基于深度随机森林算法的短期用户负荷预测——以金华地区为例

Short-Term User Load Forecasting Based on Deep Random Forest:Take Jinhua City as an Example

通过网络爬虫获取天气数据,并结合金华市用户负荷数据,采用深度随机森林算法对用户负荷进行短期预测。借助4种评价指标,通过对比支持向量回归算法、K近邻算、贝叶斯岭回归算法、随机森林算法以及多个深度神经网络算法,发现深度随机森林算法预测效果最佳,支持向量回归算法次之,而深度神经网络算法在该数据集上表现一般。

By crawling weather data and combining with user load data in Jinhua City,a deep random forest algorithm is introduced to implement short-term user load forecasting.With four evaluation indicators,by comparing the support vector regression algorithm,the K-nearest neighbor algorithm,the Bayesian ridge regression algorithm,the random forest algorithm,and several neural network algorithms,it is found that the deep random forest algorithm has the best performance,and followed by the support vector regression.However,the neural network algorithm performed mediocre on this dataset.