、南部及東部等四個區域分別做負載預測,以兼顧區域的負載特性和氣象因素。本研究採用迴歸搭配時間序列的多方程方式進行各區域短期負載預測,依地區分別建立小時別負載預測模型,允許迴歸模型搭配動態誤差的架構,使其能夠從負載形狀的短暫變化中相對快速地恢復。以實際負載資料驗證模型的可行性,各區域的整體平均絕對誤差百分比(Mean Absolute Percentage Error, MAPE)都在3%以下,是屬於高準確的預測模型。透過電力負載預測,期望提供未來24小時的耗電量及用電趨勢,以確保合理電力調度與系統安全運轉。
關鍵字
短期負載預測,時間序列,迴歸模型
Abatract
The object of this research is to analyze correlations between electric load and temperature in Taiwan. In order to adapt to local weather and usage characteristics, regional load forecasting models of norther, middle, southern and eastern Taiwan are separately built and explained in this study. These short-term forecasting models integrate both regression and time series methods to predict regional hourly load profiles, and dynamic error adjustments are also applied for the purpose of quick recovery when sudden load changes happen. The models are tested by real data and verified as high accuracy with MAPE less than 3%. Through this study of electric load forecasting, the total usage and profile changes in future 24 hours could be obtained to improve gird security and reasonable unit scheduling.
Keywords
short-term load forecasting, time series, regression model