類神經網路在電力負載需求預測上之應用

The Application of Neural Network for Electric Load Demand Forecasting

廖炳坤(Ping-Kuen Liaw)

指導教授 ： 鄭春生

元智大學/工程學院/工業工程與管理學系/碩士(1993年)

摘要

電力負載之預測對電力調度人員而言，是一項相當重要的課題。良好的短
期電力需求預測，使電力調度人員得以提早進行電力潮流分析，提高電力
輸送效率；對於降低電廠機組反應時間、提高電力品質也有很大的幫助。
類神經網路在電力系統上的應用極廣，例如電廠緊急狀況處置、問題診斷
、諧波辨識等等。本論文即是針對臺灣地區的用電狀況，以類神經網路建
立良好的短期電力需求預測模式，並期望對臺灣地區之短期負載需求預測
有較令人滿意的預測結果。預測模式中，根據過去的用電記錄、時間變數
及氣象變數預測隔日之尖峰用電、低谷用電及二十四小時的每小時用電量
。研究中所使用的資料為民國75年到民國79年各項電力需求及氣象記錄，
並以民國80及81年之資料來評估以類神經網路做短期電力需求預測之正確
性。研究結果顯示，以類神經網路做短期電力需求預測之絕對誤差百分比
平均小於2%，由此可知類神經網路對短期電力需求，確實能提供不錯的預
測效能。

關鍵字

類神經網路 ； 短期電力需求預測

並列摘要

Forecasting of electric load demand is quite an important
subject to load management personnel. Earlier electric flow
analysis conducted by accurate short-term load forecasting
increases the power transmission efficiency and reduces the
response time of power generators. Accurate short-term load
forecasting will be helpful for higher electric dispatching
quality. A wide variety of applications for neural network on
electric power system have been reported in the literature such
as generator emergency treatments,problem diagnosis, harmonic
wave identification, and so on. In this thesis, a neural
network model is constructed depending on the actual power
demand condition of Taiman area and is expected to be a valid
model for load demand forecasting. The proposed network is
designed to provide one-day ahead forecasting of the peak load,
valley load and 24 hours load demand based on the historical
load data and the time and the weather variables. Five years
historical data(1986-1990) was used to train the network in the
study and one and half a years data (1991-06.1992) was used to
demonstrate the effectiveness of the proposed network. The
result shows about 2% average or less mean absoluate percentage
error by the proposed model. The overall performances of the
developed networks indicates that it could be an effective
method to short-term load demand forecasting.

並列關鍵字

Neural Network ； Short-term Load Forecasting