Intelligent management of coal stockpiles using improved grey spontaneous combustion forecasting models

Gongzhuang Peng, Hongwei Wang*, Xiao Song, Heming Zhang

*Corresponding author for this work

    Research output: Contribution to journalArticlepeer-review

    Abstract

    Intelligent coal stockpiles management system is significant for the next-generation cleaner power plants. Prevention of spontaneous combustion is a key issue for such a system, both in economic and environmental terms. As many factors can influence the self heating process of coal such as moisture and ash in coal, temperature distribution and stockpiles’ shapes, the remaining ignition time is developed as an aggregative indicator to measure the tendencies of spontaneous coal combustion. Using this value, the grey models have been applied to forecast spontaneous combustion and their performances are good for systems with insufficient information. However, the forecasting accuracy of these models still needs to be improved. Therefore, the ABC-RGM(1,1) model is proposed in this work based on the rolling-GM(1,1) and the Artificial Bee Colony (ABC) optimization algorithm, which has been applied to the management system of a 4 × 600 MW power plant. The computational experiments show that the ABC-RGM(1,1) model achieves better performance than the other popular grey models and accuracy of forecast is greatly improved especially for short-term forecasts. Such an accurate model is highly important and useful for intelligent coal management systems which can improve decision making and reduce risk.

    Original languageEnglish
    Pages (from-to)269-279
    Number of pages11
    JournalEnergy
    Volume132
    Early online date21 May 2017
    DOIs
    Publication statusPublished - 1 Aug 2017

    Keywords

    • Coal management
    • Grey model
    • Optimization algorithm
    • Spontaneous combustion prevention

    Fingerprint

    Dive into the research topics of 'Intelligent management of coal stockpiles using improved grey spontaneous combustion forecasting models'. Together they form a unique fingerprint.

    Cite this