梯度提升随机森林模型及其在日前出清电价预测中的应用

RANDOM FOREST MODEL WITH GRADIENT BOOSTING AND ITS APPLICATION TO FORECASTING ON DAY-AHEAD MARKET CLEARING PRICE

发电企业竞价上网需要对未来的电价进行预测,以指导其进行报价。针对电价预测问题,提出一种新的梯度提升随机森林模型,并应用于电价预测。该方法为集成学习方法,在随机森林模型的基础上应用梯度提升算法,有效结合Bagging与Boosting两种集成学习策略的优势,提高模型预测的准确性。将该模型应用于PJM日前市场的电价预测,结果表明,该模型优于传统的随机森林模型和提升树模型,能够在其基础上进一步提高电价预测的准确性。

The power generation enterprises need to forecast the future electricity price so as to guide the bidding. To solve the electricity price forecasting problem, we proposed the innovative random forest model with gradient boosting and applied it into the electricity price forecasting. The proposed model was an ensemble learning model, which applied gradient boosting algorithm on the basis of random forest model. It effectively combined the advantages of Bagging and Boosting strategies to improve the forecast accuracy. The proposed model was applied to the forecasting on PJM day-ahead market clearing price. The results indicate that the proposed model outperforms both the traditional random forest and gradient boosting decision tree （GBDT） model, and it can improve the price forecast accuracy.