CAJ | 학술논문

为了快速、准确预测具有随机性的电力负荷，引入经验模式分解和极限学习机组合负荷预测算法。首先，利用EMD将非平稳负荷序列分解成一系列相对平稳的分量，减少不同负荷影响因素间的相互影响；然后针对各分量的不同特性，利用ELM具有预测能力强、计算时间短、计算准确性高等特点建立不同的预测模型，分别预测各分量值；最后组合ELM预测的各分量值，得到最终预测结果。仿真算例表明， EMD和ELM组合预测方法较传统单一神经网络方法在短期负荷预测精度和运算时间方面具有其独特的优势。
위료쾌속、준학예측구유수궤성적전력부하，인입경험모식분해화겁한학습궤조합부하예측산법。수선，이용EMD장비평은부하서렬분해성일계렬상대평은적분량，감소불동부하영향인소간적상호영향；연후침대각분량적불동특성，이용ELM구유예측능력강、계산시간단、계산준학성고등특점건립불동적예측모형，분별예측각분량치；최후조합ELM예측적각분량치，득도최종예측결과。방진산례표명， EMD화ELM조합예측방법교전통단일신경망락방법재단기부하예측정도화운산시간방면구유기독특적우세。
A new algorithm for short term load forecasting based on empirical mode decomposition ( EMD) and extreme learning machine ( ELM) method is presented with the random power load. The advantages of this method include short calculation time and high forecasting accuracy. Firstly, the non-stationary data is decomposed into a series of stationary components by using EMD to reduce the influence between different affecting factors. Then, dif-ferent models were built to deal with each group of data by using ELM. Finally, these forecasting results of each component are combined with ELM to obtain final forecasting result. The practical examples for short term load forecasting of power system show that the ELM method outperforms the traditional single networks on both the fore-casting accuracy and the computing speed.