Long-term Monthly Energy and Peak Demand Forecasting Based on Sequential-XGBoost

Research output: Chapter in Book/Conference paperConference paperpeer-review

Abstract

Forecasting the long-term demand (LTDF) is crucial for efficiently planning power systems. Typically, utilities perform separate long-term projections for energy consumption and peak demand, using distinct forecasting frameworks, which can lead to increased complexities. To address this challenge, this paper introduces a comprehensive LTDF model that utilizes the eXtreme Gradient Boosting algorithm with various sequential configurations. Initially, the paper develops a monthly energy consumption forecasting model for one year ahead. This model takes into account external factors such as macro-economic and climatic conditions. To capture the feature of energy consumption profile, a multi-input multi-output sequential approach is employed. Subsequently, the forecasted energy consumption becomes an influencing factor in a multivariate model for predicting the monthly peak demand one year ahead. Motivated by the great fluctuation of the monthly peak demand, the paper adopts a hybrid direct-recursive sequential configuration to address these variations effectively. Through considering the information of forecasted energy, the LTDF model achieved improved forecasting accuracy for peak demand. To validate the effectiveness of this proposed model, data from the New South Wales (NSW) power network was utilized for testing purposes.

Original languageEnglish
Title of host publication2023 IEEE 5th International Conference on Power, Intelligent Computing and Systems, ICPICS 2023
PublisherIEEE, Institute of Electrical and Electronics Engineers
Pages157-162
Number of pages6
ISBN (Electronic)9798350333442
DOIs
Publication statusPublished - 2023
Event5th IEEE International Conference on Power, Intelligent Computing and Systems - Shenyang, China
Duration: 14 Jul 202316 Jul 2023
https://ieeexplore.ieee.org/xpl/conhome/10235327/proceeding

Publication series

Name2023 IEEE 5th International Conference on Power, Intelligent Computing and Systems, ICPICS 2023

Conference

Conference5th IEEE International Conference on Power, Intelligent Computing and Systems
Abbreviated titleICPICS 2023
Country/TerritoryChina
CityShenyang
Period14/07/2316/07/23
Internet address

Fingerprint

Dive into the research topics of 'Long-term Monthly Energy and Peak Demand Forecasting Based on Sequential-XGBoost'. Together they form a unique fingerprint.

Cite this