关键词: 短期电力负荷预测, 萤火虫算法, 选择性集成学习, 气象因子, 预测模型

Abstract: With the rapid development of the power industry, the power load prediction is becoming more and more important in recent years, and short-term load prediction plays an extremely important role in dispatching and market operation of the power system. Power load prediction can effectively improve the utilization of power generation equipment. The selective ensemble learning method based on Kappa statistic and the glowworm swarm optimization algorithm (GSO) to forecast short-term load is proposed. This proposed method firstly generates multiple learners by bootstrap sampling, and then use glowworm swarm optimization algorithm to select some learners with large differences and high accuracy to participate in the integration. Compared with a single learner, the accuracy of the proposed method is significantly improved. The daily average load curves of two enterprises in Wuhan from 2015 to 2016 are used as a case study to carry out load forecasting. Comparing with other forecasting methods, the prediction accuracy of the proposed method is proved to be higher.

Key words: short-term power load forecasting, glowworm swarm optimization algorithm, selective ensemble learning, meteorological factor, forecasting model