Combining Autoregressive and Neural Network for Electric Load Forecasting
碩士 === 國立東華大學 === 資訊工程學系 === 92 === The electric power is one of the important energy that is the foundation for the economic development of a country. For electric power business, accurate load forecasting plays an important role in economic scheduling of generating capacity, scheduling of fuel pur...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2004
|
Online Access: | http://ndltd.ncl.edu.tw/handle/35475046510599534880 |
id |
ndltd-TW-092NDHU5392046 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-092NDHU53920462016-06-17T04:16:18Z http://ndltd.ncl.edu.tw/handle/35475046510599534880 Combining Autoregressive and Neural Network for Electric Load Forecasting 結合自我迴歸與類神經網路應用於電力負載預測 Hsin-Ming Liu 劉興明 碩士 國立東華大學 資訊工程學系 92 The electric power is one of the important energy that is the foundation for the economic development of a country. For electric power business, accurate load forecasting plays an important role in economic scheduling of generating capacity, scheduling of fuel purchases, planning of energy transactions, and dispatching of generation units. Its effects include the proper schedule and planning of power generation, the reducing of the operation cost, the raising of power supply stability, and the avoidance of electricity restriction or wasting of resources. This thesis targets the problem of forecasting the peak load of the subsequent days in one week. It can provide the important referential indication for the operation of the power system. General prediction models for time series that considers exogenous variable influence are too complicated for power load forecasting because the load has highly dependable factors in temperature. Therefore, our research work is a hybrid approach that combines the autoregressive and neural network approaches to forecast the power load. Firstly, the proposed hybrid method adopts the autoregressive model to find out the regulation of the electric load, and then input the found load value as a referential value to the neural network with exogenous variables to catch the nonlinear relation between these exogenous variables and the peak load. This referential input, which provides good initial predicting values, also shortens the training time of neural networks. According to our experimental results, this hybrid method outperforms both autoregressive models and neural networks. Cheng-Chin Chiang 江政欽 2004 學位論文 ; thesis 72 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立東華大學 === 資訊工程學系 === 92 === The electric power is one of the important energy that is the foundation for the economic development of a country. For electric power business, accurate load forecasting plays an important role in economic scheduling of generating capacity, scheduling of fuel purchases, planning of energy transactions, and dispatching of generation units. Its effects include the proper schedule and planning of power generation, the reducing of the operation cost, the raising of power supply stability, and the avoidance of electricity restriction or wasting of resources. This thesis targets the problem of forecasting the peak load of the subsequent days in one week. It can provide the important referential indication for the operation of the power system.
General prediction models for time series that considers exogenous variable influence are too complicated for power load forecasting because the load has highly dependable factors in temperature. Therefore, our research work is a hybrid approach that combines the autoregressive and neural network approaches to forecast the power load. Firstly, the proposed hybrid method adopts the autoregressive model to find out the regulation of the electric load, and then input the found load value as a referential value to the neural network with exogenous variables to catch the nonlinear relation between these exogenous variables and the peak load. This referential input, which provides good initial predicting values, also shortens the training time of neural networks. According to our experimental results, this hybrid method outperforms both autoregressive models and neural networks.
|
author2 |
Cheng-Chin Chiang |
author_facet |
Cheng-Chin Chiang Hsin-Ming Liu 劉興明 |
author |
Hsin-Ming Liu 劉興明 |
spellingShingle |
Hsin-Ming Liu 劉興明 Combining Autoregressive and Neural Network for Electric Load Forecasting |
author_sort |
Hsin-Ming Liu |
title |
Combining Autoregressive and Neural Network for Electric Load Forecasting |
title_short |
Combining Autoregressive and Neural Network for Electric Load Forecasting |
title_full |
Combining Autoregressive and Neural Network for Electric Load Forecasting |
title_fullStr |
Combining Autoregressive and Neural Network for Electric Load Forecasting |
title_full_unstemmed |
Combining Autoregressive and Neural Network for Electric Load Forecasting |
title_sort |
combining autoregressive and neural network for electric load forecasting |
publishDate |
2004 |
url |
http://ndltd.ncl.edu.tw/handle/35475046510599534880 |
work_keys_str_mv |
AT hsinmingliu combiningautoregressiveandneuralnetworkforelectricloadforecasting AT liúxìngmíng combiningautoregressiveandneuralnetworkforelectricloadforecasting AT hsinmingliu jiéhézìwǒhuíguīyǔlèishénjīngwǎnglùyīngyòngyúdiànlìfùzàiyùcè AT liúxìngmíng jiéhézìwǒhuíguīyǔlèishénjīngwǎnglùyīngyòngyúdiànlìfùzàiyùcè |
_version_ |
1718307380355137536 |