摘要
准确的短期电价预测有助于电力市场各个参与者选择交易策略和估算效益,因此短期电价预测受到人们广泛关注。为了解决特殊样本带来的预测误差,应用模糊C-均值聚类算法进行相似日聚类,以与预测日相似的数据构建样本集。再采用高斯过程回归来建立短期电价预测模型,对短期实时电价进行预测,得到具有概率分布及对应置信水平的区间预测结果。最后,采用美国代顿电力市场的历史数据进行实例计算,证明了该方法可有效提高模型的预测精度,与BP神经网络相比预测效果更佳,可以向电力市场参与者提供更全面的信息。
Accurate short-term electricity price forecasting is helpful for various participants in power market to choose trading strategies and estimate benefits.Therefore,the short-term electricity price forecasting is widely concerned around the world.In order to solve the forecasting error caused by special samples,the paper uses fuzzy C-mean clustering algorithm to carry out similar day clustering,so as to build the sample set with the data of the same date as the predicted date.Then,Gaussian process regression is used to establish the short-term electricity price forecasting model,and the short-term real-time price is predicted to obtain the interval forecasting results with probability distribution and corresponding confidence level.By using the historical data from Dayton electric power market in the United States for example calculation,it is proved that this method can effectively improve the forecasting accuracy of the model,and the prediction is better than that of BP neural network,providing more comprehensive information to the electricity market participants.
作者
杨颖
杨少华
张燕
雷自强
刘达
YANG Ying;YANG Shaohua;ZHANG Yan;LEI Ziqiang;LIU Da(State Grid Ningxia Electric Power Co.,Ltd.,Yinchuan750001,China;Economy Research Institute, State Grid Ningxia Electric Power Co.,Ltd.,Yinchuan750002,China;State Grid Xi’an Power Supply Company, Xi’an 710032, China;Institute of Smart Energy, North China Electric Power University, Beijing 102206, China)
出处
《智慧电力》
北大核心
2018年第12期23-29,共7页
Smart Power
基金
国家社会科学基金重大项目(15ZDB165)~~
关键词
相似日聚类
区间预测
短期电价预测
电力市场
高斯过程回归
similar day clustering
interval forecasting
short-term electricity price forecasting
power market
Gaussian process regression
